Retinal physiology and circulation: effect of diabetes

Abstract

In this review, we present a discussion of diabetes and its complications, including the macrovascular and microvascular effects, with the latter of consequence to the retina. We will discuss the anatomy and physiology of the retina, including aspects of metabolism and mechanisms of oxygenation, with the latter accomplished via a combination of the retinal and choroidal blood circulations. Both of these vasculatures are altered in diabetes, with the retinal circulation intimately involved in the pathology of diabetic retinopathy. The later stages of diabetic retinopathy involve poorly controlled angiogenesis that is of great concern, but in our discussion, we will focus more on several alterations in the retinal circulation occurring earlier in the progression of disease, including reductions in blood flow and a possible redistribution of perfusion that may leave some areas of the retina ischemic and hypoxic. Finally, we include in this review a more recent area of investigation regarding the diabetic retinal vasculature, that is, the alterations to the endothelial surface layer that normally plays a vital role in maintaining physiological functions.

Introduction

The retinal vasculature, as with most circulatory beds, has unique physiological characteristics appropriate to the function of the tissue. For the retina, this includes delivering sufficient oxygen and nutrients to cells having a high metabolic rate, but without the vasculature being so dense as to prevent light transmission. Proper vision is threatened by any interruption or significant decrease of retinal blood flow, with these disturbances occurring early in the progression of diabetes. Determining the causes of the diabetes-induced decreases in retinal blood flow is of significant clinical concern, but have yet to be fully elucidated.

The endothelium comprises the inner lining of the vasculature, and in turn, the inner surface layer of the endothelium (molecules on the cell membrane) serves as an intermediary between the blood and endothelial function. On the surface of the endothelium are junctional molecules, adhesion molecules, and a dense layer of proteoglycans and glycosaminoglycans called the glycocalyx. The endothelial surface layer of the retinal circulation is significantly altered in diabetes, with the consequences of these changes likely to participate in the progression of retinopathy, which is a relatively recent topic of investigation in the field that will be discussed in this review.

Herein we discuss some general topics of diabetes and its complications; the anatomy and physiology of the retina; the organization and function of the posterior circulations of the eye; mechanisms that contribute to diabetic retinopathy; and diabetes-induced changes in retinal blood flow and the endothelial surface layer. Abbreviations used in the review are listed in Table 1.

TABLE 1.

LIST OF ABBREVIATIONS

20-HETE	20-hydroxyeicosatetraeonic acid
ACE	angiotensin converting enzyme
AGE	advanced glycation end product
Ang II	angiotensin II
aPKC	atypical protein kinase C
AR	aldose reductase
AT	angiogensin II receptor type
ATP	adenosine triphosphate
BBB	blood brain barrier
BRB	blood retinal barrier
CAD	coronary artery disease
cGMP	cyclic guanine monophosphate
COX-2	cyclooxygenase-2
cPKC	conventional protein kinase C
DAG	diacylglycerol
DC	dendritic cell
EDCF	endothelium-derived constricting factor
eNOS	endothelial nitric oxide synthase
EPO	erythropoietin
ET	endothelin
FADH2	flavin adenine dinucleotide
FAM3B	family with sequence similarity 3 member B
GAG	glycosaminoglycan
GC	guanylate cyclase
GCL	ganglion cell layer
GEnC	glomerular endothelial cell
GK	Goto-Kakizaki
GLUT	glucose transporter
HDL	high-density lipoprotein
HIF	hypoxia-inducible factor
HUVEC	human umbilical vein endothelial cell
iBRB	inner blood retinal barrier
ICAM-1	intercellular adhesion molecule-1
IDDM	insulin-dependent diabetes mellitus
Ig	immunoglobulin g
IL-1β	interleukin-1β
ILM	inner limiting membrane
INL	inner nuclear layer
iNOS	inducible nitric oxide synthase
IOP	intraocular pressure
IPL	inner plexiform layer
JAM	junctional adhesion molecule
JIP1	c-Jun N- terminal kinase interacting protein-1
JNK	c-Jun N- terminal kinase
KKS	kallikrein-kinin system
LDL	low-density lipoprotein
L-NA	Nitro-L-arginine
L-NAME	NG-nitro-L-arginine methyl ester
L-NMMA	NG-monomethyl-L-arginine
L-NPA	N(ω)-propyl-L-arginine
MAPK	mitogen-activated protein kinase
MCP-1	monocyte chemoattractant protein-1
MCT	mean circulation time
MFS	major facilitator superfamily
MI	myocardial infarction
mmHg	millimeters of mercury
MMP	matrix metalloproteinase
NAD	nicotinamide adenine dinucleotide
NADH	nicotinamide adenine dinucleotide plus hydrogen
NADPH	nicotinamide adenine dinucleotide phosphate
NCD	non-communicable disease
NF-κB	nuclear factor kappa-light-chain-enhancer of activated B cells
NFL	nerve fiber layer
nNOS	neuronal nitric oxide synthase
NO	nitric oxide
NOD	non-obese diabetic
NOS	nitric oxide synthase
NOX	nicotinamide adenine dinucleotide phosphate oxidase
NPDR	non-proliferative diabetic retinopathy
nPKC	novel protein kinase C
oBRB	outer blood retinal barrier
OLM	outer limiting membrane
ONL	outer nuclear layer
ONOO−	peroxynitrite
OPL	outer plexiform layer
P	transmural pressure
PaCO2	arterial partial pressure of carbon dioxide
PAD	peripheral artery disease
PDGF	platelet-derived growth factor
PDR	proliferative diabetic retinopathy
PECAM-1	platelet endothelial cells adhesion molecule-1
PEDF	pigment epithelium derived factor
PGD2	prostaglandin D2
PGE1	prostaglandin E1
PGF2α	prostaglandin F2α
PGI2	prostaglandin I2
PK	protein kinase
PMA	12-myristate 13-acetate
PS	phosphatidylserine
PVD	posterior vitreous detachment
pO2	partial pressure of oxygen
r	vessel radius
RAGE	receptor for advanced glycation end product
RBC	red blood cell
RMECS	retinal microvascular endothelial cells
RPE	retinal pigmented epithelium
ROS	reactive oxygen species
S1P	sphingosine-1 phosphate
SHP-2	Src homology region 2 domain-containing phosphatase-2
SOD	superoxide dismutase
STAT3	signal transducer and activator of transcription 3
STZ	streptozotocin
T	vascular wall tension
TGF	transforming growth factor
TLR	toll-like receptor
TNFα	tumor necrosis factor-α
TNFR	tumor necrosis factor-α receptor
TxA2	thromboxane A2
VE-cadherin	vascular endothelial cadherin
VEGF	vascular endothelial growth factor
VEGFR2	vascular endothelial growth factor receptor 2
VSMC	vascular smooth muscle cell
ZO	zonula occludens

Diabetes and Its Complications

Incidence and epidemiology

Diabetes currently afflicts millions of individuals worldwide and is considered one of the greatest health emergencies of the 21^st century. According to the Center for Disease Control and Prevention, one in 10 US adults has diabetes, with a projected growth to one in three US adults by the year 2050 (2). Diabetes is one of the four non-communicable diseases (NCDs) that are responsible for around 80% of premature deaths globally, with the three other NCDs being cancer, cardiovascular, and respiratory diseases (101). The number of people diagnosed with diabetes has grown from 108 million in 1980 to 425 million in 2017, with expected growth to 629 million by the year 2045 (91). This is an alarming trend considering the massive financial burden diabetes can have on healthcare systems and individuals with diabetes. The cost of diabetes has risen from $232 billion dollars per year in 2007 to an estimated $727 billion per year in 2017, with predicted growth to $1.7 trillion for direct and indirect costs by the year 2030 (91). In the United States, there were an estimated 30 million people with diabetes (diagnosed and undiagnosed) in 2017, with expected growth to 43 million people in the year 2045 (91), making the United States the second highest number of people with diabetes following China.

Major complications of diabetes

Diabetes is often associated with significant organ damage and failure, which leads to an increase in mortality rates. Almost all diabetes-related complications can be attributed to vascular damage and remodeling due to hyperglycemia. This vascular remodeling can be divided into two main categories (Figure 1) depending on the size of the affected blood vessels: 1) macrovascular complications that affect the larger blood vessels, leading to coronary artery disease (CAD), stroke, and peripheral artery disease (PAD) and 2) microvascular complications that affect the microcirculatory blood vessels, leading to diabetic retinopathy, diabetic nephropathy, and diabetic peripheral neuropathy, with the latter attributed to abnormalities within the neuronal cells resulting from microvascular dysfunction. Patients who are diagnosed with insulin-dependent diabetes mellitus (IDDM) are likely to have macrovascular or microvascular complications within 15 years of their diagnosis, with these complications occurring more in type I than type II diabetes (638).

Figure 1. Macrovascular and microvascular complications of diabetes.

Macrovascular complications include peripheral artery disease, coronary artery disease, and stroke; microvascular complications include diabetic retinopathy, neuropathy, and nephropathy.

In both macrovascular and microvascular complications, hyperglycemia is considered an independent and major risk factor that plays a vital role in the ensuing vascular damage. Under normal conditions, protective factors such as antioxidant enzymes and insulin protect the blood vessels from damage, making them less susceptible to injury (342). However, under hyperglycemic conditions, these factors are inhibited, while factors leading to vessel damage and injury are upregulated.

Blood vessels in the macrovascular system differ from those in the microvascular system with regard to structure, function, wall composition, wall thickness, and vessel diameter, yet the vascular network shares the endothelium organ system. The endothelium contains a heterogenous population of cells that have metabolic activity (described in (157)) comprised of three major energy sources: glucose, fatty acids, and amino acids. Endothelial cells arising from different vascular beds have different sizes, anatomical compartments, and express different proteins, and their response to physiological and pathological conditions can vary remarkably (14-16).

Hyperglycemia can induce endothelial cell modifications due to the disruption of homeostasis leading to damage in both the large and small blood vessels. These modifications include altered signaling, upregulation of metabolic activity, and alterations in ultrastructure (465, 524). Different molecular mechanisms can cause hyperglycemia-induced macrovascular and microvascular complications, yet both share many components that interlink with each other leading to vascular injury. An increase in vascular permeability (608), vascular cell apoptosis (206, 559), leukocyte adhesion (196), and altered blood flow (97, 239, 559) are among the shared systemic factors that contribute to both macrovascular and microvascular complications due to diabetes. Thus, the question of whether one complication precedes the other, or if they develop concurrently, is still under investigation.

Macrovascular complications of diabetes

The macrovascular system consists of large capacity blood vessels, namely the arteries and veins, in which the primary function is to rapidly deliver blood to and from the various organs of the body. Thus, any blockage or damage to these vessels can have deleterious pathological effects, which can be terminal. Insulin resistance and poor glycemic control are associated with an increased risk in the development of macrovascular diseases in diabetes (435).

One of the major contributors to macrovascular complication hallmarks in diabetes is the development of atherosclerosis in medium and large blood vessels, leading to their partial or total blockade (572), resulting in cardiovascular (coronary artery disease), cerebrovascular (stroke) and peripheral arterial diseases. Diabetes can increase the atherogenic process in these vessels. The exact mechanism is still under investigation, but hyperglycemia is considered a significant player in this process.

The development of atherosclerosis due to hyperglycemia is a complex process that involves multiple mechanisms such as the formation of advanced glycation end products (AGEs), reactive oxygen species (ROS) as a result of altered glucose metabolism, and endothelial cell dysfunction leading to altered enzymatic activity and an increase in adhesion molecule expression (481). Moreover, hyperglycemia is a major contributing factor to vascular smooth muscle cell (VSMC) proliferation in the atherosclerotic arterial intima and the formation of the atherosclerotic fibrous cap (168). Studies have indicated a possible role of proto-oncogene serine/threonine-protein kinase Pim-1 upregulation in hyperglycemia-induced VSMC proliferation and migration via the activation of signal transducer and activator of transcription 3 (STAT3)/Pim-1 signaling. Additionally, FAM3B (Family with Sequence Similarity 3 Member B), which is co-released with insulin from pancreatic β-cells upon glucose stimulation, has been found to be upregulated under hyperglycemic conditions leading to the inhibition of miR-322-5p and proliferation of VSMC (651).

Hyperglycemia can lead to metabolic disturbances and impairment in lipid metabolism (596). In type I diabetes, there are increased levels in total and low-density lipoproteins (LDLs), accompanied by lower levels in high-density lipoprotein (HDLs), with this combination leading to an increased risk of atherosclerosis development (572, 596). Furthermore, ROS and AGEs enhance atherogenesis by promoting the retention of apolipoprotein B-containing and modified LDL in the arterial intima (591). The increased expression of endothelial cell adhesion molecules such as E-selectin, P-selectin, and intercellular adhesion molecule-1 (ICAM-1) lead to adhesion of mononuclear leukocytes such as monocytes and T-cells, and their migration to the arterial intima (271, 348) where they differentiate to macrophages and dendritic cells (DCs). Monocyte-driven macrophages take up lipoprotein and differentiate to foam cells. These foam cells and other immune cells such as DCs produce cytokines and chemokines such as tumor necrosis factor-α (TNF-α) and monocyte chemoattractant protein-1 (MCP-1) that exacerbate inflammation and recruit additional immune cells (276). Moreover, the formation of the necrotic core in advanced plaques can be a result of macrophage apoptosis, which also can contribute to the release of matrix metalloproteinases (MMPs) that break down the fibrous atheroma cap, leading to atherosclerotic plaque rupture and thrombus formation (87, 395, 488). Thrombosis enhances the risk of major macrovascular complications, such as myocardial infarction (MI), stroke, and peripheral artery disease.

Microvascular complications of diabetes

The microvascular system, which consists of arterioles, venules, and capillaries, is the basic functional unit of the cardiovascular system. These blood vessels have different cellular components and distribution than macrovessels, and have mechanisms regulating permeability and blood flow to achieve proper nutrient delivery and to maintain adequate systemic blood pressure (83, 187, 225). The microvascular complications of diabetes are the leading cause of blindness and end-stage renal disease in developed countries (212). Additionally, around 60% of patients develop neuropathy, which together with PAD can account for around 50% of non-traumatic amputations in the US (212).

Diabetic retinopathy

Diabetic retinopathy is the leading cause of visual impairment in working-age adults worldwide, and is one of the major microvascular complications of diabetes. It affects the retinal vasculature, leading to complications such as microaneurysms, cotton wool spots, abnormal angiogenesis, capillary dropout, and edema (47). Almost 100% of patients with insulin-dependent diabetes mellitus (IDDM) will develop some form of retinopathy that will continue to progress even after tight glycemic control (145), while retinopathy will develop in >60% of individuals with Type 2 diabetes (185). During the time between the initial diagnosis of diabetes and the clinical detection of retinopathy, which can be in years, various irreversible physiological and biochemical changes occur (145). This phenomenon of the continual progression of retinopathy in spite of tight glycemic control indicates a point of no return, possibly due to a ‘metabolic memory’, in which vascular and retinopathy occurrence is virtually inevitable (9, 571). Most therapies currently being developed are directed towards limiting diabetic retinopathy progression; however, the understanding of the molecular mechanisms leading to the pathology is limited, which could hinder the development of better therapies aimed at preventing its occurrence and progression.

Historical perspective of diabetes and associated retinopathy

A 3rd dynasty Egyptian papyrus written in 1552 BC by the Egyptian physician Hesy-Ra described frequent urination as a characteristic of a disease that was later identified as diabetes, with this being considered the oldest record of the disease. In ca.1500 BC, early Hindu writings mentioned an emaciating disease (possibly diabetes) where ants were attracted to the urine of those afflicted. Around 1000 years later, the link between obese individuals and high sugar content in their urine was established. The term diabetes, which originated from the Greek diabaino, was first introduced by the renowned Pneumatic School physician Aretaeus of Cappadocia, who lived in Alexandria and Rome in the 2^nd century AD (291) and who was the first to distinguish between diabetes mellitus and diabetes insipidus (332). However, some records indicate that the term “diabetes” might have been coined much earlier by Apollonius of Memphis in 230 BC (223). Diabetes was later described by 1^st century Greeks as a melting disease where the body was thought to be converted into urine.

