The most critical pathologic findings of PDR are neovascularization, hemorrhage, and fibro-vascular proliferation, leading to retinal traction and detachment and vitreous hemorrhage [63]. Oxygen-induced retinopathy (OIR) in rodents is an accurate and reproducible model of vascular proliferative changes in the retina [55]. Hypoxia-driven vascular proliferative changes seem to be similar to those seen in the retinopathy of prematurity, age-related macular degeneration, and diabetic retinopathy. OIR was developed in canine models for the first time in the early 1950s. In this model, Arnall Patz and colleagues investigated the effects of hyperoxia on retinal vessel development to study proliferative retinopathy [56,57]. To develop this model, one-day-old pups were exposed to hyperoxia for four consecutive days. In the early 1990s, this approach was introduced in rodents by Dr. Smith and her colleagues and has gained increasing popularity. In addition to OIR canines and rodents, aberrant angiogenesis has also been reported in zebrafish, rabbit, and monkey models.

The rodent OIR model is the most common approach to investigating the effect of hypoxia on the retina since it mimics the characteristics of human retinal proliferative changes [55,58,64]. Because retinal vasculature develops in the first two weeks of birth in rodents, researchers can leverage this opportunity to analyze the aberrant vascular development triggered by hypoxia. In this model, hypoxia is induced at postnatal day (P) 7 after the regression of hyaloid vessels to avoid the development of mixed hyaloidopathy. The rodent pups were then exposed to hyperoxia (75% oxygen) for five consecutive days from P7 to P12 and then observed at room air from P13 to P17 [55]. The peak changes of neovascularization are usually observed at P17, and these are resolved by P25. The C57BL/6 mice or the Sprague Dawley (SD) rats are the common strains employed in this model due to their neovascular susceptibility to hypoxia [58,64,65]. The OIR mice developed irregular blood vessels and a reduction in the retinal inner and deep plexuses at P18, mimicking retinal proliferative events triggered by hypoxia in patients with diabetic complications [66]. Downie and colleagues reported an increase in extraretinal neovascularization and impaired pericyte distribution in the OIR SD rat retinas as early as P18 [67].

Genetically modified Akimba, Akita, and Kimba mice manifest vascular dysfunction. Akimba mice were specifically developed to study the microvascular changes of DR and showed these changes at the early age of eight weeks old [47]. Thus, at eight months of age, these mice developed neovascularization, retinal edema, and detachment that progressed further through 25 weeks of age [47]. In the Kimba mice, abnormal blood vessel development was seen as early as P28, while an increase in vascular permeability and adherent leukocytes was observed at six weeks of age. Additionally, loss of retinal capillaries, neovascularization, an increased avascular area, alteration in the vessel length, and pericyte loss were reported from nine weeks to the advanced age of 24 weeks [46,68]. Vascular dysfunction in Ins2^Akita mice presents as an increase in leukocytosis at eight weeks, compromised vascular permeability at 12 weeks, microaneurysms at six months, and neovascularization at nine months of hyperglycemia [42,69].

STZ mice also show microvascular changes earlier in the course of diabetes compared to STZ-induced hyperglycemic rats. For example, vascular permeability detected by imaging the distribution of fluorescein-conjugated dextran is compromised in these animals as early as eight days post-STZ injection [70]. However, a decrease in arteriolar diameter and velocity were reported at four weeks and eight weeks post-STZ injection, respectively [26]. Later in the course of diabetes (six to nine months), the STZ-induced hyperglycemic mice manifested pericyte loss and developed acellular capillaries [71].

