DPP-4 Inhibitors Attenuate Fibrosis After Glaucoma Filtering Surgery by Suppressing the TGF-β/Smad Signaling Pathway

Abstract

Purpose: This study investigated the effect of dipeptidyl peptidase-4 inhibitors (DPP-4is) on fibrosis after glaucoma filtering surgery with clinical data and an in vitro model that used transforming growth factor-β (TGF-β) to induce human Tenon's fibroblast (HTF) fibrosis.

Methods: The medical records of 41 eyes of 35 patients with diabetes with neovascular glaucoma (NVG) who received initial trabeculectomy were retrospectively reviewed. The surgical success rate was compared between cases that received (n = 23) and did not receive (n = 18) DPP-4i treatment for diabetes. The antifibrotic effects of linagliptin (a DPP-4i) were evaluated with quantitative real-time PCR for fibrosis markers (α-smooth muscle actin, collagen Iα, and fibronectin), a scratch assay, and a collagen gel contraction assay of primary cultured HTFs treated with TGF-β1 and linagliptin. Western blotting analysis was performed to evaluate the levels of phosphorylated Smad2 and Smad3 in the presence of linagliptin.

Results: The Kaplan-Meier curve for bleb survival was higher in patients who received DPP-4is (P = 0.017, log-rank test). The in vitro experiments demonstrated that treatment with linagliptin attenuated the elevated levels of fibrosis markers induced by TGF-β1 in HTFs. Linagliptin treatment also prevented the migration and gel contraction of HTFs. Linagliptin inhibited the phosphorylation of Smad2 and Smad3, which is the canonical pathway of TGF-β signaling.

Conclusions: The current study indicates the potential effect of DPP-4is for maintaining bleb function after glaucoma filtering surgery in patients with diabetes with NVG. Our results demonstrate that linagliptin attenuates fibrotic change in HTFs by inhibiting TGF-β/Smad signaling.

Glaucoma is one of the leading causes of blindness worldwide.¹ Elevated intraocular pressure (IOP) results in visual field defects and the loss of retinal ganglion cells.^2,3 Trabeculectomy is a common glaucoma surgery for medically uncontrolled IOP and is still in use after the emergence of minimally invasive alternatives.⁴ Optimal trabeculectomy outcomes require preventing postoperative scarring-over of the bleb. Mitomycin C (MMC), which inhibits DNA synthesis, is a common adjuvant for this purpose, but MMC does not sufficiently increase the surgical success rate.^5,6 Furthermore, MMC-associated complications, such as hypotony maculopathy and a leaky bleb that is susceptible to endophthalmitis, also result in surgical failure.^7–9 Safer, more broadly targeted adjunctive antifibrotic agents would help preserve the filtering bleb and maintain good IOP control. 

Trabeculectomy is especially common in neovascular glaucoma (NVG), a refractory form of glaucoma that occasionally arises as a complication of proliferative diabetic retinopathy (PDR).^10,11 Various types of antidiabetic drugs are currently available, including dipeptidyl peptidase-4 (DPP4) inhibitors (DPP-4is), a class of oral hypoglycemic agents used to treat type 2 diabetes.¹² DPP-4is prevent destruction of the hormone incretin, which helps the body regulate insulin.¹³ DPP-4is have a lower risk of causing hypoglycemia than drugs such as sulfonylurea, and interestingly, studies have shown that DPP-4 contributes to fibrotic change in the kidneys, liver, skin, lungs, and heart; DPP-4is ameliorate this effect.^14–20 DPP-4, also known as CD26, can be either a type II transmembrane protein or can be soluble. DPP-4 is ubiquitous on the surface of many cells, including fibroblasts, keratin-forming cells, epithelial cells, and endothelial cells.^21,22 The first DPP-4i, sitagliptin, was approved in 2006 in the United States and 2009 in Japan.^23,24 DPP-4is, such as sitagliptin, vildagliptin, saxagliptin, alogliptin, and linagliptin, are currently used for diabetes treatment.²⁵ We hypothesized that the antifibrotic effects of DPP-4is might improve trabeculectomy outcomes in NVG by preventing bleb scarring. 

