Inhibition of miR-144-3p/FOXO1 Attenuates Diabetic Keratopathy Via Modulating Autophagy and Apoptosis

Abstract

Purpose: Diabetic keratopathy (DK) is a vision-threatening disease that occurs in people with diabetes. Mounting evidence indicates that microRNAs (miRNAs) are indispensable in nerve regeneration within DK. Herein, the role of miRNAs associated with DK, especially focusing on autophagy and apoptosis regulation, was investigated.

Methods: To identify differentially expressed miRNAs, we performed miRNA sequencing on trigeminal ganglion (TG) tissues derived from streptozotocin-induced type 1 diabetic mellitus (T1DM) and normal mice. MiR-144-3p was chosen for the subsequent experiments. To explore the regulatory role of miR-144-3p in DK, miRNA antagomir was utilized to inhibit miR-144-3p expression. Bioinformatic tools were used to predict the target genes of miR-144-3p, and a dual-luciferase reporter assay was then applied for validation. Autophagy and apoptosis activities were measured utilizing TUNEL staining, immunofluorescence staining, and Western blotting.

Results: Overall, 56 differentially expressed miRNAs were detected in diabetic versus control mice. In the diabetic mouse TG tissue, miR-144-3p expression was aberrantly enhanced, whereas decreasing its expression contributed to improved diabetic corneal re-epithelialization and nerve regeneration. Fork-head Box O1 (FOXO1) was validated as a target gene of miR-144-3p. Overexpression of FOXO1 could prevent both inadequate autophagy and excessive apoptosis in DK. Consistently, a specific miR-144-3p inhibition enhanced autophagy and prevented apoptosis in DK.

Conclusions: In this study, our research confirmed the target binding relationship between miR-144-3p and FOXO1. Inhibiting miR-144-3p might modulate autophagy and apoptosis, which could generate positive outcomes for corneal nerves via targeting FOXO1 in DK.

Diabetes mellitus (DM) is a metabolic disease featured by long-lasting hyperglycemia that can cause nerve damage in multiple organs. The Diabetes Atlas, launched in 2021, hints that over 500 million adults suffer from diabetes globally.¹ Among the various complications caused by DM, diabetic keratopathy (DK) is the leading ocular disorder.² Neurons originating from ophthalmic sections of the trigeminal nerve constitute most corneal nerve fibers.³ These corneal nerve fibers cooperate with epithelial cells to generate neuropeptides and neurotrophic factors and maintain the homeostasis of the cornea.^4,5 Clinically, DK is characterized by changes in tear film, delayed epithelial regeneration, attenuative sensitivity, and corneal ulceration, which might ultimately lead to sight-threatening consequences.⁶ Recent studies indicate the involvement of DNA methylation, microRNA (miRNA) action, and sympathetic nerve overactivation in the corneal epithelial wound healing process in DK.^3,7 Moreover, dysregulation of growth factor, neurotrophin, inflammation, and autophagy signaling pathways is considered to be related to the pathogenesis of DK.^8,9 However, even if there is a pressing need for effective therapeutic interventions to promote corneal nerve regeneration, the existing treatments have limited effects, and the pathogenic pathways involved in DK remain unclear.¹⁰ The necessity of elucidating the mechanism and searching for potential targets in DK served as the impetus for this work. 

Through microarray analysis and high-throughput sequencing, novel regulatory and biomarker molecules for various diseases could be identified effectively. In terms of neurodegenerative disorders, machine learning techniques and high-throughput sequencing have identified 90 independent risk-associated genetic variants in Parkinson's disease.¹¹ In addition, the matrix metalloproteinase-12 was shown to upregulate in diabetic peripheral neuropathy by microarray and protein alteration assessment in type 2 diabetic mice.¹² In mapping the anterior segment cell types of the human eye, single-nucleus RNA sequencing was combined with Mendelian inheritance patterns to reveal over 900 human ocular disease genes.¹³ Moreover, the microarray analysis was also utilized to identify 29 differentially expressed miRNAs (DEmiRNAs) in the human diabetic and normal corneas.¹⁴ 

