VEGF production and signaling in Müller glia are critical to modulating vascular function and neuronal integrity in diabetic retinopathy and hypoxic retinal vascular diseases

Abstract

Müller glia (MG) are major retinal supporting cells that participate in retinal metabolism, function, maintenance, and protection. During the pathogenesis of diabetic retinopathy (DR), a neurovascular disease and a leading cause of blindness, MG modulate vascular function and neuronal integrity by regulating the production of angiogenic and trophic factors. In this article, I will (1) briefly summarize our work on delineating the role and mechanism of MG-modulated vascular function through the production of vascular endothelial growth factor (VEGF) and on investigating VEGF signaling-mediated MG viability and neural protection in diabetic animal models, (2) explore the relationship among VEGF and neurotrophins in protecting Müller cells in in vitro models of diabetes and hypoxia and its potential implication to neuroprotection in DR and hypoxic retinal diseases, and (3) discuss the relevance of our work to the effectiveness and safety of long-term anti-VEGF therapies, a widely used strategy to combat DR, diabetic macular edema, neovascular age-related macular degeneration, retinopathy of prematurity, and other hypoxic retinal vascular disorders.

1. Introduction

Müller glia (MG), the major macroglia in the retina, extend from ganglion cell layer (GCL) to photoreceptor inner segment area and are surrounded by neurons in a highly organized manner (1). This cellular localization is designed for MG to play major supporting roles in physiological and pathological responses, including retinal metabolism, function, maintenance, and protection by responding to stresses, providing trophic factors, removing metabolic wastes, controlling extracellular space volumes and ion and water homeostasis, participating visual cycles, releasing neurotransmitters, modulating vascular function, and regulating innate immunity (for review, see (2)). During the progression of diabetic retinopathy (DR), a neurovascular disease and a leading cause of blindness, MG play a key role in developing a wide range of pathological characteristics, which was reviewed extensively in this DR Special Edition (Coughlin et al., 2017). MG regulate retinal inflammation, neovascularization, and vascular leakage and lesion, key DR-associated pathological features by producing a major angiogenic factor, vascular endothelial growth factor (VEGF or VEGF-A). On the other hand, VEGF signaling is critical to MG viability under hypoxic and diabetic conditions, which in turn, provides necessary protection to neurons in DR. In this article, I will 1) summarize our work on delineating the role and mechanism of MG-modulated vascular function through the production of VEGF and on investigating VEGF signaling regulated MG viability and neuroprotection in diabetic animal models, 2) explore the relationship among VEGF and neurotrophins in protecting Müller cells (MCs) with in vitro models of diabetes and hypoxia and its potential implication for neuroprotection in DR and hypoxic retinal vascular diseases, and 3) discuss the relevance of our work to the effectiveness and safety of long-term anti-VEGF therapies, a widely used strategy to combat DR, diabetic macular edema (DME), neovascular age-related macular degeneration (AMD), retinopathy of prematurity (ROP), and other hypoxic retinal vascular disorders.

2. Contribution of MG-derived VEGF to blood-retina barrier breakdown

VEGF, a heparin-binding homodimeric glycoprotein, is a potent mitogenic factor for stimulating endothelial proliferation and migration and tube formation during vasculogenesis, embryogenesis, and angiogenesis (3, 4). Its potency is reflected by haploid insufficiency that global deletion of a single Vegf allele in mice causes embryonic lethality during organogenesis (5, 6). VEGF mediates its function through a large group of receptors (Rs) and co-receptors, including VEGFR1, VEGFR2, VEGFR3, neuropilin-1, neuropilin-2, vascular endothelial cadherin, and integrin (for review, see (7)). Intensive investigations on the mechanisms of DR, neovascular AMD, ROP, and other hypoxic retinal diseases suggest that VEGF plays a major role in inducing vascular leakage and retinal neovascularization in these disorders (8–12), which has led to the development of anti-VEGF therapies as a major strategy to combat these leading causes of blindness.