However, it was not until the mid-19^th century when the link between diabetes and eye disease was established, starting with the findings by the French ophthalmologist Apollinaire Bouchardat. Visual loss progression was observed and reported in patients with diabetes even with the absence of cataract by Bouchardat in 1846 (82, 285). Vision was found to be improved through exercise and diet (291). In the now famous painting by the Austrian ophthalmologist Eduard Jäger, diabetic macular changes were recorded with the help of the ophthalmoscope he built in 1855 (180). By the second half of the 19^th century, several studies further confirmed the link between maculopathy and diabetes. In 1869, an article published by Henry Noyes supported this link (432), which was further described by Edward Nettleship’s paper in 1872 detailing the first proof of diabetes-induced histopathological changes to the macula (291).

In 1876, Wilhelm Manz reported the retinal changes occurring during the proliferative stage of diabetic retinopathy, including degeneration of the optic disk, vitreous hemorrhage, and retinal detachment (183), which further supported the link between diabetes and retinal damage. By the end of the 19th century, an extensive description of the characteristics of diabetic retinopathy was available; however, the mechanisms and explanations for these characteristics were lacking. The first classification of diabetic retinopathy into stages was by the German ophthalmologist Julius Hirschberg, where he classified the pathology into 1) ‘retinitis centralis punctuate’, 2) hemorrhagic form, 3) retinal infarction, and 4) hemorrhagic glaucoma (569). Additionally, the role of retinal microvascular dysfunction in diabetic retinopathy was established by the English ophthalmologist Arthur Ballantyne (41). These and other early discoveries of the link between diabetes and changes to the retina have initiated and contributed substantially to the current understanding of diabetic retinopathy.

Anatomy

Eye

The eye is one of the most complex organs of the body. It contains multiple sections, with tissue that is among the most metabolically active and vascularized in the body. The eye can be divided into three main layers (Figure 2): (1) the sclera and the cornea that form the external layer, (2) the intermediate layer that is divided into an anterior part encompassing the iris and ciliary body, in addition to the posterior part encompassing the choroid, and (3) the retina, which forms the internal layer. Additionally, three fluid chambers exist in the eye: (1) the anterior aqueous chamber (filled with aqueous humor), (2) the posterior aqueous chamber (also filled with aqueous humor), and (3) the vitreous chamber (filled with vitreous humor). Anterior to the vitreous humor sits the lens, a transparent body that is suspended by ligaments known as the zonula of zinn. The lens can change its shape in order to focus light by the contraction of the ciliary muscles that control the zonula of zinn. Once the focused light hits the retina, it is converted to neurochemical signals that are transported to the brain via the optic nerve to be processed.

Figure 2. Anatomy of the eye.

Ocular tissue can be divided into three layers: 1) an outer layer consisting of the sclera and the cornea, 2) an intermediate layer consisting of the iris, the ciliary body, and the choroid, and 3) an internal layer consisting of the retina.

Retina

The retina is a thin tissue that is comprised of a complex network of neurons and supporting cells that provide sensory function, and is considered part of the central nervous system (108). Moreover, the retina shares the high metabolic demand of the brain, and thus, it is one of the body’s most metabolically active tissues (108, 579). Additionally, the blood-brain barrier (BBB) and the BRB share many other characteristics (390).

The retina can be divided into distinct layers, three containing nerve cell bodies, and two containing neuronal synapses (Figure 3). Beginning from the outermost layers (farthest from the light source and adjacent to the choroid and the sclera), the following are the layers of the retina:

Figure 3. Layers of the retina.

The layers of the retina starting from the outermost to the innermost layer are the 1) retinal pigmented epithelium (RPE), 2) photoreceptors (PR), 3) outer limiting membrane (OLM), 4) ) outer nuclear layer (ONL), 5) outer plexiform layer (OPL), 6) inner nuclear layer (INL), 7) inner plexiform layer (IPL), 8), ganglion cell layer (GCL), 9) nerve fiber layer (NFL), and 10) inner limiting membrane (ILM).

1. Retinal pigmented epithelium (RPE)

The RPE contains highly specialized neuroectodermally-derived columnar/cuboidal epithelial cells that produce and store melanin, and sits between the neurosensory retina and the choroid (504). It has essential roles in maintaining visual functions including phagocytosis, limiting the scattering of light due to melanin, maintenance of the outer retinal barrier, and transport of metabolites (503, 504).

2. Photoreceptor Layer - outer and inner segments

Photoreceptors are neuroepithelial cells known as the rods and cones which are part of the photoreceptor layer (outer and inner segments) and the outer nuclear layer (ONL) (381). Adjacent to the RPE cells is the rod outer segment which contains the protein rhodopsin, which is responsible for detecting photons. The metabolic demands required for the release of the rod neurotransmitter glutamate are met by the high density of mitochondria found in the inner segment of the rods (504). In the human retina, rods constitute 95% of the photoreceptors, and provide high sensitivity under dark conditions (11). Cones, on the other hand, constitute only 5% of the human retina (11). Cones facilitate high acuity color vision during the daytime, and are not evenly distributed throughout the retina, where they reach a maximum density in the fovea (110), and like rods, release glutamate in the outer plexiform layer (OPL) (381).

3. Outer limiting membrane (OLM)

The OLM is a basal lamina layer that separates the rod and cone bodies from their nuclei. It provides structural support to the retina via its mechanical properties. The junctional proteins occludin, zonula occludens-1, and junctional adhesion molecules (JAMs) have been found to be expressed at the OLM, suggesting it may play a role in the maintenance of the retinal blood barrier (441).

4. Outer nuclear layer (ONL)

The ONL contains the cell bodies, which contain the nuclei of rods and cones (381).

5. Outer plexiform layer (OPL)

The OPL lies between the outer nuclear layer and the inner nuclear layer. Synapses between the dendrites of horizontal cells, bipolar cells, and rod and cone axons are formed within this layer (379). These synaptic interactions lead to the formation of the ON and OFF pathways, where visual signals are split into two channels for detection of objects that are either lighter or darker than the background, thus providing visual contrast (112).

6. Inner nuclear layer (INL)

The INL contains cell bodies of horizontal cells, bipolar cells, and amacrine cells. Horizontal cells are excitatory interneurons that connect bipolar cells, and are responsible for the ability to adjust vision under light and dark conditions (84). Bipolar cells are excitatory neurons that connect the photoreceptors to ganglion cells, and transform the photoreceptor signals in term of polarity, chromatic preference, and kinetics (167). Amacrine cells are inhibitory interneurons that connect bipolar and ganglion cells; they mainly feedback-inhibit bipolar cell terminals, and feedforward-inhibit ganglion cells (40, 84, 112).

7. Inner plexiform layer (IPL)

The IPL is the location where dendrites of ganglion cells, dendrites of amacrine cells, and axons of ganglion cells interact and form synapses with each other (487, 504).

8. Ganglion cell layer (GCL)

The GCL contains the cell bodies of the ganglion cells, which are the output neurons of the retina. The intermediate bipolar and amacrine neurons transmit signals from the photoreceptors to the ganglion cells, where image-forming and non-image-forming signals are transported to the brain via their axons through the optic nerve (487, 507).

9. Nerve fiber layer (NFL)

The NFL contains unmyelinated ganglion cell axon fibers, which run through the vitreal surface of the retina toward the optic disc, penetrating the sclera, and forming the optic nerve (487, 504).

10. Inner limiting membrane (ILM)

The ILM is composed of astrocytes, footplates of Müller cells, and basal lamina, and is considered the boundary between the retina and the vitreous body (487).

Retinal vascular anatomy

The retinal vascular system is easily accessible making it the most studied vascular system in the human body (181). Due to the retina’s high metabolic activity as a result of the continual conversion of light into neuronal signals, its oxygen demand is high (even more so in the dark than in the light). To meet this high oxygen demand, and to prevent the interference of light transmission by a dense vasculature, two distinct vascular networks are utilized to supply the retina with nutrients and oxygen: 1) the choroidal microcirculation that branches from the posterior ciliary artery, which supplies the outer one-third to one-half of the retina (Figure 4), and 2) the retinal microcirculation that branches from the central retinal artery, and supplies the inner one-half to two-thirds of the retina.

Figure 4. Oxygen sources to the retina.

The inner half of the retina receives oxygen from the retinal circulation, while the outer half receives oxygen from the choroidal circulation.

The ophthalmic artery, which originates from the internal carotid artery, branches into the central retinal artery, the short and long ciliary arteries, and the anterior ciliary arteries (Figure 5). The central retinal artery divides into the following four superficial arteriole branches which supply quadrants of the retina: superior nasal artery, inferior nasal artery, superior temporal artery, and inferior temporal artery. The superior and inferior temporal arteries wrap around the thinner avascular macula (236). These superficial vessels branch into the retina creating two distinct capillary layers: the intermediate capillary layer and the deep capillary layer (450) (Figure 6). The superficial vasculature lies in the ganglion cell layer, and contains arterioles, venules, and their branches, mainly supplying the ganglion cells. The arteriolar branches become more narrow as they continue deeper into the retina, where they divide into the intermediate capillary layer. The intermediate capillary layer is a distinct network of capillaries that lies at the border between the IPL and the INL. Capillaries then continue further into the retinal tissue, to finally form the deep capillary layer, which is a dense network of vessels that are highly interconnected, at the border between the INL and the OPL. The architecture results in zones of tissue lacking capillaries, with these zones being between the superficial vessels and intermediate capillary layer, between the intermediate and deep capillary layers, and between the deep capillary layer and the choroid (471). The retinal capillaries drain into the retinal veins, which lie deeper to the retinal arteries, and ultimately drain into the central retinal vein (236).

Figure 5. Arterial branches leading to the eye.

The blood vessels supplying blood to the inner and outer retina, the iris, and the ciliary body originate from the ophthalmic artery, which is a branch of the internal carotid artery. The ophthalmic artery further divides to give rise to the anterior ciliary arteries, the long and short posterior ciliary arteries, and the central retinal artery.

Figure 6. Retinal microvascular layers.

The retinal microcirculatory system can be divided into three distinct layers: 1) the superficial arterioles and venules, 2) the or intermediate capillary layer, and 3) the deep capillary layer.

The inner retina (i.e., NFL to the outer portion of the INL) receives nutrients from the retinal circulation, while the chroidal circulation supplies the outer retina. Arteries exit and venules enter the optic disc in an alternating fashion, with the pattern being radial in rodents (240, 382, 450), but more of a C-shape around the macula in humans. Additional differences found between the human and rodent retina are discussed in greater detail elsewhere (111, 450, 603). Smooth muscle cells surround the retinal arterioles prior to branching into retinal capillaries (540). Retinal capillaries consist of non-fenestrated endothelial cells (109) connected by tight junctions which form the inner blood-retinal barrier (109, 440, 541). These capillary endothelial cells are surrounded by pericytes (540).

Choroid structure and vascular anatomy

With the exception of the endothelium of the blood vessels and innermost layer of Bruch’s membrane, the choroid is derived from neural crest cells. The choroid is a highly vascularized structure that sits on the outer side of the retina with the inner choroid (i.e. Bruch’s membrane) connected to the retinal pigment epithelium (from the ora serrata to the optic nerve), while the outer choroid is attached to the sclera. The choroid has one of the highest flow rates per gram of tissue in the human body, and combined with the anatomical arrangement, the choroid is positioned to meet the metabolic demands of the outer retina and retinal pigment epithelium (226).

Branching off the ophthalmic artery, the short posterior ciliary arteries penetrate the sclera, providing localized blood supplies by forming into fan-shaped lobules. These arteries pierce the sclera to form three layers of the choroidal microcirculatory system (302). The outermost layer is known as Haller’s Layer which contain larger blood vessels surrounded by smooth muscle cells. The middle layer is Sattler’s layer, which contains medium-sized blood vessels. The vessels in Haller’s and Sattler’s layers are not fenestrated (226). The innermost layer is the choriocapillaris and is located adjacent and inner to Sattler’s layer. The endothelial cells of the choriocapillaris are fenestrated with most fenestrations on the inner side, which allows for oxygen and nutrient supply to the outer retina (109, 226). The long posterior ciliary arteries and the anterior ciliary arteries supply the iris and the ciliary body (302). The choroidal circulation is regulated by the sympathetic and parasympathetic nervous systems (133, 369).

Retinal Oxygenation and Metabolism

The main function of the retina is to convert light signals to neuronal signals that can be carried by the optic nerve to the brain for processing. This continual activity necessitates a tight control of retinal blood flow to ensure proper function (mechanisms discussed in a later section), in part through adequate oxygen delivery.

Oxygen tension profile in the retina

The oxygen tension profile from the front to the back of the retina is unique, and is influenced by the presence of the retinal and choroidal microcirculation networks. These vascular networks are critical for sufficient oxygen supply to the retina, but differ in their capillary density, arterio-venous oxygen distribution, and autoregulatory control. The oxygen tension profiles differ among animal species due to variations in retinal and choroidal circulations and retinal thicknesses. For example, the rat and mouse retina have similar capillary layers (i.e. intermediate and deep layers) (645, 646) as found in the human retina, except in the macula, which is the thinnest part of the human retina and does not contain any vasculature. Additionally, the guinea pig does not contain a retinal circulation and receives its entire oxygen supply from the choroidal circulation (647).

The oxygen tension profile of the choroid is highly dependent on the supply coming from the choroid. The choroidal circulation receives 80-85% of ocular blood flow (25), and has one of the highest blood flows per unit mass in the body (25, 219). The choroid supplies 79% of the oxygen consumed by the retina (26), but with a low oxygen extraction (24, 99, 159, 578). Yu et al. (645, 648) measured the retinal oxygen partial pressure in rats and mice using microelectrodes, and observed a significant decrease in the oxygen tension starting from the retinal tissue near the choroid (~45 mmHg), with a decrease in oxygen tension to ~10-15 mmHg in rats (646) and ~4-6 mmHg in mice (645) just anterior to the photoreceptors, consistent with the high metabolic rate of these neurons. Similar trends were observed in pigs and cats (607). This decrease in oxygen tension reflects the high metabolic rate of these neurons, and reflects the inability of the choroidal microcirculation to supply oxygen to the entire inner retina, with the exception of rabbits and guinea pigs. This is evidenced by the oxygen profiles in the inner retina being similar with or without inspiration of 100% oxygen following laser occlusion of the retinal arteries (648).

In contrast to the choroidal circulation, blood flow in the retina is not as high (25); however, inner retinal oxygenation by the retinal microcirculation is vital with the arteriovenous difference in oxygen partial pressure being approximately 40% (252, 578), and with possibly one-third of the oxygen consumed by the retina being supplied by the retinal circulation. To demonstrate the retinal circulation contribution to inner retinal oxygenation, Yu et al. (648) conducted experiments where retinal ischemia was initiated by laser occlusion, followed by oxygen tension measurements, with the oxygen tension decreased to ~0 mmHg throughout the inner retina. By performing this experimental protocol, the researchers were able to model the inner retinal oxygen consumption, which was feasible due to the elimination of flowing vessels that could interfere with this protocol and analysis. The researchers concluded that the inner retinal oxygen consumption is substantial, and equals that of the outer retina. Moreover, the unique distribution of the retinal microvessels and their sparsity compared to the choroidal microcirculation permits light transmission through the retinal layers. However, the sparsity of the retinal circulation can also contribute to its susceptibility to vascular disease (644).

Retinal metabolism

Three dominant layers have been identified as sites of oxygen consumption in the retina, with this identification established with oxygen profile analyses in rats. The three layers include: 1) the photoreceptor inner segment, which contains the mitochondria, 2) the outer plexiform layer, and 3) the deeper region of the inner plexiform layer (644). Moreover, activity of cytochrome oxidase, the enzyme involved in the final step of the mitochondrial respiratory chain, is the most intense in retinal cross-section staining in these three layers (331), which indicates a significant oxygen consumption. This oxygen consumption is needed to produce ATP and facilitate synaptic transmission (648), phototransduction (28), Na-K ATPase activity, and other activities that require significant amounts of energy, including guanosine triphosphate synthesis and the generation of the dark current (607). Glycolysis produces a small amount of energy even though a large amount of glucose is consumed. The transition from steady light to dark increases oxygen consumption with the demand thought to be met by oxidative metabolism instead of glycolysis (28). The energy demands during light flashes, alternating with darkness, are thought to be met by glycolysis (28), which is consistent with an abundance of glycolytic enzymes found in the inner retina (368).

Darkness may induce complete retinal tissue anoxia in some parts of the cat retina, as a result of the increase in oxygen consumption (358). Interestingly, this phenomenon is not a general characteristic of all mammalian retinal tissue, as reported by Yu & Cringle (643), who demonstrated that oxygen tension in the outer retina of dark adapted rats does not decrease below 5 mmHg despite a 50% increase in oxygen consumption. This difference was explained by the increase in oxygen supply by increases in retinal blood flow in dark adapted rats, with this increase also occurring in humans (171).

