Hypoglycemia Induces General Neuronal Death, Whereas Hypoxia and Glutamate Transport Blockade Lead to Selective Retinal Ganglion Cell Death In Vitro

Abstract

purpose. To examine the impact of experimental ischemia and interruption of glutamate transport on retinal neuronal cell, especially retinal ganglion cell (RGC), survival in vitro.

methods. Cell cultures were prepared from adult pig retinas and maintained under different experimental conditions of increasing hypoglycemia, environmental hypoxia (delayed postmortem period or atmospheric PO₂ <2%), or chemical hypoxia (potassium cyanide), or in the presence of glutamate transporter blockers l-trans-pyrrolidine-2,4-dicarboxylic acid (tPDC) and l(−)-threo-3-hydroxyaspartic acid (THA), or the glutamine synthetase inhibitor methionine sulfoximine (MS). After 48 hours, cells were returned to standard culture conditions and allowed to develop for 5 days, when they were fixed and immunostained with different retinal neuronal phenotypic markers.

results. Control normoxic cultures contained large numbers of immunocytochemically identified photoreceptors (PRs), bipolar cells (BCs), amacrine cells (ACs), and RGCs after 7 days in vitro. A 24-hour postmortem delay before culture led to significant reductions in all types (40%–70%), proportionately greater in ACs and RGCs. Lowering of sugar levels also led to increased losses in all cell types, whereas potassium cyanide treatment deleteriously affected only ACs and RGCs. Ambient hypoxia led to consistent reductions only in the number of RGCs, which were exacerbated by addition of high concentrations of glutamate. Inclusion of glutamate receptor antagonists had a partial protective effect against RGC loss. Treatment with tPDC and THA also led to selective RGC death, but MS had no effect on any cells.

conclusions. Different components of the ischemic pathologic process (hypoxia, hypoglycemia, glutamate transport failure) lead to distinctly different patterns of neuronal loss in adult retina in vitro. RGCs are especially vulnerable, corresponding to their in vivo susceptibility. These data may suggest neuroprotective strategies for limiting retinal damage during ischemia.

Much evidence suggests that alterations and imbalances in metabolism of the excitatory neurotransmitter glutamate (Glu) play a pivotal role in central nervous system (CNS) disease. ^{1 2 3} Under normal physiological conditions, extracellular levels of this amino acid are maintained at low levels through active uptake by glutamate transporters (GluT). ⁴ At present, five distinct Na²⁺-dependent GluTs have been identified, localized to either the neuronal or glial processes surrounding glutamatergic synapses: GLAST (EAAT-1), GLT-1 (EAAT-2), EAAC-1 (EAAT-3), EAAT-4, and EAAT-5. ^{4 5} Maintaining low extracellular Glu is essential for returning glutamate receptors (GluR) to their inactive state to permit physiological synaptic function and for avoiding pathologic overstimulation of GluR, which leads to excessive depolarization and massive entry of Ca²⁺ into the neuron, often culminating in death. This pathologic phenomenon, termed excitotoxicity, is frequently observed in various disease states, including congestive heart failure, stroke, neurologic disorders, and diabetes. ^{1 3} Additional injurious aspects of ischemia-reperfusion include reductions in oxygen and sugar delivery, reduced clearance of waste products, and build-up of free radicals. ^{6 7}  

The retina, a peripherally located region of the CNS, exhibits all the features we have described. Glu is the major excitatory neurotransmitter, four retinal GluT (GLAST, GLT-1, EAAC-1, and EAAT-5 ^{8 9 10} have been identified, and excitotoxicity is thought to be a leading factor in neuronal death in retinal ischemia, diabetic retinopathy, and glaucoma. ^{11 12 13} Much data have been obtained on the molecular mechanisms underlying the pathogenesis of excitotoxic neuronal death, using the retina as a model. ^{11 14 15} Ionotropic GluR are defined pharmacologically as N-methyl-d-aspartate (NMDA)–, kainic acid (KA)–, or α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)–preferring subtypes. Especially NMDA GluRs have been implicated as principal mediators of excitotoxic damage, ^{11 16} although there is also evidence that AMPA-KA GluRs play important roles. ¹⁵ Many of these data have been compiled by using direct application of Glu or one of its nonphysiological agonists, either in vivo or in vitro. ^{14 17 18} Additional support for Glu imbalances underlying retinal ischemic damage has come from experimental induction of ischemia in vivo accompanied by treatment with GluR antagonists. ^{19 21} The approach of studying the impact on the sensitive neuronal populations through impeding Glu transport has been less studied, although recent reports indicate such treatments lead to Glu release ²² and retinal ganglion cell (RGC) damage. ²³  

