Introduction

Diabetic retinopathy is a leading cause of blindness worldwide, affecting patients with both type 1 and type 2 diabetes mellitus. Diabetic retinopathy develops in over 95% of patients with type 1 Diabetes Mellitus and 60% in those with type 2 Diabetes Mellitus over 25 years (Ronald Klein et al., Reference Ronald Klein, Knudtson, Lee, Gangnon and Klein2008). Approximately 93 million people have diabetic retinopathy worldwide including both nonproliferative and proliferative forms, those involving neovascularization, with 17 million having proliferative diabetic retinopathy (Yau et al., Reference Yau, Rogers, Kawasaki, Lamoureux, Kowalski, Bek, Chen, Dekker, Fletcher, Grauslund, Haffner, Hamman, Ikram, Kayama, Klein, Klein, Krishnaiah, Mayurasakorn, O’Hare, Orchard, Porta, Rema, Roy, Sharma, Shaw, Taylor, Tielsch, Varma, Wang, Wang, West, Xu, Yasuda, Zhang, Mitchell and Wong2012). Indeed, diabetic retinopathy remains a leading cause of blindness among working age adults (Fong et al., Reference Fong, Aiello, Gardner, King, Blankenship, Cavallerano, Ferris and Klein2003) and in the United States up to one third of the total adult population could have diabetes by the year 2050 (Boyle et al., Reference Boyle, Thompson, Gregg, Barker and Williamson2010).

Diabetic retinopathy has long been described as a microvascular disease with disease progression classified by retinal vascular changes. Indeed, characterization of the disease depends on observations based on the Airlie House classification scale of clinically visible vascular abnormalities which in nonproliferative diabetic retinopathy include lesions such as: microaneurysms, venous beading, and intraretinal microvascular abnormalities, and in proliferative diabetic retinopathy involves observation of neovascularization (1991). Changes to the barrier permeability are believed to contribute to macular edema (reviewed in [Antonetti et al. Reference Antonetti, Klein and Gardner2012]). Normally, the retina promotes a well-developed junctional complex and strict control of transcellular transport in the retinal blood vessels and across the retinal pigmented epithelium creating the blood-retinal barrier (BRB) that helps to maintain the proper neural environment. Loss of this BRB in diabetic retinopathy contributes to increased retinal vascular permeability and edema leading to vision loss (Klein et al., Reference Klein, Klein, Moss and Cruickshanks1998; Moss et al., Reference Moss, Klein and Klein1998) Clearly, the use of anti-vascular endothelial growth factor (anti-VEGF) therapies helping to control both macular edema and neovascularization point to the importance of these vascular changes to vision loss in diabetic retinopathy (Ajlan et al., Reference Ajlan, Silva and Sun2016; Li et al., Reference Li, Veerappan, Mittal and Do2020). However, these clinical studies also reveal anti-VEGF therapies are effective in only approximately half of the patients treated. Clearly new approaches are still needed.

Longitudinal studies of patients with diabetes mellitus suggests important neural changes that may occur before or separately from vascular abnormalities. Thinning of the nerve fiber and ganglion cell layer, termed retinal neurodegeneration, are observed without evidence of vascular pathology (Sohn et al., Reference Sohn, van Dijk, Jiao, Kok, Jeong, Demirkaya, Garmager, Wit, Kucukevcilioglu, van Velthoven, DeVries, Mullins, Kuehn, Schlingemann, Sonka, Verbraak and Abramoff2016). Changes in retinal function also have been documented before clinically observable vascular pathology (Gardner et al., Reference Gardner, Abcouwer, Barber and Jackson2011; Lynch & Abramoff, Reference Lynch and Abramoff2017). These clinical observations in conjunction with the identification of neural retinal cell death within a month after streptozotocin induction of diabetes in rats (Barber et al., Reference Barber, Lieth, Khin, Antonetti, Buchanan and Gardner1998) have led the scientific and clinical community to increasingly recognize both neuronal and vascular changes that occur to the retina during diabetes and has led to studies focused more on understanding the retina as a complex functional unit with a high degree of interaction between neuronal and vascular function.