Our previous work showed that some specific noncoding RNAs (ncRNAs) can regulate diabetic corneal disease in different ways.^15–17 Among them, miRNAs are evolutionarily conserved and act as regulators of various signaling pathways by inducing mRNA degradation or suppressing translational progress.¹⁸ For example, miR-34c and miR-146a were found to affect nerve recovery and inflammatory responses during diabetic corneal epithelial wound healing.^15,19 In addition, miR-223-5p was shown to contribute to corneal epithelial and nerve healing through modulating inflammation and proliferation under hyperglycemic conditions.²⁰ The long noncoding RNA Rik was proven to be a competitive endogenous RNA (ceRNA) that can bind to miR-181a-5p and improve diabetic corneal epithelial wound healing.²¹ These studies consistently report the presence of regulatory miRNAs in DK and the necessity of further study. 

In the present study, we combined next-generation sequencing and bioinformatic analyses of DEmiRNAs obtained from the trigeminal ganglion (TG) tissues of normal mice and diabetic mice to efficiently identify novel targets. After preliminary validation, the miR-144-3p/Fork-head Box O1 (FOXO1) axis was determined for further investigation. Our data reveal that inhibition of miR-144-3p might improve re-epithelialization and nerve regeneration in the diabetic cornea by promoting autophagy and suppressing apoptosis via FOXO1. 

Materials and Methods

Animals

Male C57BL/6 mice (6-8 weeks old) were purchased from SPF Biotechnology Co., Ltd. (Beijing, China), and resided in the Fujian Medical University Laboratory Animal Center. All animal studies were approved by the Biomedical Research Ethics Review Committee of Fujian Medical University (IACUC FJMU 2021-0454). Establishment of the type 1 diabetic mouse model was conducted by injecting streptozotocin (STZ, 50 mg/kg; Sigma-Aldrich, St. Louis, MO, USA) intraperitoneally for 5 successive days. Blood glucose levels in the tail veins and body weight were assessed. Mice with blood glucose levels exceeding 16.67 mmol/L after 16 weeks of STZ injection were identified as diabetic mice and used in the subsequent study. 

Corneal Sensitivity Measurement

The corneal sensitivity of unanesthetized mice was surveyed by a Cochet-Bonnet esthesiometer (Luneau Ophtalmologie, Chartres Cedex, France). Initially, measurement started from 6 cm (maximum extension length) nylon monofilament. When there was no positive response, it was decreased by 0.5 cm. The longest filament length that elicited blink response was considered as the corneal sensitivity threshold. 

Whole-Mount Staining of Corneal Nerve Fibers

Five days after corneal epithelial debridement, the eyeballs of the mice were gathered and fixed in Zamboni fixative. The wounded corneas were dissected from the eyeballs, including the limbus. PBS containing 2% goat serum, 0.1% Triton X-100, and 2% BSA was applied to block for an additional 2 hours after obtaining the corneas. Staining was carried out overnight by anti-β-tubulin III antibody (Merck Millipore, Darmstadt, Germany) in PBS at 4°C. The next day, radial cuts were made to divide each cornea into six petals and washed six times in PBS. The total amount of corneal nerve fibers was then inspected via a confocal microscope (Zeiss, Gottingen, Germany) and the area of β-tubulin III-positive staining area was calculated by ImageJ through a modified version of the method used in the previous study.^15,22 

MiRNA Sequencing and Data Processing

To perform miRNA sequencing, samples of TG tissues were collected from three diabetic and three age-matched normal mice. In brief, after removing the skin from the skull and the top of the skull, we made a cut between the forebrain and the olfactory bulbs and lifted the front and top portions of the brain carefully. The optic nerves and crania nerve V were exposed and then cut. We removed the rest of the brain from the skull and collected TG tissues gently for the following experiments. The miRNA sequencing and preliminary data analysis were performed by KangChen Bio-tech (Shanghai, China). For the following bioinformatics analysis and other experiments, a P < 0.05 and |log2-fold change ≥ 0.585 were considered as thresholds for DEmiRNAs. TargetScan, miRDB, miRWalk, Gene Ontology (GO), and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases were utilized for predicting target genes of DEmiRNAs and conducting functional enrichment analysis. The visualization of the results of bioinformatics analysis was presented by Cytoscape and Chiplot (https://www.chiplot.online/). 