We have developed a cell culture system derived from adult mammalian retina in which all major neuronal types are represented in approximately the same proportions as in vivo. ^{24 26} This culture model has recently been used to investigate the effects on retinal neurones of manipulating Glu levels within the medium, and these treatments have led to specific, dose-dependent losses in RGCs. ²⁷ In the present study, we have examined the effects on identified retinal neurons of mimicking ischemia through chemical or environmental manipulation. The data show that paradigms involving hypoglycemia led to widespread cell loss, whereas those involving hypoxia induced more specific damage to amacrine cells (ACs) and especially RGCs. Addition of GluR antagonists had only limited protective effects, suggesting alternative routes of cell death, such as free radical damage. 

Materials and Methods

Tissue Collection and Cell Culture

Porcine eyes were obtained from a local abattoir, and globes were removed from the animals within 5 minutes of death. The eyes were transported to the laboratory in cold CO₂-independent Dulbecco’s modified Eagle’s medium (DMEM-CO₂) on crushed ice (average delay before dissection: 2 hours). Retinal cultures were prepared as we have described previously. ^{26 27} Retinas were isolated into fresh DMEM-CO₂ by circumferential sectioning of the cornea and removal of the anterior chamber. Major blood vessels were excised, and retinas were then chopped into small fragments; washed in Ringer’s solution without Ca²⁺ or Mg²⁺, supplemented with 0.1 mM EDTA; and incubated in 0.5 ml 0.2% activated papain (Sigma-Aldrich, Saint Quentin Fallavier, France) in the same buffer for 20 minutes at 37°C. The tissue was dissociated by repeated gentle trituration and seeded in DMEM-Ham’s F12 (except the experimental hypoglycemia series that was maintained in DMEM only; Life Technologies SARL, Cergy Pontoise, France) supplemented with 5% fetal calf serum (FCS; Life Technologies) and penicillin-streptomycin (10 IU/l) into 24-well tissue culture plates containing coverslips previously coated with polylysine (2μ g/cm² for 2 hours) followed by laminin (1μ g/cm² overnight; both from Sigma-Aldrich). Microscopic inspection using trypan blue exclusion showed that viability was more than 90% at the time of seeding. Seeding was performed at an initial density of 5 × 10⁵ viable cells/cm², and cells were incubated at 37°C in a humidified atmosphere of 5% CO₂-95% air, unless otherwise stated. 

Postmortem Delay

Under standard conditions, cultures were established immediately after eyes arrived from the abattoir (maximum time elapse between enucleation and cell seeding, 2 hours). In one set of experiments, eyeballs were left intact and stored in DMEM-CO₂ at 4°C for 24 hours before dissection and cell seeding. Cell viability was adjusted at the time of seeding to control for losses already incurred by the 24-hour postmortem period. Cultures were maintained for 7 days in vitro before fixation and immunolabeling. 

Induction of Hypoglycemia In Vitro

Cells were seeded in DMEM in which glucose levels were reduced to 20%, 5%, and 1% of the normal medium concentration—4.5 mg/ml (25 mM), reduced to 0.9 mg/ml (5 mM), 0.23 mg/ml (1.25 mM), and 0.05 mg/ml (0.25 mM)—controlled by mixing appropriate volumes of high-glucose DMEM with glucose-free DMEM. Cells were maintained under these regimens for 48 hours and then the medium changed to normal DMEM and the cultures maintained for a further 5 days in vitro. 

Chemically or Environmentally Induced Ischemia In Vitro

Immediately after cell seeding, 1 mM potassium cyanide (KCN) was added to retinal cells (50 μl/well). Control wells received 50 μl medium alone. After 48 hours’ treatment, cultures were rinsed twice with serum-free DMEM, replenished with normal DMEM and maintained in vitro for a further 5 days (total of 7 days). 

Immediately after cell seeding (defined as 0 days in vitro), plates were transferred to a controlled-atmosphere incubator in which oxygen levels were reduced to less than 1% normal partial pressure (i.e., a 5% CO₂-94% N₂-1% O₂ mix). In some experiments, in addition to maintenance under hypoxic conditions cultures were incubated with the NMDA GluR antagonist (+)-5-methyl-10,11-dihydro-5H-dibenzo-(α,β)-cyclohepten-5,10-imine maleate (MK-801; 10 μM) or the AMPA-KA GluR antagonist 6-cyano-7-nitroquinoxaline-2′3-dione (CNQX; 50 μM) alone, exposed to high concentrations of Glu (1 mM), or exposed to both 1 mM Glu and either CNQX or MK-801. Cultures were maintained under hypoxic and excitotoxic conditions for 48 hours, after which the media were changed to fresh DMEM-5%FCS and the plates returned to a normoxic (5% CO₂-95% air) atmosphere for a further 5 days. The cultures were then fixed in 4% paraformaldehyde and processed for immunocytochemistry. 

