growth factor (VEGF) (15–24). These changes may alter retinal neurons in ways that are not yet fully identified and understood, but that likely compromise the functional output of the retina.

DIABETES REDUCES RETINAL FUNCTION

Diabetic retinopathy is associated with multifactorial deficits of vision (25–28) and impaired electroretinographic (ERG) responses (29–34). Although the development of proliferative retinopathy and diabetic macular edema are the most severe visionthreatening characteristics of diabetic retinopathy (35), even patients without clinically detectable retinal vasculopathy can experience subtle vision impairment (36, 37). Psychophysical testing is one method of assessing distinct aspects of patients’ visual function and monitoring vision changes over time, even when patients are not consciously aware of impaired vision. It also provides an opportunity to determine when diabetes-induced deficits of vision, which range from mild impairment to blindness (38), first become apparent at a perceptive level. In diabetic retinopathy, many pathological cellular and molecular processes occur in the early stages of the disease, and may inflict subtle changes on retinal function before retinopathy is anatomically evident. Routine clinical evaluation of patients’ sensory capacity has the potential to expose deficits of vision that individuals may be unaware of in daily life. Several clinical studies have described a loss of visual function in diabetic patients in the very early stages of diabetes, and often in the absence of classically recognized symptoms of diabetic retinopathy (32, 39). These deficits include decreased hue discrimination and contrast sensitivity, delayed dark adaptation, abnormal visual fields, and decreased visual acuity (28, 37, 40–42). Spatial resolution and/or contrast sensitivity are two of the most commonly studied aspects of vision consistently altered by diabetes. Early reports suggest a high frequency of declining contrast sensitivity that is more strongly correlated with disease duration that with lack of metabolic control (26). A similar study detected significant visual dysfunction in 60% of patients with background retinopathy, but also in nearly 40% of patients with no detectable signs of vascular disease (43). Defects in color perception have also been reported in patients with minimal or no retinopathy, and involve both blue–yellow and red–green discrimination (26, 43). More recent research consistent with these findings has reported loss of color and contrast sensitivity in patients after a short duration of diabetes, and with no discernable signs of retinopathy, by fundus examination or fluorescein angiography, demonstrating a generalized loss of central vision and hue discrimination, as well as significant decreases in both static and dynamic spatial resolution (25, 44). Together, these changes suggest a broad impairment of retinal neuronal function.

The electroretinogram (ERG) is used to assess the electrophysiological response of the retina to various visual stimuli, and potentially provides a physiological basis for the loss of function observed in psychophysical studies. The ERG wave represents the characteristic sequence of neuronal activation that forms this response, reflecting the duration and amplitude of functional output throughout the retina. The a-wave, an initial negative deflection of the ERG signal, represents photoreceptor hyperpolarization resulting from the cessation of the photocurrent that flows continuously in the absence of light. The second major component of the ERG is the positive b-wave,

which is thought to originate from depolarization of rod–bipolar neurons and Müller glia. A series of small wavelets, termed oscillatory potentials (OPs), contribute to the ascending edge of the b-wave, and are attributed partially to amacrine cell activity. The OPs and the b-wave reflect the inner-retina function, and exhibit abnormalities in the form of decreased response amplitudes and delayed peak-times in patients with diabetes, even with good glycemic control and no detectable vascular lesions (29, 33, 34, 37, 45, 46). These data suggest that diabetes affects multiple neuronal subtypes throughout the retina, including photoreceptors, bipolar cells, and amacrine cells, while more advanced ERG techniques reveal similar abnormalities in ganglioncell responses. ERG deficits may contribute to loss of visual sensitivity in patients with diabetes, which ranges from mild to severe, and which includes reductions in overall acuity, dark adaptation, scotopic vision, contrast and contour sensitivity, and hue discrimination (25, 27, 42–44, 47).

