Reentrant spiral waves of spreading depression cause macular degeneration in hypoglycemic chicken retina

Abstract

Spreading depression (SD), a slow diffusion-mediated self-sustained wave of depolarization that severely disrupts neuronal function, has been implicated as a cause of cellular injury in a number of central nervous system pathologies, including blind spots in the retina. Here we show that in the hypoglycemic chicken retina, spontaneous episodes of SD can occur, resulting in irreversible punctate lesions in the macula, the region of highest visual acuity in the central region of the retina. These lesions in turn can act as sites of origin for secondary self-sustained reentrant spiral waves of SD that progressively enlarge the lesions. Furthermore, we show that the degeneration of the macula under hypoglycemic conditions can be prevented by blocking reentrant spiral SDs or by blocking caspases. The observation that spontaneous formation of reentrant spiral SD waves leads to the development of progressive retinal lesions under conditions of hypoglycemia establishes a potential role of SD in initiation and progression of macular degeneration, one of the leading causes of visual disability worldwide.

Current research has revealed a number of molecular mechanisms underlying the pathophysiology of macular degeneration (1), a disease that involves damage to the central retina, where the higher density of photoreceptors results in maximal visual acuity (2). Macular degeneration is a leading cause of visual disability (1, 2). However, the reasons for the particular sensitivity of the macula remain speculative. A number of patients report the presence of monocular scotomas that gradually enlarge and move across the retinal field, and this phenomenon, which can lead to loss of vision, has been ascribed to retinal migraine (3, 4). Although the occurrence of retinal migraine has been disputed (5), the likely cause is spreading depression (SD), a phenomenon initially described in the central nervous system by Leão (6). SD is a slow diffusion-mediated self-sustained wave of generalized depolarization of gray matter or retina that does not respect synaptic connectivity (7) and severely disrupts neuronal function. SD results in a temporary collapse of transmembrane ionic gradients and membrane potential (8, 9), and is characterized by a negative field potential 10–20 mV in amplitude and 1–2 min in duration, propagating at a velocity of a few millimeters per minute (6–10). The advancing wave front is accompanied by a burst of action potentials and K⁺ efflux into the interstitial space, followed by electrical silence, which is associated with the negative field potential (11, 12). Na⁺ and Cl⁻ enter the cells, causing an influx of water, swelling, and reduction in the volume of the extracellular space. In addition, proinflammatory compounds are released, often accompanied by edema and extravasation of blood proteins (13). Recovery can occur in minutes, but repeated episodes of SD increase neuronal loss surrounding traumatic brain lesions, and have been implicated in other central nervous system (CNS) pathologies, including migraine (14). SD has been demonstrated in a number of taxa, including fishes, amphibians, reptiles, birds, and mammals (8, 9).

The retina is an energy-demanding tissue and is very sensitive to interruption of its supply of oxygen and glucose in both avian and mammalian tissue (15, 16). Hypoglycemia, a condition that facilitates the initiation of cerebral SD in rats (17), appears to lead to retinal cell death and gradual loss of visual acuity in aging mice (15). Here we used the ex vivo chicken retina, an established system in which waves of SD can be visualized as areas of increased light scattering (8). We found that in hypoglycemic retina, a small number of spontaneous or evoked episodes of SD cause irreversible punctate lesions in the macula. These lesions in turn can act as sites of origin for self-sustained reentrant spiral waves of SD that progressively enlarge the lesions. Furthermore, the macular degeneration is not observed when SD is blocked or when activation of caspases by SD is prevented. These observations of macular damage under conditions of hypoglycemia establish a potential role of SD in macular degeneration, particularly in diabetes.

Results

Elicitation of Radial SD Waves in Retina Under Control Conditions.