We used retrospective clinical data analysis to investigate the effect of DPP-4is in maintaining bleb function after trabeculectomy for NVG in patients with diabetes. We also performed an in vitro study to determine whether linagliptin could attenuate fibrotic change in human Tenon's fibroblasts (HTFs) stimulated with transforming growth factor-β (TGF-β). 

Methods

Patients

We retrospectively reviewed the medical records of 41 eyes with NVG of 35 patients with diabetes who underwent initial trabeculectomy with MMC between May 2012 and June 2017 at Tohoku University Hospital. We collected postoperative follow-up data for 1 year. The trabeculectomy procedures were performed as previously reported.²⁶ Additionally, we recruited patients with primary open-angle glaucoma (POAG) to obtain Tenon's tissue samples during trabeculectomy for two purposes: to establish primary HTF cultures and evaluate the expression of DPP-4. Written informed consent was obtained from the subjects and the study was approved by the Institutional Review Board of Tohoku University Hospital (No. 2021-1-265 and No. 2022-1-475) and followed the tenets of the Declaration of Helsinki. 

Surgical Outcomes

Surgeries were defined as successful if IOP was maintained below 18 mm Hg without bleb needling, additional glaucoma surgery, or use of glaucoma medications for 1 year postoperatively. IOP measurements taken within 2 weeks after surgery were excluded from the analysis because IOP fluctuations are normal after trabeculectomy and may not necessarily indicate surgical failure. Clinical characteristics, including the use of DPP-4is, were compared between surgical success and failure groups. A Kaplan-Meier survival analysis of the probability of surgical success was also performed to compare the groups that did and did not use DPP-4is. 

Evaluation of Bleb Structure With Anterior Segment Optical Coherence Tomography

Images of the bleb were obtained with anterior-segment optical coherence tomography (AS-OCT; CASIA system, SS-1000; Tomey, Nagoya, Japan) 2 weeks postoperatively to evaluate bleb structure in the early postoperative stage, and the reflectivity of the bleb wall was measured in the images using a method that we previously reported.^27,28 Briefly, B-scan images were obtained of a slice that lay tangentially to the limbus in the horizontal direction. The reflectivity of the bleb in this slice was calculated as the average of the measured value in each pixel. 

Elastica-Masson Staining and DPP-4 Immunostaining of Tenon's Tissue Obtained From Patients With POAG

Tenon's tissue samples were obtained during trabeculectomy from 6 patients with POAG who did not have diabetes and were not taking DPP-4is. They were aged between 56 and 82 years; 3 were patients undergoing initial trabeculectomy and 3 were patients undergoing trabeculectomy re-operation. The samples were fixed in 4% paraformaldehyde and embedded in paraffin. Collagen was detected in 2-µm sections with Elastica-Masson staining. DPP-4 immunostaining with rabbit anti-DPP-4 (#NB100-59021; Novus Biologicals) in 4-µm sections was performed according to the manufacturer's protocol. The sections were incubated overnight with rabbit anti-DPP-4 (1:200) followed by HRP-conjugated anti-rabbit IgG (1:2500; Sigma-Aldrich) for 1 hour. The sections were counterstained with hematoxylin. 

Culturing of Primary Human Tenon's Fibroblasts and Cell Treatment

Primary HTFs were established from subconjunctival Tenon's capsule samples that were explanted during trabeculectomy; the samples were obtained from patients with POAG who underwent initial trabeculectomy and had no past history of inflammatory eye disease and no surgery for at least 1 year prior to sampling. The explanted subconjunctival Tenon's capsule tissue was cultured in a medium containing Dulbecco's modified Eagle's medium (DMEM; #041-29775; Wako Purechemical Industries) supplemented with 10% fetal bovine serum (FBS; Thermo Fisher Scientific), penicillin, and streptomycin (#15140122; GIBCO BRL) at 100 U/mL and 100 µg/mL, respectively. The cultured cells were incubated at 37°C/5% CO₂ in a humidified incubator. HTFs at passage 4 or 5 were used in all experiments. To induce a fibrotic response, the cells were stimulated with 5 or 10 ng/mL recombinant human TGF-β1 (#240-B-002; R&D Systems) for a certain period of time for each assay. Before TGF-β1 stimulation, the cells were starved overnight in serum-free DMEM. TGF-β1 stimulation was performed with or without a DPP-4i (linagliptin; #AG-CR1-3618; Adipogen). 

