Dr Bang V Bui Department of Optometry and Vision Sciences The University of Melbourne Parkville VIC 3010 AUSTRALIA E-mail: firstname.lastname@example.org
Glaucoma is a leading cause of blindness. It is a multifactorial condition, the risk factors for which are increasingly well defined from large-scale epidemiological studies. One risk factor that remains controversial is the presence of diabetes. It has been proposed that diabetic eyes are at greater risk of injury from external stressors, such as elevated intraocular pressure. Alternatively, diabetes may cause ganglion cell loss, which becomes additive to a glaucomatous ganglion cell injury. Several clinical trials have considered whether a link exists between diabetes and glaucoma. In this review, we outline these studies and consider the causes for their lack of concordant findings. We also review the biochemical and cellular similarities between the two conditions. Moreover, we review the available literature that attempts to answer the question of whether the presence of diabetes increases the risk of developing glaucoma. At present, laboratory studies provide robust evidence for an association between diabetes and glaucoma.
Diabetes is a leading health priority, as it can cause severe systemic complications, such as retinopathy, neuropathy and nephropathy. This metabolic disorder results from either altered secretion of insulin by the pancreas (type 1 diabetes) or insulin resistance at a cellular level (type 2 diabetes). These changes produce widespread abnormalities in glucose uptake, resulting in hyperglycaemia, as well as abnormal metabolism and formation of lipids (Figure 1), which impact on neurons, blood vessels and glial cells. Indeed, this journal carries an excellent review on glial dysfunction in diabetic eyes.1 Many clinicians will be aware of diabetic retinopathy and its classification, as laid out by the National Health and Medical Research Council Guidelines for the Management of Diabetic Retinopathy.2 The purpose of this review is not to dwell on the specifics of diabetic retinopathy but to consider whether diabetes predisposes the eye to other disorders, in particular glaucoma.
Diabetes has been estimated by the World Health Organization to affect more than 180 million people worldwide. Despite increasing awareness and advances in management of this disease, it is anticipated that the number of sufferers of diabetes will rise to over 360 million by 2030.3 Diabetic eye disease is a leading cause of blindness4–6 in those of working age (aged from 30 to 69 years).7 The Wisconsin Epidemiologic Study of Diabetic Retinopathy reports that both the duration of the disease and its type (1 or 2) affect the chance of expressing retinopathy.8,9 The prevalence of retinopathy after five years (17 to 29 per cent) increases after 15 years to 78 to 100 per cent.
Diabetes is known to increase the risk of other health complications,10 in particular, cardiovascular problems11–13 and cerebral ischaemia.14,15 Of particular interest to eye-care practitioners is the question of whether we should adjust risk estimates for eye disease, such as glaucoma, if a person has diabetes. Epidemiological and laboratory studies provide conflicting evidence of the presence of a relationship between diabetes and glaucoma. This article will review these data to consider whether there is evidence for an increased risk of glaucoma in people with diabetes.
Glaucoma is the third leading cause of worldwide blindness. Current estimates suggest that more than 60 million people are affected by glaucoma.16,17 Many risk factors have been associated with glaucoma, including family history (first degree relatives), age, race, central corneal thickness, intraocular pressure (IOP), myopia and diabetes.18–21 Of these risk factors, the one that remains controversial is diabetes.
MULTIPLE FACTORS LEADING TO RETINAL GANGLION CELL APOPTOSIS AND OPTIC NERVE DAMAGE
Intraocular pressure is maintained by a balance between aqueous production and outflow,20,22,23 with an imbalance leading to elevated eye pressure.18,20,24 An elevation in IOP can manifest at the level of the retinal ganglion cell or lamina cribrosa. Very high levels of IOP will subject retinal cells to ischaemia as well as mechanical stress. This can lead to posterior deformation of the lamina cribrosa, a principal site of injury in glaucomatous neuropathy.25 Mechanical stress18,26,27 at the level of the lamina cribrosa28,29 impairs retrograde transport of neurotrophins,26,27,29–31 such as brain-derived neurotrophin factor (BDNF), to retinal ganglion cell soma, thus activating the cell death (or apoptotic) cascade.18,32 Impaired anterograde transport of neurotrophic factors to the lateral geniculate nucleus is also thought to contribute to the onset of apoptosis.20 Long-term vascular and mechanical stresses can produce remodelling of the extracellular matrix, thus altering the compliance of the lamina cribrosa, leading to further injury at the optic nerve.25,33
In addition to mechanical injury, IOP elevation can impair ocular blood flow,34,35 reducing perfusion pressure to retinal neurons. In a normal eye, vascular autoregulation compensates to maintain normal blood flow by changing vascular diameter. Glaucoma pathogenesis may involve abnormal autoregulation,36–39 which produces glial cell activation and retinal ganglion cell loss due to local hypoxia. It is clear that elevated IOP is not the whole story, as from 25 to 60 per cent of glaucoma sufferers appear to have ‘normal’ IOP.40,41 Additionally, many people with ocular hypertension never develop glaucoma. Thus, other risk factors must contribute to the risk of developing glaucoma. Diabetes may be one such risk factor,42 as will be discussed.
