Identification of O-GlcNAcylation Modification in Diabetic Retinopathy and Crosstalk with Phosphorylation of STAT3 in Retina Vascular Endothelium Cells

Instances of diabetic retinopathy (DR) are increasing at an alarming rate, becoming a leading cause for visual impairment and blindness in diabetic patients worldwide [1, 2]. Visual loss primarily occurs due to increased permeability of retinal vessels (diabetic macular edema) [3] and retinal neovascularization (proliferative diabetic retinopathy) [4, 5]. As DR pathogenesis is multifactorial [6-8], precise molecular mechanisms therein are still not well understood. However, blood-retina barrier (BRB) impairment is a common pathological change, as hyperglycemia induces retina vessel endothelial cell loss, particularly in the early phases of the disease [9]. Protecting retina vessel endothelial cells during hyperglycemia is vital to reducing DR.

Previous studies have reported that altered O-GlcNAcylation is a complication of insulin resistance, leading to diabetic pathologies [10, 11] such as DR [12]. O-GlcNAcylation is one of the most common post-translational protein modifications, dynamically regulated by a single pair of enzymes: O-GlcNAc transferase (OGT) and O-GlcNAcase (OGA) [13]. OGT catalyzes O-GlcNAcylation, and OGA facilitates hydrolytic cleavage of the post-translational modification. O-GlcNAcylation participates in various biological functions including transcription, translation, protein degradation, cell cycle control, and apoptosis [13]. Previous reports have implicated O-GlcNAcylation in neurological degenerative diseases, and cardiovascular and cerebrovascular diseases, as well as type 2 diabetes [14]. Recent investigation of this modification in DR [15, 16], revealed its contribution to increased vascular endothelial growth factor (VEGF) expression [12], retinal neovascularization [17], reactive oxygen species (ROS) generation [18], and BRB impairment [19], mostly in the late stages of the disease. The effects and underlying mechanisms of O-GlcNAcylation on endothelial cells, however, remain poorly understood. In the present study, we focus on the underlying targets of O-GlcNAcylation, such as signal transducer and activator of transcription proteins 3 (STAT3), to investigate mechanisms and roles of O-GlcNAcylation in DR, particularly in the early stages of the disease.

STAT3 is an important participant in tumor angiogenesis [20] and the expression of anti-apoptotic proteins [21, 22] in DR [23]. STAT3 activation is also one of the main facilitators of BRB breakdown through increased VEGF expression [24] and downregulation of endothelial tight junction protein expression [25]. STAT3 proteins are targets of phosphorylation [24] focused on the sites of Tyr705 and Ser727 [26]. Tyr705 phosphorylation is mediated by the Janus kinase (JAK) family, while Ser727 phosphorylation is catalyzed by extracellular signal-regulated kinases (ERKs).

Previous reports have shown a negative relationship between p705STAT3 and p727STAT3 [18, 27]. Tyr705 phosphorylation is essential for STAT3 activation during cell apoptosis through the JAK-p705STAT3 pathway [28]. Whether STAT3 is modified by O-GlcNAcylation, and the effect of O-GlcNAcylation on STAT3 activity and retinal vascular endothelial cells, however, has yet to be defined. We are the first to investigate the relationship between O-GlcNAcylation and STAT3 phosphorylation in DR and its effect on retinal vascular endothelium cell apoptosis, which could provide new insights into DR disease mechanisms.

Reagents used are as follows: HIF1α antibody (Bethyl Laboratories, Inc., Montgomery, TX, USA); OGT antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA); CTD110.6 antibody, anti-STAT3, anti-pSTAT3, anti-p705STAT3, anti-p727STAT3, anti-pJAK2, anti-JAK2, α-Tubulin, and β-actin purchased from Cell Signaling Technology (Danvers, MA, USA); anti-von Willebrand Factor antibody (vWF) (Abcam, Cambridge, MA, USA; VWF, ab6994); Dulbecco’s modified Eagle’s medium (DMEM; high glucose medium, 4.5 g/L glucose, and low glucose medium, 1 g/L glucose) (Gibco, Carlsbad, CA, USA); 4’,6’-diamidino-2-phenylindole hydrochloride (DAPI), Thiamet G, Alloxan, AG490, Cucurbitacin I, and bovine serum albumin (BSA) (Sigma); plastic tissue culture flasks (Costar, Cambridge, MA, USA); and immobilon-NC transfer membrane (Millipore, Billerica, MA, USA). A Leica confocal laser scanning microscope (Leica Microsystems, Mannheim, Germany) was used for scanning.