The concept of a neurovascular unit is attributed to the first Stroke Progress Review Group meeting of the National Institute of Neurological Disorders and Stroke of the National Institutes of Health (July 2001) (Iadecola, Reference Iadecola2017). The NVU originally focused on the neuronal and vascular interaction encompassing the concept of barriergenesis or promotion of the blood–brain barrier and neurovascular coupling, or the regulation of blood flow to regions of neural activity. The concept has broadened overtime to include the interactions of neurons, glia, and microglia with pericytes, and vascular endothelium, in order to maintain the highly specialized environment required for neuronal signaling. This concept easily translates to the retina and has become a focus of research exploring the changes induced by diabetes. Studies of the NVU include understanding how changes in signaling from the neural tissue regulate vascular growth and barriergenesis and how these are altered by diabetes. Research has advanced to help identify some of the signaling components necessary for formation of the BRB. The canonical Wnt signaling pathway has been demonstrated to be critical to formation of the BRB with gene deletion analysis demonstrating Norrin produced from glial cells is required to stimulate canonical Wnt signaling through Frizzled 4 receptor complex and barriergenesis (Wang et al., Reference Wang, Rattner, Zhou, Williams, Smallwood and Nathans2012; Zhou et al., Reference Zhou, Wang, Tischfield, Williams, Smallwood, Rattner, Taketo and Nathans2014). Indeed, Norrin replacement therapy can reverse diabetes or VEGF induced permeability in rats (Diaz-Coranguez et al., Reference Diaz-Coranguez, Lin, Liebner and Antonetti2020). Additional signals for barrier induction come from pericytes. Endothelial recruitment of pericytes through platelet derived growth factor B signaling and pericyte coverage of retinal vessels is required for proper retinal barrier formation (Lindblom et al., Reference Lindblom, Gerhardt, Liebner, Abramsson, Enge, Hellstrom, Backstrom, Fredriksson, Landegren, Nystrom, Bergstrom, Dejana, Ostman, Lindahl and Betsholtz2003; Daneman et al., Reference Daneman, Zhou, Kebede and Barres2010). In diabetes, preventing loss of the barrier by providing anti-vascular endothelial growth factor therapies such as blocking antibodies, or receptor traps can reduce edema, prevent neovascularization, and provide protection from further vision loss or improve vision in approximately half of treated patients (Simo et al., Reference Simo, Sundstrom and Antonetti2014; Ajlan et al., Reference Ajlan, Silva and Sun2016). While the importance of proper vascular function has been demonstrated in diabetic retinopathy, how vascular abnormalities may lead to altered neuronal function and loss of vision remains an area of much needed research.

The use of targeted genetic studies in mice provides one approach to begin to unravel the complex interaction of the cells of the neurovascular unit. By using conditional gene deletion of specific cell types the response of the neurovascular unit may be measured and the interdependency of the neurovascular unit can begin to be unraveled. The dependency of the neuronal function on vascular barrier properties has begun to be investigated with targeted tight junction protein gene deletion. For example, targeted adeno-associated virus-mediated suppression of claudin-5 expression in adult mice led to impairments in learning, memory, and anxiety-like behavior with eventual seizures and death by 3–4 weeks (Greene et al., Reference Greene, Kealy, Humphries, Gong, Hou, Hudson, Cassidy, Martiniano, Shashi, Hooper, Grant, Kenna, Norris, Callaghan, Islam, O’Mara, Najda, Campbell, Pachter, Thomas, Williams, Humphries, Murphy and Campbell2018) demonstrating the requirement for the blood brain barrier in neuronal function and, coupled with genetic association, suggests changes in the vascular component may alter neural function in schizophrenia. Conversely the role of regulation of oxidative stress in a cell specific manner has been investigated in the retina. Genetic studies in mice reveal knock out of NF-E2-related factor 2 (NRF2), a transcription factor that is a master controller for expression of a broad array of anti-oxidant genes, increased the loss of vessels and the ischemic region as well as increased neovascularization in the oxygen-induced retinopathy model (Wei et al., Reference Wei, Gong, Xu, Thimmulappa, Mitchell, Welsbie, Biswal and Duh2015). Further, the effect could be recapitulated using neuronal specific deletion of NRF2. These studies went on to demonstrate a signaling requirement of semaphorin (Sema) 6A (Wei et al., Reference Wei, Gong, Xu, Thimmulappa, Mitchell, Welsbie, Biswal and Duh2015). Interestingly, both Sema 3A (Cerani et al., Reference Cerani, Tetreault, Menard, Lapalme, Patel, Sitaras, Beaudoin, Leboeuf, De Guire, Binet, Dejda, Rezende, Miloudi and Sapieha2013) and Sema 4D were shown to contribute to vascular permeability in diabetic retinopathy (Wu et al., Reference Wu, Li, Chen, Hong, Zhang, Wang, Zhou, Li, Wang, Mao, Xia, He, Jin, Yue and Hu2020). Sema 3A required neuropilin receptor for induction of permeability as demonstrated in knockout studies (Cerani et al., Reference Cerani, Tetreault, Menard, Lapalme, Patel, Sitaras, Beaudoin, Leboeuf, De Guire, Binet, Dejda, Rezende, Miloudi and Sapieha2013). Sema 4D was elevated in animal models and in aqueous fluid of eyes from patients with diabetes and specifically inversely correlated in patients that responded poorly to anti-VEGF treatment (Wu et al., Reference Wu, Li, Chen, Hong, Zhang, Wang, Zhou, Li, Wang, Mao, Xia, He, Jin, Yue and Hu2020). Gene deletion studies revealed Sema 4D contributes to diabetes induced vascular permeability and targeted antibodies against Sema 4D were additive with anti-VEGF antibodies in effectively reducing diabetes induced vascular permeability in animal models (Wu et al., Reference Wu, Li, Chen, Hong, Zhang, Wang, Zhou, Li, Wang, Mao, Xia, He, Jin, Yue and Hu2020). Semaphorins act in guidance of the developing nervous system and these studies reveal a role in vascular function in the retina and a potential role in diabetic retinopathy (Fu et al., Reference Fu, Sun, Cakir, Tomita, Huang, Wang, Liu, Cho, Britton, Kern, Antonetti, Hellstrom and Lois2020).