ASSOCIATIONS BETWEEN DIABETES MELLITUS AND OTHER DISEASES
It is clear that those with diabetes are at increased risk of cardiovascular complications and mortality.10,43,44 Whether condition A (for example, diabetes) influences the risk of developing condition B (for example, glaucoma) is defined by the relative risk, which is the ratio of the chance of developing B in groups with or without condition A. Wong and colleagues13 report that patients with diabetic retinopathy and hypertension have a greater relative risk of stroke (2.83) than individuals without diabetes (2.61) and without hypertension (1.97). People with diabetes are two to three times more likely to develop coronary heart disease (CHD).11,12,45 It is of interest that diabetic vascular complications in the eye (including microaneurysms, soft exudates and haemorrhages) mirror those throughout the body, as such ocular changes can be predictive of systemic disease.13,45
HOW MIGHT DIABETES AND GLAUCOMA BE LINKED?
Early population studies suggested that those with diabetes are more likely to have elevated IOP and glaucoma.46,47 Previous epidemiological studies have shown that patients with diabetes have a higher IOP (by 2 to 3 mmHg) than those without diabetes.47–51 Ellis and colleagues52 showed that on average their diabetic cohort had more people with ocular hypertension (IOP greater than 21 mmHg). Oshitari and associates53 suggest that IOP elevation (approximately 2.5 mmHg) has a modest linear correlation with the severity of diabetic disease (HbA1c and IOP: r = 0.15). In contrast, Bankes54 found no IOP differences between the diabetic and non-diabetic groups. Although any IOP increase in diabetic eyes is small, it might contribute to increased stress on the ganglion cells.
Several larger scale clinical trials, including the Beaver Dam,50 the Rotterdam55 and the Blue Mountains Eye56 studies, report an increased risk of open-angle glaucoma in people with diabetes, however, no evidence for an association was reported57 in the Baltimore Eye Survey,48 the more recent Rotterdam Study19 and the Framingham Eye Study.58,59 Thus, there is inconsistency among epidemiological studies from the past four decades (Table 1).
Table 1. Summary of epidemiological trials considering the association between diabetes mellitus and open-angle glaucoma. NA = not applicable; CL = confidence limits.
Why might such discrepancies exist? It is difficult to reconcile these studies, as they have varying diagnostic, assessment criteria and statistical approaches. For example, a recent reanalysis of data from the Rotterdam eye study19 shows that diabetes is not a risk factor for glaucoma. This contradicts the earlier finding from the same population that there was increased risk.55 These conflicting outcomes arise due to a redefinition of primary open angle glaucoma (POAG), which substantially reduced the number of participants. A similar inconsistency has been identified by the authors of the Ocular Hypertension Treatment Study (OHTS). Initial multivariate analysis from the OHTS suggested that diabetes was protective against conversion from ocular hypertension to glaucoma,60 however, more recent analysis suggests that both the European Glaucoma Prevention study (EGPS) and the OHTS were underpowered to find an effect.61 The OHTS authors acknowledge that patient self-reporting of diabetes and their exclusion criteria may have produced a diabetic cohort that was atypical,62 leading to the observed protective outcome.
As many of the above studies do not specifically set out to consider a link between these two conditions, their selection and diagnostic criteria (self-report, glucose tolerance testing) differ substantially. As such, the presence of opposing outcomes may not be surprising. Given this background of conflicting epidemiology, we consider whether there is other clinical and laboratory-based evidence for a relationship between diabetes and glaucoma.
LABORATORY EVIDENCE LINKING DIABETES AND GLAUCOMA
Anatomical changes in diabetes and glaucoma
The presence of hyaline bodies particularly in the ganglion cell and retinal nerve fibre layer in post-mortem eyes of patients with diabetes63 suggests that there is inner retinal cell death. Increased cellular apoptosis has been well documented in diabetic eyes using DNA terminal dUTP nick end labelling (TUNEL),5,64–70 which stains damaged or fragmented DNA. In addition, many studies have shown significant increases in pro-apoptotic factors, such as Bax.71 Ganglion cell axon atrophy has been observed in BB/W-rats, a naturally occurring model of type 1 diabetes.72 Similarly, NeuN (neuron-specific nuclear protein) labelling was found to be significantly reduced in streptozotocin (STZ)-induced diabetic rats, which indicates significant neuronal losses.73 STZ-induction is a common animal model of type 1 diabetes, as this toxin is employed to target and destroy pancreatic β-cells, leading to loss of insulin production and a resultant increase in blood glucose levels (Figure 1).
Barber and coauthors68 showed in STZ-diabetic rats that the ganglion cell, inner plexiform and inner nuclear layers are significantly thinned, whereas the outer plexiform and outer nuclear layers are not. Bui and collaborators74 showed that structural changes were present at the optic nerve (Figure 2), including a reduction in myelin sheath thickness (control: 0.25 versus STZ: 0.22 µm), nerve fascicle area (control: 0.22 versus STZ: 0.20 mm2) and increased blood vessel/connective tissue ratio in diabetic rats 12 weeks after STZ-induction. At this time there were no significant differences in the gross structural integrity of STZ-retina (Figure 2). These optic nerve changes might account for the presence of ganglion cell dysfunction in STZ rats, as measured using the electroretinogram (ERG, Figure 3A).