Subconjunctival Injection of Adeno-Associated Virus and Antagomir

Subconjunctival injection is widely applied in clinical as well as ophthalmological experimental studies. The eyes of anesthetized mice received topical anesthetic with proparacaine hydrochloride prior to subconjunctival injection. Subconjunctival injection was performed using a microliter syringe. Mice were then subconjunctivally injected with 10 µL of miR-144-3p antagomir (20 µmol/L; RiboBio, Guangzhou, China) or miRNA antagomir negative control (NCT; 20 µmol/L; RiboBio, Guangzhou, China). For FOXO1 overexpression, an AAV vector for mouse FOXO1 expression and empty vector were provided by Hanbio Biotechnology (Shanghai, China). The mice were injected subconjunctivally with AAV- FOXO1 (20 µmol/L) or empty vector (20 µmol/L). Corneas and trigeminal ganglion tissues were collected for subsequent experiments when appropriate. 

Quantitative Real-Time PCR

We collected the TG tissues of mice and cut them into pieces. Total mRNA and miRNA were obtained by TRIzol reagent (Invitrogen, Waltham, MA, USA). Complementary deoxyribonucleic acids (cDNAs) were synthesized by Thermo Scientific Revertaid First Strand cDNA Synthesis kit (Waltham, MA, USA). Quantitative real-time PCR (qRT-PCR) was performed by 2 × PCR master mix (Arraystar, Rockville, MD, USA). The β-actin gene and the U6 gene were utilized as an internal control. 

Western Blot Analysis

We collected TG tissues from mice and lysed them with RIPA reagent in order to extract the proteins. Sample proteins were run on SDS-PAGE and transferred using PVDF membranes. Then, 5% skimmed milk was applied to the block. Primary antibodies include anti-FOXO1 (ab179450; Abcam, Waltham, MA, USA), anti-P62 (P0067; Sigma), anti-LC3B (ab192890; Abcam), anti-BCL2 (ab182858; Abcam), and β-actin (Sigma). Membranes were then incubated with corresponding antibodies overnight. The next day, we incubated membranes with the secondary antibodies horseradish peroxidase (HRP; ab6721; Abcam) for 1 hour after rinsing the membranes with TBS + Tween (TBST) thrice. Finally, enhanced chemiluminescence (ECL) solution (Millipore, St. Louis, MO, USA) was utilized to visualize the protein bands. 

Dual-Luciferase Activity Assay

HEK293T cells were proliferated to 50% to 70% confluence after 24 hours of culture. Cells were collected and then co-cultured for 6 hours with wild-type or mutant reporter plasmids and miR-144-3p mimic or negative control. Next, we replaced the transfection medium. After 24 hours, the cells were collected, and luciferase activity was quantified utilizing the Dual-Luciferase system E2920 (Promega, Madison, WI, USA) following standard procedures. Both transfection experiments and luciferase analysis experiments were performed three times. 

Immunofluorescence Staining

TG tissues of mice were fixed in optimum cutting temperature (OCT) compound (Tissue-Tek, Torrance, CA, USA). Seven-micrometer frozen sections were prepared. The samples were then fixed in 4% paraformaldehyde, permeabilized with 0.3% Triton X-100, blocked with 5% BSA, and incubated with primary antibody. The antibodies were as follows: anti-FOXO1 (ab179450; Abcam), anti-P62 (P0067; Sigma), anti-LC3B (ab192890; Abcam), and anti- BCL2 (ab182858; Abcam). Sections were then incubated with Alexa Fluor–conjugated secondary antibodies (Proteintech, Wuhan, China). Finally, after counterstaining with DAPI (Beyotime, China), the stained sections were visualized under a DMI8 inverted fluorescence microscope (Leica, Wetzlar, Germany). 

