The role of inflammation and neurodegeneration in diabetic macular edema.

The pathogenesis of diabetic macular edema (DME) is complex. Persistently high blood glucose activates multiple cellular pathways and induces inflammation, oxidation stress, and vascular dysfunction. Retinal ganglion cells, macroglial and microglial cells, endothelial cells, pericytes, and retinal pigment epithelium cells are involved. Neurodegeneration, characterized by dysfunction or apoptotic loss of retinal neurons, occurs early and independently from the vascular alterations. Despite the increasing knowledge on the pathways involved in DME, only limited therapeutic strategies are available. Besides antiangiogenic drugs and intravitreal corticosteroids, alternative therapeutic options tackling inflammation, oxidative stress, and neurodegeneration have been considered, but none of them has been currently approved.

Introduction

Diabetic macular edema (DME) is the leading cause of vision loss in diabetics, and it may complicate any stage of diabetic retinopathy (DR). In a large prospective study enrolling 366 patients with type 1 diabetes (T1DM) and 15,030 with type 2 diabetes (T2DM), the annual incidence of DME was similar between the two groups (2.68% in T1DM and 2.22% in T2DM), while the sum incidence at 9 years was slightly higher in patients with T1DM (8.46% versus 6.36%, respectively). Despite increased awareness of the risk factors and the implementation of screening practices in diabetic patients, DME is still a considerable burden in the working-age population. DR and DME share high glycosylated hemoglobin (HbA1c), long duration of the disease, poor glycemic control, and hypertension as risk factors.

The retinal neurovascular unit (RNU) refers to the complex interaction between retinal neurons (ganglion cells, bipolar cells, horizontal cells, and amacrine cells), vascular cells (endothelial cells and pericytes), glial cells (astrocytes and Müller cells), and immune cells (microglia). The RNU is essential for maintaining homeostasis, modulating neuronal function, and coordinating the retinal vascular flow with metabolic activity. Persistently high blood glucose levels and oxidative stress trigger a chronic intraretinal inflammatory response and neuronal degeneration, which eventually lead to RNU dysfunction and blood–retinal barrier (BRB) breakdown.

The therapeutic options available for DME are limited; understanding the DME pathogenesis may help guide the treatment in the future with a patient-oriented therapeutic approach. This article aims to review the pathways involved in DME, focusing on the cellular, molecular, and biochemical basis of inflammation and neurodegeneration. Then, the imaging biomarkers and the therapeutic choices on the horizon are discussed.

Inflammation

Inflammation is a standardized response against different agents (pathogens, traumas, toxic substances) and involves several cells and chemical mediators, with a cascade mechanism. In normal conditions, pro-inflammatory and anti-inflammatory stimuli mutually balance, and healing processes are promoted. This homeostasis is altered in diabetes, with sustained inflammation leading to subtle, irreversible tissue damage.

The cellular basis of inflammation

Microvascular pathology

The blood vessels supply nutrients and oxygen to neurons and eliminate metabolic wastes and carbon dioxide; the vascular endothelial cells make a semi-selective monolayer at the inner surface of the vessels, known as the inner BRB. Reduced expression of the tight junction proteins that form the inner BRB has been observed in human retinal endothelial cells exposed to hyperglycemic conditions and diabetic animal models.

The retinal capillaries are enveloped by pericytes, which maintain the integrity of the inner BRB, and an extracellular matrix known as endothelial basement membrane (EBM), which provides mechanical stability and interaction between endothelial cells and pericytes.[9-11] The pericytes regulate the blood flow and secrete inflammatory mediators promoting immune cells adhesion, extravasation, and migration into the extracellular matrix. The pericytes express the major histocompatibility complex (MHC) class 1; they can express MHC class II in selected circumstances, supporting their role as antigen-presenting cells.[12,13]

EBM thickening, pericytes and endothelial loss, and leukostasis are the main histopathology changes occurring in DR, and all have been observed since the early stages of the disease. All lead to retinal vessel weakening and increased vascular permeability. Retinal leukostasis promotes capillary occlusion and thrombosis, worsening the clinical picture.[15,16] Furthermore, inflammation promotes EBM thickening by increasing the expression of extracellular matrix proteins in the basement membrane. As a proof of concept, inhibition of intercellular adhesion molecule (ICAM)–mediated leukocyte adhesion leads to downgrading or regression of DR in mice with diabetes.