Inhibition of Glutamate Transport and Metabolism In Vitro

Cultures were treated with two different competitive Na-dependent GluT blockers: l-trans-pyrrolidine-2,4-dicarboxylic acid (tPDC) and l(−)-threo-3-hydroxyaspartic acid (THA; Tocris Cookson, Ltd., Bristol, UK), at 25 and 50 μM each, added at 0 days in vitro for 48 hours. The cultures were then returned to normal DMEM and maintained for a further 5 days before fixation. 

The glutamine synthetase inhibitor methionine sulfoximine (MS; 500μ M; Tocris Cookson, Ltd.) was added to cultures at the day of seeding for 48 hours, after which cells were gently washed twice, replenished with normal culture medium, and allowed to grow for a total of 7 days. 

Immunocytochemical Characterization of Adult Pig Retinal Neurons In Vitro

After 7 days in vitro, cells were washed twice in phosphate-buffered saline (PBS; pH 7.3) and fixed with 4% paraformaldehyde in PBS for 15 minutes at room temperature. Cells were rinsed twice in PBS, permeabilized with 0.1% Triton X-100 for 5 minutes, and incubated in PBS containing 0.1% bovine serum albumin, 0.1% Tween 20, and 0.1% NaN₃ (buffer A) for 15 minutes. Different primary antibodies directed against neuronal cell type–specific antigens were chosen for this study, as used and characterized previously ²⁷ : anti-neurofilament (NF) 68-kDa subunit monoclonal antibody and anti-NF 200-kDa subunit polyclonal antibody (Sigma-Aldrich) to RGC and horizontal cells ²⁸ ; anti-neuron specific enolase (NSE) polyclonal antibody to total neuron ²⁹ ; protein gene product 9.5 (PGP9.5; ubiquitin decarboxylase) polyclonal antibody to RGCs and ACs ³⁰ (both from Chemicon Intlernational, Temecula, CA); anti-protein kinase Cα monoclonal antibody (Sigma-Aldrich) to rod bipolar cells (BCs) ³¹ ; and anti-arrestin polyclonal antibody (generous gift of Yvonne de Kozak, Institut National de la Santé et de la Recherche Médicale [INSERM] U450, Paris, France) to rod and cone photoreceptors (PRs). ³²  

For the purposes of the present study, RGCs were identified through their intense NF immunoreactivity and distinctive morphology (large, rounded cell body, laterally displaced nucleus, and extensive neurites). ²⁷ Subsequent to fixation and permeabilization, coverslips were incubated with primary antibodies (10 μg/ml) for 2 hours, washed six times with PBS buffer, and exposed for 1 hour to goat anti-mouse IgG/fluorescent dye (Alexa 488; 10 μg/ml; for monoclonal primary antibodies) or goat anti-rabbit IgG/fluorescent dye (Alexa 594; 10 μg/ml; for polyclonal primary antibodies; both from Molecular Probes, Eugene, OR). Nuclei of cultured cells were stained with 4,6-diaminodiphenyl-2-phenylindole (DAPI; 10 μg/ml; Sigma-Aldrich), incubated with fluorescent secondary antibodies. Coverslips and sections were finally washed thoroughly, mounted, and viewed under a photomicroscope (Optiphot 2; Nikon, Tokyo, Japan) equipped with Nomarski differential interference optics and epifluorescence illumination. 

Assessment of Cellular Injury In Vitro

Neuronal cell survival was assessed, as described previously, ^{27 32} by counting the immunolabeled cells after 7 days in vitro, whereby cells exhibiting continuous plasma membranes with no signs of vacuolation and well-developed neuritic processes were scored as viable at the time of fixation. Cells showing perikaryal swelling, nuclear pyknosis, or fragmentation, as visualized by DAPI staining, were considered dead and were excluded from the counts. Data are expressed as total cell number per coverslip (total numbers and ranges of NF-immunopositive RGCs in control wells for each experimental series are given in the respective figure legends) or as mean number of cells per optical field (NSE-immunopositive neurons, PGP9.5-immunopositive ACs, protein kinase Cα–immunopositive BCs, and arrestin-immunopositive PRs (total numbers and ranges of each neuronal type in control wells for each experimental series are given in the respective figure legends). For cells other than RGCs a minimum of 20 randomly chosen optical fields per coverslip were counted under ×20 objectives. Each experiment was repeated separately a minimum of three times with triplicate treatments per experiment. 