TUNEL

Frozen TG tissue sections of mice (7 µm) were prepared. DNA fragmentation, a characterization of apoptosis, was detected by a Colorimetric TUNEL Apoptosis Assay Kit (C1098; Beyotime, China). The samples were fixed in 4% paraformaldehyde and permeabilized with 0.3% Triton X-100. Endogenous peroxidase blocking buffer (P0100A; Beyotime) was used to inactivate endogenous peroxidase in samples for 20 minutes. Then, the samples were incubated with Biotin-dUTP Solution and stained by the Streptavidin-HRP. Hematoxylin staining solution was applied to observe the nuclei. The sections were visualized and acquired by a Leica DMI8 inverted fluorescence microscope (Leica, Germany). 

Statistical Analysis

Statistical analyses were performed by GraphPad Prism version 9.0 software (San Diego, CA, USA). Data are presented as the mean ± standard error (SEM). Each experiment mentioned above was conducted individually not less than three times. Significant differences between two groups were analyzed utilizing an unpaired two-tailed Student's t-test. One‐way ANOVA and multiple comparison analysis were applied for multiple groups. When P < 0.05, the difference was considered statistically significant. 

Results

Hyperglycemia Impairs Corneal Re-Epithelialization and Nerve Regeneration in Mice

The diabetic mice exhibited hyperglycemia and significantly lower body weight than the control mice (Figs. 1A, 1B). In order to evaluate corneal neurological function, we measured the corneal sensitivity of mice, and our results suggested that diabetic mice exhibited decreased corneal sensitivity (Fig. 1C). Then, the mouse corneal epithelium was debrided. Fluorescein staining indicated that the epithelial healing rate between the diabetic corneas and control corneas at 12, 24, 36, and 48 hours after scraping presented a significant difference (Figs. 1D, 1E), and the processed images that identify the positive areas were also presented (Supplementary Fig. S1A). Five days after the corneal debridement, the density of regenerated sub-basal corneal nerve fibers was analyzed, and diabetic mice exhibited significantly lower nerve regeneration (Figs. 1F, 1G). Together, our results confirmed that hyperglycemia impairs corneal re-epithelialization and nerve regeneration in mice. 

Figure 1.

 

Hyperglycemia impairs corneal re-epithelialization and nerve regeneration in mice. (A) Body weight, and (B) blood glucose of normal (Ctrl) and diabetic (DM) mice. (C) Corneal sensitivity measurement. (D, E) Corneal wound healing process stained with fluorescein observed by a slit lamp at 0, 12, 24, 36, and 48 hours after debridement. The residual epithelial defect percentage was calculated by ImageJ. (F, G) Corneal nerve fiber staining by β-tubulin III in normal and diabetic corneas. Comparison of corneal nerve density between the two groups. Data were analyzed by unpaired t-test (n = 6 per group). *P < 0.05, **P < 0.01, ****P < 0.0001.

Bioinformatics Analyses of Differentially Expressed miRNAs in Diabetic TG Tissues

We obtained the TG tissues of mice, and next-generation sequencing analysis was applied to determine the expression profile of miRNAs in control and diabetic mice. The hierarchical clustering analysis of DEmiRNAs is presented in a heatmap (Fig. 2A). Six hundred twelve (612) DEmiRNAs were detected. After further filtering, we identified 56 miRNAs with |log2 fold change ≥ 0.585 that were significantly dysregulated, including 39 upregulated miRNAs and 17 downregulated miRNAs (Fig. 2B). Moreover, the top five upregulated and downregulated miRNAs were detected by qRT-PCR, and the miRNA expression levels with significant differences were shown (Fig. 2C). To explore the potential function of DEmiRNAs, we predicted their target genes by miRWalk, TargetScan, and miRDB. We then performed functional enrichment and pathway analysis of these target genes (Figs. 2D, 2E). MiR-144-3p had the most significant expression level difference between normal and diabetic mice. Additionally, miR-144-3p has been linked to zinc-induced insulin resistance, delayed wound healing in diabetic foot ulcers, and retinal degeneration.^23–26 Thus, miRNA-144-3p was selected for further study. 

Figure 2.

