Biochemical Activity of Reactive Oxygen Species Scavengers Do Not Predict Retinal Ganglion Cell Survival

Abstract

purpose. Retinal ganglion cells (RGCs) die as a result of axonal injury in a variety of optic neuropathies, including glaucoma. Reactive oxygen species (ROS) act as intracellular signaling molecules and initiate apoptosis in nerve growth factor–deprived sympathetic neurons and axotomized RGCs. Determination of the role of specific ROS relies on the use of small molecule or protein scavengers with various degrees of specificity. The pro- or anti–cell-death effect of several ROS generating and scavenging systems in cultured RGCs was correlated with their activity in cell-free assays.

methods. Neonatal rat retinas were dissociated and incubated with ROS-generating systems for hydroxyl radical, superoxide anion (O₂ ⁻), and H₂O₂. Scavengers tested were catalase, polyethylene glycol-superoxide dismutase (PEG-SOD), manganese (III) tetrakis(1-methyl-4-pyridyl)porphyrin (MnTMPyP), 6-hydroxy-2,5,7,8-tetramethylchroman-2-carboxylic acid (Trolox), deferoxamine, and U-74389G. Viability of retrogradely labeled RGCs was determined with calcein-AM 24 hours after plating. O₂ ⁻ and H₂O₂ scavenging in cell-free assays was measured with dihydroethidium and Amplex Red (Invitrogen, Carlsbad, CA), respectively.

results. Systematic differences were found between ROS scavenging in cell-free assays and the ability of scavengers to protect RGCs in cell culture. Furthermore, many ROS scavengers lost specificity and protected against various ROS, whereas others failed to protect against their unique ROS target. These activities stray from commonly recognized specificities of individual ROS scavengers or generating systems and are important in understanding ROS biology. In addition, antioxidant defense mechanisms used by RGCs and other retinal cells interfere with responses expected from ROS scavengers in well-defined systems. Last, H₂O₂ induced intramitochondrial O₂ ⁻, whereas paraquat produced O₂ ⁻ outside of the mitochondria, and these areas of generation can mislead interpretations of ROS scavenger activity and effectiveness.

conclusions. There is discordance between ROS effects in cultured RGCs and cell-free assays, with several mechanisms accounting for this divergence. To identify the roles of ROS signaling in cell death accurately, several approaches should be used. These include using a panel of ROS scavengers and generators, testing the panel in primary neuronal cultures, and quantifying ROS with cell-free assays.

Retinal ganglion cells (RGCs) die as a result of axonal injury, including glaucoma. ^{1 2 3 4 5 6} There are multiple biochemical and molecular changes that affect the RGCs after axotomy, including deprivation of neurotrophic factors from the target or other tissues, increases in intra-axonal Ca²⁺, influx of microglia with the nerve, accumulation of excess retrogradely transported macromolecules, induction of p38 MAP kinase and other signaling molecules, and caspase activation. ^{7 8 9 10 11 12 13} Our work and that of others is consistent with an increase in reactive oxygen species (ROS) acting as an intracellular signaling molecule and initiating apoptosis, ^{14 15 16} similar to what occurs in nerve-growth factor–deprived sympathetic neurons. ^{17 18}  

However, much of the evidence for identification of a specific ROS in a complex biological system relies on the use of small molecule or protein scavengers with various degrees of specificity. Often, these scavengers are assumed to be highly specific and to control the generation of a particular ROS. The interaction of the multitude of intrinsic cellular ROS scavengers and modulators with extrinsic molecules would probably blur the specificity of the latter. We therefore compared the pro- or anti-cell death effect of several ROS and scavenging systems in cultured RGCs with their activity in cell-free assays. We found a significant dissociation between ROS effects in these neuronal cultures and cell-free assays and describe several mechanisms that would account for this divergence. 

Methods

Animals

All experiments were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and with institutional, federal, and state guidelines regarding animal research. 

Materials

Cell culture reagents were obtained from Invitrogen-Gibco (Grand Island, NY). The retrograde fluorescent tracer 4′,6-diamidino-2-phenylindole (DAPI), the fluorescent viability agent calcein-AM, dihydroethidium (HEt), a superoxide indicator (MitoSOX Red), and a hydrogen peroxide/peroxidase assay kit (Amplex Red) were obtained from Invitrogen (Eugene, OR). Papain was obtained from Worthington Biochemical (Freehold, NJ). The ROS-scavengers manganese (III) tetrakis(1-methyl-4-pyridyl)porphyrin (MnTMPyP), 6-hydroxy-2,5,7,8-tetramethylchroman-2-carboxylic acid (Trolox), and U-74389G were obtained from Axxora (San Diego, CA). Unless otherwise noted, all other reagents were obtained from Sigma-Aldrich (St. Louis, MO). 

RGC Labeling and Culture

RGCs were retrogradely labeled by stereotactic injection of the fluorescent tracer DAPI dissolved in dimethylformamide into the superior colliculi of anesthetized postnatal day 2 to 4 Long-Evans rats. DAPI is taken up by synaptic terminals of the RGC and transported to the soma, where it binds to nuclear DNA. Injection into the superior colliculi did not cause any change in animal behavior or feeding. At postnatal days 11 to 13 the animals were decapitated, the eyes enucleated, and the retinas dissected in sterile Hanks’ balanced salt solution (HBSS) with sterile instruments. After two incubations in enzyme solution containing papain (3.7 U/mL), each for 30 minutes at 37°C, the retinas were gently triturated with a Pasteur pipette and plated on poly-l-lysine–coated 96-well flat-bottomed tissue culture plates (0.32 cm² surface area/well) at a density of approximately 2000 cells/mm². Culture platings were performed in a sterile laminar flow hood. The cells were cultured for 24 hours in Neurobasal-A (Invitrogen) and B27 supplement lacking antioxidants in a humidified 5% CO₂ incubator at 37°C. 