Ganglion-cell function can be measured more specifically with the multifocal pattern ERG, which uses a complex stimulus pattern rather than a flash of light (30). In patients with diabetes but without overt vascular complications, ganglion-cell responses to mediumand large-grated stimuli have been reported to be reduced (31), and delayed in the macular region (30, 33). Less frequently, defects in photoreceptor response are reported in similar patient cohorts exhibiting impaired activation of the blue-cone system and increased a-wave latency (45, 47, 48) although other studies report no functional changes in the outer retina (34, 37). These data provide a physiological basis for decreased performance in psychophysical testing, and implicate diabetes in the disruption of the functional output of the various neurons and synaptic connections that mediate normal vision. In addition, this sensitive technique often detects the abnormal responses of retinal neurons in the absence of vascular lesions. A number of researchers suggest that ganglion-cell electrophysiology shows promise as a supplemental diagnostic tool usable in patients with subclinical diabetic retinopathy, correlating the underlying electrophysiological impairment with loss of vision (49–51), and currently the predictive capacity of the multifocal ERG is being investigated as a potential early diagnostic tool (52).

The application of ERG studies to experimental rodent models of diabetes has yielded data similar to those obtained from clinical studies, demonstrating retinal dysfunction in the first few weeks of diabetes. In these animals, OP amplitudes are consistently depressed and delayed throughout the first 12 weeks of diabetes (53–55), with abnormalities detected after only 2 days of hyperglycemia in one recent study (56). Additional work has demonstrated reductions in b-wave amplitude after 2–4 weeks of diabetes (57, 58). Interestingly, several studies report electrophysiological changes in the a-wave of diabetic rodents (53, 56, 57), while photoreceptor response is more variable in patients with diabetes. A degree of variability exists in rodents particularly due to differences in retinal pigmentation, sex, and strain. Despite discrepancies in reported amplitude and latency measures, however, these studies are in general agreement that diabetes compromises the functional integrity of the neural retina, resulting in both innerand outer-retinal dysfunction. The pairing of psychophysical and ERG data suggests a broad dysfunction of retinal neurons, involving both rod and cone photoreceptors, bipolar cells, amacrine cells, and ganglion cells. ERG abnormalities in rodents, as in humans, develop in the early stages of diabetes when the retina appears anatomically

normal (54, 57). Rodent models of diabetes provide clinically relevant ERG data with the added benefit of enabling investigation into potential mechanisms contributing to retinal dysfunction.

DIABETES INDUCES NEURODEGENERATION IN THE RETINA

It has long been suggested that diabetic retinopathy involves a neurodegenerative component (59). In the 1960s, both Wolter and Bloodworth identified degenerating or “pyknotic” neurons in postmortem retinas from human donors with diabetes (60, 61). Since then, several studies have investigated the effects of diabetes on retinal neurons, with the recognition that at some level, these elements must contribute to visual impairment in diabetic retinopathy, and may explain the electrophysiological abnormalities in humans and rodents with diabetes. Perhaps the most basic observation has been that elements of the retina, distinct from the vasculature, undergo apoptosis in experimental diabetes. An early study demonstrated retinal-cell apoptosis in rats after a short period of diabetes (62, 63). Work elaborating on this finding established that diabetes significantly increased retinal apoptosis, detected by TUNEL, resulting in a significant reduction in cell counts in STZ-diabetic rats (64). Since then, this finding has been confirmed in multiple rodent models of diabetes including the streptozotocin (STZ) rat, the STZ mouse, and the spontaneously diabetic Ins2Akita mouse (17, 65, 66). Additional studies have demonstrated degeneration and cell loss in multiple types of retinal neurons in both the inner and outer nuclear layers, including photoreceptors (57, 67), dopaminergic and cholinergic amacrine cells (65), and bipolar cells (19, 68–70). Consistent with the theory that diabetes induces neuroretinal apoptosis, proapoptotic molecules, including activated caspase-3 have been detected in retinas from humans and rodents with diabetes

(17, 21, 65, 66, 71, 72).

