Adaptation of neuronal cells to chronic oxidative stress is associated with altered cholesterol and sphingolipid homeostasis and lysosomal function


Address correspondence and reprint requests to Angela B. Clement and Christian Behl, Institute for Pathobiochemistry, University-Medical-Center, Johannes Gutenberg-University Mainz, Duesbergweg 6, Mainz 55099, Germany. E-mail: and


Chronic oxidative stress has been causally linked to several neurodegenerative disorders. As sensitivity for oxidative stress greatly differs between brain regions and neuronal cell types, specific cellular mechanisms of adaptation to chronic oxidative stress should exist. Our objective was to identify molecular mechanisms of adaptation of neuronal cells after applying chronic sublethal oxidative stress. We demonstrate that cells resistant to oxidative stress exhibit altered cholesterol and sphingomyelin metabolisms. Stress-resistant cells showed reduced levels of molecules involved in cholesterol trafficking and intracellular accumulation of cholesterol, cholesterol precursors, and metabolites. Moreover, stress-resistant cells exhibited reduced SMase activity. The altered lipid metabolism was associated with enhanced autophagy. Treatment of stress-resistant cells with neutral SMase reversed the stress-resistant phenotype, whereas it could be mimicked by treatment of neuronal cells with a specific inhibitor of neutral SMase. Analysis of hippocampal and cerebellar tissue of mouse brains revealed that the obtained cell culture data reflect the in vivo situation. Stress-resistant cells in vitro showed similar features as the less vulnerable cerebellum in mice, whereas stress-sensitive cells resembled the highly sensitive hippocampal area. These findings suggest an important role of the cell type-specific lipid profile for differential vulnerabilities of different brain areas toward chronic oxidative stress.

Abbreviations used:



ATP-binding cassette 1


Alzheimer’s disease


fetal calf serum




heat-shock protein 70


lysosomal-associated membrane protein


light chain 3B


low-density lipoprotein receptor-related protein




Niemann-Pick disease, type C


neutral sphingomyelinase


phosphate-buffered saline


sodium dodecyl sulfate

Oxidative stress is caused by the accumulation of reactive oxygen species which damage cellular components like proteins and lipids and, therefore, disturb cellular processes. Stress induced by oxidative stimuli has been linked to a variety of neurodegenerative disorders including Alzheimer’s disease (AD), where elevated markers for oxidative stress were found in affected brain regions (Hensley et al. 1995). In AD, neuronal cell death is observed mainly in pyramidal neurons of the entorhinal and the inferior temporal cortex, the hippocampus and the amygdala whereas neurons in other brain regions such as the cerebellum are spared (Hensley et al. 1995; Gomez-Isla et al. 1996, 1997). The molecular basis of the differential sensitivity and especially of the resistance to oxidative stress of certain neurons is largely unknown.

One plausible explanation would be a specific ability of cells to adapt to high reactive oxygen species levels. It is well known that oxidative stress has dramatic influences on biological membranes by the peroxidation of lipids, which in turn results in alteration of membrane fluidity and function of membrane embedded proteins (Mecocci et al. 1996). Thus, specific membrane compositions could render neurons more or less resistant to oxidative stressors. In this regard, cholesterol, sphingomyelin, and their derivatives play important roles for the regional organization of transmembrane and membrane-associated proteins thereby modulating intracellular transport, signal transduction, and metabolism (for review, see Sprong et al. 2001).

The goal of this study was to determine the role of the cellular lipid composition and its influence on cellular transport processes in oxidative stress resistance. For our studies, we used a clonal hippocampal cell line resistant to H2O2-induced oxidative stress, HT22H2O2. HT22H2O2 cells were established by long-term treatment of the parental HT22 (HT22WT) cells with sublethal doses of H2O2. These cells show a cross-resistance to other oxidative stressors, such as glutamate, but not to non-oxidative neurotoxins like staurosporine, sphingosine, and Gramicidin A. Glycogen synthase kinase 3β has been identified as one important mediator of stress resistance in these cells (Schafer et al. 2004). Increased activities of antioxidant enzymes like glutathione peroxidase and catalase might also partially account for oxidative stress resistance (Behl et al. 1994; Zitzler et al. 2004).