 

Bioinformatics analyses of differentially expressed miRNAs in diabetic TG tissues. (A) Heatmap showing DEmiRNA expression between normal mice (Ctrl) and diabetic (DM) mice. (B) DEmiRNAs are exhibited by volcano plots. (C) Expression of miRNAs in Ctrl and DM mice was preliminarily validated by qRT-PCR (n = 6 per group). (D) KEGG and GO analyses of potential target genes of upregulated miRNAs. (E) KEGG and GO analyses of potential target genes of downregulated miRNAs. The biological processes (BP), cellular component (CC), and molecular functions (MF) of genes were revealed by GO analyses. *P < 0.05, **P < 0.01.

Inhibition of miR-144-3p Ameliorated Diabetic Corneal Re-Epithelial and Nerve Regeneration

We injected miR-144-3p antagomir or miRNA antagomir NTC subconjunctivally and built a corneal re-epithelial wound model to study the function of miR-144-3p during re-epithelial and nerve regeneration. The strategy of reducing miR-144-3p expression exhibited a protective effect by promoting the corneal epithelial wound healing rate in diabetic mice (Figs. 3A, 3B), the images identifying the positive regions were also shown (Supplementary Fig. S1B). Five days after epithelial debridement, corneas were collected for whole mount staining. The miR-144-3p antagomir-treated diabetic mice show a significantly higher density of corneal nerve fibers (Figs. 3C, 3D). Therefore, we considered miR-144-3p as a prospective regulatory candidate during healing and nerve regeneration procedures in diabetic corneal epithelial wounds. 

Figure 3.

 

Inhibition of miR-144-3p ameliorated diabetic corneal re-epithelial and nerve regeneration. (A, B) Corneal wound healing process of normal mice (Ctrl), diabetic mice (DM), diabetic mice treated with antagomir NTC (DM+ miRNA antagomir NTC), or miR-144-3p antagomir (DM+ miR-144-3p antagomir) by a slit lamp at 0, 12, 24, 36, and 48 hours after debridement was recorded by fluorescence staining assay (n = 3 per group). The residual epithelial defect percentage was calculated by ImageJ. (C) Representative images of the overall corneal nerve fiber staining by β-tubulin III in four groups. (D) Corneal nerve density of the four groups was compared. Data were analyzed by 1-way ANOVA. Multiple comparisons were then performed when there was significance. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

FOXO1 Is a Target Gene of miR-144-3p

MiRNAs can suppress protein translation by binding to the 3′UTR of miRNA response elements (MREs) in target genes.^27–29 Genes related to miR-144-3p were predicted by miRWalk, TargetScan, and miRDB (Figs. 4A, 4B). Among them, 16 genes were validated by qRT-PCR, and the mRNA expression of SIRT1, FOXO1, and PTEN was significantly downregulated in diabetic mice (Supplementary Fig. S2, Fig. 4C). Accumulating evidence suggests that FOXO1 can serve as an important prognostic marker and mediator of nerve injury in neurodegenerative diseases.^30,31 In addition, FOXO1 protein can regulate cell proliferation, differentiation, senescence, apoptosis, autophagy, and metabolism.³² Therefore, the miR-144-3p/FOXO1 axis was further investigated. By utilizing TargetScan, we found that mature miR-144-3p was quite conserved among different species, and the binding site of miR-144-3p was predicted (Figs. 4D, 4E). The dual-luciferase experiment further validated that transfection with miR-144-3p significantly decreased luciferase activity in HEK293T cells, whereas the mutant mimic failed to do so (Fig. 4F). Western blot results in TG tissues of diabetic mice showed that the FOXO1 expression was increased by the miR-144-3p antagomir significantly, which also identified that miR-144-3p targeted FOXO1 (Figs. 4G, 4H). 

Figure 4.

 

FOXO1 is a target of miR-144-3p. (A, B) Target genes of miR-144-3p were predicted. (C) The qRT-PCR preliminarily verified the expression of three mRNAs between diabetic mice and normal mice (n = 3 per group). (D) Mature miR-144-3p sequences within different species. (E) The binding site between miR-144-3p and the 3′UTR of FOXO1 was predicted, and the mutation site (red) was constructed. (F) Transfection with miR-144-3p apparently restrained luciferase activity in HEK293T cells (n = 3 per group). (G, H) The levels of FOXO1 in trigeminal ganglia (TG) tissues of diabetic mice were greatly elevated via miR-144-3p antagomir, as suggested by Western blot (WB) results (n = 3 per group). Data were analyzed by 1-way ANOVA. Multiple comparisons were then performed when there was significance. *P < 0.05, **P < 0.01, ***P < 0.001.