Immunopanning for the Purification of Ganglion Cells

For experiments requiring ganglion cells in isolation from other retinal cell types, ganglion cells were prepared by sequential immunopanning, as described by Barres et al. ¹⁹ Briefly, retinas were harvested from rats at postnatal days 11 to 13 and digested with papain. The resultant suspension was incubated in the presence of ovomucoid (US Biological, Swampscott, MA) and anti-rat-macrophage antiserum (Accurate Chemical, Westbury, NY). After sequential incubations on two panning plates coated with affinity-purified goat anti-rabbit IgG (H+L chain-specific; Jackson ImmunoResearch, West Grove, PA), nonadherent cells were filtered and transferred to a panning plate coated with affinity-purified goat anti-mouse IgM (μ chain-specific; Jackson ImmunoResearch) to which was bound monoclonal IgM antibody against mouse Thy1.1 (T11D7e2; American Type Culture Collection, Manassas, VA). After a 30-minute incubation, nonadherent cells were removed and the plate was washed eight times with phosphate-buffered saline. Adherent cells were removed from the plate by treatment for 8 minutes with trypsin (0.125%) in Earle’s balanced salt solution, followed by gentle trituration. Trypsin inactivation was achieved by addition of a 25% solution of fetal calf serum. Recovered cells were centrifuged, resuspended in medium, and plated in 384-well plates (BD Biosciences, Franklin Lakes, NJ). 

Ganglion Cell Identification and Counting

RGCs were identified by the presence of DAPI, which appears blue when viewed with appropriate filters under epifluorescence. Cell viability was determined by metabolism of calcein-AM, which produces a green fluorescence when viewed with fluorescein filters. Cells were incubated in a 1-μM solution of calcein-AM in phosphate-buffered saline (PBS) for 30 minutes, after which the medium was replaced with fresh PBS. RGC viability was assessed at 24 hours. Wells were counted in duplicate. 

ROS Assays

Hydrogen Peroxide.

An H₂O₂ assay (Amplex Red; Invitrogen) was used to determine the presence of H₂O₂ after scavenging. On addition of peroxidase, the reagent reacts 1:1 stoichiometrically with H₂O₂, resulting in the production of resorufin. The red fluorescent oxidation product, resorufin, has fluorescence emission maxima of 571 nm and 585 nm. 

Working solutions of the red dye and horseradish peroxidase were made fresh for each assay. The dye (50 μM Amplex Red; Invitrogen) was added to wells to a final volume of 100 μL. Scavengers and fluorescent dye working solutions were added to wells, and the reaction was initiated with the addition of H₂O₂. Samples were incubated for 30 minutes in the dark in a 96-well black plate with clear well bottoms. Fluorescence readings were obtained (Wallac 1420 VICTOR 2 T Multilabel Counter; PerkinElmer, Inc., Wellesley, MA) with excitation at 485 nm and emission at 580 nm. Wells were counted in triplicate. 

Superoxide.

Superoxide assays were modified from Zhao et al. ²⁰ HEt (1 mM) reacts with O₂ ⁻ to form an oxidized product with specific fluorescence emission characteristics different from HEt or ethidium. Xanthine, scavengers, and HEt were added to a 96-well black plate with clear well bottoms. The reaction was initiated by the addition of 0.05 U/mL xanthine oxidase to a final volume of 100 μL. The plate was incubated for 30 minutes, protected from light. Fluorescence readings were obtained using the multilabel counter with excitation at 485 nm and emission at 580 nm. Wells were counted in triplicate. Precise measurements of O₂ ⁻ scavenging activity by MnTMPyP may be underestimated because Mn²⁺ may act as a partial electron donor to the fluorophore. ²¹  

RGC Treatments

ROS-Generating Systems.

Standard systems were used to generate ROS in vitro. ROS generated by the experimental procedures were: H₂O₂ (30 μM), O₂ ⁻ (30 μM paraquat or 1 mM xanthine and 0.05 U/mL xanthine oxidase), and a hydroxyl radical (2 μM CuSO₄, 2 μM phenanthroline, and 100 μM ascorbic acid; CPA). These concentrations were based on preliminary dose–response experiments and were chosen to yield estimated survival rates at 24 hours of approximately 50% compared with untreated cultures. Hydrogen peroxide was prepared by dilutions from a 30% stock. Paraquat and xanthine/xanthine oxidase were used to generate superoxide anion O₂ ⁻. Paraquat undergoes single electron reduction via oxidation of NADPH or NADH to form paraquat free radical, which reacts with oxygen to form O₂ ⁻. The hydroxyl radical was generated by the Fenton reaction, with a combination of copper, 1,10-phenanthroline, and ascorbic acid. 

ROS Scavenging.