Quantitative Real-Time PCR

HTFs at a cell density of 2.0 × 10³ cells/100 µL of DMEM were plated in a 96-well plate and incubated for 3 days, followed by serum starvation overnight. Then, the cells were stimulated with 5 ng/mL of TGF-β1, either combined with 100 nM of linagliptin or excluding linagliptin, for 24 hours. RNA extraction and cDNA synthesis were performed using the SuperPrep cell lysis and RT Kit for qPCR (#SCQ-101; Toyobo) according to the manufacturer's protocol. Predesigned TaqMan primers and probes were used as follows: Acta2 (encodes α-smooth muscle actin [α-SMA] Hs00426835_g1), Col1a1 (Hs00164004_m1), FN1 (Hs01549976_m1), and Gapdh (Hs02758991_g1), which were all purchased from Thermo Fisher Scientific. Quantitative real-time PCR was performed with a 7500 Fast Real Time PCR System (Applied Biosystems), as previously reported.²⁹ Relative transcriptional levels were measured with the comparative Ct method and normalized with the corresponding Gapdh Ct values. 

Scratch Assay

HTFs were seeded in a 24-well plate with DMEM and allowed to attach and grow to more than 90% confluence. The cells were then scratched with a pipette tip and the growth medium was exchanged for a serum-free medium containing 10 ng/mL TGF-β1 with or without 100 nM linagliptin and incubated at 37°C for 12 hours. Phase-contrast images were obtained with a microscope (BZ-X800; Keyence Corporation). Then, the dimensions of the wound area surrounded by the migrating cell borders were measured in software. The data were expressed as the percentage change in the dimensions of the wound area at 12 hours compared to the baseline wound area at 0 hours. 

Collagen Gel Contraction Assay

A collagen gel contraction assay was performed using the Collagen Gel Culturing Kit (#638-00781; Fujifilm Wako Chemicals, Osaka, Japan) according to the manufacturer's protocol. Briefly, type I collagen, 10 × MEM, reconstitution buffer, and cultured HTFs (2.2 × 10⁶ cells/mL) were mixed at a ratio of 7:1:1:1. The mixture was prepared either with TGF-β1 (10 ng/mL), linagliptin (100 nM), or a combination of both. Each mixture (500 µL) was poured into the wells of a bovine serum albumin-coated 24-well plate and incubated at 37°C/5% CO₂ for 60 minutes to induce the formation of a collagen gel. Subsequently, the gels were released from the walls of the culture wells using a 23-gauge needle. The gel dimensions were measured every 24 hours for 7 days with a microscope and its software (BZ-X800; Keyence Corporation). The data were expressed as the percentage change in the dimensions of the gel at each time point compared to the baseline gel dimensions on day 0. 

Western Blotting

HTFs were seeded in a 6-cm culture dish and incubated overnight, followed by serum starvation for 24 hours. The cells were then treated with 5 ng/mL of TGF-β1, either combined or not combined with 100 nM of linagliptin, for 10 minutes. Subsequently, the cells were collected in 50 µL of PTTS buffer (0.5% Tx-100, 0.1% Tw-PBS, and 0.05% SDS in PBS)³⁰ containing 100 × Halt Protease Inhibitor Cocktail (Thermo Fisher Scientific) and a phosphatase inhibitor cocktail after PBS washing. Western blotting was performed as previously described.³¹ The primary antibodies were rabbit anti-phospho-Smad2/3 (1:2000, #MAB8935-SP) and goat anti-Smad2/3 (1:2000, #AF3797-SP), purchased from R&D Systems, and rabbit anti-β-actin (1:1000; Sigma-Aldrich). The protein expression levels were expressed as the ratio of phosphorylated Smad2/3 to Smad2/3. 

Statistical Analysis

Data analysis was performed using JMP Pro software, version 14 (SAS Institute, Japan). Values are shown as the mean with standard deviation. Univariable between-group analyses was performed with Fisher's exact test and the Mann-Whitney U test. A Kaplan-Meier survival analysis, followed by a log-rank test, was used to compare the probability of surgical success between the groups with and without DPP-4is. A multivariate Cox proportional hazards model was used to confirm the effect of DPP-4is on bleb survival. Comparisons among groups were analyzed with the Tukey-Kramer multiple comparisons test. P < 0.05 was considered statistically significant. 