Retinal macroglia and microglia

Retinal glia is mainly composed of three cellular types: astrocytes, Müller cells, and resident microglia.

Müller cells and astrocytes are the primary cells of the retinal macroglia; glial cells actively maintain retinal homeostasis and are involved in retinal metabolism.[19,20] Retinal glia shares a common origin with the brain glia; both Müller cells and astrocyte damage have been detected since the early stages of DR.[21,22] Müller cells are radially oriented and span from the external to the internal limiting membrane, interconnecting neurons, and vascular cells. Astrocytes encircle the neuronal axons and the vessels, and support the inner BRB. Similar to Müller cells, astrocytes modulate neuronal and the vascular function and regulate the vascular blood tone.

The retinal macroglia contributes to metabolic support, electrolyte balance, and protection against oxidative stress. Müller cells produce interleukins (ILs), chemokines, and vascular endothelial growth factor (VEGF) and contribute to local immune surveillance.[20,24,25] These cells also regulate the expression of aquaporins, which are essential in fluid homeostasis. Müller cells undergo reactive gliosis in hyperglycemia, which presents as cellular hypertrophy, proliferation, and increased production of intermediate filament protein (such as nestin and glial fibrillary acidic protein). Müller cells reactive gliosis is initially advantageous, as it aims at preventing tissue damage; eventually, reactive gliosis induces irreversible glial scars.

Microglial cells are involved in the homeostasis of the retina in normal conditions; under chronic insults, such as diabetes, microglial cells become activated, modify their appearance into an ameboid form, and acquire the ability to migrate into the retina. Activated microglial cells also migrate from the retina to the choroid by a transcellular route, through an atypical protein kinase Cζ (PKCζ) expressed by the retinal pigment epithelium (RPE). Accumulation of microglial cells in the outer retina has been observed in T1DM mouse models, with evidence of proliferation and reactivity in the choroidal stroma. In response to tissue damage, the microglia releases anti-inflammatory cytokines, such as IL-4, IL-10, IL-13, and transforming growth factor–beta (TGF-β); activated microglia, instead, produce pro-inflammatory and cytotoxic factors, such as tumor necrosis factor (TNF)-α, IL-1β, reactive oxygen species (ROS), and reactive nitrogen species. These mediators contribute to neuronal cell dysfunction and pericytes and endothelial cell injury, resulting in the inner BRB breakdown.

Diverse factors contribute to the activation of the microglia, including prolonged hyperglycemia and other sources of tissue stress. Recently, the renin–angiotensin–aldosterone system (RAAS) has also been involved in the activation of retinal microglia (see ‘RAAS’ section). Although there is no specific treatment reverting the physiologic role of microglia, the pathways involved in its recruitment and activation may become an attractive therapeutic target in the future.

RPE and choroid

The RPE is a cell monolayer that transports nutrients, ions, and water between the photoreceptors and the choriocapillaris. The RPE is also responsible for the phagocytosis of photoreceptors’ outer segments, the conversion of all-transretinal into 11-cis-retinal, the absorption of light, and the protection against photo-oxidation. The RPE intercellular tight junctions, in association with the choriocapillaris, the Bruch’s membrane, and the extracellular matrix, form the outer BRB. On the contrary, the choroid supplies blood to the outer retina, the photoreceptors, and the RPE, and it is responsible for the retinal metabolic exchanges.

Diabetic animal models showed a reduction in RPE-mediated active transportation of ions and water from the subretinal space to the choroidal bloodstream. Hyperglycemia reduces the expression of the tight junction proteins between RPE cells.[37,38] In addition, it stimulates RPE cells to upregulate VEGF and other inflammatory mediators in diabetic conditions, contributing to the development and progression of DME.

Impaired choroidal blood flow has been reported in both diabetic patients and animal models.[39,40] Loss of choriocapillaris, tortuous blood vessels, microaneurysms, drusenoid deposits in the Bruch’s membrane, and neovascularization occur in the diabetic choroid. Altogether, these changes have been addressed as diabetic choroidopathy. The diabetic choroid overexpresses leukocyte adhesion molecules, promoting the migration of leukocytes into the retina.[42,43]

Recent studies conducted on T2DM animal models showed increased choroidal thickness with decreased vascular density in the innermost choroid, increased VEGFR2 immunoreactivity in the retina, and increased inflammatory cell density in the outer retina. Differently, in T1DM animal models, no changes were demonstrated in choroidal thickness and choroidal vascular density during the early stages of the disease.