The clue that MG-derived VEGF may have a significant role in stimulating blood-retina barrier (BRB) breakdown came from in situ hybridization and immunohistochemistry (IHC), in which both VEGF mRNA and protein were localized to the cell bodies of the inner nuclear layer (INL) in a mouse model of ROP (13). The VEGF-positive INL cells appeared to have the distinct MG morphology (13), which was confirmed independently (14). To determine if MG-derived VEGF is a major contributor to retinal neovascularization and vascular leakage in ROP and DR, we made a Cre-driver line and generated MG-specific VEGF knockout (KO) mice by breeding the Cre-driver mice with floxed VEGF mice (15–17). The conditional VEGF KO mice did not appear to have detectable phenotypic changes in the eye and elsewhere in the body under normal conditions. Immunoblotting (IB) analysis showed that the conditional VEGF KO mice had a near 50 percent reduction of total retinal VEGF under normal conditions, and in models of oxygen-induced retinopathy (OIR) and streptzotocin (STZ)-induced diabetes (15, 18), which was corroborated by the data from IHC analysis (15, 18). These results support previous observations that MG are a major VEGF producer in the retina (13).

To examine the contribution of MG-derived VEGF in retinal neovascularization in diabetes and hypoxia, we utilized OIR (19), an accepted rodent model for retinal neovascularization in DR and ROP, and analyzed the changes in retinal vasculature and related parameters in conditional VEGF KO mice (Table 1). The OIR conditional VEGF KO mice showed 40 percent and 30 percent reductions in retinal neovascularization and vaso-obliteration, respectively, as judged by fluorescein angiography (15). This result was consistent with a 34 percent reduction of pre-retinal neovascular endothelial cells in haematoxylin & eosin (H&E) stained retinal sections (15). As a result of these vascular changes, the level of OIR-induced vascular leakage was reduced by 56 percent, measured by IB for retinal albumin (15). Likewise, the diabetic MG-specific VEGF KO mice showed a near 59 percent reduction in the level of retinal albumin 6 months after STZ-injection (Table 2), which was supported by the decrease of retinal fluorescence in animals injected with fluorescein-labeled albumin intravenously (18). The loss of MG-derived VEGF also resulted in a 45 percent reduction in acellular capillaries 6-month after STZ injection (Table 2)(18).

Table 1.

Alteration in OIR-induced retinal changes in MG-specific VEGF KO mice at P17.

Pathological changes	degree
Retinal albumin	Decrease (56%)
Retinal occludin	Increase (35%)
Pre-retinal neovascularization	Decrease (34%)
Neovascularization area	Decrease (40%)
Vaso-obliteration area	Decrease (30%)

Table 2.

Alteration in retinal protein modification and expression, inflammation, and vascular lesion and leakage, major DR-associated characteristics, in STZ-induced diabetic MG-specific VEGF KO mice.

Phenotypical changes	Degree
Retinal nitrotyrosine	Decrease (19%, 2 mo after STZ-injection)
Retinal pNF-κB p65	Decrease (48%, 2 mo after STZ-injection)
Retinal ICAM1	Decrease (62%, 2 mo after STZ-injection)
Retinal TNF1α	Decrease (53%, 2 mo after STZ-injection)
* Leukostasis	Decrease (75%, 2 mo after STZ-injection)
Retinal occludin	Increase (60%, 6 mo after STZ-injection)
Retinal ZO1	Increase (130%, 6 mo after STZ-injection)
* Retinal albumin	Decrease (59%, 6 mo after STZ-injection)
* Vascular leakage	Decrease (60%, 6 mo after STZ-injection)
* Acellular capillaries	Decrease (60%, 6 mo after STZ-injection)

^*DR-associated pathological changes.