In addition to cell loss, morphometric analysis has identified significant thinning of the retina after varying durations of diabetes. The thickness of the inner and outer nuclear layers is reduced after 10 weeks of diabetes in mice and 24 weeks of diabetes in rats (64, 66, 67), suggesting a chronic depletion of neurons. Further, diabetes is associated with decreased nervefiber diameter and degeneration of the nerve fiber layer, which is likely due to a loss of ganglion cells or ganglion cell axons (73–75). Diabetes is associated with atrophy of retinal axons in humans (76,77), and the processes of retinal neurons, including nerve terminals, ganglion cell dendrites, centrifugal axons, and horizontal cell axons appear dystrophic in diabetic rodents (57, 67, 78, 79). One of the most consistent findings in retinas of diabetic rodents is a decreased thickness of the IPL, which contain synaptic connections between bipolar, amacrine, and ganglion cells. In a seminal publication, Barber et al. (1998) reported a 22% reduction in thickness of the IPL after 7.5 months in STZ-diabetes in rats (64). Subsequent work elaborating this finding reported a 9.9% decrease in the thickness of the IPL after 1 month of diabetes in STZ-diabetic Sprague-Dawley rats, and a 15% decrease after 1 month of STZdiabetes in Brown-Norway rats (57). More recently, a third study reported a 10% decrease in inner plexiform thickness of STZ-diabetic Sprague-Dawley rats after 6 months of diabetes (67). Although one study using STZ-diabetic mice did not observe inner plexiform thinning in the first 8 weeks of diabetes (66), characterization of the spontaneously diabetic Ins2Akita mouse revealed that after 22 weeks of hyperglycemia,

the IPL was reduced by 16.7 and 27% in the central and peripheral retina, respectively (17). Although the timing and severity of decreased thickness of the IPL varied in these reports, overall the data clearly demonstrate that diabetes induces a generalized neurodegeneration of the retina. While the changes summarized here are well-characterized, work to determine the molecular mechanisms leading to neuronal dysfunction and apoptosis is far from complete.

Reports of anomalous biochemical composition of retinal neurons indicate that the pathological processes which impact retinal function begin early in diabetes. One study suggested that diabetes upregulates whole-retina neuronal nitric oxide synthase (NOS), which was accompanied by an increase in the number of NOS-containing bipolar cells (69). In contrast, a study using a shorter duration of diabetes reported an early depletion of NOS-positive neurons that was sustained through later time points (80). Consistent with this finding, a recent study found a decrease in NOS-immunoreactive amacrine cells and a redistribution of nitric oxide precursors after a short period of diabetes (81). Tyrosine hydroxylase protein content is also reduced in dopaminergic amacrine cells concomitant with an overall decrease in brain-derived neurotrophic factor (82). Glutamate receptors and calcium-binding proteins are upregulated in the inner and OPL, as well as in ganglion, bipolar, and amacrine cells of one-month STZ-diabetic rats (83). Additional work also indicates that diabetes alters the retinal expression of neuronal transcription factors, ion channels, and GABA receptors (84). These studies reveal that several aspects of neuronal biochemistry are altered by diabetes.

Despite the potential that synaptic degeneration and impaired neurotransmission are likely contributing components of diabetes-induced retinal dysfunction and visual impairment, little is known concerning the effects of diabetes on retinal synapses. Several reports and comprehensive reviews, however, propose impaired synaptic transmission as a potential mechanism underlying functional deficits evidenced by abnormal ERG responses (83). Alterations in visual function precede the emergence of clinical diabetic retinopathy in humans, and significant depletion of retinal neurons and the development of vascular lesions, which often do not appear until 3–6 months in rodent models of diabetes (16). This implies that subcellular changes in retinal neurons may play a role in early retinal dysfunction in a manner similar to the pathological changes in cells of the vasculature that are apparent soon after the onset of diabetes (15, 18, 22, 68, 85, 86).