To elucidate additional factors that may mediate stress resistance, we compared ΗΤ22Η2Ο2 cells with the stress-sensitive HT22WT cells, focussing on the cellular lipid profile and on the influence of lipid profile alterations on the endosomal/lysosomal system and vice versa. We found major differences in the lipid composition and lysosomal function between HT22H2O2 and HT22WT cells and verified these findings in another oxidative stress-resistant cell line that was established by long-term treatment with glutamate and that shows cross-resistance to H2O2 (Schafer et al. 2004). Finally, mouse hippocampus as a model tissue for a stress-sensitive brain area was compared to relatively stress-resistant cerebellar tissue to address the physiological relevance of our data.

Experimental procedures


All utilized antibodies and their sources are described in Table S1.

Cell culture

HT22 cells (here called HT22WT) are cloned murine hippocampal nerve cells that are very susceptible to oxidative stress (Li et al. 1997). Initial isolation of HT22H2O2 and HT22Glu cells has been presented in detail elsewhere (Schafer et al. 2004). Cells were cultured in Dulbecco’s modified Eagle’s medium containing 10% fetal calf serum (FCS), 1 mM sodium pyruvate, and 1x penicillin/streptomycin (Invitrogen, Karlsruhe, Germany). To maintain the stable phenotype, 450 μM H2O2 (Sigma, Deisenhofen, Germany) or 40 mM glutamate (Sigma) was added twice a week to HT22H2O2 or HT22Glu cells, respectively. Prior to when experiments were performed, cells were cultured for 10 days without toxins and medium was exchanged daily to remove residual toxins.

Quantitative real-time reverse transcription-PCR analysis

Total RNA from subconfluent cell cultures was extracted using the NucleoSpin RNA II Kit (Macherey-Nagel, Dueren, Germany). Reverse transcription was performed on 1 μg total RNA in a reaction volume of 20 μL containing 2 μL reverse transcriptase buffer, 2 μL (5 mM) dNTPs, 4 U Omniscript Reverse Transcriptase (all Qiagen, Hamburg, Germany), 2 μL oligo(dT)23 primer (10 μM; Sigma) and 10 U RNasin (Promega, Mannheim, Germany). Synthesis of cDNA was carried out for 60 min at 37°C. Quantitative real-time PCR was performed in a 25 μL reaction volume containing 1 μL cDNA, 0.5 μL (100 pmol) sense and antisense primers (Table S2; MWG-Biotech AG, Ebersberg, Germany), and 12.5 μL of 2x SYBR Green PCR buffer consisting of 2x PCR buffer (Qiagen), 1 U HotStarTaq Polymerase (Qiagen), 4 mM MgCl2, 0.5x SYBR Green I (Sigma), 0.4 mM dNTPs, 20 nM fluorescein (Invitrogen), 1.6% glycerol, and 0.3% Triton X-100. PCR was performed and monitored in the iCycler real-time thermocycler (Bio-Rad, Munich, Germany) for 35 cycles of amplification following an initial denaturation step at 95°C for 15 min. PCR cycle conditions were 95°C for 15 s, 60°C for 20 s, and 72°C for 30 s. The PCR cycle number that generated the first fluorescence signal above threshold (CT) was determined. The generation of specific PCR products was confirmed by melting curve analysis. The relative expression ratio R of target genes in HT22H2O2 cells compared with HT22WT cells was calculated with the following formula: inline image, with inline image of target genes and inline image of reference genes. L19 was used as reference gene. Data obtained from quantitative real-time PCR analysis were applied to the relative expression software tool (Pfaffl et al. 2002) to test for significance by a randomization test. Statistical significance was accepted at a level of < 0.05.

Measurement of cholesterol, cholesterol precursors, and metabolites

Cultured cells were washed three times with ice-cold phosphate-buffered saline (PBS) and cholesterol was extracted with chloroform/methanol (2 : 1; v/v). Analysis of cholesterol, its precursors, and metabolites were performed by high-specific and selective combined GLC–mass spectrometry as described previously (Lutjohann et al. 2002).