In albino Wistar–Kyoto rats, the blood retinal barrier (BRB) disruption occurs as early as two weeks post-STZ injection. Several studies reported early neovascular changes such as adherent leukocytosis and thickening of the basement membrane occurring at 8 and 12 weeks, respectively [8,72,73]. Gong et al. noted that neovascularization in STZ-injected SD rats can be observed at three to four months after induction of hyperglycemia. An increase in VEGFR1 and VEGFR2 expression levels was associated with neovascularization in STZ-induced rats [74]. Similar findings were observed in the Alloxan-induced diabetic rats; leukocytosis and neovascularization were reported at two and nine months after induction of hyperglycemia, respectively. At two months of sustained hyperglycemia, the authors observed pericyte loss, the formation of acellular capillaries, and basement membrane thickening [75,76]. In contrast, several other studies reported that BB rats with autoimmune T1D manifested these retinal changes as early as four months, while this model as well as genetic ZDF and obese OLETF rat models demonstrated BRB breakdown and pericyte loss at six to eight months [48,49,51,53,77,78]. Overall, these studies imply that the observed vascular dysfunction could vary in rat models of DR triggered by different insults. In addition to rats, hyperglycemia induced by a high-fat diet in db/db mice with T2D also leads to an increase in vascular permeability and basement thickness at 13–14 weeks of hyperglycemia [79,80]. Moreover, these mice also demonstrate pericyte loss, blood retinal barrier disruption, and vascular leakage at 12 months of age [39].

Neuronal cell death and gliosis are observed in the diabetic retina of animals with diabetes. Thus, in hyperglycemic rats, GFAP activation has been reported. STZ injection results in an increase in GFAP immunoreactivity in the retina as early as six to seven weeks [81] and as late as 8–16 weeks post-injection [81,82]. Retinal cell loss and functional changes have also been reported as early as two weeks and as late as 24 weeks post-STZ injection. Moreover, an increase in apoptotic cell death in the ONL, INL, and RGC layers resulting in a decrease in the total retinal thickness has been detected between 12 and 16 weeks post-STZ injections in rats [82,83]. In contrast, necrotic RGC death was reported at four weeks post-STZ treatment in rats [83]. These rats also manifested severe loss of photoreceptors at 12 and 24 weeks, [83] while in WBN/ Kob rat retinas, photoreceptor degeneration occurs earlier, at four weeks of age [50]. Our recent study also confirmed RGC function loss and cell death in STZ-induced hyperglycemic mice at 32 weeks post-injection [84] and tree shrews at 16 weeks post-injection [31]. In addition to retinal neurons, RPE degeneration was reported in diabetic retinas. Thus, in four-month-old diabetic BB rats, hyperglycemia induces RPE degeneration through focal necrosis [85]. In hyperglycemic OLETF rat retinas, the decrease in the thickness of the RPE layer along with a reduction in the INL and ONL thicknesses occurs later, at nine months after induction of hyperglycemia [53]. Much later, at 50 weeks post-hyperglycemia induction, retinal detachment and fibrous proliferation occurs in Torii (SDT) rats with spontaneous diabetes [54]. In other model of spontaneous diabetes, ZFD rats, extensive glial activation along with photoreceptor outer segment (POS) degeneration occurs in 32-week-old retinas [86].The latter agrees with multiple studies demonstrating the thinning of the INL and IPL in OIR rat pups at P18 [67,84,87,88]. In addition, the thinning of the inner limiting membrane (ILM) is observed in STZ-induced SD retinas [85].

In STZ-induced diabetic mice, RGC loss occurs between 6 and 12 weeks [89]. RGC death occurs through apoptosis. The number of RGC apoptotic positive cells measured by TUNEL is 25% higher than that in control retinas [90]. These data are similar to our observation of about a 30% RGC death with this model, [84] although another study reported that the RGC density across the retina varies at 20 weeks post-STZ treatment [91]. A few studies with Ins2^Akita mice detected early cone photoreceptor cell loss at three months. The authors observed a significant reduction in the IPL and INL thicknesses along with a diminishing number of RGCs at 22 weeks and 36 weeks of hyperglycemia [42,92]. Similarly, the OCT analysis of 16- and 28-week-old diabetic db/db mice retinas showed thinning in the NFL and RGC layer at a rate of 0.104 μm per week, resulting in a reduction of the total retinal thickness by 28 weeks [91,93]. The 28-week-old diabetic db/db mice also showed TUNEL positive photoreceptor cells and reduction in the ONL thickness. STZ-induced hyperglycemia in mice also leads to GFAP overexpression in retinal astrocytes at five weeks post-STZ treatment, while Müller cell gliosis are not seen even after 15 months of DM [71,94]. In contrast, the OIR mice demonstrated a reduction in the total retinal, INL, and IPL thicknesses, as well as distorted photoreceptor OS, neuronal loss, hyperactivity of Müller cells, and microglial activation at P18 [66].

