Differential production of superoxide by neuronal mitochondria

Background Mitochondrial DNA (mtDNA) mutations, which are present in all mitochondria-containing cells, paradoxically cause tissue-specific disease. For example, Leber's hereditary optic neuropathy (LHON) results from one of three point mutations mtDNA coding for complex I components, but is only manifested in retinal ganglion cells (RGCs), a central neuron contained within the retina. Given that RGCs use superoxide for intracellular signaling after axotomy, and that LHON mutations increase superoxide levels in non-RGC transmitochondrial cybrids, we hypothesized that RGCs regulate superoxide levels differently than other neuronal cells. To study this, we compared superoxide production and mitochondrial electron transport chain (METC) components in isolated RGC mitochondria to mitochondria isolated from cerebral cortex and neuroblastoma SK-N-AS cells. Results In the presence of the complex I substrate glutamate/malate or the complex II substrate succinate, the rate of superoxide production in RGC-5 cells was significantly lower than cerebral or neuroblastoma cells. Cerebral but not RGC-5 or neuroblastoma cells increased superoxide production in response to the complex I inhibitor rotenone, while neuroblastoma but not cerebral or RGC-5 cells dramatically decreased superoxide production in response to the complex III inhibitor antimycin A. Immunoblotting and real-time quantitative PCR of METC components demonstrated different patterns of expression among the three different sources of neuronal mitochondria. Conclusion RGC-5 mitochondria produce superoxide at significantly lower rates than cerebral and neuroblastoma mitochondria, most likely as a result of differential expression of complex I components. Diversity in METC component expression and function could explain tissue specificity in diseases associated with inherited mtDNA abnormalities.


Background
Maternally inherited mitochondrial DNA (mtDNA) codes for several mitochondrial proteins, primarily components of the mitochondrial electron transport chain (METC). Point mutations in mtDNA are causes of several diseases, including mitochondrial encephalomyopathy with lactic acidosis and stroke-like episodes (MELAS; position 3243 and others); neuropathy, ataxia, and retinitis pigmentosa (NARP; position 8993); and Leber's hereditary optic neuropathy (LHON; positions 11778, 3460, and 14484). LHON is the most common hereditary optic nerve disease causing severe visual loss. Most primary mutations code for proteins in METC complex I, although some secondary mutations are in complex III [1]. A fundamental paradox of diseases caused by mtDNA point mutations is that although all mitochondria in the body have the same mutation (when homoplasmic), the disease is usually expressed in a restricted number of tissues. In the case of LHON, the disease is almost exclusively manifested in retinal ganglion cells (RGCs), the axons of which make up the optic nerve. The reason for this cell-type specificity is unknown, but presumably is related to a particular susceptibility of RGCs to a consequence of these mutations.
The effect of LHON mutations has been studied by producing transmitochondrial cybrids in human cell lines, demonstrating upregulation of some mtDNA transcripts, e.g. aldose reductase [2]. LHON cybrids forced to use the METC as their primary energy source by plating on glucose-deficient galactose media, experience apoptotic death and increased cytochrome c release [3]. Common LHON mutations reduce the rate of respiration in LHON cybrids and oxygen consumption [4,5]. Most importantly, LHON cybrids have increased superoxide production compared to wild-type cells [6].
Given that METC complexes I and III are the main sources of basal superoxide production [7,8], it is possible that aberrant production of superoxide from mutated METC components would cause cell death. But why would abnormalities in superoxide production be a mechanism for the relatively specific effect of LHON mtDNA mutations on RGCs? One explanation is the finding that RGCs use superoxide as an intracellular signal for initiating the apoptosis program after apoptosis [9,10], probably by oxidizing critical sulfhydryls in signaling macromolecules [11,12]. Supporting this theory, RGC death is triggered when mitochondrial superoxide levels are increased by knocking down mitochondrial superoxide dismutase (SOD-2) [13]. Given that RGCs use superoxide for intracellular signaling, and that they are specifically affected in a disease of specific mtDNA mutations, we examined superoxide production in isolated RGC mitochondria, using the RGC-5 RGC cell line. We show that mitochondria derived from RGCs differ from cerebral and neuroblastoma mitochondria with respect to superoxide production in the presence of complex-specific METC substrates and inhibitors. Furthermore, we showed distinct patterns of METC component expression. Differences in superoxide production between neuronal cell types could prevent aberrant apoptosis signaling [10], and its disturbance with mtDNA mutations [6] could account for tissue-specific expression of disease phenotype.