Scavengers tested were catalase (0.5–500 U/mL), polyethylene glycol–conjugated superoxide dismutase (PEG-SOD; 3–300 U/mL), MnTMPyP (5–100 μM), Trolox (1–100 μM), deferoxamine (2–200 μM), and the 21-aminosteroid U-74389G (3–100 μM). Concentration ranges were chosen based on published data and to avoid toxicity to RGCs. ROS scavengers and ROS-generating systems were simultaneously added to the wells at the time of plating. There is a slow and steady rise in O₂ ⁻ in RGCs after dissociation lasting up to 168 hours after plating, ²² and therefore we chose to perform the studies acutely, as no particular recovery period could be practically chosen. 

Measurement of RGC Superoxide In Situ.

Intracellular superoxide levels were measured by quantifying the fluorescence of the oxidation product of HEt. Oxidation of HEt by superoxide converts HEt, which exhibits weak blue fluorescence, to an ethidium derivative that exhibits peak fluorescence in the rhodamine spectrum (excitation 480 nm, emission 586 nm). ²⁰ HEt, which is cell permeant, enters the cell and, after oxidation, accumulates in the nucleus where it binds DNA, with a small shift in its emission spectrum to 567 nm. ²⁰  

At 1 or 24 hours after plating, the chambered coverslip cultures were placed on an inverted-stage microscope (Axiovert 135; Carl Zeiss Meditec, Inc., Dublin, CA) with a 100× oil-immersion lens and a heated stage, to maintain the culture at 37°C. RGCs were identified in mixed retinal cultures by the presence of DAPI staining, which appears bright blue with appropriate filters (excitation 330 nm, dichroic 400 nm, emission 450 nm) under epifluorescence. Cells were treated with 3.2 mM HEt in Neurobasal A without phenol red for 15 minutes. After 15 minutes had elapsed, RGCs were located and images were acquired with a cooled charge-coupled device (CCD) camera (Roper Scientific, Trenton, NJ) using software (MetaFluor; Universal Imaging, Corp, West Chester, PA) and filters appropriate for the superoxide-HEt product (excitation 480 nm, dichroic 505 nm, emission 580 nm long-pass). The fluorescence levels of RGCs were measured and recorded for as many cells as could be identified and imaged in 15 minutes, and then were treated with an ROS-generating system. After 40 minutes had elapsed, the same RGCs were located and fluorescence levels measured and recorded. Images were acquired at a binning of 2, an exposure time of 200 ms, and a 1× gain. Fluorescence measurements after pharmacologic treatment were normalized to pretreatment levels. 

Some experiments used the superoxide indicator MitoSOX Red (Invitrogen) to measure superoxide levels in the cell. It is a recently developed fluorescent probe that is targeted to the mitochondria and highly selective to oxidation by superoxide. It exhibits fluorescence similar to HEt (excitation 510 nm, emission 580 nm) and binds to mitochondrial DNA after oxidation. Cells were treated with 5 μM MitoSOX Red in Neurobasal A without phenol red (Invitrogen), which was replaced after 10 minutes with medium alone. Imaging of cells proceeded in the same manner using the settings described for HEt, and the data were analyzed similarly. 

Statistical Analysis

All RGC viability calculations were normalized to the control (no treatment) condition by dividing the mean number of living RGCs in an experimental condition by the mean in controls. Because the amount of RGC death differed between oxidative insults, we needed a standardized way to compare efficacy of scavengers in preventing this death. To do this—that is, to determine the efficacy of a particular ROS scavenger in preventing cell death from a specific ROS-generating system—we calculated the concentration that rescued either 50% (C ₅₀) or 90% (C ₉₀) of RGCs. In other words, the C ₅₀ was calculated as the value satisfying the following formula:  

\[n(C_{50})\ {=}\ 0.5\ n(\mathrm{no\ treatment})\ {+}\ 0.5\ n(\mathrm{ROS\ only}),\]

and the C ₉₀ was calculated as the value satisfying  

\[n(C_{90})\ {=}\ 0.9\ n(\mathrm{no\ treatment})\ {+}\ 0.1\ n(\mathrm{ROS\ only}),\]

where n represents the number of living RGCs in the corresponding condition. Interpolations were performed geometrically. Comparisons were by unpaired t-test. 

Results

Effect of ROS on RGC Viability

An increase in ROS is sufficient to induce an irreversible death cascade in RGCs. ^{23 24 25} To test which ROS were important in initiating cell death, we cultured RGCs in various generating systems. Paraquat and xanthine/xanthine oxidase were used to test the effects of intracellular and extracellular production of O₂ ⁻, H₂O₂ was used directly, and CPA was used to generate hydroxyl radical by the Fenton reaction. 

As expected, the presence of paraquat (30 μM), 1 mM xanthine with 0.05 U/mL xanthine oxidase, H₂O₂ (30 μM), or CPA (2 μM CuSO₄, 2 μM phenanthroline, and 100 μM ascorbic acid) greatly reduced RGC survival (Fig. 1) . 

ROS Scavenging in Cultured RGCs

We expected that the addition of ROS-specific scavengers, determined by published reports of their activity in biochemical assays, would rescue RGCs from the corresponding ROS generated in culture. For example, we would expect MnTMPyP to protect RGCs from O₂ ⁻ generated by paraquat and xanthine/xanthine oxidase. To quantify the activity of ROS generation and scavenging activities in RGC cultures, we incubated RGCs with predetermined concentrations of ROS-generating systems and various concentrations of scavengers. The concentration of each scavenger that rescued 50% and 90% of the RGCs for a fixed concentration of generated ROS was calculated using the dose–response curves (Table 1) . 