Consistent with findings in diabetic animal models, Sugimoto and associates75 employed high resolution optical coherence tomography and found a significant reduction in retinal nerve fibre thickness, even though total retinal thickness was increased in human diabetic eyes without retinopathy. Using a similar approach, van Dijk and co-workers76,77 have found that people with type 1 diabetes (minimal retinopathy) showed a significant reduction in ganglion cell (paracentral retina) and retinal nerve fibre (at the macula) layers, with little change in outer retinal layers. Moreover, ganglion cell layer thickness was found to show a strong correlation with the duration of diabetes.77 Thus there is evidence that diabetic neuropathy has a greater effect on the inner retina, as is the case in glaucoma.78,79
Functional changes in diabetes and glaucoma
In addition to the anatomical evidence, the presence of neuropathy can be quantified using functional measures, such as the ERG, of the eye's response to light. This approach has been employed extensively in studies of diabetes and glaucoma.
A temporary decline of rod photoreceptoral function has been observed as soon as two days after the induction of STZ-diabetes.80 This returns to normal after two weeks and thereafter shows a more gradual attenuation between four and 12 weeks. It is believed that some of this outer retinal dysfunction is caused by impaired sodium/potassium (Na+/K+)-ATPase activity81,82, which produces a smaller dark current and thereby a smaller a-wave. Altered retinal function can arise from diabetic dyslipidaemia (Figure 4) as supplementing omega-3 fats by dietary means prevents some diabetic-induced neuronal dysfunction.83 Diabetic rats fed a diet rich in omega-3 fatty acids (fish oil), had normal polyunsaturated fatty acids (PUFA) profiles and returned functional responses similar to citrate-treated controls animals (control-omega-3 fed; -560 ± 31 versus STZ-omega-3 fed; -567 ± 32 mV) as shown in Figure 5. Those STZ-diabetic rats fed an omega-6 rich diet show an exaggerated functional deficit (STZ-omega-6 fed; -351 ± 38 versus STZ-omega-3 fed; -567 ± 32 mV).
INNER RETINAL RESPONSE
Oscillatory potentials (OPs) are high frequency, small amplitude wavelets that reside on the b-wave (ON-bipolar cell function) of the ERG.84,85 Oscillatory potentials are thought to arise from inner retinal feedback pathways involving amacrine cells. Studies in humans reveal that oscillatory potentials are reduced early in diabetic retinopathy,86–88 well before ERG a-wave (photoreceptoral function) or b-wave changes.87,89 Uccioli and colleagues90 reported that oscillatory potentials were reduced 10 years after the onset of diabetes, without any a-wave or b-wave loss. Other studies have associated oscillatory potential attenuation91–94 with an accumulation of γ-amino butyric acid (GABA) in the inner retina.92 Studies employing pharmacological inhibitors of the GABAergic pathway report modulation of oscillatory potential.95,96 In addition, inhibition of glycinergic pathways modify the oscillatory potentials.97,98 Taken together, these studies suggest that the inner retinal inhibitory pathways are damaged by diabetes. Consistent with this possibility, Gastinger, Singh and Barber99 have found that the dopaminergic and cholinergic amacrine cells (also GABAergic) are lost early in diabetes.
In agreement with findings in people with diabetes, STZ-diabetic rats show delayed oscillatory potential implicit time (at 12 weeks control; 28 ± 0.5 versus STZ; 33 ± 0.9 ms)80,91 and reduced (at 12 weeks control; 107 ± 7 versus STZ; 76 ± 7 mV) oscillatory potential amplitude,80,100,101 as shown in Figure 6. Layton, Safa and Osborne102 and Bui and coauthors74 reported a reduction in both oscillatory potential and b-wave amplitudes (Figure 3A), which implies that the changes in oscillatory potential may reflect losses of ON-bipolar cell b-wave response. Of note, Yee and associates83 showed that anomalies of the oscillatory potential were correlated with high blood glucose (rs= -0.53, p < 0.05), whereas the b-wave was not (rs= -0.29, not significant) and attenuation of oscillatory potentials could not be reversed by omega-3 fatty acid supplementation despite improvements in the b-wave (Figure 5, lower panel). This observation suggests the possibility of two lesions in the retina, with one being lipid dependent (outer retina) and the other being blood glucose related (inner retina).83
GANGLION CELL AND VISUAL PATHWAY
As discussed above, there is evidence for inner retinal damage in diabetes. Whether this manifests in terms of ganglion cell function has been assessed using skin electrodes placed above the occipital lobes to measure visual evoked potentials (VEP).103 Parisi and Uccioli104 showed that people with type 1 diabetes have reduced VEP amplitudes and slower implicit times compared with age-matched controls. Similarly, the photopic negative response (PhNR), a cone-driven ganglion cell response,105–107 has also been shown to be affected by diabetes, with smaller amplitudes associated with greater disease severity. Interestingly, a reduction in the PhNR has also been observed in patients with open-angle glaucoma and in a primate model of experimental glaucoma.105,106,108 The pattern ERG (pERG), recorded to a high contrast reversing checker board stimulus, is an excellent indicator of ganglion cell function.86,106 The pERG is abnormal in optic nerve diseases and glaucoma,106,108 as well as in type 1 diabetic retinopathy.86,104 Although the above studies show that there is ganglion cell dysfunction in diabetes, this might simply arise due to changes in the outer retina.