DIABETES ALTERS THE FUNCTION OF GLIAL CELLS

IN THE RETINA

Pathological changes in retinal glial cells occur with diabetes and likely impair their functions in neuronal and vascular support, which include providing nutritional support to neurons, processing glutamate and other neurotransmitters released from the synaptic cleft, and maintaining the blood–retina barrier. Diabetes disrupts retinal glutamate metabolism, a process particular to Müller cells in the mammalian retina. Early work demonstrated a 45% reduction in the conversion of glutamate to glutamine in rat retina after 3 months of STZ diabetes, which was accompanied by a 1.6-fold increase in total retinal glutamate (87). Glutamate also increases in the vitreous of patients with proliferative diabetic retinopathy (88). While the de novo synthesis of amino acids is unchanged, glutamate oxidation is significantly reduced in rat retina by diabetes.

Further, the activity and content of the enzyme that converts glutamate to glutamine, glutamine synthetase, are significantly decreased in retinas from diabetic rats (89). These data indicate an early failure of the Müller-cell glutamate metabolism, in which both the transamination of glutamate to alpha ketoglutarate and the amination of glutamate to glutamine are significantly impaired after a short period of experimental diabetes. Similar findings have been obtained from studies of cultured Müller cells isolated from retinas of control and diabetic rats. In these cells, the activity of the electrogenic glutamate transporter was dysfunctional after 4 weeks of diabetes, and was reduced by 67% after 13 weeks (90). More recently, it was demonstrated that a short period of diabetes increases immunoreactivity to D-aspartate, a nonmetabolisable substrate for the GLAST glutamate/aspartate transporter,in retinal Müller cells, and that these cells undergo gliosis in the early stages of diabetes (91).

In the healthy retina, glial fibrillary acidic protein (GFAP) is largely restricted to astrocytes, where it provides structural integrity and stability to the cytoskeleton of glial filaments (92). In diabetes, GFAP expression is reduced in astrocytes but is significantly induced, in Muller cells indicative of gliosis. Early studies demonstrated this phenomenon using rodent models of diabetes, including the streptozotocin (STZ)-induced diabetic rat and the spontaneously diabetic BB/Wor rat, and demonstrated a fivefold increase in Müllercell GFAP expression, concomitant with a decrease in astrocyte GFAP expression, that occurred progressively from 1 to 4 months of hyperglycemia (15,87). Additional work has confirmed this finding, demonstrating upregulated Muller cell GFAP content after short periods of diabetes ranging from 6 to 12 weeks (18,62,79,93,94), as well as after chronic hyperglycemia in rodents (19), in conjunction with decreased astrocytic GFAP. This finding has been confirmed in human retinal tissue from patients with approximately 10 years of diabetes, which demonstrates increased GFAP expression in Müller cells distributed throughout the entire retina, as well as whole-retina increases in GFAP content (95). The GFAP expression by Müller glia has been similarly reported to increase with disruption of the retinal pigment epithelium and the blood–retina–barrier, as well as in response to neuronal loss (96). Although the specific function of GFAP in retinal Müller cells remains to be fully identified, this protein is implicated in the stability of Müller-cell processes and the inner limiting membrane (97). The drastic redistribution of GFAP in retina after short durations of diabetes likely indicates the progression of retinal gliosis in response to the disease (98).

Increased production of VEGF by retinal glia is another consequence of diabetes (99). The VEGF induction is not only implicated in the aberrant growth and increased permeability of retinal blood vessels that contribute to the classic vascular symptoms of diabetic retinopathy (100), but also functions as a trophic factor for multiple types of retinal cells (101, 102). The VEGF mRNA and protein are expressed by neurons and Müller glia in mammalian and human retina (103), which has been attributed to the role of these cells in influencing maintenance of the blood–retina barrier. Growing evidence supports the role of retinal neurons and glia in increased VEGF production in both humans and rodents with diabetes (99, 104, 105). In retinas from human donors with proliferative and nonproliferative diabetic retinopathy that exhibit increased Müller-cell GFAP expression, VEGF and its inducible angiogenic cofactor and inducible NOS have been reported to increase in Müller glia (106, 107). Studies of diabetic-rat retina have demonstrated similar effects, with increased VEGF protein and mRNA occurring predominantly in the Müller cells (108). Due to the close interaction of Müller cells