A wave of SD evoked by either gently touching the retina with a tungsten needle or by focal application of a high-K⁺ solution (20 mM) was seen as an expanding circle of decreased transparency centered on the site of initiation (Fig. 1A and Movie S1). The increase in light scattering was accompanied by a slow negative shift in the extracellular potential (Fig. 1B). The field potentials were largest in the inner plexiform layer (Fig. 2) (8) and inverted close to the outer plexiform layer. Under normal oxygen and glucose conditions, SD waves could be elicited every 15 min over 4 h without either tissue damage or a change in the field potential (Figs. 1B, 3A, and 4A). The propagation velocity of the consecutive SD waves was 4.0 ± 0.2 mm⋅min⁻¹ in different preparations (n = 16) at 34 °C. SD velocity was uniform throughout the central and peripheral retinal regions, and did not differ in either the nasotemporal or dorsoventral axes. This spread velocity falls into the range of propagation velocities observed for SDs in other regions, species, and preparations (1.7–9.7 mm/min) (6–9). In the wake of a wave of SD, the retinal tissue remained refractory to further waves for a period of 120 ± 4 s (n = 15 measurements; n = 5 retinas; Fig. 1D). However, when the concentration of K⁺ was increased from 4 mM (replacing NaCl to maintain isosmolarity), the number of spontaneous radial SDs/h progressively increased (Fig. 1D) whereas the absolute refractory period decreased (Fig. 1E). The minimum partial oxygen (pO₂) of 28.5 ± 3 mmHg (measured with a microelectrode; Materials and Methods; n = 11 measurements; n = 6 retinas) was reached in the inner plexiform layer at a depth of ∼100 μm during SD waves, thus ruling out the occurrence of transient focal hypoxia during SD propagation (Materials and Methods). pO₂ of 140 mmHg (corresponding to ∼20% oxygen) was maintained in the gas phase directly above the retinas. Under our experimental conditions, 6 mM MgCl₂ (n = 20) completely suppressed SD, as reported (18).

Fig. 1.

Evoked SD waves in ex vivo normoxic chicken retina. (A) A circular wave (arrow) triggered by focal application of K⁺ (20 mM) was seen as an opalescent circle expanding outward from the site of initiation (arrow; left eye). There was an ∼5-mm depth of saline over the retina. D, dorsal; N, nasal; pec, pecten; T, temporal; V, ventral. (B) During each wave of SD, a transient negative field potential about 20 mV in amplitude and 1 min in duration was recorded by an extracellular electrode positioned in the inner plexiform layer ∼100 μm beneath the retinal surface. The arrows indicate representative SD potentials in response to the 1st, 5th, 9th, 13th, 15th, and 17th (end) stimuli, recorded every 15 min over a 4-h test period. (C) A spiral SD wave propagation (panels 1–4; 1, arrows indicating wave front in this and subsequent figures) made possible by a 1-mm lesion in the macula (1, between the asterisks). The pecten had been removed in making the retinal circle. Images are at 15-s intervals (counterclockwise rotation, from 1 to 4) selected from ∼900 images of the same field (Movie S2). (D and E) In retina under isosmotic conditions (replacing NaCl with KCl) and normal glucose, increasing concentrations of K⁺ led to an increase in the number of spontaneous radial waves of SD per/h (6 mM KCl vs. 11 mM KCl; *P < 0.01) (D) and a decrease in the refractory period (4 mM KCl vs. 8 mM KCl; *P < 0.01) (E).

Fig. 2.

Distribution of SD extracellular potentials recorded (negative up) in different layers of the chicken retina (submerged in ∼5 mm saline). (A) Control normoxic retinal section processed with cresyl violet showing approximate electrode locations for records in B. GCL, ganglion cell layer; INL, inner nuclear layer; IPL, inner plexiform layer; NFL, nerve fiber layer; ONL, outer nuclear layer; OPL, outer plexiform layer. Depth recordings were performed using a stereotaxic device equipped with a multibarrel microelectrode for recording and dye marking to confirm the recording site. Results were typical of experiments in four retinas. The SD-associated wave was largest in the IPL (23 ± 1 mV; n = 12) and was smaller toward the outer retina [OPL (1.5 ± 0.25 mV; n = 10); P < 0.001]; the polarity was inverted in the ONL.

Fig. 3.