Results

Patient Characteristics in the Surgical Success Group Versus the Surgical Failure Group

The clinical characteristics of the surgical success and failure groups after trabeculectomy for NVG are shown in Table 1. Among 41 cases that satisfied the inclusion criteria, 24 cases were successes and 17 cases were failures 1 year postoperatively. Among antidiabetic medications used before trabeculectomy, only DPP-4is had a significantly higher percentage of use in the success group (70.1%) than the failure group (35.3%, P = 0.03). Additionally, the following factors were statistically different in the success group: age was older (P < 0.01), pre-operative IOP was lower (P = 0.03), and the duration of diabetes treatment was longer (P = 0.02). The use of glaucoma eye drops before trabeculectomy was not statistically different, as shown in Supplementary Table S1. 

Table 1.

 

Clinical Characteristics of Surgical Success and Failure Groups Among NVG Cases

 

Comparison of Clinical Characteristics in Patients With and Without DPP-4is

The percentage of cases using DPP-4is was significantly higher in the success group than the failure group 1 year postoperatively, as shown above. Thus, we compared the clinical characteristics between cases with and without DPP-4is (Table 2). Age was higher in cases with DPP-4is, but this was not statistically significant (P = 0.053). The following factors were statistically different in the cases with DPP-4is compared to the cases without DPP-4is: a history of vitrectomy was less frequent (P < 0.008), and diabetes treatment with biguanide was more frequent (P = 0.03). 

Table 2.

 

Clinical Characteristics of Groups With or Without DPP-4is Among NVG Cases

 

Table 3.

 

Multivariate Cox Proportional Hazards Model to Determine the Risk Factors for Bleb Failure

 

Bleb Survival Analysis and a Bleb Structural Evaluation With AS-OCT Revealed That the Group That Used DPP4-is had Better Surgical Outcomes Than the Group That Did Not Use DPP4-is

Based on the finding of more frequent use of DPP-4is in the success group, we performed a survival analysis with Kaplan-Meier curves, which showed that estimated bleb survival rate (i.e. surgical success) was significantly higher in the patients that used DPP4-is (73.9%) versus those that did not (38.9%) 1 year after surgery (log-rank test: P = 0.017; Figure 1A). Then, we aimed to show that the use of DPP-4is was an independent contributor to surgical success with a Cox proportional hazards model. Although there were other potential factors that could have influenced surgical outcomes, as shown in Table 1, we started by including age and pre-operative IOP as variables, based on the results of a univariable analysis, as well as considering the relatively small sample size. Those factors (i.e. younger age and higher pre-operative IOP), can be prognostic for surgical failure according to previous reports.^32–34 This analysis revealed that the use of DPP-4is was an independent contributor to surgical success after adjustment for other potential prognostic factors (i.e. age and pre-operative IOP (relative risk [RR] = 0.29, P = 0.021; see Table 3)). Furthermore, we used a Cox proportional hazards model with anti-VEGF intravitreal injection as the fourth variable, as shown in Supplementary Table S2. This analysis also demonstrated that the use of DPP-4is contributed to surgical success (P = 0.008). The types and doses of DPP-4is are also shown in Supplementary Table S3; the doses of DPP-4 inhibitors used in the cases were within recommendations and were not excessively high. We also evaluated bleb structure with AS-OCT. Previous studies have shown that higher AS-OCT-measured reflectivity of the bleb wall, implying bleb scarring, was associated with higher postoperative IOP.^35,36 Moreover, we previously demonstrated that AS-OCT-measured reflectivity of the bleb wall 2 weeks postoperatively may predict bleb outcomes.²⁸ In the current study, AS-OCT-measured bleb reflectivity (measured in arbitrary units [AUs]) 2 weeks after trabeculectomy was significantly lower in the group that used DPP-4is than the group that did not use DPP-4is (118.3 ± 3.7 AU vs. 130.4 ± 4.1 AU, P = 0.033; Figure 1B). Representative AS-OCT images of blebs 2 weeks after surgery in cases that did or did not use DPP-4is are shown in Figure 1C. This analysis indicates that DPP-4is might improve trabeculectomy outcomes by preventing bleb scarring from an early postoperative stage. 

Figure 1.