Vitreous humor samples were collected from 10 patients with proliferative diabetic retinopathy (PDR) during initial pars plana vitrectomy. Control samples were obtained from 10 retinal detachment (RD) patients without diabetes. All patients included in this study signed an informed consent form before surgery. Undiluted vitreous samples (0.5 - 1.0 mL) were aspirated under standardized conditions, immediately transferred to sterile tubes, and stored at -80 °C prior to analysis

We used Sprague-Dawley (SD) rats as an animal model for diabetes. Age-matched, female SD rats (6–8 weeks) weighing between 200 to 250 g were obtained from Shanghai Laboratory Animal Center, CAS, in Shanghai, China. Animals were housed under standard animal care conditions (20 - 22˚C; humidity 40 -60%) with a 12 h/12 h light/dark cycle.

Prior to STZ injection, rats were weighed, and baseline blood glucose was measured via tail vein using the OneTouch Glucose Monitoring System (Johnson & Johnson, New Brunswick, NJ, USA). Rats were administered 50 mg/kg of streptozotocin (STZ) in 10 mM sodium citrate, intraperitoneally. Blood glucose was measured after 3 days, a blood glucose concentration of ≥ 16.7 mmol/L was considered to be diabetic. All procedures were approved by the Animal Ethics Committee of Tongji University, Shanghai, China. Eyes were enucleated and retina tissues immediately dissected; then kept at -80˚C for further analysis.

Primary bovine retinal vascular endothelium cells (BRVECs) were cultured as previously described [16]. Human retinal vascular endothelium cells (HRVECs) were purchased from the American Type Culture Collection (ATCC) (Manassas, VA, USA) and passaged as described previously [16]. These cells were confirmed by the cell-specific marker vWF. All experiments were carried out in accordance with the Association for Research in Vision and Ophthalmology Statement.

O-GlcNAcylation is the post-translational modification of serine (Ser) or threonine (Thr) residues manipulated by OGT and OGA. Total levels of O-GlcNAcylation were downregulated by 2.5 mM Alloxan and upregulated by 2.5 µM Thiamet G for 24 h.

HRVECs were treated with 80 µM AG490 for 24 h and 0.25 µM Cucurbitacin I for 2 h, then cells were lysed for Western blot (WB) analysis. AG490, a janus kinase 2 (JAK2) inhibitor, was used to reduce JAK2 activity. Cucurbitacin I (JSI-124) was used to reduce STAT3 phosphorylation both in p727STAT3 and p705STAT3.

Equal amounts (30 µg) of proteins from cell extracts and animal retina samples were separated on a 10% acrylamide gel and subsequently transferred onto nitro-cellulose (NC) membranes at 200 mA for 1.5 h. After blocking in 5% (w/v) BSA for 1 h at room temperature (RT), blots were incubated overnight at 4˚C with primary rabbit monoclonal antibody against anti-STAT3, anti-pSTAT3, anti-p705STAT3, anti-p727STAT3, anti-pJAK2, anti-JAK2, CTD110.6, α-Tubulin, and β-actin (all antibodies were used at a 1: 1000 dilution). Blots were washed with TBS-T (0.1% Tween-20 in TBS) 3 times prior to incubation with the secondary antibody for 1 h at RT. The membrane was washed in TBS-T buffer and visualized using Odyssey (LI-COR Biosciences, Lincoln, NE, USA). Densitometric analysis was performed using Image J software (version 1.43, Broken Symmetry Software, Bethesda, MD, USA). For each experiment, the measurements were repeated 3 times.

Immunoprecipitation of STAT3 was carried out using anti-STAT3 antibody conjugated to agarose beads. The collected lysate (equivalent to 500 μg total protein) was incubated with 1 μg of anti-STAT3 antibody conjugated to agarose beads overnight at 4°C with gentle shaking. After washing the resin three times with lysis buffer, the beads were incubated with 40 μL 1×SDS-PAGE loading buffer and then centrifuged at 2, 000g for 1 min to collect eluted antigen. The eluent was run on Tris-Glycin 10% gradient gels, and analyzed by WB as described above.

Adversely, immunoprecipitation of O-GlcNAcylation protein was carried out using the 10 μL wheat germ agglutinin (WGA) beads (Santa Cruz Biotechnology) overnight at 4°C with gentle shaking, then washed, centrifuged, and analyzed by WB as described above.