In hypoglycemic retina, spontaneous sustained spiral SD waves result in damage to the macula and surrounding tissue. (A) In normoglycemic retina, 11 waves of SD triggered by focal application of K⁺ (20 mM) during a 4-h period did not result in a macular lesion (right eye; panel 1, evoked SD; panel 2, lack of lesion; representative of n = 16 retinas). (B) In hypoglycemic retina, after four episodes of spontaneous SD waves originating within the macula, a punctate lesion developed (Left, red arrow). The lesion was PI-positive with more staining at the edge (Inset, right eye; representative of n = 12 retinas). The lesser PI labeling in the center of the lesion may be because the cells there had lost membrane integrity. The lesion acted as the site of origin of additional waves of SD (Right, white arrow). (C) A clockwise spiral wave of SD in a different preparation (right eye; the larger arrows indicate the wave front at the edge of the lesion; the smaller arrows indicate the wave front at some distance from the lesion). The lesion, larger than that in B, supported reentrant SDs. The time interval between panels is 75 s. (D) Western blots for caspase 3. Equal loading is indicated by the β-actin bands. (Left) Levels of cleaved caspase 3 in a control (normoglycemic) retina. (Right) Left lane: Levels of caspase 3 in hypoglycemic central retina after three SD waves. Center lane: Formation of caspase 3 was prevented by 10 mM MgCl₂ (Mg), which blocks SD. Right lane: Formation of caspase 3 was also blocked by QVD, a broad-spectrum caspase inhibitor that does not block SD. (E) Densitometric analysis of caspase-3 levels normalized to β-actin in the same sample and to control (control versus SD, P < 0.001; control versus Mg or QVD, P > 0.05).

Fig. 4.

In hypoglycemic retina, spontaneous sustained spiral SD waves result in damage to the macula and neighboring tissue. Sections of the retinas illustrated in Fig. 3 were fixed in Bouin's solution and stained with cresyl violet. (A) In normoglycemic retina, successive waves of SD evoked by focal application of K⁺ (20 mM) at 10-min intervals during a 4-h period did not result in a retinal lesion. (B) Blockade of SD during hypoglycemia by MgCl₂ prevented macular damage. (C) In hypoglycemic retina, repeated episodes of SD caused damage in the macular region of the ganglion cell and inner nuclear layers with pronounced tissue distortion. (D) In the periphery of the same retina as in C, 5 mm from the center, there was no apparent damage.

Elicitation of Spiral SD Waves in Retina.

To investigate the dynamics of spiral SD reentry and the smallest possible loop in which the impulse could continue to circulate, we elicited spiral waves in retinas in control medium (20 mM glucose) by making a slit around which the wave propagated. In isolated circles of the retina (7 mm in diameter) centered on the macula, we made short linear incisions ranging from 0.5 to 3 mm in length and evoked SD waves by focal application of 20 mM KCl close to the slit on one side. To prevent the resulting wave from going in both directions and colliding on the other side of the slit, we blocked the progression in one direction by local perfusion with MgCl₂ (6 mM, n = 20). Then, rapid washout of the MgCl₂ allowed maintained propagation of the wave front coming around from the other side.

Using visual and field potential measurements, we found that, under control conditions (i.e., 20 mM glucose), ∼1 mm was the minimum slit length required to support reentrant spiral SD waves for up to 30 min (Fig. 1C; in this case, counterclockwise rotation); lesions longer than 1.5 mm could support continuous propagation of SD waves for at least 2 h. The spiral SD waves propagated with a velocity of 1.95 ± 0.05 mm/min for a lesion length of 1 mm and 3.15 ± 0.1 mm/min for a lesion length of 3 mm (n = 10 for each slit length) at 34 °C (Movie S2); that is, the velocity was greater for the lower frequency resulting from the longer minimum path length.

Hypoglycemia Favors the Occurrence of Spontaneous Radial and Reentrant Spiral SD Waves.