In vivo human studies, relying on optical coherence tomography (OCT), have revealed contradictory results regarding the choroidal thickness in different stages of DR. Thus, the choroidal thickness is not yet a reliable parameter of diabetic choroidopathy in clinical practice. Other quantitative parameters, such as the choroidal volume or the choroidal vascularity index, have been proposed, but their clinical validity must be conclusively demonstrated.

The molecular basis of inflammation

The cellular alterations seen in DME are the result of inflammatory cytokines secreted by glial cells, RPE, macrophages, and activated leukocytes. The knowledge of the cytokines involved in DME is relevant from the therapeutic perspective, as periocular or intravitreal agents might be specifically designed for their inhibition.

Higher figures of VEGF, IL-1, TNF-α, endothelin-1, IL-6, IL-8, interferon (IFN)-γ-induced protein-10, monocyte chemotactic protein-1, C-C chemokine receptor types 2 and 5 have been measured in the vitreous samples and the aqueous humor of eyes with DR, compared with non-diabetic patients.[46,47] This cytokine imbalance starts from subclinical DR and increases steadily with worsening retinopathy; a significant difference in inflammatory mediators’ concentration has been found comparing diabetics with no DR and healthy population and diabetics without DR with those with mild DR.

Eyes with DME have higher aqueous and vitreous levels of inflammatory and pro-angiogenic cytokines compared with healthy controls or diabetic patients with no DR.[48-51] The levels of these molecules correlate with DME severity, retinal thickness on OCT, and the amount of leakage on fluorescein angiography.[49,50] For instance, the placental growth factor (PGF) is a homolog of VEGF that binds to VEGFR-1 and promotes angiogenesis by inducing the growth and migration of endothelial cells. PGF modulates inflammation and induces the chemotaxis of monocytes and macrophages. Higher levels of PGF levels in aqueous humor correlate to DME severity. Although intravitreal aflibercept inhibits PGF secretion and its effect, no approved antiangiogenic agent exclusively targets this pathway.

Recently, scientific interest has also been focusing on the angiopoietin/Tie2 pathway. Angiopoietin 2 (Ang-2) competes with Angiopoietin 1 (Ang-1) for the binding to the Tie2 tyrosine-kinase receptors of the endothelial cells. While Ang-1 stabilizes the blood vessels, Ang-2 reduces Tie2 activation, enhances the VEGF response, and promotes vascular leakage and neovascularization.[58,59] Pharmacological agents blocking Ang-2/Tie2 activation have theoretical antifibrotic, neuroprotective, anti-inflammatory, and vascular-stabilizing properties. Nevertheless, only a molecule, Faricimab, is currently in a phase III trial for commercial approval (see ‘Anti-VEGF’ section).

The biochemical basis of inflammation

Advanced glycation end products

Advanced glycation end products (AGEs) are biological macromolecules (proteins, lipids, or DNA) that become glycated after exposure to sugars. The formation of AGEs in diabetic patients increases in response to high glucose blood levels; AGEs are also induced by activation of the RAAS. Additional sources of AGEs are smoke and diet, with heated foods and high lipid and protein content.

In normoglycemic animals, blood-infused AGEs accumulate in the retina, induce diabetic-like retinal lesions, and upregulate VEGF, promoting angiogenesis. When they are in excess, AGEs cause protein structural alterations (e.g. cross-linking), affecting enzymatic activity, receptor recognition, and physiological protein turnover. AGEs damage the endothelial junctional molecules (occludins and cadherins), directly activate leukocytes, increase the production of pro-inflammatory cytokines and chemokines (such as IL-6, IL-1β, TNF-α, and monocyte chemoattractant protein-1 [MCP-1]), and promote the upregulation of the vascular cell adhesion molecule 1 (VCAM-1) and intercellular adhesion molecule 1 (ICAM-1) on the endothelial cell surface.

AGEs act on retinal cells either by a receptor-independent or a receptor-dependent pathway. Vascular endothelial cells, pericytes, microglia, Müller cells, and RPE cells constitutively express the receptor for advanced glycation end products (RAGEs); however, RAGEs are upregulated in diabetic patients. Ligand binding activates a network of intracellular signaling pathways based on Ras, mitogen-activated protein kinases (MAPKs), and nuclear factor-κB (NF-κB), resulting in inflammation, neurodegeneration, microvascular dysfunction, and apoptosis. The AGE-RAGE activation induces ROS formation also through activation of the reduced nicotinamide adenine dinucleotide phosphate (NADPH) oxidase enzyme and impairs antioxidant systems; AGEs production increases under oxidative conditions, amplifying this mechanism.