Statistical Analysis

Statistical analysis was performed by computer with an analysis of variance (ANOVA) software package followed by the Tukey multiple comparison test. P < 0.05 was considered statistically significant. For the statistical treatment of the 24-hour postmortem delay, Student’s t-test was used. These statistical treatments were performed on raw data, whereas graphic representation of results was normalized to control values (expressed as 100%) for each experimental series, to minimize variation. 

For studies on RGC neurite morphology after each experimental treatment, between 100 and 200 individual immunolabeled RGCs were examined under high-power microscope optics, and the length of their longest neurites determined with a calibrated graticule inserted in the microscope eyepiece. Data are expressed relative to cell body diameters, placed in bins of increasing size, and calculated as frequency histograms. 

Results

Effect of Increased Postmortem Delay and Hypoglycemia

Delaying culture preparation by 24 hours after enucleation led to large decreases in neuronal survival after 7 days in vitro. NSE-immunopositive neurons were reduced by 40%, reflected in decreases in specific populations of PRs (45% loss), BCs (30% loss), and especially ACs (70% loss) and RGCs (60% loss; Fig. 1A ). 

Progressive reduction of glucose concentrations in the culture medium under normoxic conditions also led to progressive increases in neuronal loss. Glucose levels as high as 5 mM induced statistically significant decreases in PRs compared with control levels, whereas further reduction to 1 and 0.2 mM glucose led to significant reductions in every neuronal type, in order of severity: ACs (40% loss), RGCs (50% loss), BCs (60% loss), and PRs (70% loss; Fig. 1B ). 

Effects of Chemical and Environmental Hypoxia

Inclusion of 1 mM KCN in the culture medium at 0 days led to significant reductions specifically in AC and RGC cells, which exhibited 25% and 35% losses, respectively. No other neuronal population examined demonstrated detectable decreases (Fig. 2A) . 

Maintenance of cultures under environmental hypoxia (final PO₂ ∼1%) for 48 hours also led to specific RGC losses after 7 days in vitro. RGCs decreased by more than 30% (Fig. 2B) , whereas, as observed in KCN-treated cultures, neither PRs nor BCs showed detectable losses. Statistically significant decreases were also not observed in ACs under these conditions. To see whether hypoxia-induced death of RGCs involves excitotoxic stimulation of GluR, the specific antagonists CNQX or MK-801 were included in the culture medium during hypoxia. MK-801 reduced hypoxic RGC loss by approximately 40%, whereas CNQX had no effect (Fig. 3) . An alternative approach to determine whether Glu imbalance is involved in hypoxic neuronal death was to increase exogenous Glu concentrations simultaneously with increased hypoxia. Whereas there was still no effect on ACs, PRs, or BCs, RGC loss was aggravated by combined hypoxia-Glu treatment, with nearly 60% reductions in number of surviving cells. Simultaneous addition of MK-801 again led to a partial block of cell loss, whereas CNQX had no effect (Fig. 3) . 

The neurite morphology of those RGCs surviving initial exposure to hypoxic conditions was markedly modified compared with that observed in control normoxic cultures. Whereas control RGCs after 7 days in vitro exhibited long, smoothly branching neurites (Fig. 4A) , residual posthypoxic RGCs revealed shorter, stunted neurites (Fig. 4B) . Addition of either CNQX or MK-801 had no effect on the morphology of hypoxic RGC neurite outgrowth (Fig. 4C) , whereas combined treatment with hypoxia and 1 mM Glu produced further shortening of RGC neurites (Fig. 4D) . 

Quantification of neurite morphology in RGCs remaining after the experimental treatments confirmed these immunocytochemical data: frequency distribution analysis of major neurite length in RGCs revealed that whereas more than 75% of control untreated RGCs possessed a major neurite exceeding 10 cell body diameters in length, this value was only 3% after hypoxia, with the majority (>60%) of RGCs under these conditions exhibiting neurites measuring between 3 and 6 cell body diameters (Fig. 5) . As indicated earlier, addition of GluR antagonists during hypoxic treatment did not modify the overall distribution compared with hypoxia alone, whereas simultaneous exposure to hypoxia and elevated Glu further compromised neurite length, with almost 40% RGCs exhibiting neurite length less than two cell body diameters (Fig. 5) . 

In contrast, the neurite morphology of posthypoxic PRs or BCs did not differ from that of untreated control cultures (Figs. 6A 6B 6C 6D) , nor did that exhibited by ACs, despite the reduced number of cells (Figs. 6E 6F) . 