To create conditions favoring the occurrence of spontaneous SD waves (8), we maintained ex vivo retinas in normoxia (20% O₂, 5% CO₂, 75% N₂) but low glucose (4 mM) under normosmotic conditions (Materials and Methods). Under hypoglycemic conditions, random episodes of SD arose spontaneously primarily within the macula (Fig. 3 B and C). After at least two spontaneous episodes (3 ± 0.25 episodes; ranging from 2 to 5; n = 18), a punctate lesion, which was propidium iodide (PI)-positive, indicative of increased membrane permeability and possible cell death, formed within the macula close to or at the site of initiation of the SD waves (Fig. 3B, Inset). Subsequent spontaneous SD waves resulted in additional damage to the macula and surrounding tissue, but more peripheral retina was spared (Figs. 3 B and C and 4 C and D). Lesion extent and cellular identity were confirmed by histological studies using cresyl violet and NeuN in sections of retinas previously stained ex vivo with PI. Under hypoglycemic normoxic conditions, a nondecremental spiral wave arose only after the primary lesion reached a length greater than 1 mm, consistent with the minimum size of cut needed to allow reentrant spiral waves under control conditions (Fig. 3 B and C and Movie S3). SD waves propagated around the lesions at a speed of 2.35 ± 0.20 mm/min (n = 12) at 34 °C, ranging from a velocity of 1.8 mm/min (circumference of the lesion: 5.7 mm) up to 3.6 mm/min (circumference of the lesion: 8.75 mm). The period of rotation of the spirals was 3.5 ± 0.2 min per turn around the lesion (n = 12 retinas), and the waves were self-sustaining for 20 min up to at least 2 h (4–36 cycles). The time for one cycle around the lesion increased with lesion size, and the velocity also increased because of the reduction in refractoriness. Whereas the prominent reentrant spiral wave propagated around the lesion, near the periphery of the retinal circle, the edge of the wave could initiate local, small spirals, perhaps in association with the injury associated with isolation of the tissue (Movie S3).

Mechanisms Whereby Hypoglycemia Leads to Reentrant Spiral SD Waves.

An SD wave starts out from somewhere on the edge of the macular lesion, moves in all directions, and then dies out going in one direction around the lesion, thus permitting the still-propagating wave front to complete the rotation and continue around again. Parameters that may affect propagation include osmolarity and K⁺ concentration. Under our experimental conditions, increasing NaCl (above the control value of 100 mM) in the hypoglycemic solution affected SD; 126 mM (n = 8) decreased the velocity of propagation of evoked waves of SD by ∼10%; and 139 mM (n = 8 retinas) blocked both evoked and spontaneous waves. Similarly, adding 25 mM mannitol to the low-glucose bathing solution (n = 20 retinas) prevented spontaneous radial and spiral waves of SD and the development of macular lesions for up to 4 h, suggesting that lowering extracellular osmotic pressure by removing glucose contributed to SD initiation and propagation in hypoglycemic retina. However, low glucose plus 25 mM mannitol neither blocked evoked SD waves nor prevented macular degeneration associated with multiple evoked SD waves (n = 11 retinas). Thus, hypotonic hypoglycemia was deleterious both in increasing the number of spontaneous SD waves and in increasing sensitivity to the metabolic stress.

We hypothesize that in the hypoglycemic (and hyposmotic) retina in which the threshold for SD is reduced, depolarization by K⁺ efflux associated with substrate deprivation reaches that threshold and initiates a wave of SD. To test this hypothesis, SD waves were elicited by brief localized K⁺ applications (1, 3, 5, 10, 15, and 20 mM) in hypoglycemic and in normal glucose (Materials and Methods). The concentration of KCl, at which 50% of the trials triggered SD waves (EC₅₀ values), was ∼15 mM in normal glucose (n = 5 trials per concentration; 30 retinas) and 10 mM in hypoglycemic Ringer (n = 5 trials per concentration; 32 retinas), indicating a lower threshold for SD in low glucose, possibly resulting from increased K⁺ leak. In normal glucose, the maximal response (E_max = 100%) was reached at a concentration of 20 mM KCl.