There were unexpected significant discordances between predicted ROS scavenger specificities and those in RGC cultures. As expected, we found that cultures exposed to H₂O₂ were rescued by the H₂O₂ scavenger catalase. The Fe³⁺ scavenger deferoxamine, the nonspecific ROS scavenger (water-soluble vitamin E analogue and lipid peroxidation chain breaker) Trolox, the O₂ ⁻ scavenger PEG-SOD, and the ONOO⁻/O₂ ⁻ scavenger MnTMPyP also rescued RGCs. 

Paraquat generates O₂ ⁻ in metabolically active cells, and predictably we found that PEG-SOD rescued cultures incubated with paraquat, but unexpectedly MnTMPyP did not. Deferoxamine, catalase, and Trolox also prevented cell death in paraquat-treated cultures. 

CPA generates OH·. As expected, MnTMPyP, catalase, and Trolox, protected cultures exposed to CPA. Unexpectedly, PEG-SOD did not rescue cells exposed to CPA. In summary, we established that ROS scavengers protected against various ROS, but in ways that differed from their usually described pattern of activity. 

ROS Scavenging in Cell-Free and Cell Culture Systems

Because neuronal viability in RGC cultures exposed to various ROS diverges from the predicted biochemical activity of each scavenger, we measured the cell-free activity of each scavenger against specific ROS, focusing on O₂ ⁻ (generated intracellularly by paraquat in RGCs and xanthine/xanthine oxidase in cell-free assays) and H₂O₂. We could not use paraquat for O₂ ⁻ generation in cell-free assays, as it inherently requires cell-generated NADH or NADPH. ^{26 27} Furthermore, NADH has autofluorescent properties that would have prevented detection of O₂ ⁻ using our methods. We measured O₂ ⁻ and H₂O₂ levels, by using HEt and Amplex Red assays in a microplate format, and derived dose–response curves after incubations of substrate with ROS for 30 minutes (Fig. 2) . 

We then tested whether the ability of various scavengers to protect against ROS-induced RGC death correlates with how well they scavenge particular ROS. For scavengers against H₂O₂, we tested C ₅₀ (the concentration of scavenger that had rescued 50% of RGCs) and C ₉₀ (the concentration of scavenger that had rescued 90% of RGCs; Table 1 ) against an initial H₂O₂ concentration of 30 μM. Figure 3depicts the H₂O₂ concentrations after incubation with the C ₅₀ and C ₉₀ concentrations of each scavenger for 24 hours. Significant reductions in H₂O₂ were seen with those concentrations of MnTMPyP and catalase that had previously been shown to rescue RGCs, and with high concentrations of PEG-SOD. There was no detectable H₂O₂ scavenging effect seen with concentrations of U-74389G, which had been shown to rescue RGCs. Together, this suggests that there is a dissociation between ROS scavenging effects measured in cell-free assays and cultured RGCs. 

In studying O₂ ⁻ scavenging in cell-free systems, we were limited because paraquat, which we had used to generate intracellular O₂ ⁻, does not produce O₂ ⁻ without the presence of NADH or NADPH, while exogenously generated O₂ ⁻, being a charged molecule, does not cross cell membranes well. We therefore adopted a two-pronged approach. First, we determined the concentrations of ROS scavengers which rescued 50% and 90% of RGCs from 30 μM paraquat, a concentration that we had previously shown (Fig. 1)killed approximately 50% of cultured RGCs. Next, we used xanthine/xanthine oxidase to generate O₂ ⁻, and HEt to measure O₂ ⁻ levels. Xanthine/xanthine oxidase generates both O₂ ⁻ and H₂O₂, ²⁸ and we therefore chose concentrations by assaying O₂ ⁻ alone using HEt. Dose–response curves of xanthine/xanthine oxidase concentrations versus produced O₂ ⁻ were generated, and a concentration of xanthine/xanthine oxidase (1 mM / 0.05 U/mL) was chosen that resulted in near-maximum O₂ ⁻. Finally, the ROS scavenger concentrations that rescued RGCs from paraquat were tested against the previously chosen concentration of xanthine/xanthine oxidase. 

As expected, the O₂ ⁻ scavengers MnTMPyP and PEG-SOD significantly reduced levels of O₂ ⁻ in cell-free assays, whereas other scavengers did not (Fig. 3B) . We compared the cell-free scavenging activity with the ability of these scavengers to protect RGCs from O₂ ⁻ generated by xanthine/xanthine oxidase. As expected, MnTMPyP and catalase rescued RGCs (Fig. 3) . 