Mitochondria can be isolated in bulk from RGC-5 cells
Because RGCs are post-mitotic and are present in relatively small (10 5 cells/retina in the rat) numbers, it is impractical to biochemically study RGC mitochondria in bulk. Instead, we used the RGC-5 cell line, which when Immunoblotting Measurement of Cytochrome c Oxidase in Differentiated and Undifferentiated RGC-5 Mitochondria Figure 1 Immunoblotting Measurement of Cytochrome c Oxidase in Differentiated and Undifferentiated RGC-5 Mitochondria. RGC-5 cells were differentiated with staurosporine. Isolated mitochondrial samples from undifferentiated and differentiated mitochondria standardized for protein content were compared to corresponding mitochondria depleted samples for the presence of the mitochondrial transmembrane protein cytochrome c oxidase (COX), complex IV. Samples were immunoblotted with antibody to subunit 1 of complex IV. There was considerable purification of mitochondria compared to the mitochondria depleted samples for differentiated and undifferentiated RGC-5 cells. Differentiated RGC-5 cells had 2.5 times more cytochrome c oxidase than undifferentiated RGC-5 cells.
differentiated are phenotypically similar to RGCs [14]. RGC-5 cells were grown in tissue culture and mitochondria isolated. To assess purity of mitochondrial isolation, mitochondrial-enriched and cytosolic fractions were immunoblotted for cytochrome c oxidase, demonstrating significant enrichment ( Figure 1). Degree of mitochondrial purification was similar among cell-types, based on quantitation with the fluorophore MitoTracker Green FM, which reacts with mitochondrial free sulfhydryls. Mitochondrial samples exhibited similar relative fluorescence values per mg protein ( Figure 2).

Cerebral and RGC-5 cells differ in superoxide production with complex I substrates and after complex I inhibition
Superoxide was indirectly measured by detecting H 2 O 2 generation over time as a result of spontaneous dismutation of superoxide by mitochondrial SOD-2 [15,16]. All results are expressed as the mean ± SEM (in nmol/min/mg protein) of 5 independent experiments, each in duplicate. A typical experiment is depicted in Figure 3. The basal level of superoxide production by RGC-5 mitochondria in the absence of substrate was 0.030 ± 0.004 nmol/min/mg protein, approximately one-seventh that of cerebral mitochondria (0.208 ± 0.033; p = 0.0017). Mitochondria were then incubated with glutamate (10 mM) and malate (5 mM), which yields NADH and serves as a substrate for complex I. In the presence of glutamate/malate there was a small but significant increase in superoxide production in cerebral (0.262 ± 0.028; p = 0.0042 compared to no substrate) and RGC-5 mitochondria (0.046 ± 0.004; p = 0.0005 compared to no substrate). When samples were then treated with the complex I inhibitor rotenone (6.7 µM) there was an insignificant change in the rate of superoxide production in both cerebral mitochondria (0.279 ± 0.035; p = 0.40 compared to glutamate/malate alone) and RGC-5 mitochondria (0.048 ± 0.004; p = 0.24) after the addition of rotenone. Nonetheless the rates of superoxide production per mg protein in all conditions were significantly lower in RGC-5 mitochondria than in cerebral mitochondria (Table 1).

Cerebral and RGC-5 mitochondria differ in superoxide production with complex II substrate and after complex III inhibition
Mitochondria were incubated with succinate (10 mM), which yields FADH 2 and serves as a substrate for complex II. In the presence of succinate, there was a small but significant increase in superoxide production in cerebral mitochondria (0.193 ± 0.027 to 0.216 ± 0.022; p = 0.04) and RGC-5 mitochondria (0.028 ± 0.005 to 0.031 ± 0.005; p = 0.024). Subsequent treatment with the com-   The rates of superoxide production were determined after the addition of the complex I substrates glutamate (10 mM) and malate (5 mM) and the complex I inhibitor rotenone (6.7 µM). Superoxide was indirectly measured using the H 2 O 2 probe Amplex Red in the presence of horseradish peroxidase (HRP). The values are given in nmol H 2 O 2 /min/mg protein ± standard error of the mean (SEM), with n being the number of independent samples for each mitochondria type.