In another series of experiments, preparations were perfused with Ringer solution containing 1 mM MgCl₂, 6 mM KCl, and 20 or 4 mM glucose (in the latter adding 8 mM NaCl to keep osmolarity constant). The number of spontaneous SD waves/h increased from 1 ± 0.5/h in control (n = 8; 20 mM glucose) to 5.2 ± 0.8/h in low glucose (n = 8; 4 mM glucose) (P < 0.001), suggesting that in hypoglycemia, the threshold to trigger SD is reduced to the point where SD arises spontaneously with increased frequency.

Spiral SD Waves Increase Caspase 3 in the Central Retina.

As reentrant spiral waves of SD continued in low-glucose solutions, the size of the PI-positive lesions progressively increased. The average lesion area after 3 h of spiral SD was 4.85 ± 0.5 mm² (n = 12). We assayed for apoptosis in these hypoglycemic retinas by determining the level of (cleaved) caspase 3, a critical executioner of apoptosis (19), detected by blotting of protein in an ∼5-mm diameter circle from the central retina. Caspase-3 levels in these regions of hypoglycemic retinas increased over control normoglycemic retinas after as few as three SD waves (Fig. 3D, Left: control; Right, first lane: SD with hypoglycemia; quantitation in Fig. 3E; control, n = 9; SD with hypoglycemia, n = 3; P < 0.01, control vs. SD with hypoglycemia). In contrast, levels of caspase 3 in hypoglycemic retinas treated with either 10 mM MgCl₂ (n = 6), which blocks SD, or 10 μm QVD [quinolyl-valyl-O-methylaspartyl-(2,6-difluorophenoxy)-methylketone] (n = 3), a cell-permeant broad-spectrum caspase inhibitor (20) that does not block SD, were not significantly different from control normoglycemic retinas (Fig. 3D, Right, second and third lanes; Fig. 3E; n = 6; P > 0.05 vs. control). QVD presumably acted by blockage of caspase 9 or another caspase that cleaves procaspase 3.

We examined retinas histologically after SD under hypoglycemic conditions and observed marked damage in the macula but not in the peripheral retina (Fig. 4). To determine mechanisms and potential protective measures, we evoked SD in hypoglycemic retinas in the presence of tetrodotoxin (TTX; 1 μM; sodium channel blocker), halothane (2.5 mM; gap junction blocker), MgCl₂ (10 mM; SD blocker; blocker of transmitter release), or QVD (10 μM). TTX, which does not block SD (8), did not prevent SD-induced cell death (n = 8). Both MgCl₂ and halothane blocked initiation of SD waves (18, 21) and prevented development of punctate lesions and lesion expansion (Figs. 3 B and C and 4B), as well as activation of caspase 3 (Fig. 3 D and E). In contrast, QVD treatment did not block either initiation or propagation of SD waves (a similar number of spontaneous SDs occurred with and without QVD), but did prevent the SD-induced punctate lesions, increase in caspase 3 (Fig. 3 D and E), and subsequent onset of spiral waves and increase in size of the macular lesions. The frequency of spontaneous SDs observed in the presence of QVD (n = 6.75 ± 0.6 waves/h; n = 8 retinas) would have caused at least punctate lesions in the absence of QVD. The efficacy of QVD in conferring neuroprotection, but not suppressing SD, suggests that SD-mediated neuronal death and lesion expansion require activation of one or more caspases.

Discussion

Although spontaneous spiral waves have been observed in many biological systems, their emergence, dynamics, and potential function in retina remain uncertain. The results presented here demonstrate that radial SD waves evolve into reentrant spiral waves following damage in hypoglycemic macula, and that these waves lead to further cell mortality and lesion expansion. We not only analyzed the conditions that promote initiation and propagation of SD waves in hypoglycemic retinas but also obtained evidence that loss of neurons can be prevented by blocking SD or caspase activation.