Diabetic Mice Displayed Defective Autophagy and Excessive Apoptosis in TG Tissues

Given its important function of regulating autophagy and apoptosis activities in several diseases, we explored the potential role of FOXO1 in DK through analyzing the gene expression levels associated with autophagy and apoptosis, including FOXO1, P62, LC3B, and BCL2.^32,33 Western blot results implied a significantly higher level of P62 protein expression, while the levels of FOXO1, LC3B, and BCL2 were downregulated in diabetic mice (Figs. 5A, 5B). A lack of autophagy tends to lead to the accumulation of P62, and the decreased LC3B expression level indicated that the formation of autophagosomes was inhibited.^34,35 Meanwhile, proteins of the BCL2 family were determined to serve as various cytotoxic signal transducers and hence have critical roles in cell homeostasis. In addition, the decreased expression of the anti-apoptosis protein BCL2 suggested the promotion of apoptosis in diabetes.³⁶ Similar patterns were observed through immunofluorescence staining (Figs. 5D–H). An elevation of TUNEL-positive cells also appeared in diabetic mouse TG tissues (Fig. 5I). Collectively, these results consistently indicated lower expression of FOXO1, impaired autophagy, and enhanced apoptosis in the TG tissue of diabetic mice. 

Figure 5.

 

Diabetic mice displayed defective autophagy and excessive apoptosis in TG tissues. (A, B) Protein levels of P62 were significantly higher, while protein levels of FOXO1, LC3B, and BCL2 were downregulated in DM mice than in Ctrl mice. (C) Histopathological observation of TG tissue and trigeminal ganglion neurons (black arrow). Scale bar = 200 µm, 100 µm, and 50 µm. (D–H) Immunofluorescence analyses of FOXO1 (cytoplasm), P62 (nucleus and cytoplasm), LC3B (cytoplasm), and BCL2 (cytoplasm) proteins in DM mice and Ctrl mice. Scale bar = 75 µm. (I) Representative images of TUNEL staining in DM mice and Ctrl mice (nuclei = blue-purple and TUNEL-positive = brown-yellow). Scale bar = 75 µm. Data were analyzed by 1-way ANOVA. Multiple comparisons were then performed when there was significance (n = 5 per group). *P < 0.05, **P < 0.01.

Overexpression of FOXO1 Promoted Autophagy and Reduced Apoptosis in Diabetic TG Tissues

To further characterize its function in DK, an AAV vector for FOXO1 was constructed to explore the role of FOXO1 in regulating autophagy and apoptosis. Western blot results indicated that FOXO1 was upregulated in the AAV-FOXO1 group after 5 days of AAV infection compared to the other 2 groups. We also found that the expression of P62 protein was considerably suppressed along with increased levels of LC3B and BCL2 (Figs. 6A, 6B). The immunofluorescence results aligned with this conclusion (Figs. 6C–G). TUNEL-positive cells in the TG tissue were decreased by AAV-FOXO1 treatment in diabetic mice (Fig. 6H). In summary, we believe that upregulation of FOXO1 positively regulated autophagy and constrained apoptosis in vivo. 

Figure 6.

 

Overexpression of FOXO1 promoted autophagy and reduced apoptosis in diabetic TG tissues. (A, B) Protein levels of P62 were markedly suppressed, whereas protein levels of FOXO1, LC3B, and BCL2 were increased in subconjunctivally AAV-FOXO1 injected diabetic mice (AAV-FOXO1) compared with diabetic mice (DM) and empty vector injected diabetic mice (AAV-NTC; n = 3 per group). (C–G) Immunofluorescence analyses of FOXO1 (cytoplasm), P62 (nucleus and cytoplasm), LC3B (cytoplasm), and BCL2 (cytoplasm) proteins in the 3 groups (n = 3 per group). (H) Representative images of TUNEL staining in the three groups (nuclei = blue-purple and TUNEL-positive = brown-yellow; n = 3 per group). Scale bar = 75 µm. Data were analyzed by 1-way ANOVA. Multiple comparisons were then performed when there was significance. *P < 0.05, **P < 0.01.