HRVECs were incubated with 2.5 µM Thiamet G for 24 h and 0.25 µM Cucurbitacin I for 2 h. Cells were suspended in Annexin V binding buffer (BD Biosciences) at a concentration of 1 x 10⁶ cells/mL. Annexin V-FITC (BD Biosciences) was then added, followed by incubation in a 100 µL cell suspension for 15 min. Propidium iodide (PI) was spiked into 400 µL Annexin V binding buffer and added to the cell suspension immediately, then analyzed on a FACScan flow cytometer (Becton-Dickinson, Franklin Lakes, NJ, USA).

Experiments were repeated at least 3 times. Quantitative results were expressed as mean ± SEM. ANOVA and t-tests were used for statistical analysis, with p < 0.05 considered significant.

All procedures used in this study were approved by the Medical Ethics Committee of the Shanghai Tenth People’s Hospital. Principles of human subject research and cell research were conducted in accordance with the Declaration of Helsinki. Informed consent was obtained from all patients.

We investigated the OGT expression in PDR and control vitreous samples (RD patients). Results showed increased OGT expression in PDR vitreous samples compared with controls (Fig. 1B, p < 0.01), indicating increased O-GlcNAcylation levels in PDR patients.

We detected expression of O-GlcNAcylation and pSTAT3 in diabetic rat retinal tissues and primary BRVECs in a high glucose environment. Increased O-GlcNAcylation and elevated pSTAT3 expression were detected in diabetic rat retinas (Fig. 2B, ¹_VS², p < 0.05; Fig. 2D, ¹_VS², p < 0.05). Similarly, increased O-GlcNAcylation and pSTAT3 expression were observed in primary BRVECs under high glucose conditions (Fig. 3B, ¹_VS², p < 0.05).

Next, we treated primary BRVECs with advanced glycation end products (AGEs) for 24 h to simulate a pathological condition, to investigate the effects thereof on O-GlcNAcylation and pSTAT3 expression. As shown in Fig. 3, AGEs significantly elevated O-GlcNAcylation and pSTAT3 levels (Fig. 3B, ³_VS², p < 0.05; Fig. 3D, ³_VS², p < 0.05). These results demonstrate elevated pSTAT3 expression consistent with augmented O-GlcNAcylation in high glucose combined with AGEs.

As AGEs might directly influence O-GlcNAcylation expression or endothelial cell apoptosis [29], we utilized Thiamet G, an OGA inhibitor, to increase O-GlcNAcylation levels in BRVECs, and Alloxan, an OGT inhibitor, to decrease O-GlcNAcylation levels, to investigate the effects of O-GlcNAcylation on pSTAT3 expression.

In addition to hyperglycemia, hypoxia caused by capillary abnormalities is another vital factor to DR [30]. We cultured primary BRVECs in high glucose combined with hypoxic environment (3% O₂). Our results showed that O-GlcNAcylation levels increased significantly in high glucose compared with normal glucose (Fig. 4B, ¹_VS², p < 0.05), and that the elevation was more pronounced when combined with hypoxia (Fig. 4B, ⁶_VS², p < 0.05), which could be manipulated by Thiamet G and Alloxan treatment.

To investigate the specific effects of STAT3 sites involved in DR under high glucose conditions, levels of different STAT3 sites were explored. High glucose increased p705STAT3 expression (Fig. 5C, ⁶_VS³, p < 0.05), but had no effect on p727STAT3 levels (Fig. 5B, ⁶_VS³, p > 0.05). Low oxygen alone did not affect p727STAT3 or p705STAT3 expression (Fig. 5B, ³_VS¹, p > 0.05; Fig. 5C, ³_VS¹, p > 0.05), but combined with high glucose was found to upregulate p705STAT3, although it did not significantly affect p727STAT3 (Fig. 5C, ⁴_VS¹, p < 0.01; Fig. 5B, ⁴_VS¹, p > 0.05). This led us to conclude that Tyr705 is the site sensitive to high glucose, suggesting that Tyr705 is a site of importance for DR.