Rotating spiral waves occur in a wide range of chemical, biological, and physical systems that can be considered as effectively 2D excitable or oscillatory media in which the local nonlinear dynamics exhibits threshold behavior, separating a resting or recovered state from an excited state (22). In neocortex in vivo, spiral waves occurred during pharmacologically induced oscillations, activity mediated synaptically rather than by SD (23). Spiral waves have been documented in the heart, where they are based on propagation of sodium-dependent action potentials that become reentrant through gap junctions; in humans, spiral waves are a possible basis of ventricular arrhythmia (24–26). In an earlier study of chicken retina under control conditions, anodal or cathodal polarization applied during the falling phase of an SD wave could induce a spiral wave that continued for ∼30 min (27). These waves did not require a central lesion around which to propagate. Low-Mg²⁺ solutions also led to spontaneous generation of spiral waves. No information was reported about damage to the retina resulting from this activity. SD, which occurs only in the CNS (including retina), and cardiac spiral waves are analogous in being actively propagated waves, but the mechanisms, as well as time courses, are profoundly different. The impact of SD waves may be understood on the basis of the minutes-long depolarization associated with the collapse of ionic gradients across excitable membranes and changes in the extracellular:intracellular ratio of water volume. Although precise mechanisms of propagation are unclear, intercellular spread does not involve normal synaptic connectivity (with the possible exception of electrical coupling) (7). In contrast, cardiac arrhythmias prevent the heart from pumping effectively, and damage arises from insufficient blood reaching the brain and other vital organs (including the heart).

SD results in transient elevations of the extracellular K⁺ concentration of 30–80 mM in the intact nervous tissue ascribable to depolarization-induced K⁺ and Na⁺ activation and excitatory transmitter release (8), although the failure of TTX to block SD indicates that Na⁺ channel activation in not essential. We found that relatively small increases in extracellular K⁺—altered from 4 to 6 to 8 mM—are sufficient to increase the “excitability” of the central retina for generation of SD waves with respect to both shortening of the refractory period and an increase in the number of spontaneous radial SDs/h (Fig. 1D). Because insulin-induced hypoglycemia can induce SD (28) and the refractory tissue is dominated by glycolytic activity (28, 29), we focused on SD dynamics affected by decreased glucose concentrations, observing, under this circumstance, onset of spontaneous spiral SD waves and formation of central retinal lesions.

Initiation of a spiral wave clearly involves a nonuniformity in the retina. We show that under hypoglycemic conditions and in 20% oxygen, spontaneous reentrant SD spiral waves propagate around a small lesion for many cycles, leading to an increase in cleavage of procaspase 3 and expansion of the lesion to encompass larger regions of the central retina, which has a large metabolic load (30). Under these conditions, the initial threshold for onset of SD is reduced to the point where SD episodes arise spontaneously with increased frequency. The finding that during SD the tissue oxygen level is comparable to the in vivo situation (31, 32) indicates that hypoxia is not a cause of SD initiation or subsequent macular degeneration. Further, lack of transient anoxia during SDs does not support the suggestion that occurrence of anoxic periods may be a compromising factor leading to neuronal damage (33). Spontaneous SD activity is increased by hyposmolarity, which is ascribable to reduced extracellular space and greater accumulation of K⁺ and possibly neurotransmitters. Metabolic compromise would contribute to accumulation of K⁺ and, when a new wave starts in the refractory period of the previous wave, it may fail to propagate in one direction because of minor nonuniformities of the retina, leading to the onset of spiral SD waves (Movies S2 and S3).

Primate retina and brain are capable of supporting SD (34, 35). Cortical SD waves are believed to be involved in certain clinical neurological disorders (7), including migraine aura (8), which can be accompanied by a temporary partial or complete loss of vision over a portion of or the entire visual field (36). The typical pattern of scotomas is a small binocular blind spot that initiates in the central visual field and enlarges to occupy the whole field (36). In contrast, retinal SD causes a monocular loss of vision that can be without painful sequellae or evidence of damage to the retina (4). Conversely, retinal migraine can be associated with retinal damage and temporary blindness in some patients, and we suggest that conditions leading to SD, even on a level inadequate to initiate the full-blown wave phenomenon, are a hazard for macular degeneration.