Effect of Glutamate Transport Inhibition

Because the studies on hypoxic neuronal death suggested partial involvement of excitotoxic mechanisms, the effect of Glu imbalances on neuronal survival in vitro was examined by addition of the competitive GluT blockers tPDC and THA. Whereas addition of either antagonist did not reduce PRs, BCs, or ACs (Fig. 7A) , RGCs were adversely affected by both GluT inhibitors, more than 40% and 23% losses after tPDC and THA treatment, respectively (Fig. 7B) . Similar to hypoxic exposure, surviving RGCs displayed reduced neurite outgrowth after maintenance in tPDC or THA for 48 hours in vitro (Figs. 8A 8B) . Frequency distribution analysis of maximal neurite length revealed a strong downward shift from control values, more than 75% RGCs with neurites exceeding 10 cell body diameters, toward more than 40% with neurites under two cell body diameters (tPDC) or more than 60% with neurites between 3 and 6 cell body diameters in length (THA; Fig. 5 ). 

Finally, one more approach to experimentally manipulating Glu metabolism was through blockade of intracellular enzymatic degradation of Glu to glutamine by the underlying Müller glia. Addition of the glutamine synthetase inhibitor MS under normoxic conditions did not produce detectable cytotoxicity in any immunologically identified cell type, including RGCs (Fig. 7B) . 

Again as seen for hypoxia, neither PR nor AC morphology was altered (Figs. 8C 8E) , whereas GluT blockade led to severe reduction in neurite length in BCs (Figs. 8D) . Neurite-bearing BCs were reduced by more than 40% by tPDC treatment compared with control cultures (Fig. 6A) . 

Discussion

The present study has attempted to highlight particular pathways involved in retinal neuronal death after exposure to distinct aspects of ischemic insult. Use of several complementary approaches suggests that trauma involving hypoglycemia results in more global neuronal loss, whereas treatments involving hypoxia and perturbation of Glu uptake induce more specific deficits, especially in RGC number and morphology. 

Numerous studies examining experimentally induced retinal ischemia in animals have demonstrated partial protection of vulnerable neuronal populations by pretreatment with GluR blockers of both the AMPA and NMDA subtypes. These studies have been performed mainly through raising intraocular pressure (IOP) ^{19 20 33 34} or inducing photothrombosis, ^{35 36} leading to severe temporary or permanent loss of blood flow. In these paradigms, both oxygen and glucose supply are diminished. Analogous studies have also been performed in vitro, using retinal tissue of embryonic or newborn origin, and have again shown that both NMDA and non-NMDA components are implicated in retinal ischemic death. ^{14 17 37 38}  

Human cadaveric nervous tissues are also subjected to ischemic conditions, the severity of which presumably increases with increasing postmortem delay times. Because human donor tissue may be a valuable resource for therapeutic approaches to limit retinal degenerations such as retinitis pigmentosa or glaucoma, it was important to see whether sufficient cells could survive such episodes and by what means their survival could be optimized. We have published semiquantitative data showing considerable survival of human photoreceptors after death and of porcine photoreceptors subjected to equivalent postmortem delays. ²⁴ The present study describes the capacity for postmortem survival and maintenance in vitro of each retinal neuronal population, as well as suggesting which are the most important variables for enhancing survival. Not unexpectedly, RGCs survive the least well, because they are highly vulnerable to ischemic stress, whereas PRs and BCs exhibit less decrement. 

Tissue culture approaches facilitate separation of hypoglycemic and hypoxic episodes. The effects of glucose deprivation alone and combined with excitotoxic treatments on embryonic retinal cell cultures have shown that the reduction in cellular energy levels compromise neuronal survival, particularly in the presence of excess Glu. ³⁹ At least in dissociated cultures of chick retina, deprivation of glucose by itself did not lead to abundant cell death, ⁴⁰ although other investigators have observed neuronal swelling and cell loss in response to glycolytic inhibition. ⁴¹ Hypoglycemia was found to lead to increased production of reactive oxygen species, and accordingly neuronal degradation was not affected by GluR antagonists, nitric oxide synthase inhibitors, or Ca²⁺ channel blockers, whereas cell viability was improved by pretreatment with antioxidants such as vitamin E. ⁴²  