Studies have also begun to implicate the outer retina specifically in the pathology of diabetic retinopathy. Targeting photoreceptor signal transduction by knockout of guanine nucleotide binding protein, alpha tranducin 1 (Gnat1), provides evidence for outer retinal contribution to diabetic retinopathy. Diabetic retinopathy leads to a hallmark retinal vessel degeneration. The Gnat1 gene deletion animals led to a basal increase in vessel degeneration but prevented any further increase from diabetes (Liu et al., Reference Liu, Tang, Du, Saadane, Samuels, Veenstra, Kiser, Palczewski and Kern2019). Further, inhibition of the visual cycle and reducing outer retinal function, using retinylamine, was protective of both vessel degeneration and permeability in diabetes (Liu et al., Reference Liu, Tang, Du, Lee, Golczak, Muthusamy, Antonetti, Veenstra, Amengual, von Lintig, Palczewski and Kern2015). Studies of diabetic patients that are resistant to diabetic retinopathy has led to new insight into potential protective proteins. The visual cycle protein, retinol binding protein 3 (RBP3) expression is inversely related to diabetic retinopathy severity in humans and injection or photoreceptor driven expression of RBP3 in mice can reverse diabetes induced permeability and visual changes measured by electroretinogram (Yokomizo et al., Reference Yokomizo, Maeda, Park, Clermont, Hernandez, Fickweiler, Li, Wang, Paniagua, Simao, Ishikado, Sun, Wu, Katagiri, Pober, Tinsley, Avery, Feener, Kern, Keenan, Aiello, Sun and King2019). Collectively, these studies highlight the close association of the retinal neural function with the highly specialized vasculature and emphasize the need for careful analysis using targeted genetic approaches to begin to understand the specific interaction of the retinal neurovascular unit.

While there is a clear interaction of the retinal neural and vasculature cells for a proper functioning unit, clinical data suggest there may be patient populations with distinct disease phenotypes that may not include changes in the vessels that are easily observed by current methods. A cross-sectional study by the European Consortium for the Early Treatment of Diabetic Retinopathy found that 61% of patients without microvascular disease presented with abnormalities related to neurodegeneration assessed by multifocal electroretinogram or structural changes observed by optical coherence tomography (Santos et al., Reference Santos, Ribeiro, Bandello, Lattanzio, Egan, Frydkjaer-Olsen, Garcia-Arumi, Gibson, Grauslund, Harding, Lang, Massin, Midena, Scanlon, Aldington, Simao, Schwartz, Ponsati, Porta, Costa, Hernandez, Cunha-Vaz and Simo2017). Conversely, 32% of patients with visible microvascular disease did not present with any sign of neurodegeneration. Clearly additional studies are needed including longitudinal assessment, but these data suggest distinct subtypes of diabetic retinopathy that may require targeted therapeutic approaches.

The study of diabetic retinopathy has largely focused on understanding changes to the microvasculature. This approach has proven highly effective with anti-VEGF therapies providing preservation of vision for many patients. However, new approaches are still needed. Ultimately, the retinal microvasculature supports retinal neural function. To achieve this, the neuronal and vascular cells together form a unit with a high degree of cell communication and interaction creating the proper environment for neuronal signaling and vision. How the retinal vasculature and neuronal changes in diabetes relate to the loss of the neurovascular unit and proper visual function is an area of active investigation and further studies in this area are greatly needed to develop novel treatment options. The studies herein address more specifically the fundamental changes to the neural cells in diabetic retinopathy and begin to clarify what changes in neural metabolism and interaction may contribute to vision loss in diabetic retinopathy.