 

Bleb survival analysis and bleb structure in patients who did or did not receive DPP-4is. (A) Kaplan-Meier survival curves for the cumulative probability of surgical success. The group that received DPP-4is had a significantly higher probability of success than the group that did not receive DPP-4is (log-rank test; P = 0.017). (B) Reflectivity of the bleb wall 2 weeks after surgery. The group that received DPP-4is had significantly lower reflectivity of the bleb wall than the group that did not receive DPP-4is (Mann-Whitney U test, P = 0.033). Data are shown as the mean with standard deviation. *P < 0.05, **P < 0.01, and ***P < 0.001. (C) AS-OCT images of representative cases. The case that received a DPP-4i shows an extended, diffuse, spongy, and hyporeflective bleb. The case that did not receive a DPP-4i shows a hyper-reflective, flat bleb. DPP-4i = dipeptidyl peptidase-4 inhibitor, AS-OCT = anterior-segment optical coherence tomography.

DPP-4 Expression Increased in Scarred Tenon's Tissue From Patients With POAG During Trabeculectomy

Past studies have shown that DPP-4 expression is locally elevated in organs with fibrosis, including the skin and kidneys.^14,18 Thus, we investigated the expression of DPP-4 protein in the Tenon's tissue of patients to determine whether DPP-4is might directly inhibit DPP-4 in the filtering bleb. We obtained scarred and non-scarred Tenon's tissue samples from, respectively, patients with POAG undergoing trabeculectomy re-operations and initial trabeculectomies. Elastica-Masson staining and DPP-4 immunostaining revealed that the scarred samples (cases 1 to 3, Fig 2A) showed increased collagen and DPP-4 expression compared to the non-scarred samples (cases 4 to 6, Fig 2B). This shows that fibrosis in blebs increases DPP-4 expression. 

Figure 2.

 

Images of Elastica-Masson staining and DPP-4 immunostaining of Tenon's tissue samples obtained from patients with primary open-angle glaucoma during trabeculectomy. (A) Cases 1 to 3 show images of Tenon's tissue with scarring in samples that were obtained during trabeculectomy re-operations. (B) Cases 4 to 6 show Tenon's tissue without scarring in samples that were obtained during initial trabeculectomy. The former show more expression of collagen (shown in light green) and DPP-4 (shown in brown). DPP-4 = dipeptidyl peptidase-4.

Linagliptin Attenuated TGF-β1-Induced Fibrotic Changes in HTFs Evaluated With qPCR, a Scratch Assay, and a Collagen Gel Contraction Assay

Although our clinical study suggests the potential of DPP-4is to help maintain bleb function, our relatively small sample size of NVG cases might have been biased by many factors. Thus, we sought to confirm the hypothesis that DPP-4is could attenuate fibrosis and improve trabeculectomy outcomes with in vitro experiments that used HTFs treated with a DPP-4i (i.e. linagliptin). During the process of conjunctival scarring, activated Tenon's fibroblasts differentiate into myofibroblasts, which express α-SMA, a representative marker. These myofibroblasts synthesize extracellular matrix components, such as collagen and fibronectin. Furthermore, activated myofibroblasts promote cell migration and the development of high-level contractile capacity of the tissue.^37,38 TGF-β plays a central role in this process. Thus, in vitro experiments using HTFs stimulated with TGF-β that measure fibrosis markers with qPCR and cell migration assays (i.e. scratch assays), and collagen gel contraction assays are a standard method to evaluate novel antifibrotic agents.^39–41 Here, we followed these methods using TGF-β1-stimulated HTFs treated with linagliptin, choosing a concentration of 100 nM based on previous work.¹⁴ The relative mRNA expression levels of α-SMA, Co11a1, and FN1 in qPCR testing were significantly higher in cells treated solely with TGF-β1 than in untreated controls; these upregulated expression levels decreased when the treatment was combined with 100 nM linagliptin (Fig. 3). The scratch assay showed that TGF-β1 treatment significantly accelerated cell migration: after 12 hours, the wound area shrank significantly more in the treatment group than the control group. Moreover, the TGF-β1-induced migration of HTFs was suppressed by the 100 nM linagliptin treatment (Fig. 4). The collagen gel contraction assay revealed that after day 2, gel contraction significantly increased in HTFs cultured in the presence of TGF-β1 compared with HTFs cultured without TGF-β1. However, combining 100 nM linagliptin with TGF-β1 attenuated the increased gel contraction (Fig. 5). Thus, the results of qPCR testing, the scratch assay, and the gel contraction assay demonstrated that linagliptin prevented TGF-β1-induced fibrotic change in HTFs. 

