Advanced Glycation End-Products Contribute to Delayed Diabetic Corneal Epithelial Wound Healing via the TLR4 Signaling

Abstract

Purpose

Diabetic keratopathy (DK) is a significant ocular surface complication of diabetes. This study investigated the role of advanced glycation end-products (AGEs) in DK pathogenesis and the underlying mechanism.

Methods

Type I diabetes mellitus (T1DM) mouse model was established by intraperitoneal injection of streptozotocin to assess AGE accumulation and inflammatory factor expression in corneal tissues. A bone marrow–derived dendritic cell (BMDC) culture platform, RNA sequencing (RNA-seq), and pharmacological interventions were used to examine AGE-induced inflammatory response and signaling pathways. The impact of Toll-like receptor 4 (TLR4) signaling (blocked by TAK-242) and AGE formation (inhibited by pyridoxamine [PM]) on diabetic corneal epithelial wound healing (CEWH) was evaluated.

Results

Compared with untreated controls, AGE-challenged BMDCs exhibited an exacerbated inflammatory response and increased phosphorylation of nuclear factor kappa-B (NF-κB) p65 and interferon regulatory factor 3 (IRF3). TLR4 blockade with TAK-242 significantly attenuated this inflammation and reduced phosphorylation of p65 and IRF3. RNA-seq analysis revealed baseline low-grade inflammation and hyper-phosphorylation of p65 and IRF3 in advanced diabetic corneas with AGE accumulation. Topical TAK-242 treatment significantly reversed delayed CEWH and accelerated corneal nerve regeneration in T1DM mice. Similarly, inhibiting AGE formation with PM also expedited diabetic CEWH and nerve regeneration.

Conclusions

The AGE/TLR4 axis drives sustained corneal inflammation by activating p65 and IRF3, thereby promoting DK progression. These findings identify AGE/TLR4 signaling as a potential therapeutic target for DK.

As a complicated metabolic disorder, diabetes mellitus (DM) is estimated to affect 642 million individuals 20 to 79 years old by 2040.^1,2 DM-induced ocular tissue damage constitutes a leading cause of global blindness.³ Although diabetic retinopathy and cataract are the most well-known ophthalmic complications, about 47% to 64% of DM patients suffer from clinically relevant effects on the cornea with sight-threatening consequences, termed diabetic keratopathy (DK).^4–6 DK is primarily characterized by delayed corneal epithelial wound closure, neurotrophic keratopathy, corneal ulceration, and feeling of burning or dryness.⁵ Critically, the uncontrolled delay of corneal epithelial wound healing (CEWH) was strongly associated with corneal ulcers, microbial keratitis, perforation, and even the loss of vision.^7,8 Accumulating evidence implicates chronic inflammation, impaired neurotrophic innervation and limbal stem cells, and dysregulated growth factor signaling in the DK pathogenesis.⁹ Although chronic corneal inflammation is recognized as a key driver of DK progression,^4,10,11 the systemic inflammatory microenvironment within diabetic corneas and its underlying mechanisms remain poorly defined.

Innate immunity is a major contributor to chronic inflammation in both diabetes and its complications,¹² including cardiovascular disease, diabetic nephropathy,¹¹ and diabetic retinopathy.¹³ Single-cell RNA sequencing (scRNA-seq) studies have implicated both adaptive and innate immunity in DK, revealing cytotoxic T cells, pro-inflammatory M1-polarized macrophages, neutrophils with enhanced neutrophil extracellular trap formation capacity, and activated dendritic cells (DCs) exhibiting migratory and maturation phenotypes.¹⁴ Cell–cell communication analysis has further confirmed the hyper-activation and pro-inflammation of immune cells in diabetic corneas.¹⁴ Several sterile danger-associated molecular patterns (DAMPs) drive chronic inflammation in DM, including HMGB1, S100A8/9,¹⁵ and advanced glycation end products (AGEs).¹¹ Among these, AGEs, formed via nonenzymatic glycation, play pivotal roles in DM pathogenesis and complication progression by activating Toll-like receptor (TLR) and receptor for AGE (RAGE) signaling.¹⁶ Histopathological analysis demonstrated a significant AGE accumulation in diabetic corneas,^17,18 suggesting their contribution to the chronic inflammation and DK progression. Nevertheless, key mechanistic aspects of AGEs in DK etiology remain incompletely understood and warrant further investigation.

To elucidate the pathological role of AGEs in chronic inflammation and DK progression, we employed streptozotocin (STZ)-induced type I diabetes mellitus (T1DM) model and bone marrow–derived dendritic cell (BMDC) platforms. In this study, our findings demonstrated that the accumulated AGEs caused and sustained low-grade inflammation in diabetic corneas that was mechanistically associated with increased activation of TLR4/nuclear factor kappa-B (NF-κB) and interferon regulatory factor 3 (IRF3) signaling. Pharmacological inhibition of AGE generation or blockade of TLR4 signaling significantly accelerated corneal re-epithelialization and nerve regeneration, highlighting potential therapeutic strategies for DK.

Materials and Methods

Animals

Male C57BL/6 mice (6–8 weeks old) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China), and Sting1^gt/gt mice with C57BL/6J background (stock #017537) were obtained from The Jackson Laboratory (Bar Harbor, ME, USA).¹⁹ All mice were bred in specific pathogen-free facility at the Eye Institute of Shandong First Medical University. Animal experiments were conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Animal Ethics Committee of the Eye Institute of Shandong First Medical University (approval no. 202090-01).