Measurement of sphingomyelin, ceramide, and neutral sphingomyelinase activity

Cultured cells were washed three times with ice-cold PBS, resuspended in 200 μL PBS and disrupted by sonication. Extracts of hippocampal or cerebellar tissue of 4-month-old C57/BL6 mice were generated by homogenization of tissue samples in 50 mM Tris–Cl, pH 6.8. Sphingomyelin and neutral sphingomyelinase (nSMase) activity were determined using the Amplex-Red SMase assay kit from Molecular Probes (Eugene, OR, USA); 200 μg protein lysate in 100 μL volume were used per test. Total sphingomyelin concentrations were determined enzymatically as described (Grimm et al. 2005). For detection of ceramide 100 μg of total protein lysate were extracted with chloroform/methanol (2 : 1; v/v) and separated on TLC silica gel 60 F254 plates (Merck, Darmstadt, Germany) using toluene/methanol (7 : 3, v/v) as mobile phase. The plates were developed using phosphomolydic acid ready-to-use solution (Sigma) and charring at 160°C. Ceramide from bovine brain was run in parallel as standard (Sigma). Ceramide content was quantified by aida image software (Raytest, Straubenhardt, Germany).


To localize cholesterol, cells were labeled with filipin (Sigma), which specifically binds to unesterified cholesterol. Cells were fixed with 4%p-formaldehyde for 15 min, treated with 1.5 mg glycine/mL PBS for 10 min, and stained with 0.05 mg filipin/mL PBS/10% FCS for 2 h. For double labeling with the lysosomal-associated membrane protein marker (Lamp1), Lamp1 antibody was added to the filipin solution and detected by Cy3-labeled anti-rat antibodies. For labeling of lysosomes in living cells, cells were incubated with 50 nM LysoTracker Red (Invitrogen) in medium for 60 min at 37°C. For staining of mouse tissue, 10 nm cryosections through the hippocampus and cerebellum of 4-month-old p-formaldehyde perfused male C57/BL6 mice were blocked in PBS containing 10% FCS and stained with anti-Lamp1 antibody for 24 h at 4°C. Images were captured on a Zeiss Axiophot 200 microscope (Carl Zeiss MicroImaging GmbH, Göttingen, Germany) at 365 nm excitation and 397 nm emission for filipin and 546 nm excitation and 590 nm emission for Cy3 and LysoTracker Red.

Western blot analysis

Cells were lysed in 62.5 mM Tris, pH 6.8, 2% sodium dodecyl sulfate (SDS), 10% sucrose, 5 mM EDTA, and protease inhibitor cocktail (Sigma). Cell extracts were sonicated and proteins were denatured for 5 min at 95°C. Protein extracts of hippocampal or cerebellar tissue of 4-month-old C57/BL6 mice were generated by homogenization and subsequent brief sonication of tissue samples in 50 mM Tris–Cl, pH 6.8, and protease inhibitor cocktail. For the determination of low-density lipoprotein receptor-related protein (LRP), rab7, heat-shock protein 70 (HSP70), light chain 3B (LC3), cathepsin L, and seladin1 20 μg protein/lane were separated by SDS–polyacrylamide gel electrophoresis on 4–12% Bis–Tris-gradient gels (Invitrogen) and analyzed by western blotting with the appropriate antibodies. For Lamp1 detection 8% Tris–glycine gels were used. Western blot signals were quantified by aida image software (Raytest). Actin, tubulin, or glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were used to ensure equal protein loading of cell lysate samples.

Determination of the autophagic activity

The autophagic activity in HT22WT and HT22H2O2 cells was measured by monitoring the turnover of the autophagosome marker LC3II, a protein of the outer phagosomal membrane. LC3II is generated from its cytosolic precursor LC3I by conjugation with phosphatidylethanolamine. The more autophagosomes are formed, the more LC3II is degraded in autolysosomes, and therefore, turnover of LC3II can be used as a measure for autophagic activity (Tanida et al. 2005). To analyze the lysosomal degradation of LC3II, we incubated HT22WT and HT22H2O2 cells for 2 h at 37°C in the absence or presence of 10 μg/mL E64d, a general cysteine protease inhibitor (to control for lysosomal LC3II degradation) and subsequently analyzed the extent of LC3II degradation in the 2 h treatment period by western blot. To calculate the amount of degraded LC3II (as a measure for autophagic activity), for each cell line the amount of LC3II in the absence of E64d was subtracted from the corresponding amount of LC3II in the presence of E64d (LC3II values were normalized to tubulin).