Several studies with diabetic rats have reported ERG findings. First, there is a delay in the implicit time detected at four to seven weeks post-STZ. Second, a decrease in the a-wave of the scotopic ERG amplitude was detected at 10 weeks, while the b-wave amplitude was found to be reduced at 25 weeks after the induction of hyperglycemia [81,95,96,97]. Similar ERG findings were observed in SDT rats at 44 weeks post-STZ treatment [98,99] and mice and rats with proliferative retinopathy at P18 [66,67,87,88]. In the STZ-treated mice, retinal functional test showed a decrease in the implicit time for OP at 4-6 weeks, reduction in the scotopic ERG a and b-wave amplitudes at six months and diminished photopic ERG negative amplitudes at eight months after the induction of hyperglycemia [84,100,101,102]. Moreover, db/db (Lepr ^db), Ins2^Akita, and high fat diet-induced diabetic mice manifested similar retinal function changes detected by ERG at 6, 9 and 12 months, respectively [39,69,92,103].

Basal insulin receptor (IR) signaling has been extensively studied in the STZ-induced diabetic SD rat retina. It has been observed that the phosphorylation of insulin receptor (IR) in hyperglycemic retinas remained unchanged up to eight weeks post-injection, whereas PI3K activity was reduced by 25% compared to the controls. At 12 weeks post-STZ injection, both kinase activity and auto-phosphorylation of the IR were significantly decreased, suggesting that the basal IR activity is diminished in the diabetic retina. It was also demonstrated that Akt1 kinase activity was significantly diminished at eight weeks post-STZ injection, suggesting compromised glucose flux [104]. Kondo and colleagues observed important differences in insulin signaling between STZ-induced hyperglycemic mice and db/db mouse models developing DR. Specifically, IR expression and tyrosine phosphorylation were upregulated the first week post-STZ treatment in mouse retinas, but no changes were observed in 8- to 10-week-old db/db mice. Moreover, IRS-1 expression was unaltered, while IRS-2 expression was increased in both db/db and STZ-induced diabetic mouse retinas. In contrast, a few studies have reported a reduction in IR phosphorylation and an increase in the activity of the protein tyrosine phosphatase-1B (PTP1B) in the rod’s inner segments one-week post-STZ injection in mice [105]. An analysis of phosphorylated PTP1B in these mouse retinas point to PTP1B as a promising therapeutic target to delay neurodegeneration in diabetic retinas [106]. Reduced IR kinase activity at 12 weeks of hyperglycemia was also reported in a study with Ins2^Akita mice [42]. The potential contribution of excess glucose to local impairment of retinal insulin receptors and AKT activity has been proposed [104]. Our recent study confirmed an excess of glucose in diabetic retinas [84]. Moreover, other studies have reported reduced AKT phosphorylation as an early event in diabetic retinas with T1D [104,107] and T2D [108,109].

Endoplasmic reticulum (ER) stress is one of the important features of the molecular pathobiology of the diabetic retina. Three independent UPR arms became activated during the ER stress response in diabetic retinas, including PKR-like ER kinase (PERK) and eukaryotic translation initiation factor 2α (eIF2α); inositol-requiring protein 1α (IRE1α)-X-box binding protein 1 (XBP1); and activating transcription factor 6 (ATF6)[110]. Activation of PERK kinase signaling results in phosphorylation (p) of eukaryotic translation initiation factor 2α (eIF2α), leading to global translational arrest and upregulation of activating transcriptional factor 4 (ATF4), C/ERB homologous protein (CHOP), and tribbles homolog 3 (TRIB3) proteins. The triggering of ATF6 is associated with its autophosphorylation, translocation to Golgi, and cleavage leading to active p-ATF6 transcriptional factor. After autophosphorylation, IRE1α, possessing both the RNAse and kinase activities, trims the Xbp1 transcriptional factor, leading to the formation of activated transcriptional factor, which controls a variety of gene expressions. Cellular stresses such as hypoxia and glucose imbalance can trigger UPR. ER stress markers are upregulated in diabetic rat retinas as early as eight weeks after the onset of diabetes induced by STZ in SD rats [111]. Apoptotic protein caspase 12, CHOP, and phosphorylated c-Jun N-terminal kinase 1 (MAPK) were dramatically upregulated in these retinas. In addition, the elevation of MAPK kinase was detected in RGCs [111]. However, the differences in the expression of the upstream and downstream mediators of PERK signaling, Grp78 and Atf4 genes, respectively, were not significant in this study. These findings suggest that AFT4 might not be the only signaling molecule responsible for the increased VEGF level in diabetic retinas [112]. Using immune-histochemical detection, the investigators reported the elevation of HIF-1α, ATF6, XBP1, and CHOP proteins in STZ-induced diabetic rat retinas at two and four months [82]. This elevation was accompanied by a decrease in the autophagy marker LC3B-II levels, indicating a potential reduction in autophagy in the diabetic retinas of mice with four months of hyperglycemia [82]. Pro-apoptotic BAX was detected in hyperglycemic ZFD rat retinas at six weeks of age [113].