Establishment of T1DM Model and Corneal Epithelial Debridement Model

The murine T1DM model and corneal abrasion model were established as previously described.¹¹ Briefly, STZ (50 mg/kg; Sigma-Aldrich, St. Louis, MO, USA) dissolved in the citrate–citric acid buffer was intraperitoneally injected into C57BL/6 mice for 5 consecutive days. The T1DM model was considered successfully induced when blood glucose levels exceeded 16.7 mmol/L. The mice with diabetic duration ≥ 5 months were used in subsequent experiments. The intact and untreated corneas from both nondiabetic and diabetic mice were harvested for RNA-seq (three corneas pooled per sample, and three samples per group).

After anesthetization using 0.6% sodium pentobarbital, corneal epithelial debridement (2.5 mm in diameter) was performed in diabetic mice using Algerbrush II corneal rust ring remover (Alger Co., Lago Vista, TX, USA). Corneal epithelial defects were dynamically monitored by applying 0.25% fluorescein sodium and examining with a BQ 900 slit lamp (Haag-Streit, Köniz, Switzerland). Residual epithelial defects areas were quantified using ImageJ (National Institutes of Health, Bethesda, MD, USA). The age-matched nondiabetic mice were used as controls.

Treatment

To determine the role of TLR4 signaling on diabetic CEWH, the TLR4 signaling inhibitor TAK-242 (6 µL, 0.375 mg/mL; MedChemExpress, Monmouth Junction, NJ, USA) was administrated via subconjunctival injection immediately following abrasion, as previously described.²⁰ Corneal epithelial defects were dynamically observed using fluorescein sodium. To examine the impact of AGEs on CEWH and nerve regeneration, diabetic mice received pyridoxamine (PM, 1g/L; MedChemExpress) in drinking water for 2 months to inhibit AGE production, based on our previously established protocol.¹¹ CEWH outcomes were evaluated by slit-lamp examination. For subsequent experiments, three corneas were pooled per sample

BMDC Generation and Treatment

Bone marrow was harvested from femurs and tibiae of wild-type (WT) and Sting1^gt/gt mice. Single-cell suspensions were prepared and incubated in RPMI 1640 media containing IL-4 (10 ng/mL; PeproTech, Cranbury, NJ, USA) and granulocyte-macrophage colony-stimulating factor (GM-CSF, 20 ng/mL; PeproTech) at 37°C in a humidified incubator with 5% CO₂. After 7-day induction, BMDCs were exposed to bovine serum albumin (BSA, 200 µg/mL; Sigma-Aldrich, St. Louis, MO, USA) and AGE–BSA (200 µg/mL; Abcam, Cambridge, UK) for 12 or 24 hours, respectively. TAK-242 (5 µM), FPS-ZM1 (a RAGE-specific inhibitor, 1 µM; MedChemExpress), or anti-RAGE antibody (a blocking antibody for RAGE, 30 µg/mL, R&D Systems, Minneapolis, MN, USA) was added to BMDCs at 1 hour prior to exposure to AGE–BSA. The BMDCs and corresponding supernatants with different treatments were harvested for the subsequent experiments.

Western Blot

The corneal tissues and BMDCs with different treatments were homogenized in radioimmunoprecipitation assay (RIPA) lysis buffer (Solarbio, Beijing, China) containing phosphatase inhibitors (Merck Millipore, Molsheim, France) and protease inhibitors (Solarbio). After quantification, the total protein was separated using 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) gels and transferred to polyvinylidene fluoride membrane (MilliporeSigma, Burlington, MA, USA). Following blocking with 5% BSA, the membranes were incubated with primary antibodies overnight at 4°C. After incubation with horseradish peroxidase–coupled secondary antibody (ZSGB-BIO, Beijing, China), the target proteins were visualized by enzyme-linked chemiluminescence (ECL; MilliporeSigma) using the ChemiDoc Touch Imaging System (Bio-Rad, Hercules, CA, USA). The primary antibodies are listed in Supplementary Table S1.

Real-Time PCR

Based on the manufacturer's instructions, the total RNA from corneal samples or BMDCs was isolated using an EASYspin Plus RNA kit (Aidlab, Beijing, China). After quantification and reverse transcription (Vazyme, Nanjing, China), SYBR Green Master Mix (QIAGEN, Hilden, Germany) was applied for real-time polymerase chain reaction (PCR) in the ABI PRISM 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). The data were normalized using the housekeeping gene Gapdh, and calculations were based on the comparative cycle threshold (CT) method (2^−ΔΔCT). The primers are listed in Supplementary Table S2.

Immunofluorescence

The eyeballs were embedded into optimal cutting temperature compound (Sakura, Tokyo, Japan) and cut into 7-µm frozen slides. The slides and BMDCs for the different groups were incubated with primary antibodies overnight at 4°C, following fixation with 4% paraformaldehyde and permeabilization with 0.1% Triton X-100. After reaction with fluorescein-conjugated secondary antibodies, the nucleus was stained with 4′,6-diamidino-2-phenylindole (DAPI) for labeling cells. The photographs were acquired using an Eclipse TE2000-U Inverted Microscope (Nikon, Tokyo, Japan). The primary antibodies are listed in Supplementary Table S1.

