Diabetic retinopathy (DR) is a complication of diabetes mellitus (DM) affecting 22% of the world’s population ^[1]. DR was previously understood to be a microvascular disease; however, new evidence has now shown it to be a neurovascular disease with complex pathogenesis originating from oxidative stress and advanced glycation end products [2,3,4]^[2][3][4]. The accumulation of these products leads to immune dysfunction as well as neuronal and vascular disruption and results in a disease in which multiple cell types and signaling pathways of the retina become involved [3,5,6]^[3][5][6]. Vascular dysfunction has a prime role in DR and has been studied extensively. Angiogenic pathology is found to appear early in DR as microaneurysms, small outpouchings from retinal capillaries, dot intraretinal hemorrhages, and cotton wool spots ^[7]. These changes can ultimately progress to vision-threatening consequences like diabetic macular edema, neovascularization, and tractional retinal detachment [4,7]^[4][7].

Currently, non-proliferative diabetic retinopathy (NPDR) management varies depending on the presence of macular edema ^[8]. NPDR without macular edema is managed by counseling patients to maintain an HbA1c below 7%. When diabetic macular edema (DME) is present, first-line treatment of intravitreal anti-vascular endothelial growth factor (anti-VEGF) therapy is started to target the vascular permeability caused by the increase in VEGF production. In cases of proliferative diabetic retinopathy (PDR), this anti-VEGF treatment is used to control neovascularization. This is often followed by pan-retinal photocoagulation laser therapy to regress new blood vessels and ablate ischemic tissue. Vitrectomies are employed in the presence of a tractional retinal detachment or persistent, large vitreous hemorrhage. Though intravitreal anti-VEGF is by far the most frequently used therapy to treat PDR and DME, it has been shown that DME patients often have a less robust response to anti-VEGF therapy compared to patients treated for neovascular age-related macular degeneration or retinal vein occlusion ^[9]. Serum studies show that a large percentage of DME patients have a relatively low to normal level of VEGF and therefore experience persistent, vision-limiting DME despite frequent intravitreal anti-VEGF therapy ^[10]. Thus, other mechanisms of DR should be explored and potentially targeted for treatment.

Current evidence suggests the TGF-β signaling family may be a potential target for the treatment of DR, as it is crucial in a wide number of homeostatic functions in the retinal vasculature, including endothelial cell barrier function and pericyte differentiation ^[7]. Both endothelial cells and pericytes manage the vascular environment within the retina, and thus communication between both cell types is essential for function ^[10]. Communication is carried out by a multitude of signaling molecules, many of which belong to the TGF-β signaling family. A study performed on mice with a conditional deletion of TGF-β found that the resulting decrease in TGF-β led to the onset of the same retinal phenotype as DR, including microaneurysms, retinal hemorrhages, and cotton wool spots ^[7]. This shows that disturbance of the delicate balance of these pathways is implicated in DR, as well as other similar diseases. Though changes in TGF-β expression in early DR are believed to act protectively on the retinal microvasculature, these changes over time appear to lead to endothelial cell proliferation and pericyte dedifferentiation [7,11]^[7][11]. Because of its role in vascular homeostasis and pathology, as well as the recent research exploring the detrimental effects caused by the dysregulation of TGF-β in the context of diabetic retinal tissues, TGF-β is proposed to be a critical factor in microvascular abnormalities caused by DM ^[11].