Mitochondrial Quantification by MitoTracker Green
Superoxide Production in Cerebral, Neuroblastoma, and RGC-5 Mitochondria after Complex I Inhibition Figure 3 Superoxide Production in Cerebral, Neuroblastoma, and RGC-5 Mitochondria after Complex I Inhibition.
Mitochondria isolated from cerebral, neuroblastoma, and undifferentiated RGC-5 cells were standardized for protein content and 50 µl were added to a well in a 96-well plate containing 50 µL Amplex Red (AR). Prior to the addition of solutions containing mitochondria a baseline of superoxide production was established by measuring the fluorescent product of H 2 O 2 and AR, resorufin at approximately 10-second intervals. In a similar manner, production of superoxide was analyzed after the addition of the complex I substrates glutamate/malate and the complex I inhibitor rotenone. The mitochondrial superoxide production levels after the addition of glutamate/malate in cerebral cells were significantly higher than that of undifferentiated RGC-5 cells during basal metabolism, in the presence of glutamate/malate and after complex I inhibition with rotenone.  The rates of superoxide production were determined after the addition of the complex II substrate succinate (10 mM) and the complex III inhibitor antimycin A (0.5 µM). Superoxide was indirectly measured using the H 2 O 2 probe Amplex Red in the presence of horseradish peroxidase (HRP). The values are given in nmol H 2 O 2 /min/mg protein ± standard error of the mean (SEM), with n being the number of independent samples for each mitochondria type.
plex III inhibitor antimycin A (0.5 µM) resulted in a dramatic increase in the rate of superoxide production in cerebral mitochondria (0.540 ± 0.116; p = .023 compared to substrate alone), but not RGC-5 mitochondria (0.024 ± 0.004; p = 0.278). The rates of superoxide production in all conditions were substantially greater in cerebral mitochondria than in RGC-5 mitochondria (Table 2). A typical experiment is depicted in Figure 4.

Differentiation of RGC-5 cells does not normalize mitochondrial superoxide production
One explanation for the differences in mitochondrial superoxide production between cerebral and RGC-5 cells is the fact that RGC-5 cells are mitotic, while cerebral neurons are post-mitotic. To address the possibility that differences in proliferative state, metabolic activity, or degree of differentiation affected superoxide production by mitochondria, we differentiated RGC-5 cells with the broad spectrum kinase inhibitor staurosporine, which we have previously shown to induce a RGC phenotype without inducing apoptosis [14]. We then measured superoxide production rates as above (Tables 3 and 4).

RGC-5 cells generate significantly less superoxide than neuroblastoma SK-N-AS cells with complex I and complex II substrates
The differences between superoxide generation from RGC-5 and cerebral mitochondria could theoretically reflect differences in the source of cells, a cultured cell line in the former and fresh tissue in the latter. To rule out this possibility, we compared superoxide generation in RGC-5 mitochondria to another neuronal cell line, the SK-N-AS neuroblastoma line (Figures 3 and 4, Tables 1 and 2). As with cerebral cells, the basal superoxide production was much lower in RGC-5 cells compared to neuroblastoma cells (0.030 ± 0.004 vs. 0.318 ± 0.035; p = 0.0002). Superoxide generation from RGC-5 and SK-N-AS cells was measured after the addition of glutamate/malate. There was a large increase in superoxide production after the addition of glutamate/malate to SK-N-AS mitochondria (0.560 ± 0.054; p < 0.0001 compared to no substrate), similar to what was seen with cerebral and RGC-5 mitochondria. After the addition of rotenone there was a significant decrease in superoxide production in SK-N-AS cells (0.383 ± 0.050; p < 0.0001 compared to substrate alone), unlike what was seen with cerebral or RGC-5 cells.
In experiments where the complex II substrate was added to neuroblastoma mitochondria, there was a minimal change in production of superoxide (0.348 ± 0.042 to 0.307 ± 0.024; p = 0.08), similar to cerebral and RGC-5 mitochondria. The addition of antimycin A to neuroblastoma mitochondria resulted in a small but significant decrease in superoxide production (0.277 ± 0.034; p = 0.026 compared to substrate alone), similar to RGC-5 mitochondria but different from the increase in superoxide seen in cerebral mitochondria. In all cases, the rates of superoxide production by neuroblastoma mitochondria were much higher than RGC-5 mitochondria.