Augmentation of O-GlcNAcylation by Thiamet G increased p705STAT3 expression under normal glucose (Fig. 5C, ²_VS¹p < 0.05), but this modification downregulated p705STAT3 expression in high glucose (Fig. 5C, ⁷_VS⁶, p < 0.01). Even the downregulated p705STAT3 levels, however, were still much higher than in a normal glucose environment (Fig. 5C, ⁷_VS³, p < 0.01); as such, O-GlcNAcylation was shown to balance p705STAT3 expression within relatively high levels under high glucose conditions.

On the other hand, p727STAT3 expression was depressed by O-GlcNAcylation under high glucose conditions. While augmented O-GlcNAcylation decreased p727STAT3 levels significantly (Fig. 5B, ⁷_VS⁶, p < 0.05), downregulated O-GlcNAcylation increased p727STAT3 expression (Fig. 5B, ⁵_VS⁴, p < 0.01). So while O-GlcNAcylation appeared to inhibit p727STAT3 expression, it regulated p705STAT3 expression within relatively high levels under high glucose or high glucose combined with hypoxia.

Tyr705 phosphorylation is mediated by JAK2, so we examined pJAK2 levels to investigate different mechanisms of O-GlcNAcylation on p705STAT3 expression. Compared with normal conditions, pJAK2 expression was elevated during hypoxia combined with high glucose (Fig. 6B, ⁷_vs³, p < 0.01). Augmented O-GlcNAcylation can upregulate pJAK2 expression under these same parameters (Fig. 6B, ²_vs³, p < 0.05). However, upregulated O-GlcNAcylation by Thiamet G or downregulated O-GlcNAcylation by Alloxan had no significant effects on pJAK2 expression during hypoxia alone (Fig. 6B, ⁴_vs⁶, p > 0.05; ⁵_vs⁶, p > 0.05). High glucose was thereby determined to be the parameter in which pJAK2 expression was affected by O-GlcNAcylation. Enhanced O-GlcNAcylation can still increase pJAK2 expression under high glucose conditions (Fig. 6B, ¹_vs³, p < 0.05), which is inconsistent with p705STAT3 expression patterns, so p705STAT3 expression may be influenced by other factors.

To delineate the correlations between O-GlcNAcylation and STAT3, immunoprecipitation of STAT3 and O-GlcNAcylation proteins was carried out by Protein A agarose beads and WGA beads. We detected STAT3 expression in precipitated O-GlcNAc-modified proteins, and increased STAT3 levels in high glucose coupled with augmented O-GlcNAc modification compared with normal glucose (Fig. 7B).

We also detected O-GlcNAcylation expression in precipitated STAT3 proteins by anti-CTD110.6 antibody (Fig. 7C). These co-immunoprecipitation results illustrate that STAT3 is modified both by O-GlcNAcylation and phosphorylation.

As Tyr705 is sensitive to high glucose, and could be balanced by O-GlcNAcylation at a certain level, we focused our further studies on Tyr705 changes. We used Cucurbitacin I to inhibit pSTAT3 expression to investigate the role of these Tyr705 sites.

HRVECs were treated with 20 μM Cucurbitacin I under high glucose for 1, 3, or 6 h, with DMSO as negative control. Cells were lysed and subjected to WB analysis using specific antibody against pSTAT3. Cucurbitacin I was shown to inhibit pSTAT3 and STAT3 expression in a time-dependent manner (Fig. 8C, ²_VS³_VS⁴_VS⁵, p < 0.01), while DMSO had no influence on pSTAT3 or STAT3 expression (Fig. 8D, ²_VS³_VS⁴, p < 0.05).

To investigate the effects of O-GlcNAcylation and pSTAT3 on HRVEC apoptosis, we used Annexin V and PI double staining to assess cell apoptotic states. Compared with normal glucose, high glucose increased the percentage of apoptotic cells (Fig. 9, ³_vs², p < 0.01), and augmented O-GlcNAcylation (adding Thiamet G) rescued HRVEC apoptosis (Fig. 9, ¹_vs², p < 0.05), indicating a protective role. PSTAT3 inhibition by Cucurbitacin I increased cell apoptosis (Fig. 9, ⁶_vs², p < 0.001); but this increase could be partially mitigated by enhanced O-GlcNAcylation (Fig. 9, ⁴_vs⁶, p < 0.01). These results suggest that augmented O-GlcNAcylation can reduce cell apoptosis through the p705STAT3 pathway.