Because xanthine/xanthine oxidase produces both O₂ ⁻ and H₂O₂, we expected rescue with not only MnTMPyP and catalase, but also PEG-SOD. Even though PEG-SOD (3 and 30 U/mL) significantly reduced O₂ ⁻ production (50% and 90%) in cell-free assays, it failed to rescue RGCs in vitro at those doses. The most likely reason is that even with the presence of PEG to allow SOD to cross the cell membrane, there is a concentration gradient between the extracellular and intracellular concentration of PEG-SOD. In addition, high amounts of extracellular H₂O₂ will be produced by superoxide dismutation. In this instance, RGC peroxidases may not be able to handle the overwhelming increases in H₂O₂. MnTMPyP, in contrast, would not contribute to an increase in H₂O₂, as it complexes with O₂ ⁻ and does not produce ROS as a byproduct of the reaction. Our data show that MnTMPyP scavenged H₂O₂ in cell-free assays. Therefore, MnTMPyP is more appropriately able to scavenge the ROS generated by xanthine/xanthine oxidase. However, because MnTMPyP did not rescue RGCs from paraquat (intracellular O₂ ⁻) but rescued RGCs treated with xanthine/xanthine oxidase (which generates extracellular O₂ ⁻), it probably does not reach the intracellular compartment in which paraquat generates O₂ ⁻. Likewise, because catalase does not cross cell membranes adequately, its effect on neuronal protection is probably due to extracellular scavenging of released ROS. 

Lack of Contribution of Cocultured Retinal Cells to ROS Generation in RGCs

To determine whether the unanticipated effects of some of the ROS-generating systems on RGCs were due to the effects of other cell types in the mixed retinal culture, we studied the effects of paraquat-generated O₂ ⁻ on individual RGCs. RGC O₂ ⁻ was measured with HEt on several cells. Cells were then treated for 40 minutes with paraquat (30 μM), with or without SOD (300 U/mL), and fluorescence was measured on the same cells. SOD (i.e., without pegylation) does not enter cells, and scavenges extracellular O₂ ⁻ only (i.e., O₂ ⁻ generated by other cells in the culture). As expected, RGCs treated with paraquat led to a significant increase in O₂ ⁻ compared with vehicle (2.44 ± 0.18 vs. 0.96 ± 0.07; P < 0.001). Addition of SOD did not abrogate the increase (1.77 ± 0.29; P = 0.41). 

As another test of whether other cells in the culture, particularly glial cells, were responsible for the idiosyncratic effects of ROS in RGCs, we repeated the viability studies of the various ROS in highly purified RGCs. RGCs were purified by sequential immunopanning, and viability was assessed in the presence of the ROS-generating systems. Purified RGC viability in the presence of paraquat (30 μM), H₂O₂ (30 μM), and CPA (2 μM CuSO₄, 2 μM phenanthroline, and 100 μM ascorbic acid) was almost identical with that of mixed cultures, suggesting that astrocytes or other cells in the culture play a minimal role in the ROS-mediated death of RGCs. 

Sites of Superoxide Generation in RGCs

Imaging with HEt confirmed that treatment with paraquat led to an increase in intracellular superoxide levels, but did not point to a source within the cell. We used a mitochondria-specific superoxide probe (MitoSOX Red), to see whether treatment with paraquat or H₂O₂ induced the same changes in superoxide as those generated by HEt. We found that mitochondrial superoxide levels in RGCs treated with paraquat for 40 minutes were not significantly different from vehicle (1.56 ± 0.25 vs. 1.50 ± 0.37; P = 0.90). We also tested H₂O₂, which we have shown induces an increase in superoxide in axotomized RGCs 24 hours after axotomy. ²⁹ Unlike paraquat, hydrogen peroxide induced an increase in mitochondrial superoxide compared with the control (3.39 ± 0.65 vs. 1.28 ± 0.10; P < 0.05). These results suggest that the compartment for superoxide generation by paraquat is extramitochondrial and that for superoxide generated by H₂O₂ is mitochondrial. 

Discussion

Although at high concentrations ROS cause necrotic cell death by direct oxidative damage to cellular constituents, at lower concentrations they participate in the signal transduction pathway for apoptosis, as has been shown in nerve growth factor–deprived sympathetic neurons ^{17 18 30} and RGCs. ^{14 22 23 29 31} The indirect proof that specific ROS are responsible for activating signaling pathways usually relies on pathway inhibition through small molecule or protein ROS scavengers. These scavengers catalyze the destruction of specific ROS by eliminating or converting ROS into other molecules that cells can then neutralize. Our study showed that there is discordance between the predicted and actual effect of individual ROS scavengers in cell-free and neuronal cell culture conditions, which may result in the inability to determine the nature of ROS signaling pathways by using scavengers alone. 

Our study showed that pharmacological agents that would be predicted to scavenge particular ROS did not follow the expected behavior of ROS formation and destruction. For example, although it is logical that catalase, a peroxidase, would decrease cell death caused by H₂O₂ directly added to cultures, we also saw protection with MnTMPyP, Trolox, deferoxamine, and PEG-SOD, none of which scavenge H₂O₂ (Fig. 3B) . There are at least five possible mechanisms for this dissociation: (1) interconversion of ROS due to cellular enzymes—for example, there are two intracellular superoxide dismutases: SOD-1 and SOD-2 that can catalyze the dismutation of O₂ ⁻ to H₂O₂, either in RGCs or other cells within the culture; (2) failure of charged ROS to reach equilibrium across cell or organelle membranes; (3) differential ability of scavengers to cross cell or organelle membranes; (4) nonspecific activity of scavengers at high concentrations; and the (5) discrepancy between the source of production of a particular ROS and the localization of a scavenger. 