3. Mechanism of MG-derived VEGF in developing DR

To obtain the mechanistic insights of MG-derived VEGF in BRB breakdown in DR, we investigated alteration in retinal protein expression and modification. Peroxynitrite is produced by combining superoxide with nitric oxide, which is a metabolic indicator of oxidative and inflammatory stresses in the retina of DR patients and animal models (20–22). An increase in peroxynitrite levels is associated with diabetes-induced VEGF upregulation and there is a feedback between VEGF pathway and peroxynitrite production (23, 24). To examine the contribution of MG-derived VEGF in protein nitration, we analyzed a biomarker of peroxynitrite, nitrotyrosine. The diabetic MG-specific VEGF KO mice had a 19 percent reduction in nitrotyrosinylated proteins, 2 months after STZ injection (Table 2)(25), suggesting that MG-derived VEGF also contributes to oxidative stress in DR. Transcriptional factor nuclear factor-kappa-B (NFκB) is a major mediator for developing early pathological changes in DR (26, 27). To determine the contribution of MG-derived VEGF in NFκB-mediated transcriptional activity, we analyzed the level of activated (phosphorylated) form of NFκB p65 subunit. MG-specific VEGF KO mice had a 48 percent reduction in the level of activated NFκB p65 in diabetic retinas 2 months after STZ-injection (Table 2)(18).

Alteration in inflammatory responses usually occurs at early stage of DR, which upregulates the expression of inflammatory markers, such as intercellular adhesion molecule-1 (ICAM1) and tumor necrosis factor-α (TNFα). To evaluate the contribution of MG-derived VEGF in retinal inflammation, we analyzed the levels of inflammatory biomarker ICAM1 and TNFα. Loss of MG-derived VEGF resulted in 53 and 62 percent decreases of retinal TNFα and ICAM1, respectively, in MG-specific VEGF KO mice 2 months after STZ-injection. This result is supported by a 75 percent reduction of adherent leukocytes in the retinal microvasculature in in these mice in leokostasis analysis (Table 2)(18). Taken together, our data suggests that MG-derived VEGF contributed significantly to DR-associated protein expression/modification and inflammatory responses at the early stage of the disease, which are probably the trigger for more serious pathological outcomes, such as vascular leakage, vascular lesions, and neovascularization in DR (Table 2) (15, 18). A summary of MG-derived VEGF in the development of DR-associated pathological characteristics is listed in Figure 1.

Figure 1.

Schematic diagram for potential roles of MG-derived VEGF in DR. Phenotypical changes are in italic. Potential functions with no direct experimental and clinical data are indicated with a question marker. MG modulate vascular function and neuronal integrity through the production and signaling of VEGF.

4. Lesson from MG-specific VEGF KO mice: implication for anti-VEGF strategy for BRB breakdown

VEGF upregulation in MG during the progression of DR and ROP is very similar to the response by glia in the brain during ischemic attack or stroke (28), pathological events that are often accompanied by elevated VEGF and VEGFR expression in astrocytes and surrounding neurons (29). VEGF signaling in glial cells may also have therapeutic value in the treatment of neurological diseases. In a Cu/Zn superoxide dismutase transgenic mouse model of amyotrophic lateral sclerosis (ALS), systemic administration of recombinant VEGF significantly reduces gliosis and increases neuromuscular junctions (30). In a rat model of Parkinson’s disease VEGF signaling not only protects dopaminergic neurons directly but also provides additional neuroprotection by mediating glial viability (31). Functionally, these changes are largely responsible for improving vascular function, glial viability and function, and neuronal survival (for review see (32)). One “programed” goal for MG-mediated upregulation of VEGF production and signaling in diabetes and hypoxia is to increase blood circulation to support appropriate metabolic activities. Unfortunately, uncontrollable VEGF upregulation leads to an unintended consequence, BRB breakdown.

VEGF is a major pathogenic factor for BRB diseases such as ROP and DR (7). In general, anti-VEGF strategies are effective in treating BRB breakdown associated pathological changes in DR, DME, ROP, wet-AMD, and other retinal diseases. Although numerous studies demonstrated the efficacies of anti-VEGF strategy on BRB breakdown in these diseases pharmacologically, our work is among a handful studies demonstrating a cellular mechanism of VEGF in the pathogenesis of ROP and DR. Based on the collective data by others and us, it was clear that MG is a major cellular source of VEGF for retinal inflammation, neovascularization, vascular leakage, and vascular lesion in DR (Figure 1)(13, 15, 18). It is important to keep in mind that retinal ganglion cell (RGC)-derived VEGF is also a major contributor to pathological retinal neovascularization, vascular leakage, and lesions in DR and ROP, as reviewed in this DR Special Issue (Hu J. et al.). Of note, RGCs may be an important contributor to retinal vascular growth (33), which is reinforced by our observation that MG-derived VEGF did not appear to have a major role in the development of retinal vasculature (15). However, this is expected as MG are not matured at the beginning of developmental angiogenesis in the retina (34).