Suppressing miR-144-3p Promoted Autophagy and Reduced Apoptosis in Diabetic TG Tissues

Subsequently, the miR-144-3p antagomir was utilized to verify whether the positive impact of miR-144-3p suppression on diabetic corneal re-epithelialization and nerve regeneration was to regulate autophagy and apoptosis through FOXO1. Western blotting and immunofluorescence staining illustrated that the miR-144-3p antagomir downregulated P62 levels and significantly increased FOXO1, LC3B, and BCL2 levels in diabetic mice (Figs. 7A–G). The miR-144-3p antagomir also reduced the number of TUNEL-positive cells in diabetic TG tissues (Fig. 7H). When taken as a whole, our findings suggested that downregulating miR-144-3p might facilitate recovery from corneal epithelial and nerve injury by enhancing FOXO1 expression to affect autophagy and apoptosis. 

Figure 7.

 

Suppressing miR-144-3p promoted autophagy and reduced apoptosis in diabetic TG tissues. (A, B) The miR-144-3p antagomir decreased total soluble protein levels while increasing FOXO1, LC3B, and BCL2 protein levels (n = 3 per group). (C–G) Immunofluorescence analysis of FOXO1 (cytoplasm), P62 (nucleus and cytoplasm), LC3B (cytoplasm), and BCL2 (cytoplasm) proteins in normal mice (Ctrl), diabetic mice (DM), the DM subconjunctivally injected with miRNA antagomir NTC (DM+ miRNA antagomir NTC), and the DM subconjunctivally injected with miR-144-3p antagomir (DM+ miR-144-3p antagomir; n = 3 per group). Scale bar = 75 µm. (H) Representative images of TUNEL staining in four groups (nuclei = blue-purple and TUNEL-positive = brown-yellow; n = 3 per group). Scale bar = 50 µm. Data were analyzed by 1-way ANOVA. Multiple comparisons were then performed when there was significance. *P < 0.05, **P < 0.01.

Discussion

Corneal neurological integrity is vital for normal corneal wound healing after injury.³⁷ Numerous neuropathies have been detected in the diabetic epithelial area, and corneal nerve density declines early in DK.^8,38–40 It is well documented that miRNAs can exert modulatory functions in diabetic peripheral neuropathy mainly through downregulating mRNAs associated with inflammation, autophagy, apoptosis, and oxidant stress.^16,41 Therefore, we aimed to search for regulatory targets for DK and examine their role in relevant cellular activities. We hypothesized that miR-144-3p might have an impact on diabetic corneal healing and neurological recovering by regulating FOXO1. Herein, we observed that diabetic corneal re-epithelialization and nerve regeneration were quite effectively promoted after downregulating miR-144-3p expression. In addition, expression changes of P62, LC3B, and BCL2 protein levels hinted that regulating autophagy and apoptosis through miR-144-3p/FOXO1 could be essential for DK. The relationship between molecules above and DK is shown in Figure 8. Aleterations of these molecules in the corneal tissues are also detected (Supplementary Figs. S3, S4). 

Figure 8.

 

Schematic diagram of the function of mir-144-3p and related molecules. The graph summarizes the main molecules involved in this study and their roles.