We incubated RGCs with ROS and/or ROS scavengers immediately after dissociation and plating (i.e., without a recovery period). Our previous research indicated that there is an increase in superoxide associated with axotomy due to RGC dissociation within the first 24 hours after plating and continuing up to 168 hours. ²² If the optic nerve is crushed before dissociation, there is a significantly greater rise in intracellular superoxide levels. ²² Because our detection methods do not distinguish between the increase in O₂ ⁻ (or other ROS) due to dissociation and that resulting from our pharmacological treatments with ROS-generating systems, allowing the cells a recovery period after plating would only increase the background intracellular ROS levels. We therefore chose to study ROS levels at the time of plating. 

In the case of the observed protection against H₂O₂ with MnTMPyP and PEG-SOD, a likely explanation is that H₂O₂, which does equilibrate across cell membranes, can be converted into O₂ ⁻ by the reverse dismutation reaction or into hydroxyl radical by the Fenton reaction. MnTMPyP and PEG-SOD are SOD mimetics that scavenge O₂ ⁻. A shift by mass action caused by a depletion of superoxide anion could explain the lower concentration of H₂O₂ and the ability to block RGC death induced by H₂O₂. The death of RGCs from H₂O₂ may therefore reflect these radicals’ activation of a signaling pathway or the oxidation of macromolecules. Alternatively, our finding that MnTMPyP scavenges H₂O₂ in cell-free assays (Fig. 3A)suggests that this commonly used SOD-mimetic may also be a partial peroxidase. 

Our cell-free assays demonstrated that the low concentration of catalase necessary to rescue RGCs from H₂O₂ significantly reduced H₂O₂ concentrations in cell-free assays (Figs. 3A 3C) . Similarly, the large concentrations of MnTMPyP and PEG-SOD that were needed to rescue RGCs were sufficient to scavenge H₂O₂ in cell-free systems. However, Trolox, deferoxamine, and U-74389G protected RGCs from H₂O₂ at high concentrations, but were poor H₂O₂ scavengers in cell-free assays. 

Another dissociation was with O₂ ⁻ generated by paraquat, which induces apoptotic death in numerous neural cell types in culture. ^{32 33 34 35} In the presence of extramitochondrial NADH or NADPH, paraquat undergoes single electron reduction to form the paraquat free radical, which reacts with molecular oxygen to form O₂ ⁻ free radicals. ^{26 27 36 37} We would expect scavenging effects from the SOD mimetics PEG-SOD and MnTMPyP. We found that Trolox, deferoxamine, catalase, and PEG-SOD increased RGC survival in the presence of paraquat-generated (intracellular) O₂ ⁻, whereas MnTMPyP did not (see Table 1 ). 

One explanation for the apparent scavenging of O₂ ⁻ by Trolox, deferoxamine, and catalase is that O₂ ⁻ generated by paraquat is rapidly converted to H₂O₂ by intramitochondrial SOD-2 and other O₂ ⁻ dismutases. If H₂O₂ is formed, the Fe³⁺ chelator deferoxamine indirectly blocks the formation of hydroxyl anion via the Fenton reaction, since it prevents ferric iron from reducing to ferrous iron. The absence of protection with MnTMPyP, which is cell membrane permeable, probably reflects its inability to reach sufficient concentration in the intracellular compartment where O₂ ⁻ is being generated by paraquat. 

Because paraquat requires the presence of plasma membrane-associated NADPH and NADH to produce O₂ ⁻, xanthine/xanthine oxidase was used to generate O₂ ⁻ and test effects of scavenging in cell-free assays. Xanthine/xanthine oxidase forms H₂O₂ and O₂ ⁻ in the presence of O₂ and H₂O. ³⁸ There was complete protection of RGCs by MnTMPyP and catalase and no protection with PEG-SOD (Fig. 3D) . Yet catalase protected RGCs from O₂ ⁻ generated by xanthine/xanthine oxidase in culture and PEG-SOD did not, probably because H₂O₂ production via xanthine/xanthine oxidase is sufficient to induce RGC death. In addition, effective H₂O₂ concentrations are probably higher in xanthine/xanthine oxidase–treated cell cultures than cell-free systems, because cellular superoxide dismutases convert mitochondria-generated O₂ ⁻ to H₂O₂. 

OH· is formed by the Fenton reaction, in which a reduced transition metal, such as Fe²⁺ or Cu⁺, reduces H₂O₂ to form OH· and hydroxyl anion. We used the previously characterized copper-phenanthroline-ascorbate system to produce OH·. H₂O₂ is generated via ascorbate and then reacts with Cu⁺ to generate OH·. Deferoxamine would not be expected to rescue RGCs from OH· because it is a ferric iron chelator, and would not inhibit a copper-mediated Fenton reaction (Table 1) . There was rescue with catalase, MnTMPyP, and to a much lesser extent, Trolox. It is not surprising that a peroxidase such as catalase would decrease Fenton-mediated cell death, as it would decrease the concentration of H₂O₂ substrate. The prevention of Fenton-mediated RGC death by MnTMPyP is probably explained by its ability in these cultures to partly act as a peroxidase, as seen in the results from cell-free assays. However, we cannot definitively compare the effect of the scavengers on OH· with the other ROS because baseline survival was roughly 25% in the OH· conditions as opposed to 40% to 60% for other ROS. Thus, we can only infer which scavengers are effective against OH· and not that a certain scavenger has a lesser effect in OH·-treated conditions versus other ROS. 