Mitochondria isolated from RGC-5, neuroblastoma SK-N-AS, and cerebrum differ in content of METC components
Most mitochondrial superoxide is produced by the METC.
To determine if differences in superoxide production among mitochondria from different neuronal subtypes could be caused by differences in the METC, we measured the relative amounts of METC complexes for protein concentration-corrected mitochondrial isolates. Immunoblots against specific METC complex subunits were scanned by densitometry and normalized to concentrations from cerebral mitochondria ( Figure 5 and Table 5). The most surprising finding was that RGC-5 mitochondria contained only one-sixteenth as much METC complex I as was present in cerebral cells. Complex II, III, and IV components were expressed at a molar concentration in RGC-5 mitochondria about half that of cerebral mitochondria. Neuroblastoma mitochondria had similar levels of complexes I-IV, ranging from 16% to 46% that of brain.
We considered the possibility that different physical properties of cells or mitochondria might produce differential purification of mitochondria. The primary organelles copurified with mitochondria are peroxisomes. To investigate the possibility that peroxisomal contamination could decrease the number of mitochondria with the preparations, we immunoblotted protein-corrected mitochondrial isolates for the peroxisomal protein PMP70. There were similar amounts of PMP70 in the mitochondria purified from RGC-5 cells, neuroblastoma cells, and cere-bral cells ( Figure 5 and Table 5), ruling out differential peroxisomal contamination as the cause of lower METC components in RGC-5 cells. We then compared mitochondrial isolates from the different cell types for a non-METC mitochondrial protein, the voltage-dependent anion channel (VDAC). There was a small decrease in VDAC expression in mitochondria isolated from RGC-5 cells in comparison to cerebral and neuroblastoma cells, but not enough to explain the significant difference in complex I expression.
We used real-time RT-PCR to compare mRNA levels of mitochondrial electron transport chain complexes I, II, III, IV, V, SOD-2, PMP-70, and VDAC to see if cell type-specific gene transcription levels correlated with protein levels. When normalized to VDAC expression, all METC subunits were downregulated in both undifferentiated and differentiated RGC-5 cells (Table 6). In general, differ-ences in mRNA levels between cell-types correlated with differences in protein levels. Surprisingly, complex I (alpha subunit, 9) transcription levels were higher in brain than RGC-5 cells, but this difference was not seen to the same degree in protein levels, measured by immunoblotting. On the other hand, complex IV (subunit 1) transcription levels were much lower than the difference in protein levels would predict ( Figure 6). Differentiation of RGC-5 cells had a minimal effect on transcription levels of all genes studied.