Retinal and Choroidal Blood Flow

Regulation of retinal blood flow

Regulation of retinal blood flow occurs through local tissue factors, metabolic factors, and physical factors in response to changes in metabolic demands, light/dark transition, and pressure changes (i.e. systemic, intraocular, and perfusion pressure). Although the autonomic nervous system affects blood vessels in extraocular tissue (133), there is no change in retinal blood flow or preretinal oxygen tension following stimulation of the superior cervical sympathetic chain (53). In addition, there is no intraocular innervation of retinal blood vessels by adrenergic, cholinergic, or peptidergic nerves (257, 261, 262, 337, 639) despite the presence of cholinergic receptors (175). However, blood flow to the retina is regulated in part by glial cells that surround the retinal vessels (350), and by the metabolic needs of the retinal neurons being nourished by the retinal circulation. Below we review a number of factors that regulate retinal blood flow.

Regulation of retinal blood flow – blood gas

Oxygen

Oxygen concentrations help regulate retinal blood flow. Zhu et al. (657) demonstrated this regulation in a series of experiments wherein newborn pigs were provided 100% oxygen to breathe during measurements of retinal blood flow. The protocol resulted in hyperoxia and a decrease in retinal blood flow by around 40% within 5-10 minutes. Treatment with inhibitors targeted against endogenous vasoconstrictors such as 20-hydroxyeicosatetraeonic acid (20-HETE), endothelin, and thromboxane attenuated this response; however, antioxidant treatment with catalase and superoxide dismutase had no effect. The vasoconstrictive hyperoxia response also occurs in other animal models (161, 496, 552) as well as in humans (280, 493), where retinal arteriolar diameters and retinal blood flow decrease by as much as 60% as a result of 100% O₂ inhalation (280, 493). The response to hyperoxia is important to maintain a constant retinal pO₂ (472, 496). In comparison, diabetes-induced decreases in retinal blood flow rate in rodents (decreases of ~33%) have been found to be attenuated not only by targeting the vasoactive mediators thromboxane (343, 344, 622, 626), endothelin (605), 20-HETE (606), and angiotensin II (343), but also by scavenging superoxide with tempol (631), with the latter implicating ROS in retinal blood flow control during diabetes. However, in the diabetic retina, the exact mechanisms by which oxygen levels regulate the vasoconstrictor pathways is yet to be determined.

Not only does hyperoxia decrease retinal blood flow, but the reverse also is true, i.e., hypoxemia (low oxygen) increases retinal blood flow in cats (359), humans (470), and monkeys (161), presumably in an effort to maintain retinal pO₂. This mechanism has been shown experimentally when outer retinal oxygen tension decreased while inner retinal oxygen tension was maintained in experimentally induced hypoxia (359, 397).

Carbon Dioxide

Both hypercapnia (high arterial partial pressure of carbon dioxide; PaCO₂) as well as hypocapnia (low PaCO₂) affect retinal blood flow. In monkeys, retinal blow flow and preretinal vitreous oxygen tension (a measure of inner retinal oxygen availability) have been shown to decrease in hypocapnia (581, 582). Hypercapnia leads to an increase in retinal blood flow with each 1 mmHg rise in PaCO₂ resulting in a 3% rise in blood flow (582) and preretinal vitreous oxygen tension in monkeys (581). The hypercapnic response (i.e., increased retinal arteriolar diameter and flow) also has been reported in healthy adult humans (589).

Autoregulation and metabolic regulation

Retinal arterioles can maintain relatively constant tissue perfusion with moderate changes in perfusion pressure, with this mechanism known as autoregulation (133). Autoregulation occurs through a wide range of pressures in the healthy eye (25, 179, 491, 497, 500). Retinal blood flow also is regulated by the metabolic needs of the retina through the release of local factors from vascular endothelial cells and neural tissue surrounding the vessels (387, 473).

Retinal vascular control of flow through autoregulation and metabolic regulation is achieved by a complex system of vasodilators and vasoconstrictors, including nitric oxide (NO), prostacyclin, thromboxane A₂ (TxA₂), endothelin-1, and angiotensin II (Ang II) (229, 473). These potent vasoactive factors are produced by endothelial (and other) cells, and can modulate vascular tone by affecting vascular smooth muscle cells and/or pericytes. Thus, the retinal endothelium plays an important role in regulating retinal blood flow, which under pathological conditions can be disrupted due to endothelial dysfunction.

Nitric Oxide

Nitric oxide (NO), also known as endothelium-derived relaxing factor (EDRF) (520), is synthesized by NO synthase (NOS), and is formed as a result of the oxidation of L-arginine by oxygen, resulting in NO and L-citrulline (186). NO synthase is expressed in several retinal cells (81, 633), and has three isoforms: inducible NOS (iNOS), neuronal NOS (nNOS), and endothelial NOS (eNOS) (186). Non-vascular retinal cells express nNOS (123, 217, 316, 443, 590, 633), while eNOS is expressed in retinal and choroidal endothelial cells and retinal pericytes (81, 315, 388). In the non-diseased retina, iNOS is not expressed (6); however, during hypoxia (295) and in the retina of diabetic patients (6), iNOS is expressed in various retinal cell types.

Regulation of vascular tone is one of the main roles of NO. Released NO regulates vascular relaxation by modulating intracellular level of ions such as Ca²⁺ and K⁺ by activation of ion channels through activating guanylyl cyclase/cyclic guanine monophosphate (GC/cGMP) in smooth muscle (242). Many vasodilators including adenosine (114, 241, 249, 459, 592), acetylcholine (86, 298, 630), bradykinin (39, 106, 115, 122, 246, 410), histamine (309, 334, 351, 445), lactate (61, 199, 202, 247), and norepinephrine (283, 306) exert their effects by the release of NO.

Ocular blood flow and vasodilation depend on NO derived from the NO synthases in the retina. A variety of early studies investigating ocular blood flow found decreases following administration of Nitro-L-arginine (L-NA) (147), N^G-nitro-L-arginine methyl ester (L-NAME) (140, 219, 246, 247, 249, 304, 323, 410, 411, 543), and N^G-monomethyl-L-arginine (L-NMMA) (34, 220), suggesting a significant role for NOS in ocular blood flow. More specifically to the retina, acetylcholine-induced relaxation in retinal arterioles is attenuated with L-NMMA treatment (48, 637), and the effects of nicotine, histamine, and substance P on retinal vascular tone is mediated by the release of NO (48, 311, 576).

NO produced from eNOS plays an essential role in maintaining vascular functions. Laspas et al. have examined the different NOS isoforms in the ex vivo reactivity of the ophthalmic artery to acetylcholine (335), with their results pointing to the importance of eNOS, but also to compensatory NOS-independent vasodilatory pathways with a chronic loss of eNOS. Further downstream in the retinal arterioles, similar results have been found with regards to the importance of eNOS in endothelium-dependent vasodilation (208, 209), and also with regards to the upregulation of compensatory pathways in the absence of the enzyme (209).

Light flicker is known to dilate retinal arterioles, and although the precise mechanism is not completely understood, the resulting functional hyperemia (68, 147-149, 200, 255, 320, 464) is thought to be mediated by NO. Retinal NO levels increase following flicker stimuli (68, 147), as is the blood flow response in the optic nerve head (68, 320). Ocular vascular responses to light flickering is suppressed by NOS inhibition with L-NAME (68, 320, 642), L-NMMA (134), N^G-nitro-L-arginine (L-NNA) (68), and N(ω)-propyl-L-arginine (L-NPA) (642).

Studies have indicated that retinal neural activity and NO contributes to the light flicker response (320, 494, 546), which is modulated by glial toxin (148, 543, 546) or nNOS inhibitors (543, 642). Moreover, the amount of NO in retinas from eNOS-deficient mice has been found to be similar to that in wild-type C57BL/6 mice (possibly due to compensatory production of NO mentioned earlier), and distribution of nNOS in the deep retinal capillary plexus is increased in eNOS knockout mice (18).

Prostaglandins

Prostaglandins, which can be synthesized by retinal endothelial cells (499), are produced as a result of the metabolism of arachidonic acid by cyclooxygenase, leading to the formation of four main prostaglandins: prostaglandin E1 (PGE₁), prostaglandin F_2α (PGF_2α), prostaglandin D2 (PGD₂) and prostacyclin (PGI₂) (489). Prostaglandins have been shown to regulate retinal vascular tone through complex mechanisms, where they can act as vasodilators or vasoconstrictors. In studies on anesthetized rabbits, systemically administered PGE₁ and PGF_2α caused retinal vasodilation (549). However, these prostaglandins caused vasoconstriction of isolated bovine retinal arteries and miniature pig retinal arterioles (422, 470). Moreover, PGE₁ and PGI₂ increased retinal blood flow in rats when injected intravenously (398).

Thromboxane

Thromboxane A₂ (TxA₂), also synthesized as a result of arachidonic acid metabolism by cyclooxygenase (322), binds to the thromboxane-prostanoid (TP) receptor (584). Sources of TxA₂ include platelets (433) and leukocytes (254). TxA₂ does not play a role in retinal blood flow in non-diabetic mice or rats but its inhibition restores retinal blood flow in hyperglycemic rodents (343, 344, 622).

Endothelin

Endothelin-1 (ET-1), also known as endothelium-derived constricting factor (EDCF) (588), is the most potent vasoconstrictor that has been discovered to date (636). ET-1 is mainly synthesized by endothelial cells, however, other cell types have been shown to synthesize the mediator, including vascular smooth muscle cells, macrophages, neurons, and leukocytes (297). ET-1 is part of the endothelin family that also include ET-2 and ET-3, and is synthesized by the cleavage of preproendothelins to big endothelins, which is further cleaved to form ET-1 (275). Endothelin-1 binds to endothelin-A (ET_A) and endothelin-B (ET_B) receptors, both of which are expressed on retinal arterioles (245). ET-1 binding to receptors on vascular smooth muscle induces vasoconstriction. Although activation of ET_B receptors on the endothelium has the capability of inducing vasodilation mediated by NO (518), ET_B receptors have been found to play a relatively insignificant role in the porcine retinal circulation (88, 466). Under basal conditions, ET-1 expression is reduced by NO and PGI2 (228, 482); however, shear stress, norepinephrine, cytokines, and angiotensin II increase its expression (378). In healthy adult humans, retinal vessel diameter, blood flow, and velocity are significantly reduced with the administration of ET-1 (463). Moreover, dose-dependent vasoconstriction occurs in rat (70), rabbit (566), and cat (221) retinal arterioles with intravitreal injection of ET-1.

Angiotensin II

Angiotensin II is a hormone produced when the angiotensin converting enzyme (ACE) converts angiotensin I to angiotensin II. Angiotensin II binds to AT₁ receptors resulting in vasoconstriction, and to AT₂ receptors resulting in vasodilation. Ocular tissue contains the components required to produce its own angiotensin II (116, 117). In addition, the retinal vasculature contains ACE, AT₁ receptors and AT₂ receptors (174, 176, 408, 509, 611).

Lactate

Aerobic and anaerobic glycolysis produces lactate in the retina with anaerobic being greater than aerobic production. Multiple cell types in the inner and outer retina produce lactate in the normoglycemic state (615). When energy supply is sufficient, lactate produces constriction in the retinal microvasculature; yet lactate produces dilation of the retinal microvasculature during hypoxic conditions (634). Lactate-induced dilation of isolated porcine arterioles has been found to be mediated by NO synthase, activation of guanylyl cyclase, and K_ATP channels (247).

Adenosine

Adenosine is formed from the degradation of adenosine triphosphate (ATP). Adenosine is a vasodilator that has been shown to function in the retinal circulation as part of the autoregulatory process (60, 213, 214). Retinal vessel responses occur from adenosine release from neural tissue on the outside of the vessel during hypoxia or ischemia (21). The response to adenosine occurs through the A₂ receptor which stimulates the opening of K_ATP channels and release of NO (249, 281, 352).

Retinal arterio-venular communication

Arteriolar diameter can be controlled by venules that are closely paired to arterioles flowing in countercurrent directions (237, 251). In this mechanism, it is thought that vasoactive metabolites from the venules diffuse to the closely-paired arterioles and cause vasodilation or vasoconstriction (251). In the human retina, arterioles and venules are found in this type of parallel, countercurrent orientation, but whether or not this arrangement helps control arteriolar perfusion has yet to be determined. In the rodent retina, arterioles and venules extend out from the optic disk in a more radial than parallel fashion, although presumably, diffusive exchange could still occur between venules and arterioles having smallest angles between the two, with data from a few studies supporting this possibility (343, 344, 622).

Retinal blood flow - light conditions

Retinal blood flow responses to light/dark adaptation have been varied. During dark adaptation, flow velocity increases in retinal arteries and veins in healthy subjects when measured with laser Doppler velocimetry (171, 492), and retinal blood flow also increases in the dark when measured using a microsphere perfusion assay (53). However, others have reported no change during dark adaptation in flow velocity, using a laser Doppler flowmeter (495) – or in arterial diameter changes, using a scanning laser ophthalmoscope (45).

Regulation of choroidal blood flow

Choroidal blood flow is one of the highest per gram of tissue in the body (25, 219) and supplies a large fraction of total ocular perfusion (25). Despite the high blood flow, oxygen extraction is only 3-4% (24, 159, 578). The autonomic nervous system controls choroidal blood flow (52, 133): sympathetic nervous system innervation occurs through the superior cervical ganglion, and uveal blood vessels constrict upon stimulation, resulting in a decrease in choroidal blood flow (23, 52, 133, 369). The choroidal response to sympathetic stimulation is thought to be protective via the autoregulation of blood flow and perfusion pressure (53).

The parasympathetic nervous system innervation occurs through the facial nerve, oculomotor nerve, and the trigeminal nerve (ophthalmic and maxillary divisions) (133, 369), with choroidal blood flow increases occurring upon stimulation of those nerves (555-557). Environmental light increases blood flow in the choroid (192), with the opposite occurring in the dark (193). Maintenance of outer retinal temperature appears to be an important role of the choroidal blood flow response to parasympathetic nervous system stimulation (454, 455).

Choroidal blood flow exhibits autoregulation at intraocular pressures (IOP) below 25 mmHg when the perfusion pressure (mean arterial pressure - intraocular pressure) exceeds 40 mmHg (305). When IOP is held constant at resting or elevated levels, choroidal blood flow becomes more linear when mean arterial pressure is decreased (i.e., less autoregulation) (305). As IOP continues to rise to levels 3-fold above control, choroidal blood flow decreases (25). Autoregulation of the choroid occurs through the myogenic mechanism (303, 305), which is based on the law of Laplace: vascular wall tension (T) = transmural pressure (P) × vessel radius (r). As transmural pressure increases, vessel diameter will decrease in an effort to maintain wall tension, resulting in the maintenance of blood flow.

The blood retinal barrier (BRB)

The retina is part of the central nervous system, and its vascular system shares the same characteristics as the brain vasculature. The blood-retinal barrier (BRB), which can be compared in its structure and function to the blood-brain barrier (BBB), is a physiological barrier that limits and controls the circulating blood components from reaching the surrounding tissue. A compromised BRB can lead to various retinal degenerative diseases and blindness.

The blood retinal barrier (BRB) can be divided into two components, the inner (iBRB) and outer (oBRB). Adherens and tight junctions between adjacent endothelial cells, and possibly the glycocalyx, form the iBRB. On the other hand, tight junctions between the retinal pigmented epithelial cells (RPE), which segregate the choroid from the neuronal retina, form the oBRB. Occludins, zonula occludens (ZO-1, -2, and -3), and claudins are specialized proteins that are components of tight junctions, and regulate the transport of molecules and solutes through the RPE and endothelial cell layers (55, 301, 390).

Although tight junctions are common between retinal endothelial and epithelial cells, their organization, distribution, and composition are different between the two cell types. In epithelial cells, tight junctions are localized to the apical side of the cell, while tight junctions in endothelial cells are distributed between gap and adherens junctions. In endothelial cells, adherens junctions are composed of a complex of proteins that anchor to the cytoskeleton and link cells to each other. These proteins include vascular endothelial cadherin (VE-cadherin), catenins, and plakoglobin (132). Moreover, cell-cell junctions contain additional adhesion molecules, such as platelet endothelial cell adhesion molecule-1 (PECAM-1), which helps in the maintenance of retinal endothelial barrier integrity (129). In addition to cell adhesion, these proteins also contribute to cell function by acting as scaffolds, and by participating in cell signaling pathways, contact inhibition, cell growth, and apoptosis (130, 269, 270, 418, 583).