2. Association of TGF-β Signaling Pathways in DR

TGF-β signaling is an established player in maintaining retinal capillaries, and TGF-β1 has been identified as a contributing factor in the pathogenesis of DR [7,26]^[7][12]. TGF-β1 is known to activate in response to ROS, resulting in the eventual proliferation of endothelial cells, angiogenesis, and blood–retinal barrier disruption [27]^[13]. Inhibition of the TGF-β signaling pathway has been shown to decrease VEGF production following hypoxic states in vitro [28]^[14]. To understand the relationship between the stage of disease and TGF-β1 serum levels, one study obtained serum levels of TGF-β1 from those with NPDR and PDR diagnoses. Within these diagnosis groups, people were further divided into aflibercept treatment and non-treatment groups. Data analysis between these categories found TGF-β1 to be three times higher in patients with exacerbated PDR than those with controlled PDR and therefore predictive of disease severity and control [23]^[15]. Interestingly, patients with NPDR who had received aflibercept treatment in the past week showed lower levels of TGF-β1 than NPDR patients without aflibercept treatment. Additionally, higher TGFβ-1 levels were found to correlate with HbA1c levels, the duration of diabetes, and the progression of DR. These findings make TGF-β1 a potential predictor of disease progression from NPDR to PDR. Other studies have shown similar findings. A study on TGF-β1 and -β2 levels in aqueous humor also found elevated levels in patients with NPDR compared to control patients [29]^[16]. Levels of TGF-β1 are also seen elevated in the vitreous humor of PDR patients compared to controls [30]^[17]. Certain polymorphisms of the TGF-β1 gene have also been studied as potential DR risk factors: it has been found that +869T/C(L10P) polymorphisms in the TGF-β1 gene may be a strong DR risk factor, whereas the 2509T/C polymorphism is not associated with DR risk [31]^[18]. TGF-β levels have also been found to correlate with the increased expression of long non-coding RNA of myocardial infarction-associated transcript (lncRNA-MIAT) [32]^[19]. LncRNA-MIAT, a known mediator in microvascular dysfunction, has been shown to be upregulated and reduce viability in adult retinal pigment epithelial cells (ARPE-19) under hyperglycemic conditions. When these cells are treated with a TGF-β inhibitor, these effects are dampened, suggesting that TGF-β may reduce the viability of epithelial cells in the setting of diabetes. The long non-coding RNA nuclear-enriched abundant transcript 1 (lncRNA-NEAT1) has also been found to trigger TGF-β1 and VEGF expression with associated findings of apoptosis and oxidative stress in diabetic mice retinas [33]^[20]. This implies that silencing the expression of lncRNA-NEAT1 could reduce the hyperglycemic stress on retinal endothelial cells. Interestingly, the correlation between TGF-β and DR is not completely uniform. At least one study has found serum concentrations of TGF-β1 to be higher in diabetes without apparent DR than those with NPDR and PDR [34]^[21]. Likewise, studies regarding TGF-β gene expression via non-coding RNAs have revealed mixed associations. One study mirrored previous findings by measuring an increase in TGF-β related to various microRNAs in the context of proliferative DR [35]^[22]. However, in a different study utilizing the mouse model, amniotic mesenchymal stem cells migrated to hypoxic retinal tissue and reduced excessive neovascularization through the release of TGF-β1 [15]^[23]. The use of siRNA to block this pathway resulted in the negation of this effect. These results highlight that TGF-β is necessary to maintain the blood–retinal barrier, though a pathological excess expression of TGF-β may result in vasculopathy itself. TGFβ-1 is also known to affect endothelial cell proliferation and migration, and evidence suggests that a total lack of TGF-β may also be detrimental to vascular integrity ^[7]. In newborn mice, total inhibition of TGFBRII signaling was shown to produce characteristics reminiscent of DR. The same study found that a lack of retinal TGF-β resulted in the dedifferentiation of microvascular pericytes, unregulated proliferation of vascular endothelium with reduced barrier function, and reactive microglia. Poor vascular function in the mouse retina led to retinal hypoxia, the induction of angiogenic molecules, and further neovascularization. This suggests that the TGF-β maintenance mechanism fails in DR and indicates that signaling pathways must be further understood if TGF-β is to be a target for future therapy. It is possible that changes in ALK1/ALK5 expression modulate the TGF-β response in pathological conditions ^[11]. Using an ALK5 inhibitor, one study on rats found that the inhibition of ALK-5 signaling resulted in leaky vessels with characteristic features of DR in the embryo, newborn, and adult rate groups. The same ALK5 inhibitor was used on diabetic rats which showed prominent signs of DR. Therefore, TGF-β/ALK5 signaling is believed to be important for protection from hyperglycemic damage, especially in the setting of diabetes without DR. Similarly, the soluble expression of endoglin has been shown to promote TGF-β1/ALK-1 signaling and interfere with TGF-β1/ALK-5, therefore increasing fibro-neovascularization, angiogenesis, and arteriovenous malformations. Altogether, this activity promotes endothelial proliferation [7,36,37]^[7][24][25]. Under hyperglycemic conditions, BMP9/ALK1 signaling was shown to be disrupted in human umbilical endothelial cells, and, inversely, signaling through ALK1 solidifies the integrity of the vascular barrier by blocking VEGF-induced phosphorylation of VE-cadherin and by solidifying occludin junctions independently of VEGF ^[4]. These results suggest that an incremental increase in TGF-β signaling is a protective mechanism, whereas an insufficient response to TGF-β may cause disease progression ^[11]. Though TGF-β1 is the main isoform studied in DR, other isoforms of TGF-β have also been associated with DR [38]^[26]. One study found that the TGF-β2 isoform was the only detectable isoform in the healthy retina and did not change in concentration with DR progression, indicating that it may be an auxiliary mechanism for DR pathology [34]^[21]. A study on human retinal pigment epithelial cells showed that TGF-β2, both independently and in combination with TNF-α, is associated with retinal neovascularization and an increase in VEGF. Moreover, blockage of the TGF-β2 signaling pathway by miR-200a-3p, a microRNA shown to be downregulated in diabetic rat retinal tissue, suppressed DR progression in diabetic rats, further establishing the role of TGF-β2 overexpression in the pathogenesis of DR [39,40]^[27][28]. At least one study has found TGF-β3 to also be elevated in PDR [41]^[29]. Another study on diabetic rats showed elevated levels of connective tissue growth factor (CTGF), VEGF, and TGF-β2 [38]^[26]. The level of these factors was higher in more severe cases of DR. When CTGF was targeted, the levels of VEGF and TGF-β2 diminished and the apoptosis of retinal cells was reduced, providing evidence of an association between these TGF-β2 and the pathophysiology of DR. Thus, as evidenced by mixed results from different reports in the literature, the association between TGF-β and angiogenesis in DR is complex. This is likely due to the many intracellular effectors, co-receptors, cell types, and individual gene expressions with which TGF-β interacts. As DR is an inflammatory disease, additional pathways for angiogenesis may originate in the immune response [16]^[30]. Macrophages are a significant source of TGF-β1 in DR, and TGF-β signaling is not only necessary for vascular maintenance but also for the immune response in the retina [16,30]^[17][30]. Studies exploring interactions between immune cells and vasculature in the environment of a pathological retina may be beneficial in elucidating these associations.