The integrity of ganglion cells can also be assessed using very dim flashes, which produce a small ERG waveform near absolute scotopic threshold.109–114 In rodents, this response, termed the scotopic threshold response (STR), is sensitive to IOP elevation115,116 and is lost following optic nerve transection.113 In the absence of IOP elevation, Bui and colleagues74 showed that the scotopic threshold response was more affected (Figure 3A) and was affected earlier than responses arising from the outer retina in STZ-diabetic rats (Figure 7).117 More specifically, ganglion cell dysfunction (STR: -85 ± 8 per cent reduction) was greater than photoreceptoral (a-wave: -12 ± 7 per cent) and ON-bipolar cell (b-wave: -18 ± 4 per cent) dysfunction. The level of scotopic threshold response dysfunction showed some association with increased connective tissue growth factor (CTGF, Figure 3B), consistent with the idea that changes in connective tissue may be related to cell injury in diabetes (as will be discussed below).
Thus, anatomical and functional evidence exists to suggest that diabetes and glaucoma share many similarities. It appears that in both cases, inner retinal neurons may be more susceptible to these conditions than neurons that reside in more distal layers of the retina.
Studies of IOP elevation in diabetes
Vesti and Trick118 have reported that diabetes exacerbates visual dysfunction in ocular hypertensive patients. The authors showed that pERG waveforms, colour vision and contrast sensitivity were more attenuated in ocular hypertensive patients with diabetes than in people with ocular hypertension without diabetes. Quigley and coauthors119 reported an association between diabetes and nerve fibre atrophy, although they failed to show any significant difference in the severity of visual field loss between glaucoma sufferers with or without diabetes.
Several studies have attempted to address directly the issue of whether diabetes influences the risk of glaucoma, by considering the effects of IOP challenge in normal and diabetic animals. Casson and colleagues120 reported that an acute hyperglycaemic state induced either by STZ-diabetes or glucose injections in rats protected the ON-bipolar cell b-wave component of the ERG from acute, severe IOP challenge. Several days following IOP elevation to 100 mmHg, larger b-waves were measured in STZ-induced diabetic eyes than in the non-diabetic control group. More recently, Holman and colleagues121 showed that short-term hyperglycaemia (10 days post STZ-diabetes) protected retinal ganglion cells from hypoperfusion induced by ligation of the common carotid arteries. Both these studies suggest that in the short-term, an increase in glucose levels is protective for bipolar cells against ischaemic insult, most likely due to the metabolic advantage provided by this substrate during ischaemic stress. This may not be the case for retinal ganglion cells, particularly during a much milder and chronic IOP challenge, as is the case in glaucoma.
Kanamori and co-workers69 showed that chronic ocular hypertension (four weeks, from 13 to 29 mmHg), induced by cauterising three episcleral veins, was more detrimental in STZ-diabetic rat retina compared with control eyes. The authors showed that there was an increased apoptosis in the inner retina in high IOP challenged STZ-treated rats. More specifically, larger cell bodies undergoing apoptosis were at the level of the ganglion cell layer. Additionally, there were smaller cell bodies that stained for TUNEL in the inner nuclear layer, which may be Müller or amacrine cells. This outcome is consistent with the idea that diabetes exacerbates inner retinal neuronal injury in a model of chronic glaucoma. It is worth noting that IOP increased apoptosis by the same amount in STZ-diabetic and control eyes. This suggests that existing retinal ganglion cell dysfunction (Figures 3A and 7) or loss in diabetes is ‘additive’ to the ganglion cell dysfunction/death produced by IOP stress. If these data translate to humans, it may be that the presence of diabetes may produce an earlier manifestation of detectable visual field loss in glaucoma. There may also be a more aggressive disease development in diabetic eyes, as evidenced by the finding that people with diabetes are more likely to have bilateral disease in normal tension glaucoma.122 Given the above evidence, the next section considers potential mechanisms that link diabetes and glaucoma.
DISEASE MECHANISMS COMMON TO DIABETES AND GLAUCOMA
Diabetes produces abnormalities in glucose as well as lipid metabolism (Figure 1). The effects of hyperglycaemia have been studied extensively, however, less attention has been paid to lipid anomalies and the potential role that this has in promoting neuronal dysfunction and a modified inflammatory status. These factors should not be thought of in isolation, as hyperglycaemia will cause glycation of lipids123 aggravating the diabetes-induced changes.
Studies in humans and animals have shown that lipid composition and metabolism are abnormal in diabetes.83,124–127 Diabetic rats show increased liver cholesterol production.128 Cell membranes of rat retinal neurons (Figure 4) and erythrocytes in people with type 1 diabetes, show a marked decrease in polyunsaturated fatty acids at the expense of docosahexaenoic acid (DHA).83,126 Docosahexaenoic acid is an important fatty acid that promotes neural function,129,130 and polyunsaturated fatty acids derivatives (eicosanoids) give rise to inflammatory modulators and neuro-protecting substrates (neuroprotectins).131,132 Thus dyslipidaemia may be a key link between diabetes and cardiovascular complications, and potentially neuropathy (Figure 8).
It is thought that reduced insulin production or increased insulin resistance leads to elevated secretion of triglycerides from the liver as well as impaired clearance of lipid (Figure 8). This imbalance in fats leads to a reduction in plasma levels of high density lipoproteins (HDL) and an increase in low density lipoproteins (LDL).133 There is an associated increase in cellular fatty acid uptake and abnormal fatty acid metabolism134 (Figure 1).