Mechanisms Contributing to Diabetic Retinopathy

Overview of diabetic retinopathy as a microvascular pathology

Although recent work has indicated that retinal neuronal changes may occur even prior to overt microvascular pathology in diabetic retinopathy, the major changes in the vasculature have justified the consideration of diabetic retinopathy as a microvascular disease. Several events occurring in diabetic retinopathy include a breakdown of the blood-retinal barrier, vascular cell death, and leukocyte plugging of capillaries with increased vascular permeability being one of the first observations in diabetic retinopathy during the non-proliferative stage. Leaky vessels lead to the accumulation of fluid in retinal tissue and the development of macular edema, the clinical feature most clinically associated with blindness (145). In addition, many studies have indicated an increase in leukocyte adhesion in the diabetic retina of animal models (mice, rats) and in humans. The plugging of vessels by leukocytes limits the oxygenation of the tissue and vessel wall, causing the formation of acellular capillaries that are no longer perfused. The decrease in blood flow and increased inflammation will stabilize hypoxia-inducible factor-1α (HIF-1α) leading to its translocation to the nucleus and its dimerization with HIF-1β, and binding to the hypoxia response element on the promoter sites of important genes such as the one encoding for vascular endothelial growth factor (VEGF) (522). VEGF will promote uncontrolled angiogenesis in the inner retina with abnormal proliferation. In addition, these new blood vessels can be leaky and fragile leading to edema and microhemorrhages in the more serious proliferative stage of diabetic retinopathy.

Stages of diabetic retinopathy

There are two classifications of diabetic retinopathy: non-proliferative diabetic retinopathy (NPDR) and proliferative diabetic retinopathy (PDR). Non-proliferative diabetic retinopathy can be further subdivided into three phases depending on the severity of the complications: mild, moderate, and severe NPDR. These stages represent progressive damage to the retinal microvasculature that contribute to the pathology of diabetic retinopathy.

Non-proliferative diabetic retinopathy

NPDR is marked by various retinal microvascular complications that can be observed clinically, such as intraretinal hemorrhages, microaneurysms, cotton wool spots, and retinal tissue edema due to increased retinal vascular permeability. Additionally, acellular retinal capillaries start to develop due to pericyte and endothelial cell loss. Acellular capillaries are non-perfused, and initially occur in few capillaries, thus have no clinical consequences. However, as the disease progresses and more acellular capillaries develop, hypoxia will become more severe leading to localized ischemia triggering the release of growth factors and the initiation of abnormal neovascularization.

Proliferative diabetic retinopathy

PDR is the advanced and more severe stage of diabetic retinopathy and characterized by the development of new blood vessel growth in the optic disc or elsewhere in the retina. However, these blood vessels are leaky and fragile, causing worsening of intraretinal hemorrhages and microaneurysms (632) and damage to the retina. The new blood vessels can penetrate the vitreous and become leaky, leading to vitreous hemorrhages and retinal detachments (632). It is thought that hypoxia leads to the release of growth factors that promote neovascularization (618); however, there is evidence that hypoxia does not occur early in the diabetic retina (338, 624). Additionally, neovascularization limits the amount of light reaching the cone and rod layers, which further impairs vision.

Initially, the new vessels are asymptomatic, with little to no hemorrhage, and usually originate around the area of the optic disk (119, 570). These new blood vessels form networks and loops, and cross arterioles and venules in the underlying retina, making them distinguishable from other retinal blood vessels (121). Mild thickening of the optic disk and the surrounding retinal tissue also occurs due to vascular leakage and edema (121). As the condition progresses, increased hemorrhage and tissue damage occurs. Moreover, white delicate fibrous tissue starts to form adjacent to the new vessels, forming what are known as fibrous vessels (502). The new blood vessels enter a cycle of proliferation and regression, leaving behind the fibrous tissue that creates traction across the retina (121).

An additional complication of diabetic retinopathy is the weakening of the attachment between the vitreous and the inner limiting membrane of the retina (92, 416). Under normal conditions, collagen, laminin, and fibronectin anchors the posterior vitreous to the inner limiting membrane, however, under hyperglycemic conditions, adhesion between the retina and posterior vitreous weakens, leading to posterior vitreous detachment (PVD) (17). Fibrous vessels and tissue are attached to the vitreous, leading to retinal tearing, hemorrhage, and damage due to PVD.

Current therapies of diabetic retinopathy

Tight glucose control in individuals with diabetes provides some benefits such as delayed onset of retinopathy, slowed progression of retinopathy (414), and improved retinal vascular permeability (56). However, despite the benefits, the risk of developing retinopathy is not eliminated (414) with intensive insulin therapy. Interestingly, in diabetic individuals with mild retinopathy, intensive insulin therapy leads to transient worsening (414) that is mediated in part by HIF-1α upregulation of VEGF (468). Even so, with proper medical supervision, the benefits of intensive insulin control outweigh the risk of hypoglycemia (120, 414).

Panretinal photocoagulation, vitrectomy, and VEGF inhibitors are therapies that are used in the treatment of PDR (189). Panretinal photocoagulation is thought to decrease upregulation of growth factors such as VEGF by ablating regions of ischemia in the peripheral retina (94, 632). A complete review of current therapies is beyond the scope of this review with more comprehensive reviews found elsewhere (58, 94, 178).

While intensive insulin therapy has some benefits and multiple therapies exist for individuals once proliferative diabetic retinopathy occurs, there is no cure for diabetic retinopathy or therapy that prevents its development.

Retinal vascular complications in diabetic retinopathy

Hyperglycemia initiates a destructive cascade of events leading to retinal microvascular alterations. These alterations affect the integrity and functions of the retina and lead to eventual vision loss. By the time microvascular alterations are clinically observable, significant and sometimes irreversible damage has already occurred. Thus, the detection of early molecular changes is important to control the progression of diabetic retinopathy.

Microaneurysms

During the mild phase of NPDR, saccular structures outpouching from capillary, known as microaneurysms, start to form. Microaneurysms are the earliest manifestation that can be observed clinically, and can be a result of structural weakening and modification of the capillary wall. Although the exact mechanisms leading to the development of microaneurysms are not fully understood, studies have speculated that they may arise as a result of pericyte loss and subsequent capillary wall weakening (188). Microaneurysms start initially as hypercellular structures, with the lumen filled with erythrocytes and thrombi, and progress to acellular microaneurysms (650). Microaneurysms can leak plasma, contributing to retinal edema and thickening.

Pericyte dropout

One of the earliest pathological changes in diabetic retinopathy is pericyte loss in the retina (98). As retinal microvessels start to become more tortuous, early signs of pericyte dropout can be observed as empty pockets in the capillary basement membrane due to their loss (98, 232). In addition to pericytes, endothelial cells also are lost, leading to the development of acellular capillaries (188). The mechanism for endothelial and pericyte loss is still under investigation; however, many studies implicate the activation of apoptotic and necrotic pathways due to hyperglycemia, and the accumulation of toxic byproducts leading to cell death (354, 393, 580).

Pericytes are supporting mesenchymal contractile cells that surround capillaries, and are derived from progenitor mesenchymoangioblast cells analogous to smooth muscle cells in larger vessels, and thus, may play a similar role (47). Due to their role in regulating capillary function, pericytes share a basement membrane and stay in close contact with endothelial cells in the capillaries. Additionally, pericytes exist in a 1:1 ratio with endothelial cells in the retina, which is considerably higher than the ratio in the brain, perhaps indicating the involvement of the cells in the strong blood-retinal barrier (231). Multiple factors facilitate the communication between pericytes and endothelial cells, including molecules such as platelet-derived growth factor-β₁ (PDGF-β₁), sphingosine-1-phosphate (S1P), angiopoietin, and transforming growth factor-β₁ (TGF-β₁) (169, 231, 233, 264, 318). Pericyte apoptosis and dropout have been observed under pathologic conditions such as hypertension, hyperglycemia, and accumulation of advanced glycation end products, as well as with basement membrane thickening (47, 231, 587). Additionally, due to their role in regulating endothelial cell proliferation and the BRB, pericyte loss could lead to the pathologic angiogenesis of retinal capillaries, as well as the breakdown of the blood-retinal barrier.

Cotton wool spots

Cotton wool spots indicate ischemic tissue in the nerve fiber layer due to occlusion of retinal microvessels (384) and consist of axoplasmic accumulations from nerve fiber axons. Axoplasmic transport, also known as axonal flow, occurs bidirectionally between the soma (neuronal body) and the synapses transporting organelles and cellular components (385). In undamaged segments of the axon, axonal transport continues, leading to terminal swelling of the damaged axon, and the formation of cytoid bodies, which are the histological hallmark of cotton wool spots (33, 385).

Blood-retinal barrier (BRB) breakdown

One of the early events occurring in NPDR is BRB breakdown and hyperpermeability, which sometimes can precede any visible clinical manifestations of retinopathy (188, 650). BRB breakdown leads to capillary leakage and the accumulation of plasma components in retinal tissue causing increased thickening and visual impairment. Hyperglycemia-induced retinal endothelial dysfunction is a significant contributor to the disruption of BRB, due to the vital role that the retinal endothelium plays in the maintenance of BRB integrity. The molecular mechanisms leading to BRB breakdown are multifactorial and complex, with many contributing factors such as increased VEGF levels (406, 612, 619), leukocyte adhesion (286), growth factors (30, 235), cytokines (287), inflammation (523), and a reduction in tight junction proteins (30, 235). These factors, as well as the influence of the endothelial glycocalyx, will be discussed more in detail later in this review.

Molecular mechanisms leading to the pathogenesis of diabetic retinopathy

The development and progression of diabetic retinopathy can be associated with multiple pathological factors, such as hyperglycemia and genetic predisposition (13, 73). Some of the major mechanisms that have been reported in diabetes-induced retinal stress leading to microvascular damage and retinopathy include: 1) oxidative stress, 2) the polyol pathway, 3) protein kinase C (PKC) activation, 4) non-enzymatic glycation, 5) inflammation, and 6) genetic factors. These mechanisms lead to an increased expression of VEGF (7, 59, 124, 375) prior to development of proliferative retinopathy (29), stimulating the growth of new vessels, increased vascular permeability (278), and leukocyte activation and adhesion (278, 288). The next sections of this review will cover the hyperglycemia-induced enzymatic and non-enzymatic changes, and the associated pathophysiological alterations in the diabetic retina.

Glucose transport in diabetic retinopathy

Hyperglycemia is a causative factor in developing microvascular complications in diabetes (1, 195, 414). Glucose, used by retinal neurons in metabolism, has to pass through the endothelial cells that form a barrier between plasma and the underlying tissue. The correlation between poor glycemic control and diabetic retinopathy severity points to the role of chronic high glucose exposure in the retinal endothelium (313). The endothelial cells themselves are sensitive to increases in plasma glucose, with a lack of insulin regulation of glucose transport (32), and with glucose uptake by the endothelium proportional to plasma glucose concentrations (66). Depending on the capillary bed, glucose can either pass through the endothelial cell junctions (with this transport being increased in the diabetic retinal vasculature) or by the specialized glucose transporter GLUT1. The GLUT1 transporter is an insulin-independent, constitutive transmembrane protein that belongs to the major facilitator superfamily (MFS) that is heavily expressed on endothelial cells, and is the primary glucose transporter in the brain and the retina (156, 534, 640).

The multifactorial nature of glucose transport regulation and GLUT1 expression under hyperglycemic conditions need to be considered when discussing the pathological changes occurring in the retinal microvasculature. Multiple factors can regulate the expression of GLUT1, such as growth factors, hypoxia, and oxidative stress. VEGF, a cytokine that is upregulated early in diabetic retinopathy, has been shown to increase glucose transport in retinal endothelial cells (544). Additionally, hypoxia, which occurs at a later stage of diabetic retinopathy, also has been shown to upregulate glucose transport in retinal endothelial cells (565). GLUT1-induced glycosylation, possible phosphorylation, or conformational changes may also affect transport kinetics under hyperglycemic condition (141, 142, 155).

Interestingly, there are conflicting reports about GLUT1 expression levels in the retina under hyperglycemic conditions. In postpartum retinal tissues obtained from three long-term diabetic individuals with minimal or no clinical manifestation of retinopathy, a significant increase in GLUT1 expression was observed in the inner blood retinal barrier (BRB) (329). However, in 1-year-old Goto-Kakizaki (GK) rats that develop type-2 diabetes, no change in GLUT1 levels was observed (173). In another study, GLUT-1 levels were reduced in two weeks or two months of streptozotocin (STZ)-induced diabetic rats (38). These seemingly conflicting reports might be attributed to the use of different disease models, duration of diabetes, and analysis techniques.

In addition to GLUT1, other glucose transporters are expressed in various organs and tissues of the body. For example, GLUT2 is expressed in the gastrointestinal system, pancreatic islets, and the liver; GLUT3 is expressed in the central nervous system and the brain; and GLUT4 is expressed in adipose tissue and skeletal muscle (138). Of these latter three, GLUT3 has been reported to be expressed in retinal cells (609). Expression of GLUT3 remains constant or decreases in vitro under hyperglycemic conditions (314, 479), and remains constant in the retina of hyperglycemic rats (38).

Altered metabolism in the diabetic retina

There have been conflicting reports about oxygen consumption and arteriovenous oxygen differences in the retinas of diabetic rats (54, 128, 599), but with decreases of retinal oxygen consumption demonstrated in diabetic rabbits (272, 562) and cats (360). In all stages of retinopathy, arteriovenous oxygen differences in humans have been reported to decrease (230). The early decrease in retinal blood flow probably coincides with a decrease in oxygen consumption in patients with diabetes (95), which has been observed in several animal models. A blood flow reduction can result from a sustained or transient decrease in retinal oxygen consumption, which could be occurring in the diabetic retina as speculated by Small et al. (539) and Rimmer & Linsenmeier (490).

Models of retinal metabolic changes in diabetes often use rats or mice, and a difference in rat vs mouse metabolism was detailed in a study conducted by Obsorova et al. (436), where several metabolites such as lactate, ammonia, glutamate, free NAD+/NADH, and α-ketoglutarate were significantly decreased by 30-50% in STZ-induced diabetic rat retina, while only pyruvate and ammonia were decreased in the STZ-induced diabetic mouse retina.

A decrease in retinal metabolism under diabetic conditions could be explained by hyperglycemia-induced neuronal death. Increased retinal cell death has been observed in STZ-induced diabetic rats at 1-12 months of the disease (44). Additionally, in STZ-induced diabetic rats, photoreceptor apoptosis continues to increase in the time span of 4-24 weeks of hyperglycemia (453). The same phenomenon also is reported in STZ-induced diabetic mice (367, 380), and in Ins2^Akita diabetic mice (367), where retinal neuronal cell death is observed. Moreover, retinal thinning, which is reported in diabetic rodents (380, 649) and humans (50, 63, 426, 586), is attributed in part to retinal cell death.

Altered oxygen levels in the diabetic retina

An uncontrolled growth of new blood vessels as a result of retinal hypoxia is thought to contribute to the development of diabetic retinopathy (31). There are several pieces of evidence that suggest a possible decrease in oxygen tension early in the diabetic retina; however, the evidence is not conclusive and additional studies may be required to resolve the time course of either hypoxia or hyperoxia early in the progression of experimental and human diabetes.

As will be discussed in more detail later in the review, retinal blood flow rates decrease early in the progression of diabetic retinopathy both in humans and in rodent models. Additionally, there has been a reported decrease in functional capillarity (RBC perfusion) of the retina of hyperglycemic rats following streptozotocin injection (progressively decreasing over a 15-90 day period), which can be partially blocked by inhibition of thromboxane synthesis or binding of thromboxane (126, 127). In addition to the changes in retinal capillary vasculature, various animal models of hyperglycemia result in non-perfused acellular capillaries (43, 299, 300, 655, 656). Hypoxia-inducible factor-1 and -2 (HIF-1 and HIF-2) expression and direct measurements of retinal oxygen tension, both of which are as discussed below, provide methods to examine potential alterations of oxygen tension in the diabetic retina.