Corneal Whole-Mount Staining

The corneal whole-mount staining was performed as previously described.¹¹ Briefly, the enucleated eyeballs were incubated in Zamboni's fixative (Solarbio) for 1 hour after euthanasia, and corneal cups were blocked in PBS containing 2% BSA, 2% goat serum, and 0.1% Triton X-100 for 2 hours at room temperature. The corneas were then incubated with Alexa Fluor 488–conjugated neuronal class III β-tubulin mouse monoclonal antibody and anti-major histocompatibility complex class II (MHC II) antibody overnight at 4°C. Corneas stained with anti-MHC II antibody were subsequently incubated with fluorescein-conjugated secondary antibodies for 2 hours. After washing with PBS, the corneal cups were cut into six radial flaps and photographed using a LSM 880 microscope (ZEISS, Oberkochen, Germany). Detailed antibody information is provided in Supplementary Table S1. The density of nerve fiber was quantified using ImageJ, following previously established methods.²¹ Briefly, nerve terminal areas were measured using the ImageJ Analyze Particles tool. Sub-basal nerve plexuses within the observed fields were semi-automatically traced, and the total nerve area was calculated using the same software. All quantification was performed by an observer blinded to experimental group assignments.

Flow Cytometry

Single-cell suspensions from corneas were prepared following a previously described method.¹⁰ Briefly, the eyeballs were incubated overnight at 4°C with 15 mg/mL Dispase II to isolate the corneal epithelium. The epithelium was digested with trypsin, and the remaining corneal tissue was digested with collagenase A. The single cells were then incubated with PE Anti-Mouse CD11c antibody (BioLegend, San Diego, CA, USA) and PerCP/Cyanine5.5 Anti-Mouse I-A/I-E antibody (BioLegend, San Diego, CA, USA) at 4°C for 1 hour. After incubation and washing, flow cytometry (FC) analysis was performed using a CytoFLEX Flow Cytometer (Beckman Coulter, Brea, CA, USA) and data were analyzed using FlowJo 10.10.0 (Tree Star, Ashland, OR, USA). Detailed antibody information is listed in Supplementary Table S1.

Enzyme-Linked Immunosorbent Assay

Based on the manufacturer's instructions, the concentrations of IFN-β (252363; Abcam) and IL-1β (KE10003; Proteintech, Rosemont, IL, USA) in supernatants from BMDCs receiving different treatments were determined using commercial enzyme-linked immunosorbent assay (ELISA) kits.

RNA-Seq and Bioinformatics Analysis

Total RNA was extracted using TRIzol reagent (Takara, Kyoto, Japan) according to the manufacturer's protocol. RNA purity and concentration were assessed using a NanoPhotometer spectrophotometer (Implen GmbH, Munich, Germany). The sequencing libraries were then constructed using Oxford Nanopore Technologies (ONT; Oxford, UK) adaptor ligation with T4 DNA ligase (New England Biolabs, Ipswich, MA, USA) following the manufacturer's instructions. The final cDNA libraries were subjected to FLO-MIN109 flow cells (Oxford Nanopore Technologies) and sequenced on a PromethION platform (Oxford Nanopore Technologies) at Biomarker Technology Company (Beijing, China). Basecalled reads were processed using an ONT tool to generate full-length, non-chimeric (FLNC) transcript clusters. Consensus isoforms were derived through iterative polishing within each cluster by pinfish (https://github.com/nanoporetech/pipeline-pinfish-analysis). These consensus sequences were mapped to the mouse reference genome using Minimap2. Differential expression analysis was performed using the DESeq 1.10.1 package in R (R Foundation for Statistical Computing, Vienna, Austria), with genes exhibiting a false discovery rate (FDR) < 0.05 and |log₂(fold change)| ≥ 0.585 designated as differentially expressed genes (DEGs). Gene Ontology (GO) enrichment analysis was conducted using the GOseq R package with Wallenius non-central hyper-geometric distribution to correct for transcript length bias.²² Gene set enrichment analysis (GSEA) was performed using GSEA 3.0 with gene sets from MSigDB 6.1.²³ RNA-sequencing data have been deposited in National Center for Biotechnology Information (accession code PRJNA1305372, PRJNA1305646).

Statistical Analysis

The data are shown as mean ± standard deviation (SD), and each experiment was conducted at least three times. Comparisons between two groups were conducted with two-tailed Student's t-test using Prism 8.0.2 (GraphPad, Boston, MA, USA), and analysis with more than two groups were performed using one-way analysis of variance (ANOVA). P < 0.05 was considered statistically significant.

Results

Accumulation of AGEs Is Greatly Increased in Diabetic Corneas

AGEs, representing a set of heterogeneous molecules formed through nonenzymatic glycation reaction,²⁴ play central roles in the pathophysiology of diabetes and its complications, including diabetic nephropathy,²⁵ diabetic cardiomyopathy,²⁶ and diabetic retinopathy.²⁵ Immunofluorescence (IF) staining and western blot (WB) analysis revealed significantly greater AGE accumulation in diabetic corneas versus normal mice (Supplementary Fig. S1). These findings are consistent with previous observations in murine models and human diabetic corneas,^11,27 and they implicate AGEs in DK pathogenesis.