Caspase-3 is the primary facilitator of programmed cell death, so we assayed cleaved caspase-3 expression to confirm the effects of O-GlcNAcylation and pSTAT3 on HRVEC apoptosis. 10 μM, 20 μM, and 30 μM Cucurbitacin I were independently added to high glucose levels for 3 h. Inhibition of STAT3 induced by Cucurbitacin I treatment downregulated pSTAT3 and STAT3 expression, but increased cleaved caspase-3 levels in a concentration-dependent manner (Fig. 10B, ¹_vs³_vs⁴_vs⁵, p < 0.01; Fig. 10C, ¹_vs³_vs⁴_vs⁵, p < 0.01). Increased cleaved caspase-3 expression induced by pSTAT3 inhibition was consistent with the flow cytometry results that O-GlcNAcylation affected HRVEC apoptosis through the p705STAT3 pathway.

VEGF is downstream of STAT3, and acts as the main subsite contributing to DR formation. We detected an increase of VEGFA expression in high glucose, but augmentation of O-GlcNAcylation downregulated VEGFA (Fig. 11B, ²_vs⁵,p < 0.05), although it was still higher than in normal glucose conditions (Fig. 11B, ²_vs⁶, p < 0.05). This result was consistent with previous data that p705STAT3 expression patterns are regulated by augmented O-GlcNAcylation under high glucose conditions.

OGT inhibition (using Alloxan or OGT siRNA) and pSTAT3 inhibition (using Cucurbitacin I) decreased VEGFA expression significantly (Fig. 11B, ⁴vs⁵, p < 0.05; ¹vs⁵, p < 0.05). These results suggest that augmented O-GlcNAcylation is able to maintain a certain level of VEGFA via the p705STAT3-VEGFA pathway.

Hyperglycemia is one of the most important risk factors for DR, but the specific molecular mechanisms involved remain poorly understood. Previous studies observed increased O-GlcNAc modification in diabetic complications [31-33], and our present study detected elevated O-GlcNAcylation in PDR vitreous samples, diabetic rat retinas, and primary RVECs in high glucose conditions. In further experiments, we purified O-GlcNAcylated proteins to explore potential targets and underlying mechanisms related to endothelial apoptosis.

STAT3 factors influence endothelial function in DR [34, 35]. Thus, we focused on the relationship between O-GlcNAcylation and STAT3 phosphorylation, and changes to specific STAT3 sites under different O-GlcNAcylation levels. Our results showed an increased expression of O-GlcNAcylation and pSTAT3 both in vivo and in vitro in high glucose alone and when combined with hypoxia. We used Thiamet G, an OGA inhibitor, and Alloxan, an OGT inhibitor, to alter O-GlcNAcylation levels and explore the effects of O-GlcNAcylation on STAT3 phosphorylation.

Our results also demonstrated increased p705STAT3 expression under high glucose conditions, with no significant effect on p727STAT3 expression, suggesting Tyr705 is the site sensitive to high glucose. O-GlcNAcylation negatively affected p727STAT3 expression and regulated p705STAT3 expression within certain levels. The results showed that augmented O-GlcNAcylation upregulated p705STAT3 in normal glucose, but downregulated p705STAT3 in high glucose, although levels remained higher than in normal glucose. In addition, a negative relationship between p705STAT3 and p727STAT3 has been reported previously [27]. While expression of p727STAT3 decreased, p705STAT3 levels increased; this also contributed to total STAT3 activity.

O-GlcNAcylation and phosphorylation are thought to be complementary processes, namely in the Yin and Yang theory [33]. However, the relationship between O-GlcNAcylation and phosphorylation is more complicated than simple “competitive inhibition”. There are at least four different dynamic interactions between O-GlcNAcylation and O-phosphorylation: first, they competitively modify the same sites, such as in the mER-beta protein [36]; second, they competitively modify adjacent sites, such as in the C/EBP beta protein [37]; third, they co-modify the same protein at same time, such as in IRS-1 factors [38]; and fourth, modification and competitive decoration can coexist simultaneously, such as in CaMKIV proteins [39]. To explore the correlations of O-GlcNAcylation and STAT3, we investigated the expression of STAT3 and O-GlcNAcylation in precipitated proteins using immunoprecipitation.

We observed STAT3 expression in O-GlcNAcylated proteins and O-GlcNAcylation in precipitated STAT3 proteins. Our results suggest that STAT3 is co-modified by phosphorylation and O-GlcNAcylation simultaneously. As O-GlcNAcylation negatively influences pSTAT3Ser727 expression, and regulated pSTAT3Tyr705 expression within relatively high levels, we supposed that phosphorylation and O-GlcNAcylation might co-modify the Tyr705 site and compete for the Ser727 site. The interaction between Ser727 and Tyr705 sites, and O-GlcNAcylation combined with phosphorylation, can produce a great deal of molecular diversity that plays an important role both in physiological and pathological conditions.