As a result of anti-VEGF treatment, visual acuity is improved immediately after the reduction of BRB breakdown associated pathological changes. This suggests that anti-VEGF drugs are very effective in reducing retinal VEGF to a level below the pathological thresholds in these disorders with the following caveat. In our studies, the levels of VEGF in OIR and diabetic conditional VEGF KO mice were reduced to approximately 50 percent of that in OIR or diabetic wild-type (WT) controls but were comparable to that in unstressed normal WT counterparts (15, 18). These results indicate that a reduction of retinal VEGF to a physiological level is sufficient to ameliorate major pathological changes in DR and ROP. This may sound simple, but it is a practical challenge to main physiological levels with current anti-VEGF therapies for DR and other hypoxic retinal vascular diseases. Anti-VEGF clinical trials demonstrate huge variations in responses to long-term treatment, a portion of patients had a reduction in visual acuity. While this could be due to multiple factors, the so-called physiological levels of VEGF are most likely different from one patient to another, which are essential to protecting retinal cells and neurons and regulating neuronal function (35–41). In short, a suitable physiological level for one patient could potentially be harmful to retinal integrity in others. The goal of anti-VEGF therapies is to reduce pathological VEGF, which is achieved mostly through VEGF neutralizing antibodies. However, these drugs are antibodies by nature and could potentially affect VEGFRs-mediated physiological function. These concerns are being borne out by an increasing number of outcomes of anti-VEGF therapies for DR, DME, ROP, neovascular AMD, and other hypoxic retinal vascular diseases, which is the subject of detailed discussion below (Section 7).

5. Function of MG-specific VEGFR2 in DR

As VEGF may play a role in physiological functions in the retina, the long-term impact of anti-VEGF therapies on retinal integrity has been a major concern. However, it was a little surprising that disrupting MG-derived VEGF did not result in any apparent alteration in retinal morphology and in diabetes-induced neuronal degeneration (15, 18). On the other hand, VEGF is a secreted protein and a partial reduction in VEGF levels will not block VEGF signaling and is unlikely to affect the physiological function, such as MG viability and neuroprotection. We thus disrupted VEGFR2 in mouse MG by using a floxed VEGFR2 mouse and our Cre-driver mouse (17, 42), which resulted in VEGFR2-null (complete deletion) in 63 percent of MG in conditional VEGFR2 KO mice (43). Although there were no apparent changes in MG and neuronal densities and in morphology of control retinas, the conditional VEGFR2 KO mice demonstrated a gradual decrease in MG density starting approximately 4 months after STZ-injection. The reduction of MG density reached to 53 percent 10 months after STZ-injection (Table 3)(43). Consequently, there is a gradual loss of rod and cone photoreceptors and GCL neurons and reduction of retinal thickness. Ten months after STZ-injection, MG-specific VEGFR2 disruption resulted in a decrease in M-opsin-positive and S-opsin-positive cones by 43 and 41 percent, in ONL and INL thickness by 35 and 33 percent, and in GCL neurons by 57 percent (Table 3)(43).

Table 3.

Diabetes-induced alteration of retinal apoptosis, neurotrophin expression, and neuronal degeneration in MG-specific VEGFR2 KO mice. Most analyses were carried out in animals 10 months after STZ injection, except that AKT, BDNF, and GDNF expression and TUNEL-positive cells were examined in animals 4 months after STZ injection.