Cell survival assay

The influence of drug treatment on cell survival was measured via the ability of cells to reduce 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazoliumbromide (MTT). Briefly, cells were pre-treated with 25 mU/mL SMase from Bacillus cereus (Sigma) or the specific nSMase inhibitor 3-O-methyl-sphingomyelin (3-O-Me-SM; Sigma, 20 μM) in serum-free medium for 2 h, respectively. Subsequently, H2O2 (500 μM for HT22WT and 2 mM for HT22H2O2 cells, respectively) was added and incubated for additional 22 h. After drug treatment, MTT (5 mg/mL) was added to each well and incubated for another 3 h. To dissolve the reduced MTT formazan, two volumes solubilization solution (0.1 g/mL SDS, 50% dimethylformamide, pH 4.1 with acetic acid) were added to each well. After 4 h, absorbance was measured at 560 nm. Absorbance of untreated control wells was set as 100%.

Statistical analysis

Unless stated otherwise, statistical significance was determined by the unpaired two-tailed Student’s t-test using the sigmastat software (Systat Software GmbH, Erkrath, Germany); significance was set at * 0.05, ** 0.01, and *** 0.001.


Cholesterol metabolism is changed in HT22H2O2 cells

Cholesterol has been shown to be a determinant of cell stability and survival (Zhuang et al. 2005). Therefore, we performed a GLC–mass spectrometry analysis to quantitatively profile cholesterol, cholesterol precursors and metabolites in oxidative stress-sensitive HT22WT and stress-resistant HT22H2O2 cells. Levels of the cholesterol precursors lanosterol, lathosterol, follicular fluid meiosis-activating sterol, and desmosterol were significantly increased in HT22H2O2 cells. The strongest increase (∼5-fold) was observed for desmosterol (Fig. 1a). Total levels of cholesterol were also significantly increased in HT22H2O2 cells, albeit to a smaller extent (Fig. 1b). We also measured levels of oxysterols and found an increase in 24-OH-cholesterol and 27-OH-cholesterol in HT22H2O2 cells (Fig. 1c). Reduced conversion of desmosterol into cholesterol by the responsible enzyme seladin1 (also termed 24-dehydrocholesterol reductase) might be responsible for the pronounced accumulation of desmosterol in HT22H2O2 cells (Waterham et al. 2001). Indeed, we found reduced mRNA and protein levels of seladin1 in HT22H2O2 cells (Fig. S1).

Figure 1.

 Levels of cholesterol precursors, free cholesterol, and cholesterol metabolites are increased, HMG-CoA reductase, ABCA1, rab7, and LRP are down-regulated in HT22H2O2 cells compared with HT22WT cells. Estimation of cholesterol precursors (a), of total cholesterol (b), and of cholesterol metabolites (c) in HT22WT and HT22H2O2 cells by GLC–mass spectrometry analysis. Values expressed are mean ± SD (*p ≤ 0.05, **p ≤ 0.01, and ***p ≤ 0.001, n = 4–5). (d) mRNA levels of HMG-CoA reductase and ABCA1 were analyzed by reverse transcription real-time PCR. In the diagram, the reduction of HMG-CoA reductase and ABCA1 mRNA is given as mean ± SEM from three independent experiments carried out in triplicate (*p ≤ 0.05). (e) The protein levels of rab7 and (f) of the Furin-cleaved β-fragment of LRP (LRPβ) and the C-terminal LRP fragment (LRPCTF) were determined by densitometric analysis of western blots from total cell lysates using rab7 and LRP specific polyclonal antibodies. Shown are representative western blots. The corresponding diagrams show the relative protein amount as mean ± SEM (***p ≤ 0.001) from at least three independent experiments.

Oxysterols are potent repressors of genes involved in cholesterol synthesis or uptake (Tam et al. 2006; Bjorkhem 2008). Therefore, we determined mRNA levels of the rate-limiting enzyme of cholesterol biosynthesis, 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA) reductase by real-time PCR. Indeed, HMG-CoA reductase mRNA levels were reduced in HT22H2O2 cells (Fig. 1d), which is (i) in line with the repressive effects of oxysterols on cholesterol synthesis and (ii) indicates that cholesterol accumulation in HT22H2O2 cells is not caused by increased cholesterol synthesis. Reduced HMG-CoA reductase levels might rather indicate an adaptive response of HT22H2O2 cells to increased cholesterol levels.