Figure 3.

 

Effects of linagliptin on the expression of TGF-β1-induced fibrosis markers in human Tenon's fibroblasts, evaluated with quantitative real-time PCR. The upregulated relative mRNA expression levels of α-SMA, Co11a1, and FN1 after treatment with TGF-β1 significantly decreased when combined with 100 nM linagliptin (Tukey-Kramer multiple comparisons test). Data are shown as the mean with standard deviation (n = 4). *P < 0.05, **P < 0.01, and ***P < 0.001.

Figure 4.

 

Effects of linagliptin on TGF-β1-induced cell migration in a scratch assay. (A) Representative microscope images of a scratch assay after 12 hours of each treatment. The wound area, surrounded by migrating cell borders, is shown in yellow. (B) TGF-β1-induced cell migration was suppressed by 100 nM linagliptin, evaluated as the percentage change in the dimensions of the wound area after 12 hours compared to baseline at 0 hours (Tukey-Kramer multiple comparisons test). Data are shown as the mean with standard deviation (n = 4). *P < 0.05, **P < 0.01, and ***P < 0.001.

Figure 5.

 

Effects of linagliptin on TGF-β1-induced collagen gel contraction. (A) Images of collagen gel in a 24-well plate after treatment on day 0 to day 7. (B) The level of gel contraction was measured as the percentage change in gel dimension compared to baseline at day 0. The increased level of gel contraction in the presence of TGF-β1 was significantly attenuated after day 2 of treatment with 100 nM linagliptin (Tukey-Kramer multiple comparisons test). Data are shown as the mean with standard deviation (n = 4). *P < 0.05, **P < 0.01, and ***P < 0.001.

Linagliptin Suppressed the TGF-β/Smad Signaling Pathway

We next determined which fibrosis pathway was suppressed by linagliptin. TGF-β promotes fibrosis through canonical Smad-dependent and non-canonical Smad-independent signaling pathways.^37,42 We investigated if linagliptin suppressed the canonical TGF-β/Smad signaling pathway with Western blotting by evaluating the activation of Smad2/3 in HTFs after treatment with TGF-β1 and linagliptin. The ratio of phosphorylated Smad2/3 to Smad2/3 was elevated after treatment with TGF-β1, but the ratio decreased after TGF-β1 treatment combined with 100 nM linagliptin (Fig. 6). This shows that linagliptin attenuates fibrotic change in HTFs by suppressing the TGF-β/Smad signaling pathway. 

Figure 6.

 

Effects of linagliptin on TGF-β/Smad signaling, evaluated with Western blotting. (A) Western blotting bands after each treatment. (B) The elevated ratio of phosphorylated Smad2/3 to Smad2/3 with TGF-β1 treatment decreased when it was combined with 100 nM linagliptin. Data are shown as the mean with standard deviation (n = 3). *P < 0.05, **P < 0.01, and ***P < 0.001.

Discussion

This study reveals the potential of DPP-4is for improving surgical outcomes after trabeculectomy, based on a clinical data analysis of patients with diabetes with NVG and a subsequent in vitro study using HTFs. Briefly, patients that used DPP-4is had a higher surgical success rate 1 year postoperatively than those who did not. A Kaplan-Meier survival analysis and a Cox proportional hazards model confirmed this favorable result. The in vitro study used HTFs stimulated by TGF-β1 and demonstrated that linagliptin, a member of the DPP-4i family, attenuated fibrotic change by inhibiting TGF-β/Smad signaling. 