Phenotypical changes	Degree
MG density	Decrease by 47%
M-opsin+ cone density	Decrease by 43%
S-opsin+ cone density	Decrease by 41%
GCL cell density	Decrease by 57%
INL thickness	Decrease by 33%
ONL thickness	Decrease by 35%
TUNEL-positive Cells	Increase by 38%
Retinal activated AKT	Decrease by 57%
Retinal BDNF	Decrease by 63%
Retinal GDNF	Decrease by 64%

To delineate the underlying mechanisms of diabetes-induced loss of MG and neuronal densities in MG-specific VEGFR2 KO mice, we examined the behavior of primary VEGFR2-null MCs under diabetes-like conditions. The loss of VEGFR2 in primary MCs did not alter the number of apoptotic cells in cultures grown in media with normal level of glucose (43). However, there was a 38 increase in terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL)-positive VEGFR2-null primary cells grown in high glucose (HG) media (43), which is in agreement with a previous observation that pathologically high level of glucose is a significant contributor to MC apoptosis (44). Furthermore, the level of activated/phosphorylated AKT in primary MCs grown in HG media was reduced by 71 percent, which is supported by the observation that MG-specific VEGFR2 KO mice had a 57 percent reduction of activated AKT and a near two-fold elevation in TUNEL-positive apoptotic cells in the retina (43). These data suggest that VEGFR2-mediated AKT activation is at least partially responsible for MG viability under pathological conditions, such as DR, which is necessary for neuroprotection in DR.

6. Mechanism of MG-specific VEGFR2 signaling-mediated neuroprotection in DR

To determine the neuroprotective mechanism(s) of MG-specific VEGF/VEGFR2 signaling in DR further, we examined the levels of trophic factors in the retina of diabetic MG-specific VEGFR2-null mice. From a handful of trophic factors, brain-derived neurotrophic factor (BDNF), ciliary neurotrophic factor (CNTF), erythropoietin (EPO), glial cell line-derived neurotrophic factor (GDNF), and insulin-like growth factor 1(IGF-1), only the levels of BDNF and GDNF were significantly reduced 4 months after STZ injection (43). Compared with diabetic WT animals, the diabetic conditional Vegfr2 KO mice demonstrated additional 63 and 64 percent reduction of retinal BNDF and GDNF, respectively (Table 3). As diabetes reduces the levels of GDNF and BNDF, the retinal BDNF and GDNF concentrations in diabetic MG-specific VEGFR2 KO mice were actually 25 percent and 20 percent of that in both WT and/or conditional VEGFR2 KO mice under normal conditions (43). Since the dramatic reduction of retinal BDNF and GDNF occurred at a time when we did not observe any significant loss of retinal cells, MG, and neurons, VEGF/VEGFR2 signaling is most likely to affect the biosynthesis, secretion, or degradation of BDNF and GDNF in MG, which is certainly worthwhile to study further. We are actively investigating these leads. As BDNF and GDNF are trophic factors for MG (2), it is also important to reveal the relationship among BDNF, GDNF, and VEGF in promoting MG viability and neuroprotection. In cultured rat cells (rMC1, (45)), BDNF and GDNF had strong protective effects on MC viability under diabetic and hypoxic conditions (Figure 2A, B). However, they only showed marginal effects on promoting growth under normal conditions. Moreover, the effects of BDNF, GDNF, and VEGF on rMC1 cell viability appeared to be additive or synergistic under diabetic or hypoxic conditions. This result suggests that VEGF signaling may interact with BDNF- and/or GDNF-specific survival and proliferation, BDNF and GDNF production/secretion, and/or both, as shown in our working hypotheses (Figure 2C). This is not unusual as these trophic factors has been shown to act on motor neurons synergistically in Parkinson disease and ALS (46, 47). This potential synergistic effect enhances MG’s ability to sever as major supporting cells. Since MG demonstrated very similar responses to BDNF, GDNF, and VEGF in diabetes and hypoxia in vitro (Figure 2A, B), it is most likely that, with respect to support their own viability and neuroprotection, MG may utilize the same mechanism(s) in DR and hypoxic retinal vascular diseases. This feature will be very useful in conducting mechanistic studies, as hypoxic animal models require significantly shorter time to develop. Among the supporting functions, the decreases in retinal BDNF and GDNF are likely major contributors to diabetes/hypoxia-accelerated retinal neuron degeneration in MG-specific VEGFR2 KO mice (43)(Le et al. unpublished observation). This conclusion is supported by numerous observations showing the efficacies of BDNF and GDNF in neuroprotection under various stress and disease conditions (for review see (48)). It may be beneficial to provide neurotrophins and/or other trophic factors to anti-VEGF drugs for improving the safety of long-term treatment for DR and other hypoxic retinal vascular diseases.