Exogenous AGEs Potentiate Inflammatory Response in BMDCs

Although chronic inflammation has been considered an important factor driving DK progression,⁹ its underlying mechanisms remain incompletely defined. scRNA-seq analysis revealed hyperactive and pro-inflammatory states in multiple immunocytes of diabetic corneas, including cytotoxic T cells, macrophages, DCs, and neutrophils.¹⁴ IF staining showed elevated expression of MHC II in the corneas of diabetic mice compared to normal controls (Supplementary Figs. S2A, S2B). FC analysis further revealed an increase in the number of CD11c⁺ DCs in diabetic corneas, along with enhanced activation status (MHC II⁺CD11c⁺ DCs) (Supplementary Fig. S2C). We therefore hypothesized that AGE-mediated activation of corneal resident immune cells sustains chronic inflammation, ultimately promoting DK. To better understand the role of AGEs in the corneal inflammatory response, we stimulated BMDCs with AGE–BSA and performed RNA-seq.

Principal component analysis (PCA) showed distinct transcriptional clustering between AGE–BSA- and BSA-treated groups (Fig. 1A). After a series of correction and filtering, we identified 456 DEGs in AGE–BSA-challenged BMDCs (AGE) versus BSA-stimulated controls (BSA), including 272 upregulated and 184 downregulated genes (Fig. 1B). GO analysis and GSEA indicated that the upregulated DEGs were mainly associated with immune response, inflammatory response, positive regulation of IL-12 production, response to interferon-gamma (IFN-γ), positive regulation of IL-10 production, and response to IFN-β (Figs. 1C, 1D). We validated the increased expression of several selected cytokines (Cxcl10, Ifit1, Ifnβ1, Il1β, Il6, and Il12b) at 12 hours and 24 hours post-AGE exposure, with higher expression at 12 hours (Fig. 1E). ELISA confirmed increased IL-1β and IFN-β secretion in AGE–BSA-treated BMDC supernatant compared to BSA-treated or untreated controls, with dramatic elevation at 12 hours (Fig. 1F).

Figure 1.

Exogeneous AGEs significantly enhance the inflammatory response in BMDCs. (A) PCA plot depicting expression profiles across AGE and BSA groups. (B) Volcano plot illustrating the number of DEGs in AGE-treated versus BSA-treated BMDCs. (C) Significantly enriched BPs among upregulated DEGs in AGE-treated versus BSA-treated BMDCs. (D) GSEA revealed enrichment of the immune response pathway in the upregulated DEGs from AGE-treated versus BSA-treated BMDCs. (E) Quantitative real-time PCR analysis of Cxcl10, Ifit1, Ifnβ1, Il1β, Il6, and Il12b levels in BMDCs under different treatments (n = 5/group). (F) ELISA measurement of IL-1β and IFN-β in the supernatant of BMDCs with different treatments (n = 6/group). (G) The phosphorylation of p65, TBK1, and IRF3 in the different groups assessed by WB (n = 4/group). Uncropped immunoblots are shown in Supplementary Figure S7. (H) Quantification of phosphorylated p65, TBK1, and IRF3 from G using ImageJ. AGE, the BMDCs challenged with AGE–BSA; BSA, the BMDCs exposed to BSA. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; ns, no significance.

Given the critical roles of NF-κB p65 and IRF3 in pro-inflammatory cytokines and type-I interferon (IFN-I) production,²⁸ we assessed their activation. When compared to BSA-treated or untreated controls, greater increases in phosphorylation of p65, TANK-binding kinase-1 (TBK1), and IRF3 were observed in AGE–BSA-treated BMDCs at 12 hours, but there was no significant difference at 24 hours (Figs. 1G, 1H). These data demonstrate that AGE-induced inflammation amplification involves the activation of p65 and IRF3.

AGE–BSA-Stimulated Pro-Inflammation in BMDCs Depends on TLR4 Signaling

Although cyclic GMP–AMP synthase (cGAS)/stimulator of interferon genes (STING) signaling regulates inflammation via NF-κB and IRF3 activation,^28,29 its role in AGE-mediated response was investigated using BMDCs from WT and Sting1^gt/gt mice. Intriguingly, AGE–BSA treatment induced comparable expression of pro-inflammatory cytokines and target genes of IRF3 downstream between WT and Sting1^gt/gt BMDCs (Supplementary Figs. S3A, S3B). Phosphorylation of p65, TBK1, and IRF3 similarly showed no apparent alteration between WT and Sting1^gt/gt BMDCs after AGE treatment (Supplementary Fig. S3C). These findings indicate that the AGE–BSA-driven inflammation in BMDCs was cGAS/STING independent.

TLR4 signaling transduction occurs in a MyD88-dependent and TIR domain–containing adaptor-inducing interferon-β (TRIF)-dependent manner (Fig. 2A), converging on NF-κB p65 and IRF3 activation.^30,31 TAK-242 could strongly bind to the intracellular domain of TLR4 and disrupt its interactions with downstream adaptor molecules, ultimately inhibiting TLR4-induced inflammation.^20,32,33 We therefore selected TAK-242 as a specific TLR4 inhibitor to examine the involvement of TLR4 signaling in AGE–BSA-stimulated inflammation. WB analysis revealed that the TAK-242 co-treatment significantly attenuated AGE–BSA-induced phosphorylation of p65 (MyD88 dependent) and IRF3 (MyD88 independent) in BMDCs (Fig. 2B). Consistently, TAK-242 co-treatment both downregulated the transcriptional level of MyD88-dependent genes (Il1β, Il12b) and TRIF-dependent genes (Cxcl10, Ifit1) (Fig. 2C). Moreover, we also observed increased expression of MyD88 and TRIF in AGE–BSA-stimulated BMDCs, but a reduction after co-treatment with TAK-242 (Figs. 2D, 2E).

Figure 2.