Trabeculectomy was first described by Cairns in 1968.⁴³ Numerous modifications since then have included the intra-operative adjunctive use of antimetabolites, such as MMC and 5-fluorouracil (5-FU), which have improved outcomes by preventing subconjunctival fibrosis.⁴⁴ However, novel antifibrotic agents are needed to improve success rates and reduce postoperative complications, which can lead to blindness. Our group previously reported that amsacrine, a DNA topoisomerase II inhibitor, and colchicine could prolong bleb function in rabbits, possibly by inducing apoptosis of Tenon's fibroblasts.^45,46 Whereas apoptosis inducers have potential as candidates for new adjuvants, the risk of causing avascularity in the bleb, which makes it prone to leaking, may remain, as with MMC or 5-FU.^47,48 By contrast, regulating TGF-β signaling may prevent bleb scarring without causing avascularity. Previously, subconjunctival injection of anti-TGF-β2 antibodies attracted attention, because it showed promise in a rabbit model of glaucoma filtering surgery and in a small clinical trial.^49–51 However, a larger, randomized clinical trial failed to confirm this promise.⁵² Thus, agents that target molecules downstream of TGF-β, rather than TGF-β itself, have been investigated.^{40,41,53–55} DPP-4 is well known as a regulator of glucose metabolism, but it is also activated by TGF-β and plays a role in fibrosis.^14–20 Thus, DPP-4is are novel candidates for antifibrotic treatment targeting the downstream molecules of TGF-β to prevent bleb scarring. 

Past studies have demonstrated that DPP-4 may be a target for the treatment of fibrotic diseases, such as kidney fibrosis and systemic sclerosis. In a model of renal fibrosis that used mice with streptozotocin-induced diabetes, linagliptin suppressed the level of DPP-4 in endothelial cells and ameliorated kidney fibrosis.¹⁴ Additionally, the authors reported that siRNA knockdown of DPP-4 in human dermal microvascular endothelial cells resulted in suppression of the TGF-β-induced endothelial-to-mesenchymal transition and Smad3 phosphorylation. These findings are consistent with our results, because even though we did not perform knockdown of DPP-4 in HTFs, we found that linagliptin (i.e. a DPP-4i) inhibited the phosphorylation of Smad2 and Smad3 in TGF-β-stimulated HTFs. Another study by Soare et al., found that the expression of DPP-4 was elevated in the fibrotic skin of patients with systemic sclerosis.¹⁸ Furthermore, they found that DPP-4-positive fibroblasts expressed higher levels of α-SMA and collagen. Our current study also showed elevated DPP-4 and collagen expressions in scarred Tenon's tissue that was obtained from patients during trabeculectomy re-operations. The study by Soare et al. also showed that sitagliptin, another DPP-4i, inhibited TGF-β-induced phosphorylation of ERK (i.e. one of the noncanonical TGF-β pathways), but did not inhibit phosphorylation of Smad3 in human dermal fibroblasts. These results are inconsistent with our finding that linagliptin suppressed TGF-β/Smad signaling, but suggest that the effects of DPP-4is may depend on the specific class of DPP-4i or type of cell. Nonetheless, previous and current findings suggest that the expression of DPP-4 plays an important role in fibrosis in various organs and that DPP-4is are potential antifibrotic agents with versatile applications. 

Our study shows that DPP-4is, which are commonly used for diabetes treatment, have the potential to improve surgical outcomes of glaucoma filtering surgery. Generally, drug repurposing is a useful strategy for finding new treatments, because the approval process is shorter than for new drugs. Other studies have examined drugs that were repurposed for the prevention of bleb scarring. Rosiglitazone, a peroxisome proliferator-activated receptor γ agonist, which is another type of antidiabetic agent, was found to prevent subconjunctival fibrosis after glaucoma filtering surgery by inhibiting profibrotic gene expression through augmentation of autophagy in Tenon's fibroblasts.⁵⁶ Other reports have shown that suberoylanilide hydroxamic acid, a histone deacetylase inhibitor used to treat cutaneous T-cell lymphoma, might help attenuate subconjunctival fibrosis after trabeculectomy by suppressing angiogenesis, in addition to inhibiting Smad and non-Smad TGF-β signaling.^40,54 Our clinical findings reveal potential effects of DPP-4is that may give them advantages compared to other candidate repurposed drugs. In other words, we successfully performed reverse translational research to demonstrate the effect of DPP-4is. 