Discussion
Mitochondria isolated from the RGC-like RGC-5 cell line produce significantly less superoxide than cerebral and neuroblastoma mitochondria when incubated with complex I or II substrates or when treated with complex I or III inhibitors. Correspondingly, RGC-5 mitochondria contain significantly lower levels of complex I compared to Superoxide Production in Cerebral, Neuroblastoma, and RGC-5 Mitochondria after Complex III inhibition Figure 4 Superoxide Production in Cerebral, Neuroblastoma, and RGC-5 Mitochondria after Complex III inhibition. Mitochondria isolated from cerebral, neuroblastoma, and undifferentiated RGC-5 cells were normalized for protein content and superoxide production was analyzed after the addition of the complex II substrate succinate and the complex III inhibitor antimycin A. The mitochondrial superoxide production levels in cerebral cells were significantly higher than that of undifferentiated RGC-5 cells in the presence of succinate and after complex III inhibition with antimycin A. Both succinate and antimycin A elicited minimal superoxide production in undifferentiated RGC-5 cells, while in cerebral cells there was a near two-fold increase in superoxide production after the addition of antimycin A. Overall, the rate of superoxide production of RGC-5 cell mitochondria was much less than that of cerebral mitochondria for a given amount of mitochondrial protein.
Comparing the rates of production in the absence of substrates and inhibitors, basal cerebral mitochondria superoxide production was seven times of that of RGC-5 cells. The most notable differences after METC inhibition was the increased superoxide production in cerebral mitochondria in the presence of the complex III inhibitor antimycin A. After treatment with antimycin A, cerebral mitochondria had a more than 2-fold increase in superoxide production, compared to an insignificant decrease in superoxide production rate in RGC-5 mitochondria.
The difference in superoxide production is not a result of RGC-5 cells being undifferentiated mitotically active cells. We previously showed that low-dose staurosporine, a broad spectrum protein kinase inhibitor, can induce RGC-5 cells to differentiate to a phenotype that is similar to a mature RGC [14]. To see if differentiation affected ROS production, we compared superoxide production in differentiated and undifferentiated RGC-5 cells. Differentiated and undifferentiated RGC-5 produced similar amounts of superoxide in the presence of complex I substrates and inhibitors and only differed slightly in superoxide production after incubation with succinate, a complex II substrate, in the presence of the complex III inhibitor antimycin A. Differences in superoxide production between mitochondria from RGC-5 cells and cerebral cells were also not due to differences in the source of the tissue (a cell line vs. fresh tissue), because similar differences were seen when RGC-5 cells were compared to a neuroblastoma cell line. Finally, the differences were also not a result of differential purification of mitochondria, as the levels of the peroxisomal-associated protein PMP70 were similar across cell-type.
There are limitations to our study. We did not measure superoxide directly, instead indirectly measuring superoxide generation using the H 2 O 2 probe Amplex Red. Indirect superoxide measurement with a probe specific for H 2 O 2 is The rates of superoxide production were determined after the addition of the complex I substrates glutamate (10 mM) and malate (5 mM) and the complex I inhibitor rotenone (6.7 µM). Superoxide was indirectly measured using the H 2 O 2 probe Amplex Red in the presence of horseradish peroxidase (HRP). The values are given in nmol H 2 O 2 /min/mg protein ± standard error of the mean (SEM), with n being the number of independent samples for each mitochondria type. commonly performed [15,16] because mitochondrial superoxide dismutase (SOD-2) converts superoxide to H 2 O 2 , and the latter freely crosses organelle and cell membranes. Amplex Red reacts with H 2 O 2 in the presence of HRP and is converted to the fluorescent product resorufin.
In these experiments we detected changes in the levels of resorufin and used them to determine the production rates of superoxide. The production of hydrogen peroxide by complex I does not necessarily require dismutation [17]. suggesting that measuring levels of H 2 O 2 is an acceptable method for measuring superoxide production or leakage from complex I. Nonetheless, we cannot exclude the possibility that abnormally low SOD-2 activity biased our results.
We also examined superoxide production in mitochondria obtained from the RGC-5 cell line, and not from freshly isolated RGCs. It is possible that some of the differences that we measured in superoxide generation arose because the cerebral mitochondria were isolated from whole cerebral tissue and the RGC-5 mitochondria were isolated from a clonal cell line. The environment of the RGC-5 is normoxic, being exposed to atmospheric partial pressure of oxygen, compared to the relatively hypoxic environment within the brain. Also, cerebral mitochondria were isolated from a mixture of cell types, not only neuronal, but also glial cells, and this could blur differences between the cerebral and RGC-5 mitochondria.
Purifying a usable quantity of mitochondria from freshly isolated RGCs is impractical, and thus the use of a differentiable RGC cell line allows otherwise unfeasible mitochondrial studies. Difficulties in isolation of mitochondria from RGCs include the large numbers of animals necessary to obtain enough RGCs (there are approximately 10 5 RGCs per rat retina) and subsequently obtaining mitochondria from a small number of purified RGCs. It is impractical to do mitochondrial metabolic experiments from freshly isolated RGCs, as discussed above, and whole retina cannot be used because RGCs make up only a small fraction of the tissue.