3. The Role of TGF-β in DR-Related Neurodegeneration

Neurodegeneration of the retina is one of the earliest findings in diabetic retinopathy, often identifiable before vascular abnormalities are present [53]^[31]. This is a result of a myriad of cellular and molecular mechanisms, the most well-known of which include an upregulation of pro-apoptotic proteins and elevated ROS in the diabetic retinal neuronal cells. As mentioned previously, TGF-β signaling pathways are not limited to vascular control. TGF-β signaling plays important roles in immune regulation, neuronal survival, and neuronal maintenance. However, research regarding the precise mechanisms and effects of each TGF-β molecule remains inconclusive and, at times, contradictory. Experiments on RGCs under oxidative stress have been used to illustrate the mechanisms of TGF-β1 and TGF-β2 combating cellular oxidative stress [54]^[32]. These experiments demonstrate TGF-β’s ability to promote the expression of neuroprotective and antioxidative proteins like nuclear factor erythroid-2 related factor (Nfr2), Kelch-like ECH-associated protein 1 (Keap1), aldehyde dehydrogenase 3A1 (ALDH3A1), and heme oxygenase-1 (HO-1). One study found that TGF-β1 and TGF-β2 limited damage from hyperglycemia in RGCs [55]^[33]. In this study, hyperglycemic conditions increased ROS within RGCs in vitro and resulted in irreversible epigenetic changes including histone modifications, DNA methylation, and non-coding RNAs. When TGF-β1 and TGF-β2 were knocked down, RGCs proliferated less and were more sensitive to oxidative stress. This is important in the setting of diabetes because ROS is a major contributor to the cellular damage that takes place in DR. By contrast, separate studies have shown the AGE/RAGE axis seen in diabetes to increase TGF-β presence [18,56]^[34][35]. This has the downstream consequences of increased ROS production, suppressed antioxidant mechanisms, and increased expression of Nox enzymes.

4. Current Research on TGF-β’s Potential Role in DR Treatment

Understanding the mechanisms that the various members of the TGF-β superfamily play in the pathogenesis and progression of DR allows for research regarding potential pharmacologic intervention to prevent the onset or delay the progression of the disease. One molecule of interest is GDF11, a member of the TGF-β superfamily, which has previously been associated with the regulation of retinal neurogenesis and promotion of angiogenic activity in ischemic limb tissue in diabetic rats [67,68]^[36][37]. Mei et al. studied the actions of supplemental GDF11 in the retinas of diabetic rats to explore the possibility of halting the progression of DR [69]^[38]. The results of the study showed that the administration of GDF11 was protective against retinal vascular endothelial cell and retinal pericyte apoptosis, two major characteristics of DR. It was also found that pretreatment of the diabetic mice with recombinant GDF11 (rGDF11) reduced the apoptosis of retinal endothelial cells. Pretreated diabetic mice showed an increased expression of anti-apoptotic proteins like Bcl-2 and a decreased expression of pro-apoptotic proteins like Bax. Diabetic mice treated with GDF11 also displayed greater blood–retinal barrier (BRB) integrity than the non-treatment diabetic mice. Furthermore, Western blot analysis showed a greater presence of tight junction proteins in the retinas of the treatment diabetic mice, as compared to the non-treatment diabetic mice. Explorations of the possible mechanism behind these results show that GDF11 may promote the canonical TGF-β/SMAD2 pathway, as well as the noncanonical pathways of NF-κB and PI3K-Akt-FoxO. The utility of targeting TGF-β signaling pathways for the treatment of DR has also been demonstrated in vitro using acrolein, an endogenous compound which has been previously implicated in TGF-β-mediated retinal pigment epithelium (RPE) cell death in the setting of diabetes [69,70,71]^[38][39][40]. ARPE cells incubated in glucose and acrolein showed significant cell death [71]^[40]. However, when these cells were also treated with SIS3, a specific inhibitor of SMAD3, or SB431542, a TGFβR1 antagonist, these results were almost entirely prevented. Not only do these results demonstrate a TGF-β mechanism for acrolein’s damaging effects, but they suggest the value in targeting TGF-β signaling in the treatment of DR [27,71]^[13][40].