One potential flaw in our methodology is that we relied on the relative specificity of two fluorescent dyes, HEt and MitoSOX Red, as superoxide sensors. We could not use electron paramagnetic resonance spectroscopy (EPR), which is a more direct method of measuring ROS, because the number of RGCs needed to generate a sufficient signal would be impractical. Although our results with PEG-SOD and SOD, two very specific superoxide scavengers, were concordant with superoxide’s being measured by HEt and MitoSOX Red, we cannot completely exclude the involvement of other ROS. 

An explanation for the complex nature of the interaction between supposedly specific ROS scavengers and their targets in RGC cultures could be the presence of glial cells in mixed retinal cultures. Astrocytes increase survival of cocultured RGCs in ROS-generating systems. ^{39 40} Astrocytes can also modulate the toxic effects of ROS and thereby ameliorate the effects of extracellular ROS on RGC survival. ⁴¹ However, we addressed this possibility in three ways (Fig. 4) . First, we used purified RGC cultures and found that the effect of the ROS-generating systems were nearly identical with purified RGCs as on RGCs in mixed retinal cultures (vehicle, 100.0% ± 9.8%; H₂O₂, 51.6% ± 7.9%; paraquat, 44.5% ± 1.2%; and CPA, 23.9% ± 2.7%). This indicates that ROS are not being generated or converted by glial or other neural cells. Second, we performed fluorescent imaging of O₂ ⁻ generated by paraquat in individual RGCs to determine ROS levels within the cell and thereby distinguish autocrine from paracrine production. ²⁹ We did not study H₂O₂ because it is freely diffusible against cell membranes. Third, we generated superoxide with paraquat and used SOD to scavenge extracellular (but not intracellular) superoxide, finding that there was no effect on RGC superoxide levels, unlike what was observed with PEG-SOD. Together, these findings imply that our observations represent direct ROS effects on RGCs, and not those mediated by glial or neural cells in the culture. 

Finally, we ascertained sites for ROS generation in RGCs treated with paraquat and H₂O₂, using the mitochondria-specific superoxide sensor MitoSOX Red and the nonspecific superoxide sensor HEt (Fig. 4) . We found that paraquat (which generates superoxide via plasma membrane-associated NADPH) causes superoxide levels to increase in extramitochondrial compartments, but H₂O₂ (which we had previously shown leads to superoxide generation in axotomized RGCs ²⁹ ) causes superoxide generation in mitochondria. 

Conclusion

ROS scavenging by small molecules or proteins may contribute to our understanding of signaling pathways in neurons. Our data demonstrate a complex interaction of scavengers with ROS in the cellular milieu. Multiple antioxidant defense mechanisms, which occur in RGCs and other neurons, interfere with responses expected from the mechanism of action of scavengers in well-defined systems. Various approaches, which include use of scavengers for ROS and quantifying ROS, are needed in cell-free assays and in vitro cell cultures, to identify the role(s) of ROS in signaling cell death. Inferences about ROS generation and scavenging effects in cell-free assays and in vitro cell cultures must therefore be made cautiously and should be confirmed by measuring ROS levels on a single cell basis. 

This study identified many important mechanisms relevant to ROS in a biological system. ROS scavenging that occurs in cell-free assays can be compromised or altered in the presence of various cell types. In RGCs, other cell types do not affect ROS susceptibility and intracellular interactions. Last, we found that H₂O₂ induces mitochondrial O₂ ⁻ generation, whereas paraquat generates cytoplasmic O₂ ⁻. This finding highlights an important concept in ROS biology. We showed that the specificity of scavenging and ROS-generating systems is not strict and can involve the production of other ROS. For instance, certain antioxidants, such as U-74389G, have relatively little neuroprotective effect from oxidative stress. 

Therefore, conclusions regarding oxidative mechanisms of neuronal pathophysiology should only be made when a variety of ROS scavengers are used. We found that some ROS scavengers are less ROS specific than others and therefore are able to interact with several ROS and subsequent signaling pathways. Scavengers such as MnTMPyP, Trolox, and catalase are good candidates to test for neuroprotection as they demonstrated protection against multiple ROS. In models where an oxidative mechanism could be involved, these particular scavengers, alone or in combination, could be used to screen whether oxidative stress is involved in a neuronal degeneration. 

 Supported by National Eye Institute Grant R01EY12492, Retina Research Foundation, and an unrestricted departmental grant from Research to Prevent Blindness, Inc. LAL is a Research to Prevent Blindness Dolly Green Scholar.

 Submitted for publication July 31, 2005; revised January 23 and April 20, 2006; accepted June 26, 2006.

 Disclosure: C.R. Schlieve, None; C.J. Lieven, None; L.A. Levin, None

 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.

 Corresponding author: Leonard A. Levin, Department of Ophthalmology and Visual Sciences, University of Wisconsin Medical School, 600 Highland Avenue, Madison, WI 53792.

 

Figure 1.

 

Dose–response of four ROS-generating systems on RGC viability. Mixed retinal cultures were incubated for 24 hours in a ROS-generating system in defined medium. RGCs were identified by DAPI positivity and living RGCs by calcein-AM fluorescence. Results are expressed as the percentage of living compared with number of living RGCs in control conditions as the mean ± SEM. These results are representative of three independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001.

Table 1.