HIFs are transcription factors that are upregulated when oxygen tension in the cell shifts from normoxia to hypoxia (296). HIF-1 contains two subunits, HIF-1α, which is expressed during times of normoxia but rapidly degraded, and HIF-1β, which is constitutively expressed (296). As oxygen tension decreases in the cell, HIF-1α becomes stabilized in the cytoplasm and forms a dimer with HIF-1β in the nucleus following translocation. A more comprehensive review of the regulation of HIF-1 can be found elsewhere (296). Stabilization of HIF-2α occurs at higher oxygen tensions than that of HIF-1α (457). HIF-1α regulates expression of glycolytic enzymes, while HIF-2α regulates expression of transforming growth factor-α (TGF-α) and erythropoietin (EPO), and both HIFs regulate VEGF and GLUT1 (457).

Oxygen levels in the retina have been measured using oxygen microelectrodes, and by exogenous bioreductive markers of hypoxia (e.g., pimonidazole). Microelectrode measurements have provided a direct measure of oxygen tension through the layers of the retina (644), while oxygen tension within a cell must decrease below 10 mmHg for bioreductive markers to bind macromolecules in the cell (363). This threshold for bioreductive sensors results in detection in the some parts of the retina even in control, non-diabetic eyes due to low oxygen tension levels that result from high oxygen utilization (645, 646).

Interestingly, within one year of inducing diabetes in cats and dogs, no signs of retinal hypoxia were reported in studies where microelectrodes have been used to measure oxygen concentration (162, 550, 551). Similar results have been reported by our lab in studies conducted in diabetic mice and rats, where using the hypoxia probe pimonidazole, or with hypoxia-inducible factor HIF-1α immunostaining, diabetes produced no significant signs of hypoxia in the early stages of hyperglycemia (1-6 months) (624, 625, 627), even with a substantial decrease in retinal blood flow rate (20-45%) (343, 344, 605, 606, 622, 626, 631). In contrast, an increase in retinal oxygen in the tissue was detected by our lab (using pimonidazole) at three months of hyperglycemia in diabetic rats (625), which was also reported at the same time point by Lau and Linsenmeier using oxygen microelectrodes (339). Nevertheless, increased pimonidazole and/or HIF-1α staining in the diabetic retina have been reported by other groups (124, 349, 370, 469, 483) at different time points, with de Gooyer et al. (124) reporting increased pimonidazole binding in the ganglion cell layer and inner nuclear layer of the mouse retina following five months of hyperglycemia. The increased binding of pimonidazole was eliminated in a hyperglycemic mouse model having outer retinal degeneration (124). In rats, there have been reports of increases in HIF-1α in the ganglion cell layer and inner nuclear layer two weeks after the induction of hyperglycemia (467), and in nuclear extracts eight days after induction of hyperglycemia (468). HIF-1α and HIF-2α in diabetic human retinas have not been found to differ from nondiabetic human retinas post mortem (198).

With longer-term diabetes, Linsenmeier et al. (360) found a significant increase in hypoxia in cats with 6-7 years of hyperglycemia using oxygen microelectrodes, with a significant reduction of almost 50% in inner retinal oxygen tension when compared to controls, and a decrease in photoreceptor oxygen utilization reported in one out of three diabetic cats.

Reactive oxygen species (ROS) in the diabetic retina

ROS production in the diabetic retina

Superoxide overproduction has been implicated in the development of diabetic retinopathy, where multiple enzymatic pathways have been shown to contribute to the production of the radical, such as by xanthine oxidase, mitochondrial complex Q, cyclooxygenase, p450 monooxygenase, NADPH oxidase, lipoxygenase, and uncoupled nitric oxide synthase. The increase in retinal superoxide levels and production have been reported in multiple studies of rodent models of diabetic retinopathy. Increased superoxide production, primarily from the mitochondria, and small amounts from NOS and NADPH oxidase, have been reported in STZ-diabetic rats with two months of hyperglycemia (150). Moreover, an increase in superoxide production in the photoreceptor layer, both in light- and dark-adapted STZ mouse diabetic eyes, was observed following two months of hyperglycemia (153).

Additionally, activity of superoxide dismutase (SOD), which converts superoxide into oxygen and hydrogen peroxide, was decreased in STZ-induced diabetic rats of both short-term hyperglycemia of 6 weeks (437) or 2 months (326), as well as with a longer-term duration of 11 months (325). However, it should be noted that an increase in SOD activity has been reported in the retinas of STZ-induced diabetic rats (136). The resulting production of hydrogen peroxide could be further converted into water with the help of catalase, or via its interaction with glutathione with the help of the enzyme glutathione peroxidase. In the diabetic rat retina, a moderate increase in catalase has been reported, with a possible decrease in glutathione peroxidase and no change in glutathione levels (436).

Role of ROS in the diabetic retina

Over the years, a “unified hypothesis” (67, 211, 427) has been developed in which superoxide overproduction plays a central role in diabetes-induced pathophysiological complications. In this hypothesis, superoxide initiates multiple damaging signaling pathways that occur concurrently, leading to the increased activity of the hexosamine pathway, the polyol pathway, the activation of PKC, and the increased formation of AGEs and their receptors. The increase in cellular glucose levels under hyperglycemia, the increase in pyruvate as a result of glucose metabolism, and participation of pyruvate in the Krebs cycle, all lead to an increased activity of the mitochondrial transport chain, and the overproduction of superoxide. This comes as a result of the production of the electron donors NADH and flavin adenine dinucleotide (FADH₂) in the Krebs cycle, and their role in the electron transport chain. Proton extrusion increases, which causes the mitochondrial membrane voltage to reach and exceed its threshold, leading to the blockage of electron transport in complex III, and generation of superoxide by coenzyme Q donation of its electrons to O₂. Additionally, the highly reactive hydroxide anion can be produced by the reduction of hydrogen peroxide that is formed by the conversion of superoxide by SOD. The highly reactive nitrogen species peroxynitrite (ONOO⁻) can be formed with the reaction of superoxide with nitric oxide, which can be damaging to the cell.

The initial hyperglycemia-induced superoxide production by the mitochondria may trigger the activation of additional pathways that could generate superoxide, such as uncoupled NOS or NADPH oxidase (211). The blocking of NADPH oxidase (NOX) activity in the diabetic mouse retina completely abolished ROS production, in a study in which mice were treated with the NOX inhibitor apocynin, or in NOX2-deficient mice. These results indicate a significant role for NADPH oxidase in the diabetes-induce ROS production in the diabetic retina. However, in findings reported by Nishikawa et al. (428), mitochondrial superoxide was shown to be the primary inducer of the hyperglycemia-induced retinal ROS production. The link between the two mechanisms could be explained by Al-Shabrawey et al. (19)., who suggested a role for mitochondria-derived ROS in activating PKC, which in turn could activate NOX, leading to sustained ROS production. Evidence for this connection was obtained in additional studies performed to show the ability of NOX to continue ROS production that was initiated by mitochondrial ROS (308, 345, 513).

The polyol pathway

The activation of the polyol pathway has detrimental and injurious effects on retinal cells. For example, there is an accumulation of the strongly hydrophilic, polyhydroxylated alcohol sorbitol, which initiates osmotic shock leading to hyperosmolar stress (310, 366, 568). Hyperosmolar stress leads to deleterious cellular responses including apoptosis, oxidative stress, DNA damage, inhibition of transcription and translation, mitochondrial adaptation, and cell cycle arrest (69). Consequently, hyperosmolar stress affects the retina and its functions due to the damage occurring to endothelial cells, loss of pericytes, thickening of the basement membrane, and injury to the retinal pigmented epithelium (165, 194).

Another metabolic product of polyol pathway activation, fructose, can induce cellular damage due to its high glycation activity, being a more potent glycating agent than glucose (560). Fructose is phosphorylated to fructose-3-phosphate, and further metabolized to 3-deoxyglucose, with both of these products potent glycating agents that can produce advanced glycation end-products (AGEs). AGEs can alter cellular components, leading to further cellular damage and impairment of the retina (293).

Studies have indicated a role for increased aldose reductase (AR) expression and sorbitol accumulation in disruption of the BRB. Increased AR levels in vascular retinal endothelial cells and Müller cells were found to be accompanied by a disruption of the BRB in the eyes of diabetic patients in a postmortem electron microscopic study (594). Additionally, NADPH is a cofactor for AR, with an increased activity of AR reducing the availability of NADPH to contribute to other important processes in the cell, such as the reduction of glutathione disulfide to glutathione (46). Moreover, the NADH/NAD+ ratio will increase due to the consumption of NAD+ by sorbitol dehydrogenase, leading to the initiation of pseudohypoxia, a state in which the tissue mimics a hypoxic response, in spite of normal oxygen levels, with an activation of HIF-1α (216, 614). Currently, multiple therapies for diabetic retinopathy are targeted toward the inhibition of AR; however, this strategy lacks complete specificity, with drug reactions with analogous AR enzymes, or with impaired AR detoxification and further injury (22).

Advanced glycation end products

AGEs are proteins and lipids that have been post-translationally modified and non-enzymatically glycated and oxidized by exposure to aldose sugars (516). AGEs can lead to vascular damage by modifying cytokines, hormones, and extracellular matrix proteins, leading to their dysfunction (65). Additionally, AGEs cross-link key molecules in the basement membrane, and can initiate deleterious signaling pathways that can alter multiple cellular functions leading to further damage (537).

Oxidative stress, protein turnover, and hyperglycemia are detrimental factors in the formation of AGEs. Schiff bases and Amadori products act as precursors to AGEs in the Maillard reaction. The reversible chemical reaction between the amino groups of proteins and glucose creates Schiff bases, which undergo rearrangement during continuous glucose exposure, and become the more stable Amadori products (528). Protein denaturation and crosslinking occur when the carbonyl groups, which are highly reactive and result from the reorganization of the Amadori products, react with amino, sulfhydryl, and guanidine functional groups of these proteins (10, 191). On the other hand, stable AGE compounds are created from the interaction of these carbonyl groups with lysine and arginine (364).

The receptors for advanced glycation end products (RAGEs), which belong to the immunoglobulin superfamily of receptors, interact with AGEs to initiate various signaling cascades within the cells leading to increases in nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) production, increases in VEGF production, inhibition of DNA synthesis, and an increase in retinal pericyte apoptosis (374), all of which can be detrimental to the cell. Moreover, studies have indicated a role of increased N-epsilon-carboxy methyl-lysine, an AGE product, in the development of type II diabetic complications in the retina (438). Currently, multiple techniques have been developed for the detection of AGE accumulation in the body, such as pulse wave analysis and skin autofluorescence. Additionally, various studies are being conducted with the hopes of finding successful therapies to limit AGE formation and its damage, such as the use of pigment epithelium derived factor (PEDF). PEDF has been shown to inhibit the formation of ROS due to AGEs, which in turn limits diabetes-induced pericyte dropout that can lead to BRB disruption (205, 530).

Protein kinase C (PKC) pathway activation

PKCs are a group of enzymes that are involved in various signaling pathways that can participate in multiple physiological and pathophysiological processes. PKCs belongs to the AGC kinase family that are widely distributed in various mammalian tissues, and include cAMP-dependent protein kinase A (PKA), cGMP-dependent protein kinase G (PKG), and lipid-activated PKC (327, 419). PKC activation, and the subsequent phosphorylation of target proteins at their serine and threonine residues (429), trigger a pathophysiological response that may contribute to the pathogenesis of diabetic retinopathy. Studies of this pathology have shown an increase in retinal vascular permeability, pericyte dropout, and retinal blood flow alterations that were associated with PKC activation (265, 452, 613).

The increased level of glucose in endothelial cells during hyperglycemia leads to the increased de novo synthesis of diacylglycerol (DAG), an endogenous activator of PKC. Glucose, which is transported into the endothelial cell via GLUT1, is metabolized via glycolysis to initially form the metabolic intermediate glyceraldehyde-3-phosphate, which is then converted via a multi-step process to DAG (629). DAG increases the affinity of PKC for Ca⁺⁺, which results in increased PKC activity, and increased ROS and NO production (312). Additionally, enhanced PKC activity has been linked to an increase in VEGF production (480, 613) and the increase in TGF-β1 production that can contribute to the accumulation of extracellular matrix proteins (227, 628). Moreover, the overexpression of PKC under hyperglycemic conditions has been associated with an increase in retinal vascular permeability, which is possibly due to the modification of tight junction proteins and the cytoskeleton, leading to the disruption of the BRB (575).

Various isoforms of PKC have been shown to be expressed in the retina, with an increase in their activity under hyperglycemic conditions. PKC isoforms are derived from the alternative splicing of mRNA transcripts and multiple genes, have a wide variety of functions in biological systems (263), and can be divided into groups depending on their structure and cofactors. Classical or conventional PKCs (cPKCs) are activated by DAG, calcium, and phosphatidylserine (PS) or phorbol esters such as phorbol 12-myristate 13-acetate (PMA). These cPKCs include cPKC-α, -βI, -βII, and -γ (107, 402). Novel PKCs (nPKC) lack the characteristic C2 domain that is present in the classical PKCS. nPKCs are calcium-independent, but are activated by PS, PMA, and DAG, and include nPKC-δ, -ε, -θ and -η (419). In contrast, atypical PKCs (aPKC: aPKC-ı, -λ, and -ξ) differ from the two other classes of PKCs in that their regulatory domain has one zinc finger instead of two; however, aPKCs lack the C2 domain, as do nPKCs (400).

In the diabetic retina, studies have shown an increase in the activity of PKC-βI and βII relative to other PKCs (274, 531), with a possible role in hyperglycemic complications. PKC-α, -γ, and −δ activity also have been shown to increase under hyperglycemic conditions in the retina; however, this increase was not as substantial as with PKC-βI and βII (207). Additionally, hyperglycemia-induced retinal vascular leakage has been found to be attenuated with the inhibition of PKC-ξ, as are tight junction modifications due to VEGF and AGEs (545). The link between VEGF and the effects of hyperglycemia-induced PKC activation is further supported by the rapid activation and membrane relocation of PKC-α, PKC-γ, and PKC-δ with the intravitreal injection of VEGF, and the accompanying increase in retinal vascular permeability (12). These results indicate a possible role for hyperglycemia-induced PKC activation in the development and progression of diabetic retinopathy.

Inflammation

Damage to tissue and cells during injury due to hyperglycemia leads to the initiation of an inflammatory response, which involves the upregulation of inflammatory mediators and cytokines, and the activation and recruitment of leukocytes. This increase in the induction of inflammatory proteins, such as iNOS, cyclooxygenase-2 (COX-2), ICAM-1, VEGF, interleukin 1β, and tumor necrosis factor-α, occurs at the transcriptional level due to the activation of proinflammatory transcription factors, such as NF-κB (151, 567).

The recognition of diabetic retinopathy as an inflammatory disease started with the observation that diabetic patients with rheumatoid arthritis who were treated with salicylates had a lower incidence of retinopathy (474). The relationship between iNOS and diabetic retinopathy has been demonstrated by the decrease in leukocyte adhesion in the diabetic retinas of iNOS knockout mice, and in diabetic mice treated with the NOS inhibitor N^G-nitro-L-arginine methyl ester (L-NAME) (340). Moreover, diabetic patients have increased levels of neutrophils in the retina and choroid (386), and the levels of TNF-α and interferon-γ are elevated in the vitreous (597) and retina of diabetic patients (135).

Accumulation of platelets and macrophages in the retina also have been reported in experimental models of diabetes and in humans. In rats, platelet adhesion was found early in hyperglycemia; however, anti-platelet serum increased blood retinal barrier breakdown, which suggests that the platelets also play a protective role (635). Platelet aggregation is maintained in rats for up to nine months of hyperglycemia (277), and in humans with diabetes, platelet-fibrin thrombi have been observed in retinal capillaries (57). In addition, retinal tissue macrophages were found in hyperglycemic rats after two and nine months of hyperglycemia (521), and in the retinas of diabetic individuals with proliferative diabetic retinopathy (166). However, to our knowledge, the cause and effect relationship of platelets and macrophages to the development of diabetic retinopathy has not been established.

AGE formation has been found to play a role in the initiation of the inflammatory response in diabetic retinopathy. This response is possibly mediated via activation of the mitogen-activated protein kinase (MAPK) pathway due to AGE/RAGE interactions, and the resulting increases in retinal n3 polyunsaturated fatty acids (such as docosahexaenoic acid) (573) and n6 essential fatty acids (such as linoleic acid) (75). Monocyte chemoattractant protein-1 (MCP-1), VEGF, and ICAM-1 can become upregulated due to AGE/RAGE interactions, leading to increased leukocytosis and increased T-cell adhesion (273, 396).