At present, there is no approved therapeutic strategy addressing the AGE-RAGE signaling. Several molecules have been investigated, including AGE cross-link breakers, AGE inhibitors, RAGE antagonists, antidiabetic drugs, antihypertensive drugs, statins, sevelamer, nutrition, and phytotherapy. Almost all those pharmacological options showed promising results in preventing or regressing diabetic complications in preclinical studies. However, those results could not be reproduced or were partially achieved in clinical trials.

Polyol pathway

In normal conditions, glucose is metabolized through the hexokinase pathway. In diabetes, high glucose levels saturate the hexokinase pathway, and the excess glucose is shunted into the polyol pathway, which converts hexose sugars into sugar alcohols (polyols). The polyol pathway is a two-step process. Glucose is reduced to sorbitol by the aldose reductase (the rate-limiting enzyme of the polyol pathway), oxidizing NADPH to NADP+. Then, sorbitol dehydrogenase converts the sorbitol into fructose, producing reduced nicotinamide adenine dinucleotide (NADH) from NAD+.

The activation of the polyol pathway lowers the concentration of NADPH and NAD+, both required for the regeneration of glutathione, the main antioxidant tripeptide of human cells. This leads to the overproduction of ROS and increased levels of oxidative stress. Because sorbitol is strongly hydrophilic and less permeable through cell membranes, it accumulates within the retinal cells and induces osmotic damage. Moreover, part of the fructose produced in the polyol pathway is phosphorylated to fructose-3-phosphate, which, in turn, is broken into 3-deoxyglucosone, a potent glycolyzing agent and AGE promoter.

Hexosamine pathway

The hexosamine biosynthetic pathway is a branch of the glycolysis producing uridine diphosphate N-acetylglucosamine (UDP-GlcNAc) from the glycolytic intermediate fructose-6-phosphate. UDP-GlcNAc is a key substrate for macromolecule glycosylation, required for the biosynthesis of glycoproteins, glycolipids, proteoglycans, and glycosaminoglycans. In normal conditions, only a small amount of glucose is metabolized into UDP-GlcNAc. In hyperglycemia states, the hexosamine pathway is overactive, with an excess of protein glycosylation.

Increased O-GlcNAcylation of p53 leads to pericyte loss and microvascular dysfunction, while increased O-GlcNAcylation of NF-kB promotes retinal ganglion cell apoptosis. UDP-GlcNAc may compete with phosphorylation at post-translation modification sites of transcription factors, altering the regulation of anti-inflammatory mediators. Finally, the glycosylation of RNA polymerase-II transcription factors may lead to the dysregulated translation of proteins involved in DR and DME pathophysiology.

PKC pathway

Hyperglycemia causes the accumulation of diacylglycerol (DAG) in the retina. DAG activates different isoforms of PKC. Hyperglycemia activates PKC also through the AGE-RAGE pathway and the polyol pathway.[78,79]

The activation of the PKC pathway contributes to retinal vascular dysfunction and pericyte loss.[80,81] Furthermore, the PKC signaling pathway alters the extracellular matrix and induces angiogenesis and leukocyte adhesions. PKC activation decreases retinal blood flow through a mechanism mediated by endothelin A, a potent vasoconstrictor. Finally, the PKC pathway can aggravate the oxidative stress levels and downregulate essential survival signaling pathways, promoting apoptosis.[81,84]

Poly(ADP-ribose) polymerase pathway

Poly(ADP-ribose) polymerase (PARP) is a family of proteins involved in DNA repair, genomic stability, and apoptosis. PARPs detect single-strand DNA breaks and initiate a cellular response of DNA repair. Oxidative stress induced by hyperglycemia may damage DNA through a ROS-mediated mechanism. The excessive activation of PARP leads to cellular depletion of NAD+ and changes in transcriptional regulation and expression of several genes implicated in DR and DME.