TLR4 signaling mediates AGE–BSA-induced inflammation in BMDCs. (A) Schematic illustrating TLR4 signaling pathways. (B) WB analysis of the phosphorylated p65 and IRF3 in BMDCs after 12 hours of treatment (n = 4/group). Uncropped immunoblots are shown in Supplementary Figure S8A. (C) Quantitative real-time PCR analysis of Il12b, Il1β, Cxcl10, and Ifit1 expression in BMDCs after different treatments (n = 3/group). (D) WB analysis of MyD88 and TRIF in BMDCs after different treatments for 12 hours (n = 4/group). Uncropped immunoblots are shown in Supplementary Figure S8B. (E) Quantification of MyD88 and TRIF in the different groups by ImageJ. Con, BMDCs without any treatment; AGE, AGE–BSA-stimulated BMDCs; AGE+TAK-242, BMDCs co-treated with AGE–BSA and TAK-242; AGE+FPS-ZM1, BMDCs stimulated with AGE–BSA in the presence of FPS-ZM1. *P < 0.05, **P < 0.01, ***P < 0.001.

Multiple evidence supports the involvement of AGE/RAGE signaling in promoting inflammatory response.³⁴ Therefore, we further examined the role of RAGE signaling in AGE–BSA-stimulated inflammation in BMDCs. Compared with the control group, AGE–BSA-stimulated BMDCs showed increased phosphorylation of p65 and IRF3 (Fig. 2B), but RAGE inhibition by FPS-ZM1 failed to suppress p65 and IRF3 phosphorylation (Fig. 2B). When compared with THE AGE group, TAK-242 treatment remarkably reduced the expression of Il12b, Il1β, Cxcl10, and Ifit1 in AGE–BSA-challenged BMDCs, yet FPS-ZM1 treatment also failed to decrease the expression of Il12b, Il1β, and Ifit1, showing only a modest reduction of Cxcl10 expression (Fig. 2C). Furthermore, a similar trend was shown by neutralization antibody against RAGE (Supplementary Fig. S4). These results confirmed RAGE-independent AGE–BSA-stimulated inflammation in BMDCs. Collectively, TLR4 signaling contributes to the AGE–BSA-induced inflammatory response in BMDCs.

Pharmacologic TLR4 Inhibition Attenuates AGE-Induced Pro-Inflammation in BMDCs

To comprehensively assess the effect of TAK-242 on AGE-induced inflammation, we performed RNA-seq. PCA revealed that samples of the Inhib (inhibited by TAK-242) groups were distinct from the AGE groups based on their transcriptional profiles (Fig. 3A). Volcano plot analysis showed 1510 DEGs in the Inhib group versus the AGE groups, which were comprised of 622 upregulated and 888 downregulated genes (Fig. 3B). GO analysis and GSEA demonstrated that the downregulated genes were predominately enriched in immune response, cellular response to IFN-β, cellular response to IFN-γ, activation of innate immune response, response to IFN-γ, and cellular response to IFN-α (Figs. 3C, 3D). Moreover, TAK-242 co-treatment significantly reduced expression of key pro-inflammatory cytokines (Ifnβ1, Il1β, Il12b, Cxcl10, Ifit1, and Il6) in AGE-challenged BMDCs (Fig. 3E). The significant reduction of IFN-β in supernatants was also detected after co-treatment of AGE–BSA and TAK-242 (Fig. 3F).

Figure 3.

Blocking TLR4 signaling suppresses AGE–BSA-induced inflammation in BMDCs. (A) PCA plot showing the expression profiles of the Inhib and AGE groups. (B) Volcano plot of DEGs in the Inhib versus AGE groups. (C) GSEA showing enriched immune responses in the downregulated DEGs. (D) The top enriched BPs in the downregulated DEGs (n = 3/group). (E) Quantitative real-time PCR analysis of the selected inflammatory genes (Cxcl10, Ifit1, Ifnβ1, Il1β, Il6, and Il12b) in BMDCs from the different groups (n = 6/group). (F) The level of IFN-β in supernatant of the different groups as determined by ELISA (n = 6/group). (G) WB analysis of the phosphorylated p65, TBK1, and IRF3 in BMDCs with different treatments (n = 4/group). Uncropped immunoblots are shown in Supplementary Figure S9. (H) The expression and location of phospho-IRF3 (red) and phospho-p65 (green) in BMDCs from the different groups as determined by IF. Nuclei were counterstained with DAPI (blue). Scale bars: 30 µm. AGE, BMDCs stimulated with AGE–BSA in the absence of TAK-242; Inhib, BMDCs stimulated with AGE–BSA in the presence of TAK-242. ***P < 0.001, ****P < 0.0001.

WB analysis showed that TAK-242 co-treatment decreased the phosphorylation of p65, TBK1, and IRF3 in AGE–BSA-challenged BMDCs (Fig. 3G). Additionally, IF staining demonstrated that AGE–BSA stimulation in BMDCs promoted greater nuclear location of p65 and IRF3, but TAK-242 co-treatment significantly reduced this nuclear accumulation (Fig. 3H), a finding that is consistent with the phosphorylation data. Taken together, AGE drives a pro-inflammatory response through TLR4-dependent activation of NF-κB and IRF3.