Next-generation sequencing was applied to determine the DEmiRNAs between normoglycemic mice and hyperglycemic mice. KEGG analysis of target genes of DEmiRNAs revealed that they are associated with protein processing in the endoplasmic reticulum, GABAergic synapse, phospholipase D signaling pathway, synaptic vesicle cycle, FOXO signaling pathway, hepatitis C, and parathyroid hormone synthesis secretion. Next, through expression validation as well as a literature search, miR-144-3p was selected from 56 significantly DEmiRNAs between normal and diabetic mice for subsequent experiments. The expression level of miR-144-3p, confirmed by qRT-PCR, was radically higher in mice with DM. Previous studies suggested miR-144-3p might be linked to some diabetic complications and neurodegenerative diseases.^42,43 Recent studies clarified that inhibiting miR-144-3p enhanced autophagy in thyroid carcinoma and that miR-144-3p might participate in apoptosis during renal ischemia/reperfusion injury.^44,45 Additionally, the negative effects of exosomal miR-144-5p, which is derived from diabetic bone marrow-derived macrophages, on bone-forming potential and capacity for fracture recovery were effectively reversed by inhibiting miR-144-5p.⁴⁶ Our results showed that suppressing miR-144-3p improved the regeneration of corneal epithelium and nerve fibers considerably. We suggest abnormal miR-144-3p upregulation could be an essential factor in delayed corneal epithelial healing and insufficient corneal nerve regeneration in DK. 

Among the target genes of miR-144-3p, FOXO1 drew our attention due to various mechanisms the FOXO family involved in insulin metabolism, gluconeogenesis, and diabetic complications.^47–49 Dysregulation of FOXO protein family expression has been proven to participate in metabolic disorders, human longevity, and tumor suppression.⁵⁰ Further experiments also confirmed the significantly lower expression of FOXO1 in diabetic TG tissue and that miR-144-3p had the ability to modulate FOXO1 expression. FOXO1 is highly expressed in insulin-sensitive tissues, such as the pancreas, liver, muscle, and adipose tissue.⁵¹ In a diabetic cardiomyopathy mouse model, FOXO1 has been indicated to participate in left ventricular dysfunction and remodeling through inducing fibrosis, autophagy, and apoptosis.⁵² Studies show that FOXO1 could trigger autophagy and inflammation in diabetic retinopathy.^33,53 Moreover, upon encountering oxidative stress (OS), FOXO1 was reported to promote autophagy in the low OS group. In addition, upregulating transcription of proapoptotic proteins that resulted in promoting apoptosis in the high OS group.⁵⁴ 

Numerous studies have shown that the balance between apoptosis and autophagy is pivotal for biological and pathophysiological processes.^55,56 During autophagy, dysfunctional cellular components and aberrant proteins are cleared out. Meanwhile, apoptosis is an autonomous procedure that maintains normal growth and development of cells.^57–59 Many stimulative pathways elicit autophagy and apoptosis simultaneously.⁶⁰ In fact, a study showed that active autophagy in Schwann cells can be a powerful initiator of peripheral nerve injury repair.⁵⁶ However, pancreatic kallikrein could exert a protective effect on diabetic retinopathy by inhibiting apoptosis.⁶¹ Defective autophagy and excessive apoptosis have been observed in the development of DK.^17,62 Our previous studies targeting miRNAs or mRNAs associated with autophagy and apoptosis also showed their potential therapeutic value in DK.^16,17 In the present study, we revealed that inhibition of miR-144-3p and promotion of FOXO1 both enhanced the expression levels of autophagy proteins as well as anti-apoptotic proteins in diabetic mice. Combined with the improvement in the regeneration of corneal epithelium and nerve fibers by suppressing miR-144-3p mentioned earlier, these results confirmed that the miR-144-3p/FOXO1 axis could affect corneal nerve and epithelium recovery in DK by regulating autophagy and apoptosis. Further investigation is required to explore the therapeutic potential related to the miR-144-3p/FOXO1 axis specifically or combined with other pivotal molecules. 

In conclusion, our findings identified FOXO1 as a direct downstream target of miR-144-3p and showed that FOXO1 can coordinate autophagy and intrinsic apoptosis. Inhibiting miR-144-3p may ameliorate the delayed corneal epithelial recovery and nerve regeneration in diabetic mice by upregulating FOXO1 expression, providing a promising strategy for corneal nerve protection. 

Acknowledgments

Supported by Joint Funds for the innovation of science and Technology, Fujian province (grant 2020Y9060). The author(s) have no proprietary or commercial interest in any materials discussed in this article. 

Disclosure: S. Wei, None; D. Liao, None; J. Hu, None