 

Rescue of RGCs by ROS Scavengers

	H2O2 (30 μM)	Paraquat (30 μM)	CPA (Cu 2μM, P 2 μM, A 100 μM)
50% Rescue
MnTMPyP	7 μM	—	18 μM
Trolox	23 μM	3.8 μM	85 μM
Deferoxamine	120 μM	3.4 μM	—
Catalase	0.62 U/mL	0.77 U/mL	21 U/mL
PEG-SOD	30 U/mL	78 U/mL	—
U-74389G	59 μM	—	—
90% Rescue
MnTMPyP	16 μM	—	53 μM
Trolox	63 μM	10 μM	—
Deferoxamine	160 μM	17 μM	—
Catalase	1.8 U/mL	2.3 U/mL	—
PEG-SOD	300 U/mL	300 U/mL	—
U-74389G	192 μM	—	—

 Mixed retinal cultures were incubated for 24 hours in a ROS-generating system with a scavenger in defined medium. RGCs were identified by DAPI positivity and living RGCs by calcein-AM fluorescence. Results are expressed as the geometric mean of concentrations rescuing 50% and 90% of the RGCs from ROS. A dash means that rescue did not occur at the maximum concentration tested.

Figure 2.

 

Detection of H₂O₂ or O₂ ⁻ generation with a red fluorescent H₂O₂ probe (Amplex Red; Invitrogen) and HEt. Increasing concentrations of H₂O₂ (A) or xanthine in the presence of 0.05 U/mL xanthine oxidase (B) were incubated for 30 minutes with the fluorescent probe or HEt, respectively. These results are representative of five independent experiments.

Figure 3.

 

Scavenging of H₂O₂ or O₂ ⁻ in cell-free assays did not correlate with their ability to prevent RGC death from the respective ROS. (A) To detect H₂O₂ scavenging in cell-free conditions, the ROS scavengers MnTMPyP (MNP), Trolox (TRX), deferoxamine (DFO), catalase (CAT), U-74389G (U-7G), or PEG-SOD (SOD) were incubated with 30 μM H₂O₂ for 30 minutes and the final H₂O₂ concentration measured with a fluorescent H₂O₂ probe (Amplex Red; Invitrogen). Concentrations tested were those that had been shown to rescue 50% (light gray) and 90% (dark gray) of RGCs exposed to H₂O₂. (B) To detect O₂ ⁻ scavenging in cell-free conditions, the same ROS scavengers were incubated with 1 mM xanthine and 0.05 U/mL xanthine oxidase for 30 minutes and the final O₂ ⁻ concentration measured with HEt. Concentrations tested were those that had been shown to rescue 50% (light gray) and 90% (dark gray) of RGCs from intracellular O₂ ⁻ generated by paraquat. (C) To detect H₂O₂ scavenging in cell culture, mixed retinal cultures exposed to 30 μM H₂O₂ in defined medium for 24 hours were incubated with ROS scavengers. Concentrations chosen were those that scavenged H₂O₂ by 50% (light gray) and 90% (dark gray) in cell-free assays (Table 1) . (D) To detect O₂ ⁻ scavenging in cell culture, mixed retinal cultures exposed to 1 mM xanthine and 0.05 U/mL xanthine oxidase in defined medium for 24 hours were incubated with ROS scavengers. Concentrations chosen were those that scavenged O₂ ⁻ generated by xanthine/xanthine oxidase by 50% (light gray) and 90% (dark gray) in cell-free assays. *Significantly (P < 0.05) different when a scavenger was present versus absent. Different patterns of H₂O₂ scavenging in cell cultures and cell-free assays were seen for Trolox, deferoxamine, and U-74389G. Different patterns of O₂ ⁻ scavenging in cell cultures and cell-free assays were seen for catalase and PEG-SOD. These results are representative of four to five independent experiments.

Figure 4.

 

Measured RGC ROS activity is not mediated by interactions with other neurons or glia. (A, C) Retrogradely labeled RGCs in mixed retinal cultures. (B) Immunoaffinity-purified RGCs. (A) Superoxide levels were measured in RGCs using the fluorescent probe HEt (A) or MitoSOX Red (Invitrogen) (C). Measurements were taken before and after treatment with vehicle, paraquat (30 μM), paraquat and SOD (300 U/mL), or H₂O₂ (9.5 mM). (A) Treatment with paraquat produced a significant increase in intracellular superoxide (2.44 ± 0.18 vs. 0.96 ± 0.07; P < 0.001), which was not significantly reduced by the presence of SOD (1.77 ± 0.29; P = 0.09). (B) Purified RGCs were cocultured with concentrations of the four ROS-generating systems previously tested in mixed retinal cultures (paraquat, 30 μM; H₂O₂, 30 μM; or CPA). RGCs were identified by DAPI-positive fluorescence, and viability was assessed with calcein-AM. Nearly identical cell viabilities were obtained in RGC-purified cultures, compared with mixed retinal cultures (Fig. 1) . (C) Superoxide generation by paraquat did not significantly increase fluorescence of MitoSOX Red, a mitochondria-specific probe for superoxide (1.56 ± 0.25 vs. 1.50 ± 0.37; P = 0.90). Treatment with H₂O₂, conversely, did cause a significant increase in red fluorescence (3.39 ± 0.65 vs. 1.28 ± 0.10; P < 0.05). (D) Superoxide from other neuronal or glial cells would be scavenged by extracellular SOD and therefore not affect ROS levels within the RGC. Scavengers that are RGC-permeable (e.g., PEG-SOD) are able to enter the cell and lower intracellular superoxide levels.