RAAS

The RAAS is the endocrine system essential in blood pressure, fluid, and electrolyte regulation. Understanding the role of RAAS alterations in the pathogenesis of DME has important therapeutic implications, as the key components of RAAS may be targeted pharmacologically. The RAAS begins with the production of pre-prorenin by the juxtaglomerular cells of the kidney. Pre-prorenin is cleaved to prorenin and released as prorenin or further processed to active renin. Renin converts angiotensinogen (produced by the liver) to angiotensin I, which is cleaved into Ang II by the angiotensin-converting enzyme (ACE). Ang II is the main effector of RAAS and is involved in the regulation of vascular function and blood pressure; it binds to the angiotensin type 1 receptor (AT1R) and the angiotensin type 2 receptor (AT2R). Ang II stimulates the release of aldosterone from the adrenal glands, and aldosterone interacts with the mineralocorticoid receptor (MR) to control electrolytes and fluid.

The retina has its local RAAS, which is separated from the systemic RAAS by the BRB. RAAS components are expressed on retinal microvessels (endothelial cells and pericytes), glia (Müller cells, astrocytes, microglia), and neurons (ganglion cells, amacrine cells, bipolar cells, and photoreceptors); retinal RAAS components have been also identified in the choroid, the ciliary body, and the RPE.[90,91] An imbalance in the activation of the local RAAS has been found in eyes with DR and DME. Both ACE and Ang II increase the level of VEGF and stimulate local inflammation (microglia) and oxidative stress. In turn, increased RAAS activity is amplified by the AGEs. RAAS components induce the proliferation and activation of glial cells. Aldosterone induces the expression of specific sodium and potassium channel in Müller cells, and the expression of the transmembrane water channel aquaporin 4. Moreover, aldosterone is involved in the development of retinal ischemia and choroidal blood flow regulation.

AT1R blockers or ACE inhibitors prevent retinal new vessel formation in DR. These molecules also have cytoprotective effects.[96-98] RPE cells treated with aliskiren, a direct renin inhibitor, showed reduced oxidative stress and an increased expression of antiangiogenic factors. On the contrary, MR antagonists have shown a beneficial effect in retinal vascular pathologies, including DME, in animal models. Finally, melatonin, a hormone produced by the pineal gland, has shown a strong antioxidant effect and can inhibit the activation of RAAS, thus reducing its detrimental effects.

In summary, RAAS pathway has a relevant pathogenetic role in DR, stimulating blood vessel growth and leading to vascular leakage, edema formation, and inflammation.

Glucose transportation

The retina is one of the most metabolically active tissues in the body. The high energy demand of the retina requires a considerable supply of metabolites, but the glucose storage in the retina is much smaller if compared with the metabolic needs. The neuroretina relies on the systemic circulation for glucose delivery, mediated by glucose transporters (GLUTs) through facilitated diffusion. Glucose transporter 1 (GLUT1) is expressed at high levels in the luminal and abluminal membranes of the retinal endothelial cells and in the apical and basolateral membranes of the RPE.[103,104] Hyperglycemia results in high glucose levels in the retina because GLUT1 transports more glucose when blood glucose is high. GLUT1 activity has been associated with impaired secretion of the pigment epithelium-derived factor (PEDF) by the RPE, a multifunctional secreted glycoprotein with antiangiogenic, antitumorigenic, and neurotrophic functions.[106,107] Thus, regulating GLUT1 activity and promoting PEDF action may be an intriguing therapeutic target in the future.

Imaging biomarkers of inflammation

There is a growing scientific interest in the possibility of identifying in vivo the biomarkers of retinal inflammation. Multiple non-invasive imaging modalities are currently available.

Fundus autofluorescence

Although the role of fundus autofluorescence (FAF) in the clinical management of DME is limited, this technique holds potential in elucidating DME pathogenesis. Intraretinal fluid in the setting of DME corresponds to increased FAF signal, and the amount of increased autofluorescence correlates with the severity of macular edema. Increased macular FAF signal in DME can be due to the dispersion of macular pigments that normally mask foveal autofluorescence or the accumulation of oxidative by-products from the activated microglia (Figure 1(a)). In this view, increased foveal FAF may be considered a biomarker of microglial activation and increased inflammation in DME eyes.

Figure 1.

(a) Fundus autofluorescence of the right eye of a 71-year-old diabetic man with diabetic macular edema (DME) demonstrating areas of increased foveal autofluorescence where DME cysts are located, (b) infrared retinal image of the same eye, and (c) spectral domain optical coherence tomography encompassing the fovea (horizontal section) showing a subfoveal neuroretinal detachment and multiple hyperreflective retinal spots throughout all retinal layers, especially in the outer retina.