Our data suggest that prolonged hypoglycemic events (48 hours) lead to generalized neuronal damage under in vitro conditions and may be an important factor in neuronal loss during postmortem delay. Loss of cellular energy levels and membrane potential should be aggressive in neurons, irrespective of their neurotransmitter phenotype. The PRs were slightly more sensitive than other neuronal populations, perhaps correlating with their known high energy demands. ^{43 44} In support of this observation, it has been shown that readdition of glucose, ⁴⁵ or even of lactate, ^{46 47} can significantly prevent retinal damage from hypoglycemia. It is important to note that neuronal glucose utilization in vitro is significantly higher than that in vivo, and control media used in these prior studies contained much greater than physiological concentrations of glucose (25 mM versus 5.5 mM, respectively). The treatment with 20% control glucose actually corresponds to near-physiological levels in vivo, although PR damage is observed under these conditions. Hence, these data obtained in monolayer culture are not directly transposable to the in vivo situation, but may be more appropriately modeled using explanted retina. ^{14 44 45}  

Several studies using cell or explant cultures derived from newborn or embryonic retina subjected to hypoxia alone have demonstrated partial neuroprotection with NMDA antagonists and complete protection when NMDA and non-NMDA antagonists are coapplied. ^{14 17 34 41} Many studies have observed increased susceptibility to ischemic stress in inner retinal neurons, especially RGCs. ^{14 18 48 49} Much evidence indicates this selective toxicity is due to the expression in these cells of abundant GluR, ^{50 51 52 53} although additional reasons may include differential expression of cell-cycle–related genes, such as p53 ⁵⁴ and cyclin D1. ⁵⁵ Selective death of RGCs through hypoxic stress was also observed in the adult retinal culture system used in the present study, reflecting conservation of selectivity under in vitro conditions. 

That inclusion of Glu antagonists could partially prevent hypoxic death supports the hypothesis that damage results from Glu imbalances, due presumably to failure of energy-dependent Glu uptake by the underlying Müller glia, which normally rapidly clear Glu from the extracellular space. ²⁹ This is reflected in the largely similar results (both on RGC survival and neurite collapse) obtained in the presence of GluT inhibitors, which would also result in high extracellular Glu levels. However, the data suggest there may also be additional mechanisms: First, GluR antagonists only partially prevent RGC loss due to hypoxia (approximately 30% improvement); and second, addition of Glu itself to hypoxic cultures further reduces cell survival, indicating either that hypoxia activates only a subpopulation of GluR and that additive effects of exogenous Glu are due to increased recruitment of GluR, or that additional pathways are activated during hypoxia. Numerous other pathways have been implicated in RGC death under excitotoxic or ischemic conditions, including inhibitory transmitters, ⁵⁶ free radical generation, ^{57 58 59 60} lipid peroxidation, ⁶¹ nitric oxide, ⁶² and, very recently, tumor necrosis factor-α. ⁶³  

Furthermore, numerous neuroprotective strategies other than direct antagonism of GluR have been tested and demonstrated to be at least partially effective in preventing hypoxic or excitotoxic RGC damage: neurotrophic factors, ^{64 65} dopamine, ⁶⁶ preconditioning with subtoxic stress, ^{67 68} and calcium channel blockers. ^{48 69} The treatment paradigms used in our studies were quite severe, with all treatments being conducted for 48 hours and at maximum levels of stress (i.e., dose–response curves for O₂ tension were not performed), and it is therefore possible that less extreme insults would have responded more fully to GluR blockade. But it should be noted that other neurons were not affected detrimentally and thus that generalized cell death did not occur. Surprisingly, ACs were also relatively insensitive to hypoxia and GluT blockade, although they constitute a population vulnerable to ischemic insult in vivo. 

Addition of the GluT inhibitors used in these studies, tPDC and THA, affects directly three major populations, the Müller glia expressing GLAST, ⁹ and the PRs and bipolar cells expressing GLT-1. ⁸ Although these pharmacologic agents are relatively nonspecific competitive substrates for both GLAST and GLT-1 and may block additional retinal GluT, previous studies indicate that GLAST represents the major GluT activity within the retina, ⁷⁰ suggesting that Müller glia constitute the primary lesion site under these experimental conditions. Furthermore, application of pharmacologic blockers (tPDC and dihydrokianate) or antisense oligonucleotides specific for GLAST in the rat retina in vivo also leads to significant RGC death, ²³ and GluT levels are reduced in glaucomatous retinas compared with levels in age-matched control retinas. ⁷¹  

In contrast, a different study, in which in vivo administration of GLAST antisense oligonucleotides was used, failed to demonstrate retinal damage, whereas photofunction was perturbed. ⁷² Despite expressing GLT-1, PRs do not exhibit obvious signs of stress when treated with these drugs, indicating that these effects are not simply due to toxic overloading of cells. GLT-1 is also expressed by subpopulations of BCs, ⁸ and in the studies described herein we observed clear effects of GluT inhibition on neurite formation and stability in rod BCs. Such abnormal morphologies were not seen under hypoxic stress, and we do not know whether other subtypes of BCs are similarly affected after GluT blockade. MS, an inhibitor of glutamine synthetase, has been shown previously to exacerbate retinal excitotoxic cell death, ⁷³ although it can also have neuroprotective effects through maintaining glial intracellular glutathione levels. ⁷⁴ In our system, there were no overt effects on any retinal neuronal type of inhibiting glutamine synthetase activity. 