Although our study shows the potential of DPP-4is to maintain bleb function in cases with NVG, it remains a concerning possibility that surgical outcomes in NVG could be affected by multiple factors, including elevated cytokines, such as IL-1β, IL-6, IL-8, TGF-β, and VEGF.^56–61 In fact, a previous report found that pre-operative IL-8 elevation in the aqueous humor was associated with surgical failure after trabeculectomy regardless of the type of glaucoma (i.e. POAG, pseudoexfoliation glaucoma, or NVG), possibly due to IL-8’s pro-fibrotic properties; these involve recruiting macrophages, as seen in idiopathic pulmonary fibrosis.^58,62 Although we were not able to perform an analysis adjusting for intraocular cytokine levels in our clinical study, we demonstrated the effect of linagliptin with an in vitro model to exclude the possibility of modulation by multiple factors. Furthermore, we aimed to consider the effect of VEGF in our clinical study. Instead of adjusting for the intraocular concentration of VEGF, we included the use of anti-VEGF intravitreal injection in a Cox proportional hazards model to confirm the effect of DPP-4is, as shown in Supplementary Table S2, because the use of anti-VEGF intravitreal injection may reflect higher activity in NVG (i.e. a higher level of VEGF). Whereas the use of DPP-4is was still an independent contributor to surgical success (P = 0.008), anti-VEGF treatment also increased the risk of surgical failure, although this effect was on the margin of statistical significance (P = 0.050). However, discussion of the influence of anti-VEGF treatment is complicated, because the use of anti-VEGF treatment may indicate that NVG is more refractory, and it can effectively suppress VEGF levels and neovascularization, consequently reducing IOP in patients with NVG at the same time. Additionally, previous studies have reported conflicting results on the effects of anti-VEGF treatment on the outcome of glaucoma filtering surgery in patients with POAG or types of glaucoma other than NVG,^63–70 as well as in animal models.^71–75 Considering these factors, a future clinical study with other types of glaucoma, such as POAG, or an experimental study with animals will be needed to demonstrate the effects of DPP-4is. 

Currently, only oral DPP-4i administration has been approved. However, topical administration may be a future option. Oral DPP-4i administration has a lower risk of causing hypoglycemia than conventional drugs, but can still lead to adverse effects, including respiratory tract infection, nasopharyngitis, and headache. In contrast, topical administration via eye drops should be safer, with less risk of systemic side effects. Moreover, it has been reported that eye drops containing DPP-4is, such as saxagliptin or sitagliptin, are effective for the treatment of neurodegeneration in a mouse model of diabetic retinopathy.⁷⁶ DPP-4i eye drops reduced the extracellular concentration of glutamate by suppressing the downregulation of glutamate-aspartate transporter, consequently preventing excitotoxicity and retinal neuronal cell death. Glutamate excitotoxicity has also been proposed to contribute to retinal ganglion cell death in glaucoma.⁷⁷ Thus, DPP-4i eye drops may have the potential not only to improve surgical outcomes after trabeculectomy, but also to exert neuroprotective effects against glaucomatous neurodegeneration. Further investigation with in vivo studies of glaucoma filtering surgery and glaucoma animal models will be required to determine the effects of DPP-4i eye drops. 

This study had several limitations. First, it used a retrospective design with a relatively small number of patients, all of whom had both NVG and diabetes. NVG and POAG have important differences, and surgical outcomes for NVG can be affected by multiple factors, including cytokines, such as VEGF, or neovascularization. We included only NVG cases because it was more convenient to recruit suitable patients in a shorter period. Nonetheless, our in vivo and in vitro experiments using Tenon's tissue samples and HTFs from patients with POAG demonstrated the potential of DPP-4is as antifibrotic agents for trabeculectomy that could be used not only for NVG, but also for other types of glaucoma. Second, we only evaluated the effects of one DPP-4i (i.e. linagliptin). In a future study, we will compare the antifibrotic and cell-signaling effects of different DPP-4is. 

In conclusion, our study found that DPP-4is had the potential to maintain bleb function after glaucoma filtering surgery and that one member of the DPP-4i family, linagliptin, attenuated fibrotic change in HTFs by inhibiting TGF-β/Smad signaling. 

Acknowledgments

The authors thank Tim Hilts for reviewing this manuscript. 

Supported in part by JSPS KAKENHI Grant Number 19K18830 (S.T.). 

Disclosure: M. Yoshida, None; T. Kokubun, None; K. Sato, None; S. Tsuda, None; Y. Yokoyama, None; T. Nakazawa, None