OCT

Due to its high resolution and widespread availability, spectral domain optical coherence tomography (SD-OCT) has outweighed other imaging modalities in diagnosing and following DME. SD-OCT allows for the quantitative evaluation of the retinal thickness, fluid localization, retinal layers integrity, and segmentation. Inflammatory biomarkers visible on SD-OCT include hyperreflective retinal spots (HRS), subfoveal neuroretinal detachment (SND), external limiting membrane (ELM) disruption, and vitreous hyperreflective foci (Figure 1(b) and (c)). HRS are small, punctiform hyperreflective particles scattered throughout all the retinal layers. Their presence has been described in eyes with age-related macular degeneration, retinal vein occlusion,[115,116] and DME. HRS may represent activated microglial cells, but the differential diagnosis includes hard exudates, microaneurysms, and pigment clumping. HRS related to activated microglia have been defined as small size (<30 µm) particles, with reflectivity similar to nerve fiber layer, and absence of back-shadowing. HRS are more numerous in diabetic patients (regardless of the presence of DR) versus healthy controls, and in diabetics with DR versus diabetics without retinopathy. In the early stages of DR, HRS are located mainly in the inner retinal layers, where resting microglial cells physiologically are; with worsening of DR, HRS migrate from the inner to the outer retina or the subretinal space.[29,120] In eyes with DME, the number of HRS correlates with the aqueous concentration of pro-inflammatory mediators, such as soluble CD14. HRS could help in monitoring the response to intravitreal therapies in DME, as they are among the first features to disappear after anti-VEGF therapy.

SND occurs in approximately 15–30% of eyes with DME.[123,124] The presence of SND has been associated with higher levels of intravitreal IL-6, suggesting that inflammation may play a role in the development of SND. DME-related cell damage and attraction of scavenger cells to the retina may be a plausible source of IL-6.

Vitreous hyperreflective foci have been described in eyes with severe DR. Based on the previous findings that patients with proliferative diabetic retinopathy (PDR) had an increased number of T lymphocytes in the vitreous, it has been hypothesized that vitreous hyperreflective foci correspond to inflammatory cells. Similarly, hyperreflective macrophage-like cells (MLCs) have been recently described on averaged 3-µm OCT en face slabs located above the inner limiting membrane surface. If confirmed by future studies, the presence and the number of vitreous hyperreflective foci on SD-OCT and MLCs on the retinal surface may be used as a marker of inflammation and DR severity in diabetic eyes.

OCT angiography

OCT angiography (OCTA) depicts the degree and the extent of non-perfusion and the microvascular changes at different retinal depths (e.g. microaneurysms, intraretinal microvascular abnormalities, neovascularization elsewhere, and neovascularization of the disc). Recently, OCTA has been used to image MLCs at the vitreoretinal interface in eyes with DR. MLCs were significantly higher in density in PDR eyes compared with healthy controls, eyes without DR, and eyes with non-proliferative DR. Moreover, they showed a preferential localization in perivascular areas and on blood vessels rather than in ischemic areas in PDR. In the absence of histopathology confirmation, MLCs may be used as a biomarker of inflammation in DR eyes.

Neurodegeneration

In the last 20 years, several studies on both animal and human retinas highlighted neurodegenerative features in diabetes, which have some similarities with other chronic neurodegenerative diseases, such as Alzheimer’s or Parkinson’s disease. Neurodegeneration occurs early and independently from the vascular alterations. Patients with little to no evidence of diabetic vasculopathy may have inner retinal thinning compared with non-diabetic eyes.[134,135] Retinal nerve fiber layer (RNFL), ganglion cell, and inner plexiform layer thinning, especially in the papillo-macular bundle, have been found on SD-OCT in diabetic patients with or without DR.[136,137]

Diabetes-related neurodegenerative processes include cell apoptosis and abnormalities in synapsis morphology. Animal models showed increased dendritic arborization of retinal ganglion cells, presumably due to neuroplastic adaptation to reduced signals from apoptotic amacrine and bipolar cells. Diabetes is also associated with localized depletion of proteins involved in synaptic functions.