In conclusion, whereas hypoglycemia seems to lead to generalized neuronal deficits, hypoxia and GluT interruption induce more specific cellular lesions in RGCs in vitro. These latter observations resemble closely the effects elicited by direct Glu application in the same model ²⁷ and support the involvement of Glu imbalances in RGC toxicity. These data suggest maintenance of human cadaveric retinal tissue in media and that ensuring the supply of glucose and blockade of GluR would favor extended neuronal survival for subsequent application in ocular cell therapy. The rationale of targeting Müller glia for therapeutic approaches designed to limit toxic downstream consequences of Glu imbalance has been proposed previously, ^{23 73} and our data are in support of such strategies. Future studies will focus on the effects of GluT perturbation on glial integrity and metabolism. 

 Supported by Ciba-Vision/Novartis (XL) and Fédération des Aveugles de France, INSERM and ULP.

 Submitted for publication January 24, 2001; revised April 30 and June 18, 2001; accepted June 27, 2001.

 Commercial relationships policy: E, F (GNL), F (all others).

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked“ advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

A 24-hour postmortem delay and hypoglycemia led to generalized neuronal death in adult retina after 7 days in vitro. (A) Delaying the dissection and seeding of cells into culture by 24 hours (striped bars) led to significant reductions in all measured neuronal populations compared with control cultures seeded within 2 hours of enucleation (open bars). (B) Progressive lowering of sugar levels in vitro led to increasing losses in all neuronal populations compared with control cultures grown in DMEM-Ham’s F-12 containing 25 mM glucose (open bars). At concentrations as low as 20% glucose (or 5 mM, vertical stripes) after 7 days in vitro, PRs were significantly fewer than in control cultures; at 5% (1.25 mM, horizontal stripes) and 1% (0.25 mM, diagonal stripes), cell loss was severe in all cases examined. All values are given as means ± SD, normalized to untreated control cultures (taken as 100%). Absolute number of cells counted in control cultures (mean, range) for 24-hour postmortem effects: ACs (130, 81–157 per microscopic field); BCs (33, 18–44 per microscopic field); neurons (514, 324–754 per microscopic field); PRs (368, 178–586 per microscopic field); RGCs (191, 108–314 per entire coverslip). Absolute number of cells counted in control cultures (mean, range) for hypoglycemic effects: ACs (155, 88–246 per microscopic field); BCs (39, 27–53 per microscopic field); PRs (454, 268–591 per microscopic field); and RGCs (203, 105–298 per entire coverslip). AC, amacrine cells; BC, rod bipolar cells; PR, photoreceptors; RGCs, retinal ganglion cells. Students t-test for 24-hour postmortem data, ANOVA and Tukey multiple comparison test for hypoglycemia data: *P < 0.05,** P < 0.01, ***P < 0.001.

Figure 2.

 

Chemical and environmental hypoxia led to specific losses in AC and RGCs in adult retina after 7 days in vitro. (A) Incubation of cultures in potassium cyanide did not significantly affect PRs, BCs or total neurons, but led to a decrease in ACs and especially RGCs. (B) Maintenance of cultures in hypoxic atmosphere for 48 hours produced similar results, with no effects on PRs or BCs, and statistically significant decreases in RGCs. Under hypoxic conditions, there were no longer any statistically significant deficits in AC numbers. All values are expressed as means ± SD, normalized to untreated control cultures (taken as 100%). Absolute number of cells counted in control cultures (mean, range) for KCN’s effects: ACs (144, 81–200 per microscopic field); BCs (28, 19–48 per microscopic field); neurons (436, 260–644 per microscopic field); PRs (310, 146–590 per microscopic field); and RGCs (147, 108–204 per entire coverslip). Absolute numbers of cells counted in control cultures (mean, range) for hypoxic effects: ACs (119, 55–196 per microscopic field); BCs (23, 16–29 per microscopic field); PRs (255, 172–328 per microscopic field); RGCs (228, 110–370 per entire coverslip). ANOVA and Tukey multiple comparison test: *P < 0.05.

Figure 3.