In early DR, HRVEC apoptosis mainly contributes to BRB breakdown [9]. We observed the protective effects of O-GlcNAcylation on HRVEC apoptosis under high glucose conditions in a previous study [18]. According the influence of STAT3 on cell apoptosis, we supposed that O-GlcNAcylation might rescue cell apoptosis through the STAT3 pathway. To examine the role of STAT3 on HRVEC apoptosis and the underlying mechanisms involved therein, we applied the phosphorylation inhibitor Cucurbitacin I to downregulate pSTAT3 expression (both at Tyr705 and Ser727), and OGA inhibitor and OGT inhibitor to regulate O-GlcNAcylation levels.

The present study showed increased HRVEC apoptosis in high glucose, which could be reversed by augmented O-GlcNAcylation. Conversely, pSTAT3 inhibition by Cucurbitacin I induced higher percentages of cell apoptosis and increased cleaved caspase-3 expression in a concentration-dependent manner. While pSTAT3 inhibition increased cell apoptosis, augmented O-GlcNAcylation partially counteracted this adverse effect, defining an anti-apoptotic role for O-GlcNAcylation on HRVEC related to the STAT3 pathway.

The Tyr705 site is the target of JAK2 [40, 41], and confers protective effects on cell activity by increasing DNA binding affinity [42]. Increased STAT3 activity can promote protein dimerization and translocation to regulate expression of critical genes such as cell survival factors [42-44], signaling pathways [45, 46], and MnSOD activity in mitochondria [47]. We determined JAK2 expression provided insight into the mechanisms of O-GlcNAcylation on p705STAT3 expression. Similarly, we detected elevated pJAK2 expression in high glucose conditions corresponding to p705STAT3 expression. Augmented O-GlcNAcylation increased pJAK2 expression in high glucose, however, which was not entirely consistent with p705STAT3 changes. We theorized that O-GlcNAcylation might affect p705STAT3 expression by changing phosphorylation levels directly rather than through the pJAK2 pathway.

VEGF is downstream of STAT3 [48] and its presence has been implicated in retinal macular edema, neovascularization, and vitreous hemorrhage. While VEGF has always been considered a risk factor in DR [7, 49], VEGF is also necessary for vascular endothelial cell survival under pressure [50] and human embryo retina development [51]. Our results showed that high glucose upregulated VEGFA expression, which could be partially mitigated by augmented O-GlcNAcylation consistent with p705STAT3 changes. In addition, pSTAT3 inhibition and decreased O-GlcNAcylation significantly downregulated VEGFA expression. Thus, a relatively high VEGFA regulated by O-GlcNAcylation might exert a protective effect on retinal endothelial cells via the p705STAT3-VEGFA pathway. Many factors can affect VEGF expression, however, particularly under high glucose conditions, and the exact effects of specific STAT3 sites on VEGF expression and cell apoptosis might be explored in the future using site mutation.

Our project detected a functional relationship between O-GlcNAcylation and STAT3 phosphorylation, demonstrating that STAT3 is the target of O-GlcNAcylation. Furthermore, our results illustrated Tyr705 is sensitive to high glucose. O-GlcNAcylation could regulate p705STAT3 expression within relatively high levels, and partially mitigated HRVEC apoptosis induced by pSTAT3 inhibition and high glucose. We propose that O-GlcNAcylation protects HRVECs through the p705STAT3-VEGF pathway. Although Tyr705 is more sensitive to high glucose, Ser727 can be significantly negatively regulated by O-GlcNAcylation, and we recommend that the function of Ser727 in diabetic complications be explored in future research.

This work was supported in part by National Natural Science Foundation of China (No. 81271029, No. 81700840) and the Foundation of Shanghai Municipal Commission of Health and Family Planning (03.02.16.017, Shanghai, China). We are grateful for the assistance of Dr. Bebee and Dr. Yingbo Shui of the Department of Ophthalmology and Visual Sciences at the Washington University School of Medicine (St. Louis, MO, USA).

The authors declare to have no competing financial interests.

C. Xu and G.-D. Liu contributed equally to this work.