Low levels of superoxide production in RGCs compared to cerebral and neuroblastoma cells could be due to decreased flow of oxygen through the METC, decreased leakage of superoxide from the METC, or increased scavenging via non-SOD mechanisms (increased scavenging via SOD would increase H 2 O 2 levels in our assay). Our experimental design could not distinguish these possibilities. The explanation is not simply one of lower METC density in RGCs. Immunoblotting demonstrated that there was markedly less mitochondrial complex I in RGC-5 cells compared to cerebral and neuroblastoma cells, while levels of complexes II, III, and IV were similar. Analysis of transcription levels confirms lowered gene expression of these complexes in RGC-5 cells compared to cerebral cells. If there were also lower amounts of other METC components in RGC-5 cells, then this could explain why RGC-5 mitochondria produce less superoxide. However, if the differences in superoxide were simply a matter of lower concentration of METC components, then this could explain differences in baseline superoxide production, but would not explain differences in patterns of superoxide induction, e.g. why there was such a large increase in superoxide production in cerebral mitochondria but not RGC-5 cells after treatment with antimycin A. It has been shown that superoxide production by isolated succinate dehydrogenase is highly dependent on succinate concentration, with superoxide generation increasing, then decreasing with succinate concentration [18]. Lower expression of succinate dehydrogenase and other METC components, as suggested by our immunoblotting and RT-PCR data, may lead to succinate blocking autooxidation at a lower concentration in RGC-5 cells compared to cerebral cells. Another possibility relates to the fact that antimycin A leads to increased superoxide production by backflow of electrons into complex I. If complex I levels are lower in RGC-5 cells, then this would lead to less superoxide production after antimycin A treatment in those cells.
We speculate that decreased production of superoxide by RGCs may have a physiological rationale. We previously demonstrated RGCs use superoxide as a mitochondrialderived intracellular messenger for signaling the initiation of apoptosis after axonal injury [10]. If RGCs use superoxide to signal cell death, then it is reasonable to presume The band density of mitochondrial related proteins were compared between cerebral, neuroblastoma, and RGC-5 derived mitochondrial samples for METC complexes I-IV. A comparison of VDAC and PMP70 expression were used to determine the presence of non-METC mitochondrial components and peroxisomal contamination, respectively. Cerebral, RGC-5, and neuroblastoma values are given as ratios compared to the band density of cerebral mitochondria.
Immunoblotting of Mitochondrial Components Figure 5 Immunoblotting of Mitochondrial Components. Isolated mitochondrial samples standardized for protein content were compared to corresponding mitochondria depleted samples for the presence of the mitochondrial membrane proteins Complexes I-IV, and VDAC. Standardized mitochondrial preparations were also investigated for the peroxisomal marker PMP70. Samples were subject to an electrophoresis and were transferred overnight to nitrocellulose. Samples were incubated with a mouse a primary antibody corresponding the protein of interest, and subsequently treated with an HRP-conjugated antibody. Each mitochondrial protein shows that there is considerable purification of mitochondria compared to the mitochondria depleted samples for cerebral, neuroblastoma, and undifferentiated RGC-5 cells. There is a nearly 18-fold increase in the expression of complex I in cerebral cells compared to RGC-5. Similar levels of expression in PMP70 and complex II were seen in all three samples.
that they also tightly regulate its intracellular concentration in relevant compartments. Otherwise there would be death of a critical (and non-renewable) class of neurons.
Loss of this tight regulation of superoxide production could also explain the mechanism of RGC death from mtDNA mutations in LHON. Although it is commonly believed that LHON results from RGCs having highenergy requirements, with mtDNA mutations causing RGC-specific death because the cell cannot meet its energy demands. However, this theory does not clearly address why there is little neuronal loss in other high-energy cells such as photoreceptors or cardiac muscle (as is seen in some mtDNA deletions), nor does it explain the delayed onset of LHON. Most importantly, there is a discrepancy between the degree of ATP deficiency and the severity of the optic neuropathy in cells with LHON mutations [19].
Instead, it is more likely that differences in ROS production and sensitivity to oxidative stress among cells types explains the timing and the specificity of LHON. There are precedents for cell type-specific differences in mitochondrial superoxide production. Rat brain and liver mitochondria differ in superoxide production in the presence of METC substrates and inhibitors [20]. Cardiac mitochondria from birds and mice have dramatically different rates and sites of H 2 O 2 production [21]. Finally, LHON cybrids in neuronal-like differentiated NT2 cells have greatly increased superoxide production [6]. It makes sense that ROS would be involved in LHON because the mutations are in known superoxide production sites [22], inhibition of these sites dramatically increases superoxide synthesis [15], and neuronal cybrids with LHON mutations have increased superoxide production [6]. In cerebral mitochondria, complex I produces the majority of superoxide radicals [16], and the same may be true for RGCs [10]. If RGC mitochondria have insufficient mechanisms for reducing levels of superoxide production, particularly in response to METC inhibition, then mutations of critical complex I components could theoretically lead to catastrophic superoxide production and anomalous signaling of cell death in the absence of axonal damage [10]. Lower METC component expression and basal superoxide production in RGCs may be accompanied by a similar downregulation in SOD2 or other ROS-reducing systems, leading to a phenotype which is overwhelmed more readily than other cells in the event of aberrant superoxide generation.