Figure 2.

Potential mechanism of VEGF signaling-mediated MG viability and neuroprotection in DR and hypoxia. A, B: Effect of VEGF, BDNF, and/or GDNF on rMC1 cell viability (relative cellular density after 22 h) in high glucose (A) or hypoxia (B)(100 mM CoCl₂). NG: normal glucose (5 mM). HG: high glucose (25 mM) media. Growth factor supplement: 0.24 ng/mL. ***: p<0.001. **: p<0.01. **: p<0.05. Error bar: SD. p-values: represent pairwise comparison with vehicle by t-test. C: Putative mechanism of VEGF/VEGFR2 signaling-mediated MG viability and its role in neuroprotection in DR and hypoxic retinal diseases. Solid arrows: with supporting data. Broken arrows: direct experimental and clinical data are not available.

7. Importance of VEGF signaling to neuronal function and integrity during anti-VEGF therapies

Anti-VEGF strategies have been widely used to treat BRB breakdown associated pathological changes in DR, DME, ROP, wet-AMD, and other retinal diseases. As VEGF may also play a role in physiological functions in the retina, the long-term impact of anti-VEGF therapies on retinal integrity is a major concern at present. Unfortunately, there is not a huge wealth of information about the physiological functions of VEGF in the retina, except a few studies suggesting that VEGF is a protectant for MG, photoreceptors, RGCs, and the retinal pigment epithelium (RPE) (49–52). However, the lessons from VEGF, VEGFRs, and VEGF coRs-modulated functions in other part of the central nervous system (CNS) and in peripheral neurons could serve as important cues to the understanding of the mechanisms for VEGF as a functional regulator for retinal neurons. Gene targeting analysis suggests that VEGF and semaphorins-mediated neuropilin-2 signaling is important to the migration of gonadotropin-releasing hormone neurons in the neuroendocrine system (53). VEGF has been suggested to reduce excessive excitation of hypoglossal motor neurons by downregulating stimulus-evoked depolarization, a critical mechanism to prevent ALS (35). VEGF is capable of downregulating synaptic responses significantly in hippocampal principal neurons in rats, which may be critical to the onset of epilepsy (36). VEGF modulated synaptic transmission in hippocampus is mediated through VEGFR2 by inducing transient receptor potential of canonical Ca²⁺ channels and activation of calcium/calmodulin protein kinase II and mammalian target of rapamycin (37). Activating VEGF co-R, neuropilin-1, in the presynaptic membrane is also capable of reducing synaptic transmission in hippocampal synapses (38). These studies suggest that VEGF signaling may play a role in regulating neuronal function in the retina, an important question relevant to the effect of anti-VEGF drugs on visual acuity that has no literature to date. Therefore, it is important to keep in mind that the improvement in visual acuity immediately after anti-VEGF treatment could be contributed by VEGF’s direct effect on neuronal function.

Apart from many examples of functioning in axon guidance, which may be considered as a development issue (not a strict adult function), VEGF signaling has long been regarded as a neuroprotective mechanism for neurons in the brain. To name a few, it provides neuroprotection and prevents mitochondrial dysfunction and oxidative stress in hippocampal neurons (54). VEGF-mediated VEGFR2-AKT survival is critical to neuroprotection via hypoxic preconditioning in cerebellar granule neurons (55). VEGF has been suggested as a survival factor for neurons in the dentate gyrus and subventricular zone after focal cerebral ischemia, a stroke-like stimulus (56). The VEGF-activated VEGFR2-AKT survival is neuroprotective to RGCs after axotomy or under ischemic condition (49, 52). These examples indicate that VEGF-mediated survival pathway is essential to protecting retinal neurons under diabetic/hypoxic conditions, which is characteristically similar to that in patients with DR, DME, AMD, and ROP. Anti-VEGF drugs could block VEGF-mediated survival and promote geographic atrophy-like pathology and retinal, choroidal, and scleral thinning, which are clinical outcomes likely caused by altered VEGF signaling frequently seen in anti-VEGF trial patients(57, 58)