The metabolic pathways explained in the paragraphs above explain the higher rate of neuronal cell apoptosis observed in animal and human diabetic models.[140-142] Of these, oxidative stress plays a relevant role. A higher concentration of pro-apoptotic factors such as Bax, Fas, and caspase-3 has been found in the retinal ganglion cells of diabetic donors; similarly, an increased intracellular activity rate of these proteins has been observed in the diabetic rodent retina.

In patients with early DR, rod and cone photoreceptors’ transduction abnormalities have been found on full-field electroretinogram (ERG). The functional deficits could be explained, at least in part, by variations in local neurotransmitter concentrations, namely, glutamate, gamma-aminobutyric acid (GABA), and dopamine. Glutamate is the primary neurotransmitter in the retina; its total concentration is abnormally high in the vitreous of patients with PDR. Possible explanations include impairment in the activity of glutamine synthase (an enzyme that catalyzes the condensation of glutamate and ammonia to form glutamine) and reduced uptake by Müller cells.[147,148] Excessive glutamate has excitotoxic effects on the retinal neurons. GABA is an inhibitory neurotransmitter produced by the amacrine cells and increases in diabetes, with possible inhibitory consequences on the ERG wave amplitude. Finally, a premature dopamine depletion may elucidate the loss of contrast sensitivity in DR.

Different neurotrophic factors have been identified from the therapeutic standpoint, including insulin, insulin growth factor (IGF) I and IGF II, nerve growth factor (NGF),[153,154] brain-derived neurotrophic factor, and PEDF.[156,157] All these molecules have retinal neuroprotective effects, preventing ischemic damage to photoreceptors and dopaminergic neurons. Of note, the retinal cells contain a similar number of insulin receptors compared with liver cells. Insulin reduces apoptosis by activating the phosphatidylinositol 3 kinase/AKT PI 3-kinase/Akt pathway and inhibiting caspase-3. Early insulin intervention has been shown to prevent or reverse synapse loss, irrespectively to the duration of underlying diabetes.[160,161] On the contrary, NGF administration inhibits retinal ganglion cells’ and Müller cells’ apoptosis in diabetic mice.

As previously stated, retinal glial cells have a critical role in mediation between vasculature and neurons, and regulate the retinal microenvironment. Retinal glial cells participate in the inflammatory cascade and actively contribute to retinal neuron damage, which seems to precede microvascular impairment.[163,164]

Neurovascular coupling

In neural tissues, active neurons induce the dilation of local blood vessels in order to increase the blood flow, providing an adequate supply of oxygen and glucose and quickly removing metabolic wastes. This mechanism is called neurovascular coupling. Light stimulation and glial cell stimulation evoke dilation or constriction of arterioles in the mammalian retina, suggesting that glial cells play an essential role in neurovascular coupling. Nitric oxide (NO) seems to be a pivotal mediator in the regulation of these vasomotor responses. Neurovascular coupling is essential for the adaptation of the retinal vascular system to changing neural functions and increased metabolic demand. Impairment of the neurovascular coupling occurs early in DR, and this dysfunction might be involved in the development of retinal ischemia or inadequate removal of metabolic waste products. The exact mechanism of neurovascular coupling is not fully elucidated. A better understanding of the ocular blood flow regulation may shed light on the neurovascular alterations occurring in the retina of diabetic patients. The modulation of mediators regulating neurovascular coupling within the RNU may offer new therapeutical targets for the early treatment of DR and DME.

Treatment

Despite the increasing knowledge on the pathways involved in DME, only a few therapeutic strategies are currently available and the clinical response to these treatments is sometimes suboptimal. Besides adequate blood glucose control,[168-170] no therapy halts the clinical progression of the disease. Similarly, it is not yet possible to reverse the retinal damage. Alternative therapeutic strategies have been considered potential options for DME; some have reached phase III randomized clinical trials, with promising results.