Advanced Diabetic Corneas Exhibit Low-Grade Inflammation

Although much evidence indicates that chronic inflammation is a well-established driver for DK,^5,35 the inflammatory landscape in diabetic corneal tissues remains incompletely characterized. To systemically profile low-grade inflammation, we performed comparative transcriptome analysis (RNA-seq) using corneas from STZ-induced T1DM mice. Volcano plots showed a total of 81 DEGs, including 72 upregulated genes and nine downregulated genes (Fig. 4A). GO analysis of the upregulated DEGs revealed significant enrichment in biological processes (BPs) contributing to corneal inflammation in DM conditions (Fig. 4B), including cellular responses to IFN-β, cellular responses to IFN-α, cellular responses to IFN-γ, activation of the innate immune response, and positive regulation of the IFN-γ signaling pathway (Fig. 4B). Additionally, the downregulated DEGs were functionally associated with cell communication, negative regulation of IL-1β production, positive regulation of wound healing, embryonic hindlimb morphogenesis, and regulation of fat cell differentiation (Fig. 4B). Heatmap visualization showed a cluster of DEGs closely related with the above BPs, with pronounced upregulation of interferon-inducible genes (e.g., Ifi211, Ifi204, Ifi209) and downregulation of repair-associated genes (Mylk and Duox1) (Fig. 4C). GSEA plots validated the increased immune response in diabetic corneas (Fig. 4D). We further confirmed elevated expression of Tnfa, Ifi204, Ifi205, Ifi209, and Il1f6 in diabetics versus controls (Fig. 4E). Notably, Il1f6 (encoding IL-36α), as a novel IL-1 family member implicated in delayed CEWH,^36,37 may contribute to impaired diabetic CEWH.

Figure 4.

Advanced diabetic corneas displayed the enhanced inflammatory response. (A) Volcano plot of DEGs in diabetic mice corneas versus normal corneas. (B) Representative top enriched BPs for the upregulated DEGs (upper) and downregulated DEGs (bottom). (C) Heatmap illustrating the representative DEGs in diabetic corneas (n = 3/group). (D) GSEA plot showing immune responses enriched in the upregulated DEGs (n = 3/group). (E) Quantitative real-time PCR validation of selected inflammatory genes (Tnfα, Ifi204, Ifi205, Ifi209, and Il1f6) in diabetic corneas (n = 6/group). (F) IF staining analysis of phospho-IRF3 (red) and phospho-p65 (green) in DM and normal corneal sections. Nuclei were counterstained with DAPI (blue). Scale bars: 50 µm. (G) WB analysis of phosphorylated p65 and IRF3 in the different groups (n = 5/group). Uncropped immunoblots are shown in Supplementary Figure S10A. (H) Relative levels of phosphorylated p65 and IRF3 as determined by ImageJ. DM-Cornea, corneas from T1DM mice; Nor-Cornea, corneas from age-matched mice. *P < 0.05, **P < 0.01, ***P < 0.001.

We also examined the activation of p65 and IRF3 in DM corneas. More constitutive phosphorylation of p65 and IRF3 in diabetic corneas than in normal corneas was observed by IF (Fig. 4F) and WB (Figs. 4G, 4H), indicating increased baseline activation of p65 and IRF3. Therefore, our findings demonstrate that the diabetic corneas showed low-grade inflammation characterized by interferon signature activation and constitutive activation of NF-κB and IRF3 signaling.

Chronic Inflammation in Diabetic Corneas Is Probably Attributed to Accumulated AGEs

To better dissect the AGE contribution to diabetic corneal inflammation, we investigated the shared transcriptomic signatures between AGE-treated BMDCs and diabetic corneas. Venn diagram analysis showed 32 DEGs common to both systems (Fig. 5A). The heatmap visualization displayed the expression patterns of the shared DEGs (Fig. 5B). GO analysis revealed that these shared upregulated DEGs were functionally enriched in BPs, including response to bacterium, cellular response to lipopolysaccharide, innate immune response, and cellular response to IFN-β (Fig. 5C). Critically, TAK-242 treatment reversed the expression trajectory of these shared inflammatory genes in AGE–BSA-stimulated BMDCs (Fig. 5D). Together with observed AGE accumulation in diabetic corneas, these findings established AGEs as possible instigators of chronic inflammation in diabetic corneas.

Figure 5.

Shared inflammatory transcriptome in diabetic corneas and AGE-stimulated BMDCs. (A) Venn diagrams show the numbers of common DEGs between AGE-treated BMDCs and diabetic corneal samples. (B) Heatmap of representative shared DEGs in AGE-treated BMDCs and diabetic corneal samples. (C) The top significantly enriched BPs among shared DEGs. (D) Heatmap illustrating the common DEGs in AGE-challenged BMDCs after TAK-242 treatment. DM-Cornea, corneas from T1DM mice; Nor-Cornea, corneas from age-matched mice; BSA, BMDCs exposed to BSA; AGE, BMDCs exposed to AGE–BSA in the absence of TAK-242; inhib, BMDCs exposed to AGE–BSA in the presence of TAK-242.