Anti-VEGF clinical trials for DR have a shorter history and may not have sufficient information about long-term (5 year+) data for retinal integrity. However, the observation that a VEGF neutralizing antibody results in retinal degeneration in experimental diabetes (59) and that short hairpin (sh) RNA targeting VEGF in MG causes retinal thinning in an ROP model strongly suggest a potential safety risk in longer term anti-VEGF therapies for proliferative retinopathy in DR, and perhaps for ROP as well (60). As a hypoxic retinal environment is common to DR, AMD, and ROP, the lessons from long-term anti-VEGF clinical trials for neovascular AMD could mirror what occurs in DR. Recent publications on anti-VEGF clinical trials have reached a consensus that, while the drugs may be effective in reducing BRB breakdown and in improving visual acuity in some patients, average gain of visual acuities has not been maintained for wet-AMD patients after 5-year anti-VEGF treatment, in which a sizable number of individuals have a significant reduction of visual acuity at the end of 5-year trials (57, 61) Such an outcome has also been observed in two 5-year DME clinical trials (62, 63). While the loss of visual acuity in wet-AMD and DME patients after long-term anti-VEGF therapies could be the consequence of many reasons, the loss of MG-specific VEGF signaling-mediated general retinal support could be a major one, which is supported by a striking resemblance between accelerated degeneration of all retinal neurons in diabetic/hypoxic MG-specific VEGFR2 KO mice and the abnormally thin retinas (thinning in all retinal layers) in a significant portion (36 percent) of the patients after 5-year anti-VEGF treatments for wet AMD (43, 57). It is important to point out that, while diabetes/hypoxia induced neuronal degeneration in all retinal layer in the WT controls, disrupting VEGF/VEGFR2 signaling in MG caused a substantially greater loss of all retinal neurons in conditional VEGFR2 KO mice at late stage of diabetes that had a hypoxic retinal environment (43). These observations clearly elevate the significance of VEGF signaling in MG for neuroprotection during anti-VEGF treatment. As BDNF and GDNF support MG viability, providing neurotrophins and/or other trophic factors during the treatment may be a feasible strategy for improving the long-term safety of anti-VEGF therapies for DR, AMD, and other hypoxic retinal vascular diseases.

8. Concluding remarks

Work from others and our laboratory indicates the importance of MG in the development of DR. Through the production of VEGF, MG play a major role in protein alteration and modification, inflammation, neovascularization, vascular leakage, and vascular lesion in diabetic retinas (Figure 1). Our recent work also suggests an essential role for VEGF signaling in MG viability and neuroprotection in diabetic/hypoxic retinas (Table 3). Therefore, MG is a major modulator for vascular function and neuronal integrity in these diseases. It will be of great interest if including neurotrophins and perhaps other trophic factors with anti-VEGF drugs is beneficial to preserve neuronal viability in patients subjected to long-term anti-VEGF treatment for DR, DME, neovascular AMD, ROP, and other hypoxic retinal diseases.

The lessons from the work on glia in CNS and peripheral neurons suggest that VEGF signaling may be important in regulating gliosis and neuronal functional placidity, in the retina of DR and other hypoxic retinal diseases (2, 32), which has not been seriously explored. Revealing these “new” functions of VEGF signaling in MG is essential to the understanding of the mechanisms and therapeutics of DR and other hypoxic retinal diseases. With the tools in investigating MG and their relationship with neuronal viability in diabetic and hypoxic models (64–66), these questions should be addressed shortly. Our data suggest that VEGF signaling may be important to MG viability in diabetes/hypoxia. It is not clear whether such viability is solely contributed by VEGF signaling-mediated survival mechanisms, most likely that MG proliferation may also play a role in such a process (Figure 2C). In principle, mammalian MG can act as progenitors and convert themselves into various retinal neurons through dedifferentiation, proliferation, and differentiation under various stresses (67). Potentially, VEGF signaling may be critical to MG-derived neurogenesis during hypoxic and diabetic stresses in which retinal neuronal degeneration occur (68), another way to protect neurons in DR, AMD, and hypoxic retinal vascular diseases.