In addition to the direct effects of hyperglycemia and AGE/RAGE interactions, hypoxia due to capillary dropout or blockage can play a role in the initiation of an inflammatory response in the diabetic retina. Hypoxia-induced TNF-α release from activated microglia and macrophages in the retina induces the expression of VEGF, MCP-1, and interleukin-8 in endothelial cells, which further enhances the inflammatory response in the diabetic retina (641).

Kallikrein-kinin system (KKS)

Multiple studies have indicated a possible role for the KKS in the development and progression of diabetic retinopathy. The KKS has important functions in the vascular system, where it mediates blood pressure regulation, blood flow, inflammation, and coagulation (49, 362, 461). The KKS can be divided into two distinct proteolytic pathways, plasma kallikrein and tissue kallikrein. Multiple tissues express tissue kallikrein, including the eye, while plasma kallikrein originates in the liver, and forms a complex with high molecular weight kininogen in the plasma upon its release (401). In patients with proliferative diabetic retinopathy, the levels and activity of PK in the vitreous is significantly higher than normal; however, tissue kallikrein has a low to undetectable level and activity in the vitreous of patients with diabetes (462). The two major G-protein cell surface receptors for the kinin family are the B1 and B2 receptors, with B2 constitutively expressed in the retina. While B1 is not expressed under physiologic conditions, its levels become elevated in diabetic and inflammatory conditions in the retina (3, 4).

In diabetic retinopathy, the activation of the KKS leads to a disruption of the BRB, alterations in vascular diameter and blood flow, inflammatory responses, and angiogenesis. Retinal vascular leakage is induced with the administration of recombinant plasma kallikrein intravitreally, or with B1 and B2 receptor agonists (258, 477). Additionally, intravitreal administration of bradykinin dilates retinal vessels and increases retinal blood flow (317, 542). This vascular effect also has been demonstrated in ex vivo experiments, where bradykinin was shown to dilate retinal vessels isolated from rat and porcine retinas, either non-pressurized in the presence of a preconstrictor to generate tone (3, 234), or pressurized without a preconstrictor (244). Activity of the kinin receptor B1, which is elevated in the diabetic retina, has been implicated in mediating the proinflammatory effects of the KKS, being associated with the increase in neutrophil recruitment and microglial activation (158, 267).

Diabetes-induced changes in retinal & choroidal blood flow

Alterations of blood flow in the diabetic retina

Retinal blood flow in individuals with diabetes has been reported to decrease by ~30-35% in the early stages of retinopathy (95), followed by a reversal back towards and even exceeding controls in more severe stages of the disease. The mechanisms responsible for the early decrease in flow could be important to discern, inasmuch as prevention of reduced blood flow and ischemia may attenuate the subsequent harmful angiogenesis. Therefore, retinal blood flow has been measured in various experimental animal models of diabetes, using a number of different techniques. As discussed in the following sections, most of the results from these studies are consistent with the early decrease seen in the human diabetic retina. However, the studies that are in contrast to this finding are of interest as well, and as discussed below, may be informative as to potentially deleterious structural changes occurring in the diabetic microvasculature.

Microsphere Technique

One method to measure blood flow in the retina (and in other tissues of the body) is via cardiac infusion of appropriate-sized microspheres that lodge in small precapillary arterioles throughout the body in proportion to the volumetric rate of blood perfusing the tissue. Following the optimization of this technique for the rat retina established by Wang et al. (601, 602), our lab has measured retinal blood flow in control vs diabetic STZ rats, at 8 weeks of hyperglycemia for the latter, finding an approximate 33% diabetes-induced decrease (347). This decrease is in contrast to the increase found in another STZ rat study in which a non-optimal number and a non-optimal size of microspheres were used (574). The microsphere technique has been employed in the retina in other animal studies of diabetes, finding a 40-45% decrease in blood flow in dogs following five months of hyperglycemia (539), and a similar 40-45% decrease in retinal blood flow following three weeks of hyperglycemia in rats (the latter not statistically significant) (219).

Intravital Microscopy

Retinal blood flow can be measured directly in anesthetized animals with the use of microscopy by measuring arteriolar (or venular) diameters and flow velocities (239). Diameter measurements are accomplished via a high molecular weight fluorescent plasma tracer with calibrated off-line video analysis, with velocities similarly measured from the distance traveled by fluorescently labeled microspheres (or red blood cells) in single video frames. Using this method, we have found significant diabetes-induced decreases in flow in Ins2(Akita) mice following six months of hyperglycemia (627), in nonobese diabetic (NOD) mice following three weeks of hyperglycemia (343), in STZ-injected rats after three weeks (but not after 1 week) of hyperglycemia (344), and in STZ-injected mice after 4 weeks (605, 622, 626, 631).

Laser Doppler

Using laser Doppler velocimetry, retinal blood flow has been found to decrease in diabetic humans without retinopathy (170, 222), and with background diabetic retinopathy (170). In another study using this technique, retinal blood flow measured in individuals with a mean diabetes duration of 12.5 years was decreased, while retinal blood flow was increased in diabetic individuals with a mean diabetes duration of 20 years (321). Patel et al. (458) reported in diabetic individuals with background retinopathy that retinal blood flow was increased, and that flow continued to increase as retinopathy advanced. Using laser Doppler velocimetry in a rat STZ model, we have found 35-40% decreases in flow compared with controls (347), although these measurements could be affected by other vascularized ocular tissues, such as the choroid.

Mean circulation time

Another method of measuring retinal blood flow utilizes the bolus infusion of a tracer, with the analysis of its retinal transit providing an index of flow. The infusion of a fluorescent plasma marker can be recorded via intravital video microscopy to analyze the entry of the tracer into and out of the retinal arterioles and venules, with the difference in mean transit times between the arterioles and venules being the mean circulation time (MCT). A flow index is then calculated from the retinal blood volume divided by the MCT, where the blood volume is assumed to be proportional to the average arteriolar diameters squared plus venular diameters squared. However, it should be mentioned that a limitation of using this blood volume estimation when making comparisons between groups is the assumption of similar capillary bed structures.

The MCT method has been used in rat models of diabetes to estimate retinal perfusion in the STZ model, with reported increases in MCT (but no changes in diameters) (71, 95, 96, 253, 259, 564) in the first two weeks of diabetes consistent with slower flow taking a longer time to traverse the retinal capillaries. After four weeks of diabetes, however, arteriolar diameters decreased with no change in mean circulation time (253). Subsequently, at eight weeks of diabetes, our lab found that the mean circulation time was drastically shortened compared to non-diabetic controls by ~67% (with no significant change in diameters), which would be consistent with retinal flow rates unexpectedly increasing dramatically by a factor of three (347). This result from our lab is in stark contrast to our results mentioned earlier using other techniques (microsphere infusion, microscopic measurements of diameters and velocities, laser doppler velocimetry), each of which indicate a significant decrease in flow (of ~30-40%), rather than a 300% increase. With the MCT technique being the only of these techniques suggesting an increase in retinal blood flow in the early weeks of STZ-induced diabetes in rats, we have hypothesized that shunting of blood through the superficial retinal layer, rather than traversing the intermediate (inner) and deep capillary layers, could result in a quick transit of blood through the retina that could be misinterpreted as faster flow (347). This shunting could possibly be the result of precapillary constriction (Figure 7) that would force more flow through the upstream vascular thoroughfares.

Figure 7. Transit pathways through the retinal circulation.

Normally, most retinal blood flow travels through the deep capillary layer. In diabetic rats, it is possible that a decreased proportion of flow goes through the deep layer, with rapid transit times possibly explained by flow preferentially perfusing the superficial and/or intermediate capillary layers due to vasoconstriction of the vessels (indicated by the asterisks) leading away from the superficial layer.

With the MCT technique dependent on capillary bed structure, another possibility for the rapid transit of dye through the STZ retina is a diabetes-induced loss of capillary density. However, the trypsin digest results obtained by Su et al. (558) have shown that capillary density in the rat model of STZ does not change over an approximate 2-year duration of diabetes. The trypsin digest model does not quantify the number of vessels actually being perfused; therefore, we have subsequently performed a study in which we quantified perfused vessel density in the STZ rat model (at the same time point in which MCT was greatly reduced), and found no change in the density of either the intermediate or deep capillary beds (347). Therefore, we speculate that shunting, due to vasoconstriction of the arterioles leading to the deep capillary beds, is a likely explanation for the shorter transit times in the STZ rat retina.

Alterations of blood flow in the diabetic choroid

Reports of changes in choroidal blood flow in diabetic individuals and in experimental models of hyperglycemia are inconsistent, and the mechanisms involved are not completely known. Choroidal circulation measured using pulsatile ocular blood flow has shown decreased flow prior to proliferative diabetic retinopathy, but with increased flow in proliferative diabetic retinopathy (210). Others have reported increased (372), decreased (333) or no change (517) in choroidal flow as severity of retinopathy increases in diabetic individuals. Savage et al. (510) reported increased choroidal blood flow in diabetic individuals with proliferative diabetic retinopathy, which decreased below the control group following panretinal photocoagulation. Choroidal flow was decreased in diabetic individuals with proliferative diabetic retinopathy when measured with laser Doppler flowmetry (519).

Alterations of vasoactive factors in diabetic retinopathy

Nitric Oxide

Diabetic retinopathy is associated with both increases as well as decreases in the production of NO. As discussed earlier, NO is produced from several different synthases (eNOS, iNOS, nNOS), which are not all affected the same in the diabetic retina. As will be discussed below, despite an overall increase in retinal NO in diabetes, evidence exists for a decrease in vascular NO.

With respect to the human retinal vasculature, Schmetterer et al. (515) have reported that intravenous L-NMMA administration in healthy individuals decreases ocular blood flow velocity, but to a lesser extent in type I diabetic patients, suggesting a possible pathophysiological role of NO signaling in diabetic retinopathy. In investigations using a porcine model of type I diabetes (243, 248), reductions were found in bradykinin-induced retinal arteriolar dilation, which also is attenuated by NOS inhibition (246, 248, 410). Even brief intraluminal exposure to high glucose in otherwise healthy porcine retinal arterioles suppresses bradykinin-induced dilations (248). Additionally, specific inhibitors of c-Jun N-terminal kinase (JNK) and JNK-interacting protein-1 (JIP1) fully restore the porcine retinal vascular response to bradykinin with both acute and chronic hyperglycemia; however, the protective effect of JNK and JIP1 inhibitors in the retinal vascular response is suppressed by NOS inhibition, indicating that JNK and JIP1 signaling pathways are involved in a glucose-dependent impairment of NO-mediated vascular function (248).

In in vitro cell cultures, retinal microvascular endothelial cells (RMECs) in high glucose conditions produce less NO (80, 103), with or without stimulation by acetylcholine (103). In hyperglycemia, the changes of endothelial NO levels both in vivo and in vitro have been shown to be controlled by expression and activation of eNOS. In vivo, mRNA expression of eNOS in the retina is reduced in the diabetic rat (652), and in vitro, incubation of RMECs in high glucose reduces eNOS protein expression (80, 103, 600). Several genomic data analyses have shown that polymorphisms of the eNOS gene are associated with diabetic retinopathy (37, 85, 93, 250, 371, 377, 394, 413, 508, 653). A deficiency of eNOS does not prevent the formation of the retinal vasculature during development; however, eNOS knockout mice develop more severe diabetic retinopathy (353).

The endothelium and NO-dependent functional hyperemia assay of retinal vascular dilation is attenuated in diabetes, in patients without (341) or with retinopathy (172, 201, 336, 356, 376, 420, 421, 460), with the vascular response continuing to decrease with increasing stages of diabetic retinopathy (172, 356, 376). This attenuated response in diabetes also occurs with acute hyperglycemia in healthy controls, with glucose levels temporarily increased to 300 mg/dL (148). The hyperglycemia-induced attenuation of flicker-stimulated functional hyperemia is also seen in a rat STZ model of type 1 diabetes (391, 392).

Despite the decreases in retinal vascular NO and eNOS in diabetes, the overall retinal and plasma levels vary. Some studies have reported decreased plasma levels of NO in diabetes (36, 113, 328, 652), whereas other studies have suggested the opposite (8, 35, 256, 279, 407, 447, 527). A study in diabetic STZ rats has shown retinal NO levels to be decreased with very high plasma glucose levels of 500-600 mg/dL, but increased at both mild (250-400 mg/dL) and moderate (400-500 mg/dL) levels (224). Increases in NO in the diabetic retina may be derived from either iNOS or nNOS as explained below.

Nitric oxide (152, 300, 324, 655) and retinal iNOS (5, 6, 151, 152, 340, 512, 577) have been reported to increase in human diabetes and in animal models of the disease. The increases in NO are inhibited by an iNOS inhibitor (152) and in iNOS-deficient mice (655). NO can bind to elevated levels of superoxide in the diabetic retina (300), with a resulting formation of peroxynitrite that can lead to PKC activation (475). In vitro, hyperglycemia increases NO production and iNOS expression in retinal Müller cells and in RMECs, with these increases (and cell death) suppressed by iNOS inhibition (151). The iNOS inhibitor aminoguanidine improves flicker-induced arteriolar dilation in diabetic rats (391), and as explained by these investigators, high levels of tissue NO might inhibit the glial release of dilatory agents, disrupting the connection between neuronal activity and vasodilation. It is possible that iNOS inhibition decreases retinal NO levels, allowing normal neurovascular coupling.

Increases in iNOS-dervied NO in the ganglion cell and inner plexiform layers of the retina (in STZ diabetic rats) do not activate soluble guanylate cyclase (512), which is able to dilate vascular smooth muscle. Instead, the elevated levels of NO produced by iNOS in the diabetic retina may induce the deleterious production of peroxynitrite, with this production attenuated by superoxide dismutase (512). Increased NO and overexpression of iNOS in insulin-dependent (76, 340) or -independent (77) diabetic rodents have been implicated in the breakdown of the blood-retinal barrier and development of diabetic retinopathy. Junctional barrier proteins in the diabetic mouse retina are protected by non-selective NOS inhibition and iNOS knockout (340).

In addition to increases in iNOS in diabetic retinopathy, nNOS expression and activity in the retina have been found to be increased in the inner nuclear layer and inner plexiform layer (215, 451). It is possible that increases in vascular-associated nNOS activity compensate for the eNOS deficiency in the retina and might be partially responsible for the increased NO levels in diabetic retinopathy.

Prostaglandins

Prostaglandin E (PGE₂) and I₂ (PGI₂) have been found to be increased in STZ-injected rats following four weeks of hyperglycemia, with these increases attenuated by treatment with insulin (284). However, PGI₂ levels have not been found to be elevated in the vitreous of diabetic individuals (456). It is unclear how the changes in PGI₂ affect the retinal circulation during diabetes.

Thromboxane

The vasoconstrictor thromboxane A₂ (measured as the stable metabolite thromboxane B₂) has been reported to increase in platelets from diabetic rats (125-127). Inhibition of its synthesis by either a thromboxane synthase inhibitor (127), a combined thromboxane synthase inhibitor/receptor antagonist (126), or by aspirin (125), results in decreased TxA₂ levels, and improved measures of retinal red blood cell perfusion through the capillary beds of diabetic STZ rats (126, 127). This improvement could be consistent with a role for the vasoconstrictor in the decreased flow seen in diabetic rodents. Indeed, retinal arterioles that are constricted in non-obese diabetic mice, and in STZ-injected mice and rats, are more dilated following administration of a thromboxane synthase inhibitor or a thromboxane receptor antagonist (343, 344, 622, 626). Additionally, expression of the thromboxane receptor and thromboxane synthase have been reported in mouse and rat retinas; however, there appear to be no differences in these expressions between control vs diabetic rodents following four weeks of hyperglycemia (623).

Adenosine

Adenosine infusion leads to retinal arteriolar vasodilation in euglycemic rats and in STZ-injected rats. This increase in arteriolar diameters is, in part, due to activation of NO synthase and to K_ATP channel opening (412). While the implications of the adenosine response are not fully understood in relation to physiological retinal circulation, increased activity of adenosine deaminase-2, an enzyme responsible for the breakdown of adenosine, has been reported in the vitreous of individuals with diabetes. Inhibition of the increased adenosine deaminase-2 activity, via miR-146b-3p, reversed the increased permeability and leukocyte adhesion in cultured human retinal endothelial cells (506).