Targeting AGE/TLR4 Signaling Accelerates Diabetic CEWH and Nerve Regeneration

Subsequently, we evaluated the therapeutic effect of inhibiting AGE/TLR4 signaling on diabetic CEWH and nerve regeneration using a T1DM murine model (Fig. 6A). Compared to the age-matched normal controls, the diabetic mice exhibited significantly delayed CEWH post-debridement, with residual epithelial defects covering 40.79% ± 5.99% (vs. 15.40% ± 6.31%) at 24 hours and 5.91% ± 3.13% (vs. 0.14% ± 0.33%) at 48 hours (Figs. 6B, 6C). To our surprise, subconjunctival injection of TAK-242 beneficially accelerated CEWH compared with the untreated DM controls for epithelial defect areas at 24 hours (23.47% ± 7.03% vs. 40.79% ± 5.99%) and at 48 hours (0.64% ± 0.92% vs. 15.40% ± 6.31%) after abrasion, respectively (Figs. 6B, 6C). Through WB and IF, we observed enhanced phosphorylation of Akt and signal transducer and activator of transcription 3 (STAT3) in diabetic corneas after topical treatment with TAK-242 (Figs. 6D–F), both of which are reported to be involved in diabetic CEWH.³⁸ The KI67-positive cells in TAK-242–treated diabetic corneas were significantly elevated compared with diabetic corneas without treatment (Fig. 6F). As shown in Figures 6G and 6H, the nerve fiber density of diabetic mice in the corneal peripheral and central areas was pronouncedly improved after topical treatment with TAK-242 when compared with the untreated diabetic mice. Also, the negative controls under the antibody-free condition showed minimal fluorescence staining across whole corneas (Supplementary Fig. S5).

Figure 6.

Inhibition of AGE/TLR4 signaling promotes CEWH and nerve regeneration. (A) Schematic strategy depicting establishment of the murine T1DM model followed by corneal abrasion and treatment with negative control vehicle or TAK-242. (B) Representative slit-lamp images of corneal defects with different treatments at 0, 24, and 48 hours post-injury. (C) Quantification of epithelial defect area as in B determined by ImageJ (n = 6/group). (D) Western blot analysis of phosphorylated STAT3 and AKT in corneal tissues at 48 hours post-injury (n = 4/group). Uncropped immunoblots are shown in Supplementary Figure S10B. (E) Relative level of phosphorylated STAT3 and AKT in corneal tissues as determined by ImageJ. (F) Immunostaining of STAT3 (green) and KI67 (green) in corneas under different conditions. Nuclei were counterstained with DAPI (blue). Scale bars: 40 µm. (G) Representative images of regenerated corneal nerve fibers observed by β-tubulin III staining at 7 days post-injury. (Upper) Full cornea overview. Scale bar: 400 µm. (Bottom left) Central corneas. Scale bar: 200 µm. (Bottom right) Peripheral corneas. Scale bar: 200 µm. (H) Quantified nerve density in central and peripheral corneas as in G determined by ImageJ (n = 6/group). Nor, age-matched mice with corneal epithelial abrasion; DM, T1DM mice with corneal epithelial abrasion; DM+TAK-242, T1DM mice with corneal epithelial abrasion following TAK-242 treatment. *P < 0.05, **P < 0.01, ***P < 0.001; ns, no significance.

To further determine the effect of AGE/TLR4 signaling in diabetic CEWH, oral PM was applied to inhibit endogenous AGE formation for 2 months. After treatment with PM, accelerated CEWH was identified using fluorescein staining, which revealed healing rates compared to the untreated diabetic mice of 25.89% ± 3.40% versus 50.73% ± 9.14% and 0% ± 0% versus 13.01% ± 9.15% at 24 hours and 48 hours after abrasion, respectively (Supplementary Figs. S6A and S6B). Moreover, greater increases in KI67-positive cells in PM-treated diabetic corneas were observed than in untreated diabetic corneas (Supplementary Fig. S6C). When compared with the untreated diabetic corneas, PM-treated corneas showed reduced phosphorylation of p65 and IRF3 (Supplementary Figs. S6C, S6D), suggesting that inhibiting AGE formation significantly reduces the activation of p65 and IRF3. Collectively, these results demonstrate the beneficial role of targeting the AGE/TLR4 axis in diabetic CEWH and nerve regeneration.

Discussion

As a common diabetic complication in the ocular surface, DK is characterized by epitheliopathy, neuropathy, and endothelial dysfunction.^1,5,18 This condition could be exacerbated by ocular trauma and various eye surgeries (such as cataract or vitrectomy).^39–41 Impaired corneal wound healing in diabetic patients frequently leads to serious declines in visual quality or permanent vision loss, especially in cases of persistent epithelial defects.^5,6,11,42 The current therapeutic options for DK remain limited. Here, we demonstrated that AGE accumulation in diabetic corneal tissues drives low-grade inflammation through TLR4-mediated activation of NF-κB and IRF3. This inflammatory cascade results in delayed CEWH and nerve regeneration. Pharmacological TLR4 inhibition accelerates corneal re-epithelialization and nerve regeneration.

The pathogenesis of DK involves well-established mechanisms, including oxidative stress, impaired neurotrophic innervation, dysregulated growth factor signaling, chronic inflammation, and AGE accumulation.^11,17,39 However, the mechanism of diabetic corneal chronic inflammation remains poorly defined. AGEs, a range of compounds generated through nonenzymatic glycation of biomolecules, primarily exert pathological effects via RAGE signaling in diabetes and the subsequent complications.^16,43 The AGE/RAGE axis modulates oxidative stress and pro-inflammatory factor production,^44,45 impairing wound healing.⁴⁶ Pathological conditions such as hyperglycemia and oxidative stress further accelerate AGEformation.⁴⁷ Previous studies and our data have confirmed AGE deposition in diabetic corneas,^11,48 but their functional role in DK pathogenesis remains unclear. Here, we found that PM-mediated AGE inhibition significantly promoted CEWH, confirming the pathogenic role of AGE in DK development.