Membrane lipids are metabolised in the cyclo-oxygenase and lipoxygenase pathways to produce short-lived local hormones, such as leukotrienes, prostaglandins, prostacyclins and thromboxanes.135,136 The omega-3 eicosanoid derivatives are anti-inflammatory, anti-apoptotic and enhance fluid129 or blood flow compared with the omega-6 family. Diabetes produces an imbalance that favours omega-6 derivatives by about 2.5 times83 and will promote a pro-inflammatory environment. It is increasingly apparent that pro-inflammatory factors are present in glaucoma137 and many other eye diseases.138,139
Abnormalities in neuronal fatty acid uptake can modify cellular energy production as fatty acid and glucose metabolism are linked.134 The abnormal lipid processing in diabetes will increase cellular fatty acid uptake, which reduces glucose uptake (Figure 1), as evidenced by reduced glucose transporter expression in the retina140 and retinal pigment epithelial cells in vitro.141 Less cellular glucose impairs energy production, which can deplete the antioxidant capacity of neuronal tissues. Increased oxidative stress is known to be important in both diabetes and glaucoma pathogenesis.142 Finally, Bazan132 has shown that long-chain derivatives of the omega-3 family of fatty acids, called neuroprotectins, are potent anti-apoptotic molecules that down regulate Bax activity. Thus, lipid changes associated with diabetes (Figure 4 in STZ-diabetic rats) could promote apoptotic pathways.
Much of the research into diabetic eye disease has focused on vascular complications (Table 2), which manifest clinically as intraretinal haemorrhages, localised ischaemia (cotton wool patches), basement membrane thickening (vascular tortuosity) and loss of pericytes (microaneurysms).6,7,143–147 Proliferative diabetic retinopathy occurs when growth factors, such as vascular endothelial growth factor (VEGF, a potent angiogenic and vascular permeability factor148–150) promote neovascularisation,147,151,152 which can produce neovascular glaucoma.153 Those with type 2 diabetes are more likely to develop non-proliferative retinopathy,6 including sight-threatening macular and retinal oedema.7
Table 2. Stages of diabetic retinopathy, adapted from Phipps, Fletcher and Vingrys80 and Watkins7
Microvascular occlusions Venous beading and dilation
Vitreous or pre-retinal haemorrhage
Increased permeability of inner and outer blood-retinal barriers
Retinal haemorrhages and microaneurysms
Hard exudates and retinal oedema Capillary non-perfusion
These patent clinical signs are often preceded by reduced retinal blood flow,154–156 which can render tissues hypoxic and trigger the release of VEGF.157 Blood flow abnormalities have been attributed to changes in endothelial cells and intracellular microfilaments of the vascular endothelium as has been found in STZ-diabetic rats.158 Cringle and colleagues159 reported that after five weeks of STZ-diabetes, rats showed increased retinal blood flow compared with controls. This was supported by Yu and co-workers,158 who reported abnormalities in F-actin stress fibre arrangement in superficial diabetic retinal capillaries as early as five weeks following STZ induction. This change in F-actin stress fibre is consistent with the increased blood flow (Table 3).
Table 3. Summary of vascular dysregulation reported in clinical/experimental diabetes and in ocular hypertensive/glaucomatous studies
absence of F-actin cytoskeleton (5 weeks diabetic retina)158
Regardless of the absolute level of blood flow, growing evidence suggests that diabetic eyes have a reduced capacity to autoregulate blood flow, when challenged by reduced ocular perfusion pressure (the balance between blood pressure and IOP). Studies have found that people with diabetes show impaired vasodilation in response to flickering light.160,161 This suggests the presence of endothelial cell dysfunction and may involve abnormalities in nitric oxide (NO), a potent vasodilator.162 It is of interest that abnormal autoregulation is implicated in the pathogenesis of glaucoma,36,39,163 providing a potential link between the two conditions.
Reduced blood flow produces tissue hypoxia, which leads to increased VEGF production that promotes neovascularisation. In diabetic retinopathy, over expression164 and increased stability of VEGF mRNA165 have been associated with hypoxia-inducible factor-1 (HIF-1α),166–168 which accumulates when oxygen levels fall below 10 per cent. Importantly, HIF-1α is increased in ganglion cells after elevated IOP168 and in the retina and optic nerve head of human glaucomatous eyes.169
Arden and associates170 suggested that the retina is particularly prone to neovascularisation because of its high metabolic demand. The authors observed that neovascularisation does not occur in diabetic brains, which is less metabolically demanding than the retina. Similarly neovascularisation was absent in those with diabetes who also had rod photoreceptor degeneration (retinitis pigmentosa) or pan-retinal photocoagulation.171 Consistent with this idea, diabetic mice with loss of rod photoreceptors showed reduced levels of VEGF.172