Conclusion
Mitochondria from a RGC-like cell line produced superoxide at a much lower rate than cerebral or neuroblast- Changes in transcription levels of each gene studied are given in terms of the difference in threshold cycle compared to cerebral cells(∆C t ), the difference in threshold cycle relative to that of VDAC (∆(∆C t )), and the relative expression level, given as a percent of cerebral and normalized to VDAC expression.
oma mitochondria, and there was a dramatic difference in superoxide production when electrons were shunted to complex I using METC complex I and III substrates and inhibitors. These differences were mirrored by different patterns of METC component expression between cell types. Decreased superoxide production may be essential for preventing aberrant signaling of cell death in RGCs.
Mutations coding for specific components of NADH-ubiquinone oxidoreductase in LHON may disrupt the cell type-specific handling of superoxide in RGC mitochondria and lead to premature cell death. This could explain why LHON mtDNA mutations are present in all cells, but the clinical phenotype is predominantly one of RGC death.

Animals
Cerebral mitochondria were obtained from adult Long-Evans rats. All animals were used in accordance with federal, state, and institutional guidelines for the use of animals in laboratory research.

RT-PCR Analysis of Expression of Components of the Mitochondrial Electron Transport Chain
Total RNA was isolated from 0.5-1.0 × 10 6 RGC-5 cells, differentiated RGC-5 cells, and rat brain using the RNeasy Protect Mini Kit (Qiagen; Valencia, CA). Freshly prepared RNA samples were immediately transcribed to cDNA using the iScript cDNA Synthesis Kit (Bio-Rad; Hercules, CA) and quantitative real-time PCR was carried out with primers to each gene subunit analyzed by Western blotting (Table 7) using the iScript One-Step RT-PCR Kit with SYBR Green, omitting reverse transcriptase. Primers were either taken from previously published papers or designed using Primer Express (Applied Biosystems; Foster City, CA). Specificity was confirmed by searching against rat genomic and expression nucleotide databases. Samples were cycled 45 times for 15 sec at 95°C and 30 sec at 55°C, followed by 1 min at 95°C and 1 min at 55°C. Melt curve analyses were performed, and single peaks observed in all cases.
Threshold cycle (C t ) was computed automatically by the iCycler software (Bio-Rad), and compared across all conditions. Delta C t (∆C t ) was computed for each gene with respect to brain for each cell line studied, and normalized for VDAC expression by subtracting the ∆C t for VDAC in a particular cell type from the ∆C t for each other gene in that cell type (∆(∆C t )). This difference indicates the -(log2) difference in RNA copy number.

Superoxide Measurements
Superoxide was indirectly measured by assaying H 2 O 2 [15,16], the product of superoxide dismutation by mitochondrial SOD-2. The fluorescent product of Amplex Red and H 2 O 2 was analyzed in various conditions in duplicate to quadruplicate. All assays were performed in 96-well plates and read with a Wallac Victor 2 1420 multilabel counter using the appropriate filters (excitation 535 nm/ emission 580 nm). To each well Amplex red solution was added and a time-lapse fluorescent baseline was established in the absence of mitochondria. Consecutively, to the well containing Amplex red was added an equal volume of cerebral, RGC-5, or neuroblastoma mitochondria, a substrate (glutamate/malate or succinate), and an inhibitor of the METC (rotenone or antimycin A). All figures are expressed in nmol/min/mg protein, assuming full superoxide dismutation. All experiments were repeated 2-4 times.

Statistical Analysis
Comparisons between different groups were by Student's unpaired t-test, while comparisons between serial treatments of mitochondria with substrate or inhibitors were by Student's paired t-test. Significant differences required p < 0.05.