The innate immune system drives chronic inflammation in DM through multiple signaling pathways, including inflammasomes, TLR signaling, and RAGE signaling.^15,49,50 We found that AGE–BSA activates transcriptional factor NF-κB p65 and IRF3 via TLR4 signaling in BMDCs, independent of cGAS/STING and RAGE signaling. Furthermore, pharmacological inhibition of TLR4 signaling alleviated DK in our model. Notably, several studies have confirmed the pathogenic role of AGE/TLR4 signaling in other diabetic complications, such as diabetic nephropathy,⁵¹ diabetic atherosclerosis,⁵² and diabetic cardiomyopathy.⁵³ These findings align with our results and support the view that TLR4 signaling represents a key pathway in the pathogenesis of DK. Nevertheless, whether TLR4 is the primary pathway for AGE-mediated responses remains undetermined. Therefore, to identify the dominant receptors involved in AGE-induced inflammation and DK progression, genetic approaches are warranted, such as the use of Tlr4- and Rage-deficient mice.

The extensive interplay among AGEs, RAGEs, and TLR4 is well established. On one hand, several DAMPs, including AGEs, act as common ligands for both RAGE and TLR4 signaling, leading to NF-κB activation via the adaptor molecule MyD88.^54,55 On the other hand, AGE/RAGE signaling upregulates TLR4 expression, and together these receptors synergistically amplify inflammatory cascades.⁵⁶ Although we speculated that the differential response of AGE–BSA to TLR4 versus RAGE signaling in BMDCs might stem from their distinct affinities for AGE–BSA, we cannot exclude a potential role of RAGE signaling in DK progression due to the structural heterogeneity and diversity of AGE species present in DM scenarios,⁴³ as well as the cellular heterogeneity within corneas. Therefore, further studies are necessary to clarify the respective pathological roles of these receptors in DK and their potential crosstalk.

Although multiple studies have demonstrated NF-κB p65 activation via TLR4 signaling in diabetic conditions,^49,53 the role of IRF3 activation in diabetic wound healing has received little attention. In this study, we observed increased IRF3 phosphorylation in both diabetic corneal samples and AGE-treated BMDCs, alongside elevated expression of interferon-inducible (IFI) genes. This finding suggests the involvement of IFN signaling in DK. Previous studies highlighted the critical role of IFN signaling in wound repair, including cutaneous wound healing,⁵⁷ diabetic wound recovery,⁵⁸ and gastrointestinal injury repair.⁵⁹ Using Aire knockout mice, we demonstrated that blocking IFN-I/Janus kinase (JAK)/STAT1 signaling accelerated CEWH and suppressed squamous metaplasia.⁶⁰ Nevertheless, the precise role of IFN signaling in DK pathogenesis warrants further investigation.

Accumulating evidence confirms the presence of resident immunocytes in homeostatic corneas,⁶¹ including DCs, macrophages, and lymphocytes. Because DCs are potent antigen-presenting cells located in the corneal superficial layers,⁶² their depletion reduces corneal sensitivity and delays CEWH following injury.⁶³ Clinical confocal microscopy has revealed a positive correlation between corneal nerve damage and increased DC density during DK progression,^64,65 suggesting a potential causal link. Although reduced DCs in diabetic corneas have been reported to impair CNTF production and nerve regeneration,^5,66 broader evidence indicates that DC dysfunction plays a significant role in the pathogenesis of DM and its complications.^67–70 In this study, the observed NF-κB p65 and IRF3 phosphorylation in AGE-treated BMDCs and diabetic corneas implicates DCs in chronic corneal inflammation under DM conditions. However, in vitro experiments cannot fully replicate the resident corneal immune cells in a DM microenvironment. Future investigations are therefore necessary to elucidate the role of resident DCs in DK progression, including characterization of immunological properties and regulatory networks via scRNA-seq and other experimental methods, as well as mechanistic exploration using DC-targeted genetic approaches such as CD11c-DTR/GFP transgenic mice and mice with DC-specific Tlr4 depletion.

This study has several limitations. First, although the critical role of AGEs in driving inflammatory response is well established, comprehensive time–course analyses of cytokine production, transcription factor activation, and CEWH in diabetic mice are needed to define the optimal therapeutic window. Second, due to the structural heterogeneity of AGEs (such as CML- and MGO-derived AGEs),⁴³ further investigations are required to determine the inflammatory specificity of distinct AGE structures, their differential contributions to DK progression, and the underlying mechanisms involved. Third, although TAK-242 and PM effectively ameliorated diabetic corneal inflammation and promoted CEWH by inhibiting TLR4 signaling and AGE formation, further optimization of DK therapy is warranted, including testing TAK-242/PM combination treatment, investigating other AGE inhibitors, and evaluating additional in vivo therapeutic efficacy and mechanistic validation using pharmacological and genetic approaches.

In summary, our findings establish that the AGE/TLR4 signaling pathway is an important driver of corneal chronic inflammation and subsequent DK development (Fig. 7). We demonstrated that blocking TLR4 signaling accelerated CEWH and nerve regeneration, highlighting TLR4 signaling as a promising therapeutic target for DK.

Figure 7.

AGE/TLR4 axis drives chronic inflammation contributing to delayed diabetic CEWH. The deposition of AGEs in diabetic corneal tissues activates TLR4 signaling, triggering chronic inflammation that impairs CEWH. Pharmacological targeting of TLR4 signaling accelerates CEWH and nerve regeneration.