There is a fine balance between the levels of reactive oxygen species (for example, O2-, H2O2, OH) and the eye's ability to protect against them (anti-oxidants). This imbalance is known as oxidative stress173 and in diabetic eyes,162 it can be caused by hypoxia,174 hyperglycaemia-induced formation of advanced glycation end-products (AGEs),175 glycosphingolipid metabolic dysfunction,176 over activation of aldose reductase177 and mitogen-activated protein (MAP) kinases,178 as well as mitochondrial abnormalities.71,142 In addition, abnormal lipid metabolism will impair mitochondrial function, leading to compromised antioxidant systems.134 It is pertinent that increased oxidative stress, as evidenced by nitric oxide overproduction, has been reported in both glaucoma and diabetes (Table 4).162,179
Table 4. Summary of evidence for oxidative stress reported in studies of clinical/experimental diabetes and in ocular hypertensive/glaucomatous conditions
NOS = nitric oxide synthase; iNOS = inducible NOS; eNOS = endothelial NOS; nNOS = neuronal NOS
accumulation of HIF-1α associated with VEGF mRNA over expression164
POLYOL-SORBITOL PATHWAY ABNORMALITIES IN DIABETES
Under normal circumstances (that is, normoglycaemia), glucose is metabolised to glucose-6-phosphate by hexokinase. Non-phosphorylated glucose is converted to sorbitol, a reaction catalysed by aldose reductase in the polyol-(sorbitol) pathway (Figure 9). Sorbitol is then converted to fructose via sorbitol dehydrogenase. In a hyperglycaemic environment, the polyol-sorbitol pathway becomes overwhelmed by high levels of sorbitol,177,180,181 resulting in a metabolic imbalance.182 One manifestation is an increased ratio of nicotinamide adenine dinucleotide (NADH)/nicotinamide adenine dinucleotide phosphate (NAD+).182,183 This change in redox state impacts many metabolic pathways, which can increase nitric oxide formation,183 leading to endothelial cell182 and glial cell dysfunction.67 Consistent with this idea, diabetic mice with normal aldose reductase showed oxidative stress (elevated superoxide formation) and neuronal apoptosis (increased c-Jun NH2-terminal kinase activity, DNA damage).184 On the other hand, mice lacking aldose reductase showed little oxidative stress and neuropathy when rendered diabetic. Studies by Winkler and co-workers185 and Diederen and collaborators186 challenge the ‘pseudohypoxia’ hypotheses, as these authors failed to find a significant difference in NAD+ to NADH ratio in STZ-diabetic retina. Nevertheless the above studies are in agreement that in the diabetic eye, there is increased anaerobic metabolism, as evidenced by a six- to seven-fold increase in lactate production.174,185,186 Impaired energy production would make neurons less able to cope with stress.
PROTEIN KINASE C PATHWAY
Increased vascular permeability and changes in retinal blood flow observed early in diabetes have been associated with the activation of diacylglycerol and protein kinase C (PKC).81 PKC is an intracellular signalling protein and has a regulatory role in modulating neural and vascular activities. Excess PKC leads to thickening of vascular basement membranes in diabetic eyes.151 This impairs oxygen exchange producing relative hypoxia and oxidative stress and ultimately VEGF release.187 Increased PKC can also produce vasodilation,81 which may contribute to abnormalities in blood flow reported in some diabetic retinas.154,155,159
Changes in PKC are also known to affect extracellular matrix regulation,188 which is pertinent in two ways. First, PKC activation leads to elevation of several matrix metalloproteases (MMP-1, -3, -9) and tissue inhibitors of metalloproteases (TIMP)-1 but decreased TIMP-2 in the trabecular meshwork.188 A change in the trabecular meshwork can increase impedance to aqueous outflow producing IOP elevation. Khurana and colleagues189 showed that a PKC inhibitor (GF109203X) produced a 46 per cent increase in aqueous outflow in perfused porcine eyes. Second, over expression of MMP-9 has been associated with activation of astrocytes as well as structural changes at the optic nerve head in patients with diabetes.190,191 Structural changes at the optic nerve head are seen in STZ-diabetic rats (Figure 2). PKC-induced changes to connective tissues could influence lamina cribrosa compliance, making the optic nerve more susceptible to the mechanical effects of IOP elevation.
ALTERATIONS TO NA+/K+-ATPASE ACTIVITY
Sodium/potassium (Na+/K+)-ATPase is a ubiquitous plasma membrane protein. It acts to maintain transmembrane potential by pumping potassium in and sodium out of cells in an energy dependent manner. Na+/K+-ATPase activity can be altered by the lipid composition of cell membranes192,193 or elevated PKC and sorbitol,81,82 all of which occur in diabetes.
A reduction in Na+/K+-ATPase activity can impair ionic homeostasis and the activity of retinal194–196 and peripheral nerves,80,197 as well as the retinal pigment epithelium.198 PKC can modulate Na+/K+-ATPase activity,199 thereby affecting the dark current and thus, the ability of photoreceptors to transduce light to a chemical signal. Alternatively, a metabolic switch in diabetic retina from aerobic to the less efficient anaerobic pathway would substantially reduce the activity of Na/K+-ATPases. Ames and colleagues200 estimate that almost half the energy produced via aerobic pathways is needed to sustain sodium transport. A reduction in energy availability would be highly detrimental.