Linkage between low glucose levels and loss of vision highlight the importance of glycemic control in diabetics to prevent retinal injury such as macular degeneration (15). Our results suggest that spontaneous and recurrent SD may be initiated in the macula during hypoglycemia and can lead to macular degeneration. Progressive expansion of retinal lesions in an aging mouse has been reported as a result of chronic hypoglycemia induced by disruption of the glucagon receptor gene, Gcgr (15), although the time course was vastly slower than that we observed with SD in the chicken retina. Blockade of SD may represent a useful therapeutic approach not only to prevent retinal degeneration resulting from fluctuations in glucose levels in diabetes but also in other conditions, including focal brain ischemia, where the number of episodes of SD correlates with the expansion of focal ischemic lesions (37–39), spiral propagation of SD results in an expanding cortical lesion (39), and repetitive SD waves may involve hypoglycemic regions in the penumbra. The relevance of the results reported here to clinical conditions will only be established through further research, but the importance of normoglycemia to sparing neural structures during strong activation as in SD should be considered in any treatment modalities.

Materials and Methods

Details are provided in SI Materials and Methods, including three movies (Movies S1, S2, and S3) and a still image for each movie. All animal procedures were in accordance with the National Institutes of Health Guide for Care and Use of Laboratory Animals, with the principles adopted by the Society for Neuroscience (US) and approved by the institutional animal ethics committees of New York Medical College and the Federal University of Rio de Janeiro.

Ex Vivo Chicken Retina Preparations.

The experiments were performed on white leghorn chickens, 1–2 wk after hatching. After decapitation the eyes were removed, and a medial half containing the retina was transferred to a recording chamber mounted on the stage of a stereoscope (SMZ800; Nikon). The bathing Ringer solution contained 100 mM NaCl, 4 mM KCl, 1 mM CaCl₂, 1 mM MgCl₂, 30 mM NaHCO₃, 1 mM NaH₂PO₄, and 20 mM glucose (pH 7.3) and was prewarmed (34 °C), gassed (20% O₂, 5% CO₂, 75% N₂), and perfused at 1.2 mL/min. We photographed the SD waves after an equilibration period of 20 min under control conditions. SD waves were elicited with a tungsten needle or a droplet of 20 mM KCl solution.

The refractory period was defined as the time after completion of a propagated event until a second propagated event (>3 mm in diameter) could be initiated. Stimuli, a gentle touching of the retina with a tungsten needle near the site of initial stimulation, were applied at ∼10-s intervals. The effect of [K⁺]_o on the refractory period was tested at 4, 6, and 8 mM K⁺ under isosmotic conditions, replacing equimolar Na⁺ by K⁺.

The concentration–response relationship of SD waves evoked by K⁺ droplets (∼1 μL) was determined by testing 1, 3, 5, 10, 15, and 20 mM K⁺.

Osmotic Manipulations.

In most experiments, ex vivo retinas were exposed for 30 min to each of two test solutions (26 or 39 mM NaCl added to hypoglycemic solution) in order of increasing osmotic pressure. Between applications of the solutions, the retinas were allowed to recover in normoglycemic control solutions for 30 min. In a subsequent set of experiments, mannitol (25 mM) was added to the hypoglycemic solution. Mannitol was chosen because it is neither metabolized nor actively transported into cells and thus provides a relatively inert (and nontoxic) mechanism to increase extracellular osmolarity.

Measurement of Partial Oxygen Pressure.

pO₂ on the retinal surface and at 100-μm depth (inner plexiform layer) was measured in the recording chamber using an oxygen-sensitive Clark-type microelectrode [OX-5 (5-μm tip); UNISENSE] positioned by a motor-driven micromanipulator.

Western Blotting.

Control and degenerating maculae were assayed for cleaved caspase-3 levels using standard methods. Circular patches (∼5-mm diameter) were removed from central retinas that had been incubated under control and hypoglycemic conditions for the indicated times.

Immunohistochemistry and Immunocytochemistry.

Retinas were fixed for 1 h at room temperature in Bouin's solution, and cryostat sections were stained with 0.1% cresyl violet (Acros Organics, Fisher Scientific).

Statistics.

Data are presented as means ± SEM of at least three independent experiments. Fisher's exact test for unpaired data was used to evaluate the significance of the differences between various treated and untreated groups. The null hypothesis was rejected at P < 0.05.