 

Glu antagonists partially rescued RGCs from hypoxic death, and additional exogenous Glu exacerbated hypoxic RGC death. As in Figure 2 , hypoxia alone led to ∼30% decreases in RGCs. Inclusion of the AMPA-KA GluR antagonist CNQX did not alter this count, whereas the NMDA GluR antagonist MK-801 rescued ∼30% RGCs that would have died. Addition of 1 mM Glu together with hypoxia led to greater losses than either treatment alone (see Ref. ²⁷ for 1-mM Glu values). As before, this reduction was not altered by CNQX but was partly improved (∼35% rescue) by MK-801. All data are means ± SD, normalized to untreated control counts (taken as 100%). Absolute number of cells counted in control cultures (mean, range) for GluR inhibition effects: ACs (158, 152–164 per microscopic field); BCs (32, 30–35 per microscopic field); PRs (123, 90–190 per microscopic field); and RGCs (252, 235–269 per entire coverslip). ANOVA and Tukey multiple comparison test: *P < 0.05,** P < 0.01 compared with untreated control cultures.

Figure 4.

 

Hypoxia damaged neurite morphology of surviving RGCs in adult retina in vitro. Compared with typical RGCs in control normoxic cultures after 7 days, which exhibited very long, smooth neurites (A), hypoxic exposure led to marked shortening of neurites in the remaining RGCs (B). Neither CNQX (not shown) nor MK-801 (C) had any reproducible beneficial influence on neurite growth. Combined hypoxia and Glu induced severe retraction of neurites in the surviving RGC population (D). Scale bar, 100 μm.

Figure 5.

 

Frequency distribution of RGC neurite length after different experimental treatments. Between 100 and 200 randomly selected individual RGCs per treatment were scored for the length of their longest neurites, expressed as multiples of cell body diameter. Neurite lengths were grouped into bins of increasing size and calculated as percentage of RGCs exhibiting neurites of given length. Control cultures (top left) contained mostly RGCs with very long neurites, with less than 9% total RGCs possessing neurites under six cell body diameters long. Cultures exposed to hypoxia (HYP) showed drastic shifts in neurite length distribution, with only 3% having very long processes and more than 75% under six cell body diameters. Addition of GluR antagonists did not modify the hypoxic profile (HYP/MK-801), but simultaneous exposure to hypoxia and elevated Glu further reduced average neurite length (HYP/GLU) nearly 40%, or under two cell body diameters. Treatment with the GluT inhibitors tPDC and THA also produced dramatic reductions in neurite length, especially tPDC, which led to more than 40% RGCs with very short processes, and THA, with more than 80% RGCs possessing neurites of six cell body diameters or less in length.

Figure 6.

 

Hypoxia did not affect morphology of other retinal neurons after 7 days in vitro. Photoreceptors immunolabeled with arrestin antibody appeared similar, whether maintained in normoxic (A) or hypoxic (B) conditions, appearing as rounded cells with short beaded neurites; BCs also were indistinguishable whether maintained in normoxic (C) or hypoxic (D) conditions, exhibiting BC bodies with clear nuclei and short neurites; PGP9.5-immunopositive ACs were similar in normoxic (E) or hypoxic (F) cultures, showing rounded cell bodies and long, thin neurites. Scale bar, 20 μm.

Figure 7.

 

GluT blockade led to specific losses in RGCs in adult retina after 7 days in vitro. (A) PRs and ACs were unaffected by addition of either tPDC (25 μM) or THA (50 μM) to cultures; although BCs were also unchanged compared with control cultures, the percentage of BCs bearing neurites was significantly decreased. Neurite-bearing cells were scored as those possessing at least one process at least equal to or greater than cell body diameter. (B) RGCs were significantly reduced by addition of tPDC (25 μM) or THA (50 μM). Addition of either GluT inhibitor to retinal cultures led to significant losses in RGCs after 7 days in vitro. Addition of the glutamine synthetase inhibitor MS (500 μM) had no affect on survival of RGCs or other neurons. All data are means ± SD normalized to untreated control cultures (taken as 100%). ANOVA and Tukey multiple comparison test *P < 0.05. BC+, neurite-bearing BCs; BC−, cells without neurites.

Figure 8.

 

GluT blockade affected morphology of RGCs and BCs in adult retina after 7 days in vitro. Compared with control cultures (Fig. 4A) , both tPDC (A) and THA (B) induced shortening of RGC outgrowths. PRs (C) and ACs (E) resembled corresponding cells in control cultures, whereas BCs (D) showed a drastic loss in neurites (compare with Fig. 5C ). Scale bar, (A, B) 100 μm; (C–E) 20μ m.

The authors thank Abdel Jellali and Valérie Forster for technical help and Frédéric Stockel for expert photographic assistance.