Nitric oxide is a vasoactive substrate and an intracellular messenger. Excessive production of nitric oxide, via nitric oxide synthase (NOS), can induce oxidative stress promoting apoptotic cell death and impaired vascular autoregulation.162,201,202 There are several subtypes of NOS: neuronal NOS (nNOS), endothelial NOS (eNOS) and inducible NOS (iNOS). Hyperglycaemia increases iNOS activity, leading to excess nitric oxide, which is thought to contribute to diabetic retinopathy.202–206 As early as one to three weeks after STZ-induction, fewer eNOS- and nNOS-positive cells were found in diabetic rat retina compared with control retina.201,207 Abu El-Asrar et al205 found elevated iNOS immunoreactivity in Müller cells in human diabetic retina. Müller cells span almost the entire width of the retina and are important for neuronal support.1,208 Kawasaki, Otori and Barnstable209 suggest that Müller cell dysfunction may be an early contributor to neuronal injury due to their critical role in extracellular glutamate uptake and recycling.210
Excessive nitric oxide has been suggested to increase vascular resistance and reduce blood flow at the retina and optic nerve head211 in patients with primary open-angle glaucoma. Like diabetes, increased nNOS and iNOS activity has been found in people with glaucoma, particularly in astrocytes at the lamina cribrosa.179 A diabetic eye already under stress due to nitric oxide overproduction might be at risk from additional oxidative injury in glaucoma. Translating these neuroprotection strategies from the laboratory to the clinic has been unsuccessful.212
Several studies have targeted the NOS pathways for pharmaceutical intervention in glaucoma.213–215 Treatments including a NOS substrate (that is, L-arginine) and a nitric oxide donor (nitroprusside and minoxidil) have been administered in an attempt to increase ocular blood flow with some success.215 Additionally, Neufeld213 has reported that the iNOS inhibitor, aminoguanidine, provides neuroprotection in animal models of glaucoma.
Axonal transport abnormalities and neurotrophic factors
Impaired retrograde axonal transport can trigger apoptotic cell death,18,32 as neurotrophic factors fail to reach their retinal targets. Similarities between diabetes and glaucoma are summarised in Table 5. In diabetes, clinical and experimental observations have demonstrated neurotrophic deprivation,173,216 likely via impaired retrograde axonal transport.65,217 For example, nerve growth factor, insulin-like growth factor and neurotrophin-3 are reduced in diabetic animals and humans.216,218–221 Reduced brain-derived neurotrophic factor (BDNF) is thought to promote diabetic neuropathy.222,223 Decreased BDNF immunoreactivity has been found in the ganglion cell and inner plexiform layers of STZ-induced diabetic rat retina.222 This may account for degeneration of ganglion cells, dopaminergic amacrine cells and Müller cells in these eyes.222 In agreement with this idea, intravitreal administration of BDNF provided some neuroprotection, particularly for amacrine cells.64,222
Table 5. Summary axonal transport abnormalities and glial changes reported in studies of clinical/experimental diabetes and in ocular hypertensive/glaucomatous conditions
Impaired retrograde transport (at optic nerve axons)65
Impaired ortho/retrograde transport (eg: elevated IOP conditions, proportional increase)18,28,32,224,225
Axonal cytoskeletal and transport retardation changes226
↑ GFAP activation (ie: Müller cell & astrocyte dysfunction at the optic nerve head)209,241–243,247–249
Blockage of axonal transport is thought to be a critical factor in glaucoma.32 Several studies have demonstrated deficient axonal transport of radioactively-labelled amino acid 3H-leucine and horseradish peroxide in non-human primate eyes, at the lamina cribrosa, following IOP elevation.28,224,225 The degree of optic nerve head retro- and antero-grade transport impairment is proportional to the level of IOP elevation.225 Pease and colleagues29 showed retrograde transport of BDNF was reduced and its receptor (TrkB) was increased at the optic nerve head in non-human primate experimental glaucoma. This has also been found in rodent30 and porcine226 models of glaucoma.
In diabetic retina, retinal ganglion cells, microglia and pericytes show increased immunoreactivity for connective tissue growth factor,74,227–229 which mediates the activity of transforming growth factor-β (TGF-β), the action of which on connective tissue is important in contractile scarring for wound healing.230 Analogous to VEGF, connective tissue growth factor participates in angiogenesis, vascular proliferation, endothelial cell formation and migration.231 This cytokine is upregulated by increased reactive oxygen species, advanced glycation end-products, as well as tissue hypoxia and hyperglycaemia, which are all present in diabetes.227 Over expression of connective tissue growth factor causes caspase-3 dependent apoptosis.232 Bui and colleagues74 reported elevated connective tissue growth factor in STZ-treated rat eyes, which was correlated with ganglion cell dysfunction (Figure 3B). This elevation of connective tissue growth factor was associated with an increase in connective tissue at the optic nerve head (Figure 2D). In glaucoma, elevated connective tissue growth factor and TGF-β expression are thought to alter the activity of MMP and TIMP, which produce abnormal tissue remodelling and turnover.233,234 This can impact at two sites. First, abnormal trabecular meshwork connective tissue can impair aqueous outflow and increase IOP. Second, remodelling of lamina cribrosa extracellular matrix can make the tissue less compliant and susceptible to mechanical stress with elevated IOP, as has been suggested by a number of studies.25,235–238
Altered glial cell support: Müller cells and astrocytes
Apart from vascular complications, there is mounting evidence that both neurons and glial cells (including Müller cells, actrocytes and microglia) are affected early in the course of diabetes,68,71,239 prior to vascular changes as reviewed by Fletcher, Phipps and Wilkinson-Berka.1