Inducing diabetes and DR by STZ injection in mice, Chung et al. and Zhong et al. reported the activation of the ER stress response and pro-inflammatory signaling [114,115]. They found that the diabetic mouse retinas manifest increased expression of GRP78, pPERK, CHOP, VEGF, and peIF2α four weeks after STZ-induced hyperglycemia. Moreover, ATF4 deficiency resulted in altered inflammatory gene expression [115]. In addition to UPR markers, the above-mentioned study reported that MCP-1 and TNF-α expression simultaneously increased in diabetic retinas during the four-week period [114]. Interestingly, this study also highlighted that UPR signaling could be resolved later in diabetic mouse retinas, at six weeks post-STZ injection. In contrast, genetically modified Akita mice demonstrated an increase in p-elF2α and GRP78 proteins (PERK arm) in addition to elevated IRE-1 and TNFα expression at 12 weeks of age [116,117]. Elevated levels of GRP78, ATF4, and peIF2α were also found in the OIR model at P15 [117,118] and in 15-month-old db/db (Lepr ^db) mice [119]. Moreover, our study showed that limiting ATF4 expression in hypoxic retinas significantly reduced the degree of neovascularization in the OIR mouse retinas, [118] and the deficiency in ATF4 could reduce IL-1β in diseased retinas [120].

The authors' recent study with diabetic TRIB3 KO demonstrated that TRIB3 is a master regulator of insulin signaling and glucose metabolism in the retina (Figure 1). Thus, we revealed that TRIB3 is induced in diabetic retinas, leading to overexpression of HIF1a, GFAP, VEGF, GLUT1, and EGFR proteins. In turn, HIF1a regulates GLUT1 expression and, together with TRIB3, controls the uptake of glucose in the retina. Moreover, TRIB3 mediates the retinal ganglion cell fate decision, while TRIB3 KO results in neuronal survival and improvement of vascular health.

Changes in inflammatory gene expression across varied rodent models of DR have also been reported. For example, six-week-old ZFD rat retinas manifest an increase in the levels of TNF-α and NF-kB [113]. Inflammatory proteins such as clusterin, the tissue inhibitor of metalloproteinase (TIMP)-1, β-2 microglobulin, and von Willebrand factor were overexpressed in the SD rat retinas at four weeks and, particularly, at three months post-STZ injection. In addition, the overexpression of fibroblast growth factor-2 (FGF2) was detected in the ONL of diabetic rat retinas at three months post-STZ. It is also worth mentioning that inflammatory changes are strain-dependent in diabetic rat models. For example, compared to Long-Evans and Brown Norway rats, SD rats show inflammatory changes more similar to those found in human diabetic retinopathy [121]. SD rat pups with OIR were also reported to overexpress inflammatory markers at P16 [58,64].

Overall, the above-mentioned studies emphasize that alterations in cellular molecular signaling often precede retinal pathophysiological events. These findings suggest that dysfunctional insulin signaling, ER stress response, and inflammation are involved in the pathological progression of DR and can be targeted to develop novel cellular therapies for DR (Table 2).