Endothelin

Components of the endothelin system (i.e. endothelin and it receptors) are altered in ocular tissue of hyperglycemic rats (78, 79, 137) and in humans with proliferative retinopathy (439, 456); however, ET-1 levels are decreased in nonproliferative diabetic retinopathy (456). ET_A receptor antagonism and inhibition of endothelin converting enzyme have been shown to attenuate decreases in retinal blood that occur in diabetic rats (563). Infusion of ET-3 initially increases retinal blood flow in control and diabetic rats through ET3/ET_B binding; however, at later time points, ET-3 decreases retinal blood flow in control rats while flow remains elevated in diabetic rats, with flow-mediated changes found to occur via the ET-3/ET_A receptor interaction (399). Endothelin is discussed more in depth elsewhere in a review of its role in diabetic retinopathy (547).

Angiotensin II

Although limited, there is evidence that angiotensin II plays a role in vascular abnormalities in the diabetic retina. It has been reported that angiotensin II receptor antagonism enhances flow and arteriolar diameters in NOD mice, with the latter changes specifically seen for arterioles in closer proximity to venules draining the retina (343). Administration of an angiotensin converting enzyme inhibitor improved diabetic retinal changes to tight junction loss, VEGF upregulation, and alterations in vascular permeability (307). Although encouraging, the clinical implications have not been confirmed (538).

Regulatory mechanisms of blood flow

Retinal blood flow regulation may be altered in diabetes, including via autoregulation and venular control of arteriolar diameters. Concerning autoregulation, this mechanism has been found to be impaired in the diabetic retinal circulation (154, 190, 535), with one study indicating that this deficiency is more prevalent in proliferative retinopathy (535). Additionally, diameters appear to be constricted in arterioles that are closely paired to venules in the retinas of diabetic rats and mice. The decrease in closely-paired arterioles is attenuated with inhibition of an angiotensin II receptor antagonist or a thromboxane synthase inhibitor (343, 344, 622).

Diabetes-induced changes in the retinal endothelial surface layer

Platelet endothelial cell adhesion molecule-1 (PECAM-1)

PECAM-1 is a 130 kDa transmembrane glycoprotein that belongs to the immunoglobulin G (IgG) superfamily of receptors with an immunoreceptor tyrosine-like inhibitory motif (ITIM) domain. It is highly expressed at the intercellular junction of endothelial cells (417), as well as on platelets, lymphocytes, and leukocytes. PECAM-1 has an integral role in cellular adhesion and transendothelial migration of leukocytes (417). In addition, PECAM-1 has been shown to have a role in apoptosis and cell signaling. For example of the latter, PECAM-1 modulates the functions of catenins, such as β and γ-catenin, by sequestering them to the plasma membrane and limiting their transcriptional activities (268, 269). Pathophysiological alterations to the retinal microvasculature in diabetic retinopathy involves increases in vascular permeability, leukocyte plugging of capillaries, altered cell signaling, and a decrease in cell survival (Figure 8), all of which could be related to PECAM-1 loss due to its important function in the endothelium. Thus, this section will cover in more detail the function of PECAM-1, and how its loss may contribute to the pathogenesis and progression of diabetic retinopathy.

Figure 8. PECAM-1 loss may contribute to the development of diabetic retinopathy.

Diabetic retinopathy is characterized by leukocyte plugging of capillaries, a decrease in endothelial cell survival, and an increase in vascular leakage. These phenomena also could occur with PECAM-1 loss, due to its important role in vascular endothelial cell functions.

Functions of PECAM-1

Since the discovery of PECAM-1, extensive studies have investigated the role of PECAM-1 in endothelial cells, and have led to the appreciation of the essential functions PECAM-1 has in the maintenance, survival, and integrity of the endothelium, in addition to its role in mediating the inflammatory response.

Role of PECAM-1 in inflammation

PECAM-1 can mediate inflammation by facilitating leukocyte transmigration and acting as a mechanosensor that can promote the production of NF-κB. On the other hand, PECAM-1 can prevent inflammation as a result of its role in cell survival and prevention of apoptosis. These conflicting roles of PECAM-1 are due to its physical and signaling properties that are mediated through its extracellular and cytoplasmic structures.

Role of PECAM-1 in leukocyte transmigration

The classic way by which PECAM-1 facilitates inflammatory responses is by its essential role in leukocyte transendothelial migration. PECAM-1 is the main component of the lateral border recycling compartment (616), which also contains CD99, JAM-A, CD155, IQ motif containing GTPase-activating protein-1, and vimentin (404, 561, 620). Leukocyte transmigration, by paracellular and intracellular pathways, involves the homophilic binding of endothelial PECAM-1 and leukocyte PECAM-1, which was demonstrated with the inhibition of leukocyte transmigration when the IgD1 homophilic binding site on the PECAM-1 extracellular domain was blocked (405). Additionally, when CD177 on neutrophils and/or IgD6 of PECAM-1 were blocked, neutrophil transmigration was inhibited, indicating an important role of PECAM-1 heterophilic binding (505). Moreover, not only is PECAM-1 involved in leukocyte transmigration via paracellular and intracellular pathways, but also in migration across the endothelial basement membrane (355).

Interestingly, PECAM-1 plays a more significant role in regulating the transmigration of neutrophils and monocytes than lymphocytes and eosinophils, due to a preferential selectivity to leukocyte sub-types (403). PECAM-1 also can regulate the rate and direction to leukocyte migration, which further demonstrates the complexity of the PECAM-1-mediated mechanism (621).

PECAM-1 involvement in leukocyte transmigration is dependent on the stimulus and the genetic background of the disease model. In studies using mice on a C57BL/6 genetic background, PECAM-1 was not required for leukocyte transmigration as it was not affected by PECAM-1 blockade (514). Additionally, PECAM-1 recruitment in leukocyte transmigration was initiated by specific stimuli, where IL-1β, L-NAME, and H₂O_2, but not TNF-α, led to leukocyte dependence on PECAM-1 for transmigration (621). Moreover, SHP-2 binding to the ITIM domain of the PECAM-1 cytoplasmic tail is not necessary for leukocyte transmigration. However, tyrosine 663, through a yet unknown mechanism, is required for this process (118).

Role of PECAM-1 as a mechanotransducer

Endothelial cells play an important role in sensing changes in blood shear stress, cyclic stretch, and osmolarity. Rearrangement of the endothelial cell cytoskeletal and modulation of vascular tone occurs with direct force applied to PECAM-1 using magnetic beads coated with a PECAM-1 antibody (442). These changes are sensed at the apical side of endothelial cells, and translated through signaling cascades to affect the endothelial cell cytoskeleton network and junctions, and also smooth muscle cells.

Tzima et al. observed a sensory complex that includes PECAM-1, vascular endothelial cadherin (VE-cadherin), and vascular endothelial growth factor receptor 2 (VEGFR2) (583). Tensile forces exert a mechanical pull on PECAM-1, leading to tyrosine phosphorylation on its cytoplasmic tail. This initiates a global endothelial cell response, including the activation of phosphatidylinositol 3-kinase and downstream effectors, and endothelial cell cytoskeleton rearrangement (102). Upon activation of PECAM-1, VE-cadherin acts as an adaptor to relay the change to VEGFR2, although some studies implicate a more significant role for VE-cadherin as a mechanosensor and not just as an adaptor protein (105). Upon VEGFR2 activation, a signaling cascade leads to the activation of phosphatidylinositol 3-kinase, and the subsequent activation of Akt and integrin, the upregulation of NF-κB, and the initiation of a proinflammatory response (104). In addition to its association with the mechanosensory complex, PECAM-1 is also in a complex with eNOS, and upon changes in shear stress, PECAM-1 rapidly dissociates from this complex prompting eNOS activation and NO production (182).

The importance of PECAM-1 as a mechanosensor is evident from studies conducted to understand the molecular mechanisms leading to atherosclerosis. Atherosclerotic lesions occur at areas of disturbed blood flow such as branch points of aortic arches. PECAM-1 senses these changes, and initiates the signaling cascade leading to lesion development (90). Moreover, PECAM-1 null mice have a decrease in lesion areas, further confirming the role of PECAM-1 as an atherogenic molecule (90).

PECAM-1 role in cell survival

Apoptosis, or programmed cell death, is an important aspect of the endothelial cell life cycle, which is marked by an intact plasma membrane, and the expression of apoptotic cell markers on the endothelial cell surface to target it for phagocytosis. However, if left unchecked, apoptosis can contribute to a multitude of pathophysiological complications including atherosclerosis, cancer, neurodegenerative disease, and diabetes-induced vascular damage. Additionally, endothelial cell apoptosis can lead to increased vascular permeability, exposure of thrombogenic basement membrane matrix, and increased platelet and leukocyte adhesion (617).

Noble et al. provided the first evidence in supporting a role for PECAM-1 in cell survival, where a decrease in endothelial cell apoptosis was observed with the homophilic binding of PECAM-1 on endothelial cells and monocytes following serum deprivation (430). PECAM-1 can mediate its anti-apoptotic properties through its ability to transduce signals that suppress apoptosis. Additionally, PECAM-1 was able to inhibit the release of cytochrome-c upon mitochondrial damage that was induced by cytotoxic factors activating the pro-apoptotic mitochondrial pore-forming Bcl-2-associated X protein (197). Moreover, sphinogosine-1-phosphate enhanced endothelial cell survival in a PECAM-1-dependent manner (357). Interestingly, blocking PECAM-1 homophilic binding or mutations in PECAM-1 immunoreceptor tyrosine-based inhibition motif domains blunted its ability to promote cell survival, indicating the importance of both the extracellular and cytoplasmic domains in mediating PECAM-1 cell survival functions (197).

The role of PECAM-1 in endothelial cell barrier function

One of the most critical functions of the vascular endothelium is to limit the access of blood and immune system components to the underlying tissue, which is even more critical in compromised tissue such as the brain and the retina. BRB breakdown directly contributes to the development of diabetic retinopathy, causing retinal tissue edema and promoting an inflammatory response. Therefore, the understanding of the mechanisms leading to BRB breakdown, and the molecular players contributing to it, may provide therapeutic targets for preventing and limiting development and progression of retinopathy in diabetes. Retinal endothelial cells form a tight layer that controls the passage of ions, proteins, and cells to the retinal tissue. This endothelial function is achieved by specialized cell adhesion molecules that maintain and preserve the endothelium barrier property. PECAM-1 is recognized as a cell adhesion molecule (20), that along with adherens junction proteins, tight junction proteins, and nectin constitute endothelial cell adhesion junction proteins (131).

Many studies have demonstrated the contribution of PECAM-1 to the maintenance of endothelial barrier function. In an experimental animal model of multiple sclerosis, PECAM-1 knockout mice exhibited an increased permeability of the blood-brain barrier (218). Additionally, CD44 knockout mice had a decreased level of PECAM-1 accompanied by an increase in vascular permeability that was returned to normal levels with increasing PECAM-1 levels (184). Moreover, PECAM-1-PECAM-1 homophilic interactions have been found to be required for the maintenance of vascular endothelial cell barrier function (476, 610). When mice are injected with antibodies against PECAM-1, vascular leakage from hepatic and renal blood vessels increases (177). Moreover, cultured PECAM-1-deficient endothelial cells have an increased permeability when exposed to histamine (218).

Role of PECAM-1 in facilitating functions of β-catenin

One of the possible mechanisms whereby PECAM-1 can contribute to barrier function is through its association with β-catenin, with PECAM-1 acting as a reservoir for phosphorylated β-catenin. Additionally, PECAM-1 facilitates dephosphorylation of β-catenin through the activation of Src homology region 2 domain-containing phosphatase-2 (SHP-2), which renders it active and able to associate with VE-cadherin to restore the endothelial barrier.

The first observation of the involvement of β-catenin in endothelial cell barrier function was made with the isolation of proteins associated with VE-cadherin, which found β-catenin linked with α-catenin and γ-catenin (446). VE-cadherin is a critical molecule in the maintenance and restoration of cell adhesion, and the association of β-catenin with VE-cadherin suggests a role in cell adhesion as well. VE-cadherin, which is a single pass transmembrane glycoprotein, associates with β-catenin at its cytoplasmic tail. Interestingly, the association of VE-cadherin with β-catenin is constitutive, and occurs early in the endoplasmic reticulum with the newly synthesized VE-cadherin (553). The β-catenin/VE-cadherin interaction contributes to the maintenance of the endothelial barrier, and limits its cytoplasmic pool and nuclear translocation, which is the site of its signaling functions.

The sequestration of phosphorylated and unphosphorylated β-catenin by PECAM-1 at the cell membrane protects β-catenin from being recognized at the cytoplasm by the destruction complex that is composed of Axin, anaphase-promoting complex, glycogen synthase kinase 3β, casein kinase 1, and protein phosphatase 2A (266). The latter three phosphorylate β-catenin, leading to its targeting for ubiquitination and subsequent degradation. This process is inhibited by the activation of the canonical Wnt signaling pathway, which after a series of molecular events, leads to the downstream cytoplasmic protein ‘Dishevelled’ forming a complex with the rate-limiting component of the destruction complex Axin, leading to its destabilization, and to the stabilization of β-catenin due to its decreased phosphorylation (51). The increase in cytosolic β-catenin can lead to translocation to the nucleus and the initiation of gene transcription. The mechanism by which β-catenin shuttles back and forth to the nucleus, in spite of its lacking a nuclear localizing signal or nuclear exporting signal, is not fully understood. However, there is evidence of a direct interaction between β-catenin and nuclear pore complex components (526, 532).

Possible mechanisms of PECAM-1 loss in the diabetic retina

Since its discovery, PECAM-1 has been shown to have many vital functions in the endothelium that surpass its initially recognized utility of being an endothelial cell marker. These functions assure the proper functioning and survival of endothelial cells, and preserves the vascular integrity through its role in endothelial barrier function. Thus, its loss can result in severe consequences, all of which can contribute to the development and progression of diabetic retinopathy.

Loss of PECAM-1 has been reported in the diabetic retina (74, 89), however, without an elaboration on the mechanism of this loss or its consequences. PECAM-1 can be cleaved by MMPs (409) such as MMP-2, MMP-9 (294), and MMP-7 (415), and its expression can be down-regulated by the cytokines TNF-α and IFN-γ (498) that play a significant role in mediating the inflammation associated with diabetes. MMPs and proinflammatory cytokines have been shown to increase in diabetes, but to our knowledge, no studies have been performed to link their activity with the diabetes-induced loss of PECAM-1 and the effect on retinal microvascular integrity.

Matrix metalloproteinases (MMPs)

MMPs play an important role in various physiological processes including extracellular matrix remodeling, wound healing, angiogenesis, and cell signaling (448). Additionally, MMPs can participate in pathophysiological processes, such as inflammation, cancer, arthritis, and cardiovascular disease (87, 204, 604). Under hyperglycemic conditions, MMP-2 and MMP-9 levels and activity are significantly upregulated. Several studies have indicated the role of MMP activation in hyperglycemia-induced ROS formation and oxidative stress leading to deleterious consequences in diabetes. In the diabetic retina, increased activation of MMP-2 and MMP-9 leads to mitochondrial dysfunction and the induction of apoptosis, which is possibly due to the translocation of MMP-9 to the mitochondria leading to a change in its redox state, since ROS can activate MMPs. MMPs residing in the mitochondria can become activated, leading to the degeneration of the mitochondrial membrane. Our lab has recently found evidence from both in vitro as well as in vivo studies implicating MMPs (especially MMP-2) in the hyperglycemia-induced loss of PECAM-1 from rat retinal microvascular endothelial cells (163).

One of the hallmarks of diabetic retinopathy is increased vascular leakage due to BRB breakdown, which can be facilitated through the degradation of cell adhesion molecules such as PECAM-1, which has been shown to be cleaved by MMPs. Various cleavage sites by MMPs have been found on PECAM-1 using in silico analysis, indicating the possible role MMPs can play in decreasing PECAM-1 levels under hyperglycemic conditions. These results indicate the multifaceted role by which MMPs can contribute to the progression of diabetic retinopathy, where they can function as proinflammatory and proapoptotic factors, and by cleaving and remodeling the endothelial cell surface and cell adhesion proteins.