Glial cells are important for retinal1,240 and optic nerve homeostasis.241–243 They interact with other cells to maintain vascular structure and ionic balance, glutamate shuttling, metabolic support, mechanical support and the removal of invading organisms and foreign debris.73,241–243 Glial cell dysfunction can be indicated by an increased expression of glial fibrillary acidic protein (GFAP), an intermediate glial protein, which is increased early in diabetes.73,239 Increased glial fibrillary acidic protein immunoreactivity in Müller cells and astrocytes was found to be associated with higher concentrations of sorbitol and fructose in STZ-diabetic rats.67 Thus, diabetes induced changes in the polyol pathway may partly explain glial cell activation.67,233
Other glial cells may also play an important role in diabetic retinopathy. Zeng, Ng and Ling73 observed hypertrophy, as well as an increase in OX42-immunoreactive microglial cells (in the nerve fibre and ganglion cell layers) after four-months of STZ-diabetes. This suggests the presence of chronic inflammation, as these cells are involved in phagocytosis of degenerated neurons in diabetic eyes. It is of interest that an aberrant inflammatory response can increase the vulnerability of the retina to stress.244–246 This is an emerging idea in glaucoma pathophysiology.137
Glial fibrillary acidic protein immunostaining can also be detected in astrocytes at the optic nerve head,247,248 particularly in glaucomatous eyes.249 Recently, Balaratnasingam and associates250 showed that glial fibrillary acidic protein immunoreactivity decreased at six, nine and 12 hours following IOP elevation to 40 to 45 mmHg. In addition, morphological changes could be detected in astrocytes as early as three hours following this IOP insult. In another study,251 high IOP for 12 hours resulted in impaired axonal transport in the lamina and proximal post-laminar regions of the optic nerve head. These outcomes suggest that astrocytes play a role in protecting against axonal injury. Abnormal glial cells appear to be present in diabetes and glaucoma. Glial cell dysfunction, arising from hyperglycaemia and lipid changes in diabetes, might place ganglion cell axons at increased risk of injury from IOP stress.252
Glial cells rapidly take up and deactivate extracellular neurotransmitters, to prevent excessive activation of excitatory receptors. Intracellular calcium build-up via influx and the release of intracellular calcium stores can activate the apoptotic cascade.253,254 Studies in both STZ-rats and people with diabetes reveal that hyperglycaemia is associated with significantly elevated retinal glutamate levels.151,239,255,256 Li and Puro257 showed dysfunction of the glutamate-aspartate transporter in isolated Müller cells from STZ-diabetic rats. Excitotoxic injury can cause cell death, which would compound any IOP related stress.
Altered biomechanical properties of the sclera, lamina cribrosa and optic nerve
Elevated IOP can increase stress (force) and strain (localised deformation) at the optic nerve head.25,235,237 A gradual compromise in the collagen fibrils that make up the load-bearing connective tissues237 at the lamina cribrosa and the peripapillary sclera can exacerbate stress and strain on ganglion cell axons and capillary networks that transit this region.25,79,235–238,258,259 Optic nerve head connective tissue changes have been observed following experimental glaucoma in monkey258 and rat.260 These changes can occur with modest IOP elevation of 10 to 20 mmHg above normal levels in rats.260 This connective tissue remodelling alters the biomechanical characteristics of the optic nerve head, increasing the likelihood of posterior deformation of the lamina cribrosa with IOP. Axonal deviation can occur secondary to posterior bowing of the lamina cribrosa261 and posterior optic disc displacement,79,262 which can compromise axonal transport. Diabetes can exacerbate connective tissue remodelling and amplify these biomechanical changes.
Chronic hyperglycaemia has been shown to produce glycation of rat tail collagen (type 1) proteins, resulting in the formation of advanced glycation end-products.263 Application of glucose-6-phosphate into collagen gel cell cultures lead to modification of its biomechanical properties, including increased collagen cross-linking and enzymatic degradation.263 Whether diabetes-induced glycation in optic nerve and trabecular meshwork connective tissues renders diabetic eyes less able to accommodate IOP fluctuations, requires further investigation.
Diabetes mellitus and glaucoma are leading causes of blindness. The epidemiological evidence for an association between the two conditions remains inconclusive. There is a growing body of evidence that the presence of long-standing hyperglycaemia, along with lipid anomalies, may increase the risk of neuronal injury from stress. In particular, the laboratory data provide robust evidence for an association. We illustrate various pathways whereby diabetes and glaucoma might converge to produce an increased risk of neurodegeneration. These pathways are schematised in Figure 10 and can be summarised as follows.
• Altered biochemical pathways compromise cells and also increase oxidative stress.
• Vascular changes can reduce blood flow and impair oxygen diffusion. Endothelial cell injury and dysfunction can reduce the capacity for autoregulation to protect against fluctuations in eye and blood pressure. These lead to relative hypoxia.
• Glial cell activation can lead to impaired ionic support and possibly reduced glutamate uptake. This might increase the chance of excitotoxicity. Excessive glial cell activation may also contribute to chronic inflammation.
• Changes to neurons may impair their ability to function, including axonal transport, leaving these neurons, which are already vulnerable, under additional stress.
• Connective tissue remodelling might reduce compliance at the trabecular meshwork and lamina cribrosa, promoting increased IOP and greater optic nerve head mechanical stress, respectively.
Further studies are needed to directly determine mechanisms underlying any potential association between diabetes and glaucoma. Nevertheless, the above literature indicates that clinicians should consider adjusting the risk for glaucoma in patients who have chronic diabetes. Our early work with dietary modification (increased fish oil intake) indicates that some benefits can be expected in people with both diabetes and glaucoma.
GRANTS AND FINANCIAL SUPPORT
The authors are supported by NHMRC project grants 400127, 475603, 566570 and 350224.