Anti-VEGF

The first-line treatment of DME is based on anti-VEGF agents’ intravitreal injections. The molecules approved so far by the Food and Drug Administration (FDA) (ranibizumab and aflibercept) specifically tackle VEGF and other pro-angiogenic factors.[171,172] Newer anti-VEGF agents, instead, might help in addressing inflammation together with pro-angiogenesis. Faricimab is a bispecific monoclonal antibody targeting both VEGF and Ang-2 evaluated in the BOULEVARD phase II trial for DME. Patients were randomized 1:1:1 to intravitreal ranibizumab 0.3 mg, faricimab 1.5 mg, or faricimab 6.0 mg, every 4 weeks for 20 weeks followed by 16 weeks of observation. Faricimab 6.0 mg achieved greater mean gains in visual acuity from baseline to week 24 in treatment-naïve patients than anti-VEGF alone (faricimab 6.0 mg, +13.9; faricimab 1.5 mg, +11.7; ranibizumab, +10.3). The proportion of treatment-naïve patients with ⩾2-step improvement from baseline at week 24 on the diabetic retinopathy severity scale (DRSS) score was 38.6%, 27.7%, and 12.2%, respectively. There were no new or unexpected safety outcomes in the faricimab arm. These results suggest that simultaneous blocking of Ang-2 and VEGF in DME may lead to better visual gains and longer duration of treatment’s effect than anti-VEGF alone.

Corticosteroids

Corticosteroids inhibit multiple inflammatory pathways and improve the stability of the BRB.[175,176] Intravitreal injections of triamcinolone acetonide initially showed short-term benefits in visual acuity and macular thickness in patients with DME. More recent intravitreal delivery systems allow a sustained release of corticosteroids in the vitreous chamber. Corticosteroid implants, such as dexamethasone implant and fluocinolone acetonide insert, have a longer duration and require less frequent injections compared with anti-VEGF agents. The use of intravitreal steroids has proved an alternative for eyes with recalcitrant macular edema despite antiangiogenic therapy. An early switch to intravitreal corticosteroids may be helpful in a subset of patients with specific OCT features associated with increased inflammatory activity.[180-182] Nevertheless, in specific cases, additional anti-VEGF injections for residual or recurrent DME in patients with a steroid implant may be required.

Potential new therapeutic targets

Novel therapeutic targets acting on the inflammatory cascade are under investigation for DME treatment in preclinical and clinical trials. The vascular endothelial-protein tyrosine phosphatase (VE-PTP), an endothelial cell-specific receptor tyrosine phosphatase, deactivates Tie2 and inhibits its intracellular signaling pathway. A phase I investigation of a competitive inhibitor of VE-PTP administered subcutaneously in patients with DME has been shown to reduce macular edema and improve vision in some patients.

C-C chemokine receptor type 2 and type 5 (CCR2 and CCR5) are located on the surface of monocytes and are involved in the homing of inflammatory cells to their target tissues.[187,188] The inhibition of CCR2 and CCR5 may be a strategy to reduce inflammation and vascular leakage in patients with DME. In a recent phase II trial, oral administration of a CCR2/5 chemokine receptor dual antagonist showed a modest improvement in visual acuity, yet inferior to intravitreal ranibizumab.

A small phase III study demonstrated that intravenous treatment with infliximab (an anti-TNF-α monoclonal antibody) significantly improved visual acuity over placebo in patients with sight-threatening DME. However, intravitreal injections of anti-TNF-α inhibitors did not provide any benefit to eyes with refractory DME and may paradoxically induce a severe intraocular inflammatory response. Finally, ruboxistaurin (RBX), an orally administered inhibitor of PKC β, showed a beneficial effect in animal models of DR and DME.[192,193] Two phase III clinical trials suggested that RBX may reduce the relative risk of visual loss from DME compared with placebo, although statistical significance was not achieved.

Only a limited number of options are under investigation tackling neurodegeneration. Soluble CD59, which inhibits apoptotic signals through a competitive raptorial interaction, reduces the neural impairment due to diabetes in animal models. Furthermore, the concurrent activation of retinal glial cells enhances the antiapoptotic response. Other molecules, such as pigment-derived factor, somatostatin, docosahexaenoic acid, neuroprotectin D1, and brain-derived neurotrophic factor, may become key factors in future trials. However, the real value in human therapy has not been proven, as for somatostatin, or is far from being discovered.

Conclusion

Inflammation holds a pivotal role in the development of DR and DME since the initial stages of the disease, and different cell types and cytokines are involved. Parallelly, neurodegeneration has been observed since the early steps of diabetes and relies on complex cellular and molecular mechanisms. New anti-inflammatory pathways, modulation of reactivity of macroglial and microglial cells, and regulation of apoptotic stimuli could provide new therapeutic strategies in the management of DME. Imaging findings using novel non-invasive tools may also guide clinicians in choosing patient-tailored therapeutic strategies and monitoring the therapeutic response.