The proinflammatory cytokine tumor necrosis factor-α (TNF-α)

Inflammation and proinflammatory factors play a key role in the progression of diabetic retinopathy. One of the most potent proinflammatory mediators is TNF-α, which is a key inflammatory cytokine that is produced by many cells, including endothelial cells, macrophages, lymphocytes, and mast cells (373). TNF-α is initially synthesized and produced as a type II transmembrane protein and cleaved by TNF-α converting enzyme producing soluble TNF-α (260). Both transmembrane and soluble TNF-α exert their effects on endothelial cells by interacting with their receptors, TNFR1 and TNFR2, leading to increased vascular leakage, and the expression of proinflammatory molecules that further worsens the inflammatory response (598).

Serum TNF-α levels are significantly increased in patients with type I diabetes (478), which is indicative of it role in promoting the inflammatory state in diabetes. Several studies have investigated the effect of TNF-α on cell adhesion molecules such as PECAM-1 due to its important role in endothelial barrier function. TNF-α decreases PECAM-1 levels (290, 498, 501, 511, 529, 554), which can influence leukocyte transmigration and vascular leakage. These results indicate a possible role for TNF-α in the hyperglycemia-induced PECAM-1 loss in the diabetic retina, possibly leading to development and progression of diabetic retinopathy using a yet not fully understood mechanism.

Diabetes-induced changes in the retinal endothelial glycocalyx

Structure and functions of the endothelial glycocalyx

The endothelial glycocalyx is a dynamic carbohydrate-rich layer on the cell surface, with the general structure depicted in Figure 9. The glycocalyx is attached to the endothelium via its backbone molecules, mainly proteoglycans, which consist of a core protein and unbranched glycosaminoglycan chains (GAGs) covalently attached to the core protein. The membrane-spanning syndecans and glycosyl-phosphatidylinositol anchored glypicans are the primary core proteins. There are five main types of GAGs: heparan sulfate, hyaluronic acid, chondroitin sulfate, dermatan sulfate, and keratan sulfate (27, 319, 486). However, hyaluronic acid, instead of binding to core proteoglycans, binds to its receptors CD44 and the receptor for hyaluronan mediated motility (319). Additional points of attachments to the endothelium are provided by other backbone glycoproteins, such as selectins, integrins and immunoglobulins (27, 319, 486). Proteoglycans and glycoproteins form a network with both plasma and endothelium-derived soluble molecules such as albumin and orosomucoid, forming a mesh-like structure on the endothelium (27, 319, 486). The glycocalyx inhibits leukocyte and platelet adhesion to the endothelium (27, 319, 486), and also regulates the distribution of red blood cells in microvessels (72). In addition, the glycocalyx acts as a mechanosensor, regulates vascular permeability, and is a reservoir of antithrombotic factors including anti-thrombin III (27, 319, 486). As described in the following sections, hyperglycemia and diabetes can result in a loss of the glycocalyx, which could have an effect on these functions. In many ways, this mesh-like structure helps maintain the physiological and quiescent environment of endothelial cells.

Figure 9. Glycocalyx components.

The structure of the endothelial glycocalyx on the plasma side of the cell includes proteoglycans (such as syndecans and glypican-1) and glycosaminoglycans (such as heparan sulfate, chondroitin sulfate, and hyaluronic acid).

In vitro loss of glycocalyx in hyperglycemia

In an in vitro glomerular endothelial cell (GEnC) model of hyperglycemia, Singh et al. (536) observed a reduced expression of heparan sulfate in the GEnCs when treated with high glucose (25.5 mM) compared to normal glucose (5.5 mM). However, there was no change in the expression of proteoglycans (syndecan-1, syndecan-4, glypican-1, versican-1, or perlecan-1). In the same study, high glucose treatment also led to a significant increase in the passage of albumin across GEnC monolayers despite no difference in the expression of VE-cadherin, a major component of the interendothelial junction (536), with increased endothelial permeability one potential consequence of a diabetes-induced loss of the vascular glycocalyx.

Heparan sulfate helps comprise flow-sensing mechanisms on the surface of endothelial cells, with this mechanism impaired by both heparinase treatment as well as hyperglycemia in human aortic endothelial cells (64). A similar hyperglycemia-induced impairment of glycocalyx-mediated mechanotransduction has been reported for bovine aortic endothelial cells, with a loss of heparan sulfate coincident with shear-dependent eNOS activation and control of hydraulic conductivity (365). Hyperglycemia and TNF-α have been shown to have an additive effect on the shedding of the endothelial glycocalyx of human umbilical vein endothelial cells (HUVECs). TNF-α (50 ng/ml) and glucose (200 mg/dL), when applied individually to HUVECs, led to similar (22-27%) reductions in the glycocalyx thickness and shedding of syndecan-1 and hyaluronic acid (144). However, when administered together, there was 42% decrease in the glycocalyx thickness, a 4-fold increase in shedding of syndecan-1, and a 10-fold increase in shedding of hyaluronic acid.

Loss of glycocalyx in diabetes

Degradation of the systemic glycocalyx has been noted in subjects with diabetes as well as in animal models of the disease. In human studies, patients with type 1 diabetes were found to have a smaller microvascular glycocalyx in sublingual capillaries compared to controls (424, 434). This decrease was accompanied by increases in plasma hyaluronic acid and hyaluronidase under hyperglycemia (423, 424) that were even greater in patients with microalbuminuria. Another clinical study showed a reduction in systemic glycocalyx volume (and increased plasma level of hyaluronic acid) upon application of a hyperglycemic clamp in healthy males, which was reversed with N-acetylcysteine treatment (425). The glycocalyx excludes plasma molecules, in part based on size, with 70-kD dextrans excluded from the glycocalyx substantially more than 40-kD dextrans. In a db/db type 2 diabetes mouse model, investigators determined that hyperglycemia diminishes the systemic glycocalyx exclusion of 70-kD dextran, compared to normoglycemic controls (164), with the glycocalyx barrier improved with metformin treatment. However, it should be mentioned that a limitation of the measurement of systemic glycocalyx volume is that the lower molecular weight 40-kD dextran can more easily pass through the endothelium into the tissue, introducing an error in the calculation (389).

The glycocalyx is likely to be affected in many tissues throughout the body in diabetes, and has been studied not only at a systemic level, but also in specific tissues, for example, the kidneys and retina that are sites of microvascular injury in diabetes. In a db/db mouse model of type-2 diabetes, a study of kidney glomeruli found a reduction of sulfation of the glycocalyx GAG chondroitin sulfate (484). In the same study, disaccharide analysis of chondroitin sulfate showed reductions in 4-O and 6-O sulfated disaccharides, with an increase in non-sulfated disaccharide residues. The study also found a significant decrease in the mRNA expression of chondroitin 4-O sulfotransferase in the diabetic kidney.

In diabetic nephropathy, a decrease in heparan sulfate (which is found on the endothelial surface and basement membrane of the glomerular capillaries) has been observed, which has shown to correlate inversely with the level of urinary protein excretion in diabetic patients (203). The same report demonstrated decreased albuminuria and improved renal function when diabetic mice were treated with a heparanase inhibitor (203). Additionally, the loss of heparan sulfate from the glomeruli of type 2 diabetes and overt diabetic nephropathy patients is correlated with an increase in heparanase (585). These studies suggest that heparan sulfate plays an important role in glomerular protein passage, and its loss is associated with the development of proteinuria and diabetic nephropathy.

Loss of glycocalyx in the diabetic retina

The structure of the glycocalyx may differ from tissue to tissue, and as of yet, the components of the retinal glycocalyx are yet to be established. However, a few initial studies have indicated that the retinal glycocalyx is lost (by ~35-50%) in diabetic humans, rats, and mice. In humans, an in vivo clinical study observed a 50% loss of the retinal glycocalyx in type 2 diabetes patients, which was partially restored with sulodexide, a highly purified combination of heparan sulfate and dermatan sulfate (62). In STZ-induced diabetes in rats, a reduction in the retinal glycocalyx thickness was observed ex vivo in tissue sections using transmission electron microscopy (330). In a another study, which was from our lab using a genetic Ins2(Akita) model of type 1 diabetes in mice, we observed an approximate 35% reduction in the glycocalyx thickness of retinal arterioles in vivo, but not in venules (346).

Mechanisms of glycocalyx degradation

The mechanisms underlying the diabetes-induced degradation of the endothelial glycocalyx are poorly understood. Hyperglycemia is thought to not only decrease the synthesis of glycocalyx components (mainly via production and/or sulfation of GAG chains) (100, 484, 595), but also increase the rate of glycocalyx shedding. Glycosaminoglycans and proteoglycans in the glycocalyx can be cleaved by matrix metalloproteinases (MMPs) (160, 361, 485), heparan sulfate can be cleaved by heparinase (533), and hyaluronic acid can be cleaved by hyaluronidase (548). Increased concentrations of MMPs have been found in retinal tissue of diabetic rats, and also in rat retinal microvascular endothelial cells when treated with high glucose (25mM) (163). Additionally, plasma and vitreous samples from diabetes patients have shown an elevation of MMPs (42, 289, 431). Heparanase levels are elevated in plasma and urine samples of type 2 diabetes patients (525), in whom plasma heparanase is positively correlated with blood glucose levels in the early stages of diabetic nephropathy (654). Diabetes patients also exhibit higher plasma hyaluronidase levels (423, 424), although our lab did not find an increase in plasma hyaluronidase activity in type 1 diabetic Akita mice compared to controls (346). However, hyaluronidase deficiency in STZ-diabetic mice has been shown to preserve the glycocalyx structure in diabetes (146).

Besides MMPs and GAG-specific enzymes, other mediators have been implicated as playing a role in the degradation of the glycocalyx in diabetes. For example, reactive oxygen species have been shown to mediate the mechanism by which hyperglycemia enhances the shedding of the glycocalyx following trauma/shock (143). Toll-like receptor 2 and 4 (TLR2 and TLR4) also have been speculated to mediate hyperglycemia-induced changes in the glycocalyx. Hyperglycemia upregulates mRNA and protein expression of TLR2 and TLR4, reduces heparan sulfate expression on human macrovascular aortic endothelial cells, and increases endothelial release of hyaluronic acid (449). These effects were reversed when TLR2 and TLR4 signaling was blocked using inhibitory peptides and siRNA knockdown (449).

Role of the glycocalyx in RBC distribution

Several studies indicate that one of the many functions of the glycocalyx is the regulation of RBC distribution through a microvascular bed, with consequences for local hematocrit (238). RBCs are excluded from the endothelial glycocalyx in microvessels (593), which reduces the microhematocrit, and correspondingly, when the endothelial glycocalyx is degraded with either hyaluronidase (72), heparinase (139), or light-dye treatment (593), capillary hematocrit increases by 60-100%.

Not only does the glycocalyx help control microhematocrit, but it also appears to promote the homogeneous distribution of those RBCs in the various capillaries of a bed. This finding has been predicted via mathematical models in which apparent viscosity plays a role in this mechanism (383). Results from in vivo experiments agree with the mathematical models, inasmuch as the heterogeneity of capillary perfusion substantially increases following degradation of the glycocalyx with hyaluronidase (72), causing some capillaries to stop flow, and other capillaries to be perfused solely by plasma (no RBCs). A consequence of increased heterogeneity of RBC distribution through a capillary bed is a decrease in the efficiency of oxygen transport to the surrounding tissue (282, 292, 444). Given the loss of glycocalyx in diabetes, it can be speculated that RBC distribution will be affected, which may contribute to the zones of ischemia that are observed in the human retina.

Conclusion

The retina is particularly vulnerable to the deleterious effects of continued hyperglycemia in the diabetic state. Diabetes-induced changes to the retinal circulation warrant the classification of the ensuing retinopathy as a microvascular complication of the disease. It is well-known that the proliferative phase of diabetic retinopathy is associated with uncontrolled angiogenesis that can interfere with vision, but considerable evidence has been collected showing significant vascular changes occurring earlier in the disease prior to the vascular proliferation.

Among the early changes in the diabetic retina are decreases in perfusion and a possible diversion of flow away from the deep capillary bed that feeds important retinal neurons. Evidence in the literature implicates several vasoactive pathways, such as thromboxane A₂, endothelin-1, angiotensin II, and decreases in endothelial nitric oxide, in the altered perfusion. Additionally, recent studies have demonstrated significant changes in the endothelial surface layer (including a loss of platelet endothelial cell adhesion molecule-1) and the glycocalyx, with these molecules and structures serving vital functions in a healthy vasculature. Future studies on the prevention of the later stages of diabetic retinopathy should focus in part on preventing these early changes in the retinal microcirculation.

DIDACTIC SYNOPSIS.

Major teaching points:

1. The retina includes layers of several types of neurons that receive oxygen and nutrients from two different vascular beds, i.e., the retinal circulation and the choroidal circulation.

2. The retina and its circulation are particularly vulnerable to the deleterious effects of continued hyperglycemia in the diabetic state.

3. Diabetes-induced changes to the retinal circulation warrant the classification of the ensuing retinopathy as a “microvascular complication” of the disease.

4. Among the early changes in the diabetic retina are decreases in perfusion and a possible diversion of flow away from the deep capillary bed that feeds important retinal neurons.

5. Recent studies have demonstrated significant changes in the endothelial surface layer (including a loss of platelet endothelial cell adhesion molecule-1) and the glycocalyx, with these molecules and structures serving vital functions in a healthy vasculature.

DIDACTIC LEGENDS

Figure 1. Teaching points: Diabetes is associated with complications of larger vessels in the body. These macrovascular complications include peripheral artery disease, cardiovascular disease, and stroke. Diabetes also is associated with microvascular complications that especially affect the retina (diabetic retinopathy), nerves (diabetic neuropathy), and the kidneys (diabetic nephropathy).

Figure 2. Teaching points: Ocular tissue can be divided into three layers: 1) an outer layer consisting of the sclera and the cornea (also known as the fibrous coat), 2) an intermediate layer consisting of the iris, the ciliary body, and the choroid (the vascular coat), and 3) an internal layer consisting of the retina (the nervous coat).

Figure 3. Teaching points: The retina contains several different types of cells that are segregated into layers. The layers and membranes of the retina starting from the outermost to the innermost layer are the 1) retinal pigmented epithelium (RPE), 2) photoreceptors (PR), 3) outer limiting membrane (OLM), 4) ) outer nuclear layer (ONL), 5) outer plexiform layer (OPL), 6) inner nuclear layer (INL), 7) inner plexiform layer (IPL), 8), ganglion cell layer (GCL), 9) nerve fiber layer (NFL), and 10) inner limiting membrane (ILM).

Figure 4. Teaching points: The retina is a highly metabolic tissue that requires sufficient oxygen for proper function. The inner half of the retina receives oxygen from the retinal circulation, while the outer half receives oxygen from the choroidal circulation.

Figure 5. Teaching points: Ocular blood supply derives from the internal carotid artery that branches off of the aorta. The internal carotid artery then branches into the ophthalmic artery that leads to the eye. The blood vessels supplying blood to the inner and outer retina, the iris, and the ciliary body originate from the ophthalmic artery. The ophthalmic artery further divides to give rise to the anterior ciliary arteries, the long and short posterior ciliary arteries, and the central retinal artery.

Figure 6. Teaching points: The retinal arteries branch off of the central retinal artery into a superficial layer that is located in the ganglion cell layer. These arterioles branch into an intermediate capillary bed located in front of the inner nuclear layer, and then into the deep capillary bed located behind the inner nuclear layer.

Figure 7. Teaching points: Retinal blood flow can take several different pathways through the tissue, through either the superficial, intermediate, or deep capillary beds. Normally, most retinal blood flow travels through the deep capillary layer. In diabetic rats, the transit time through the retina is exceedingly fast despite slower velocity: therefore, it is possible that a decreased proportion of flow in the diabetic rat goes through the deep layer, with rapid transit times possibly explained by flow preferentially perfusing the shorter superficial and/or intermediate capillary pathways due to vasoconstriction at sites indicated by the asterisks.

Figure 8. Teaching points: Diabetic retinopathy is characterized by leukocyte plugging of capillaries, a decrease in endothelial cell survival, and an increase in vascular leakage. These phenomena also could occur with the loss of platelet endothelial cell adhesion molecule-1 (PECAM-1), due to its important role in these same related functions.

Figure 9. Teaching points: Endothelial cells (as well as other cell types) have a surface layer that consists of molecules having sugar residues. This “sugar-coat”, or glycocalyx, serves many functions. The structure of the endothelial glycocalyx on the plasma side of the cell includes proteoglycans (such as syndecans and glypican-1) and glycosaminoglycans (such as heparan sulfate, chondroitin sulfate, and hyaluronic acid).