Cholesterol accumulation could be caused by inefficient cholesterol efflux. Therefore, we assessed the mRNA levels of the cholesterol efflux transporter ATP-binding cassette 1 (ABCA1), which regulates the rate-controlling step in the removal of cellular cholesterol by transferring cholesterol and phospholipids to an apolipoprotein acceptor (Fitzgerald et al. 2004). Mutations in ABCA1, as seen in Tangier disease, result in accumulation of cellular cholesterol (Bodzioch et al. 1999). ABCA1 mRNA levels were reduced in HT22H2O2 cells (Fig. 1d), suggesting that reduced efflux could contribute to the observed cholesterol accumulation.

Altered cholesterol trafficking in HT22H2O2 cells is also indicated by the reduced expression of rab7 (Fig. 1e), a marker of late endosomes and lysososmes, that is involved in lipid transport and lysosomal maturation (Lebrand et al. 2002). In addition, LRP, which is responsible for cellular cholesterol uptake, is reduced in HT22H2O2 cells (Fig. 1f). We observed down-regulation of the LRP β-fragment, which results from furin cleavage, and its C-terminal fragment, resulting from γ-secretase cleavage. LRP down-regulation in HT22H2O2 cells may be an additional adaptive change of HT22H2O2 cells to prevent further cholesterol uptake.

Cholesterol accumulates in lysosomes in HT22H2O2 cells

To assess the localization of accumulated cholesterol in HT22H2O2 cells, we performed stainings with filipin, which specifically detects free unesterified cholesterol. In HT22WT cells, filipin was mainly detected at the plasma membrane, and only weak intracellular staining was observed. In HT22H2O2 cells, we observed a strong perinuclear cholesterol accumulation and a reduced plasma membrane staining when compared with HT22WT cells (Fig. 2a). These stainings indicate a shift of cholesterol toward intracellular compartments in HT22H2O2 cells compared with HT22WT cells. Co-immunostainings for the lysosomal marker Lamp1 revealed that the relatively large cholesterol-filled organelles in HT22H2O2 cells are lysosomes (Fig. 2b). In contrast, little colocalization of filipin with Lamp1 was detected in HT22WT cells (Fig. 2b). The pronounced perinuclear concentration of Lamp1 staining in HT22H2O2 cells instead of the normal dispersed localization throughout the cytoplasm like in HT22WT cells indicates changes in the lysosomal compartment.

Figure 2.

 Cholesterol accumulation in lysosomes correlates with increased autophagy in HT22H2O2 cells. (a) HT22WT and HT22H2O2 cells were stained by filipin to detect free cholesterol (excitation: 365 nm and emission: 397 nm; here shown in green). Note that in HT22WT cells filipin labeled mainly the plasma membrane (left) whereas in HT22H2O2 cells free cholesterol was found in enlarged vesicular structures in the perinuclear region (right); bar, 20 μm. (b) Co-immunostaining with the lysosomal marker Lamp1 (red) showed dispersed lysosomes in HT22WT cells (upper panel) which do not accumulate cholesterol as indicated by the lack of colocalization of filipin (green) and Lamp1 in these cells (merged image). In HT22H2O2 cells strong colocalization between filipin and Lamp1 staining was evident indicating lysosomal accumulation of cholesterol (yellow dots surrounding the nucleus in the merged image, labeled by a white arrow). Note that in (b) exposure for filipin staining of HT22H2O2 cells was lower than for HT22WT cells to avoid overexposure of filipin labeled structures. Therefore, the weak plasma membrane staining of HT22H2O2 cells by filipin is nearly undetectable in this picture; bar, 20 μm. (c) Western blot analysis for the determination of Lamp1 and (d) of HSP70 levels in HT22WT and HT22H2O2 cell extracts. (e) For determination of the autophagic activity the amount of LC3I and LC3II in HT22WT and HT22H2O2 cell extracts after 2 h incubation with or without E64d was analyzed by western blot employing a polyclonal LC3 antibody. For each protein, one representative blot is shown (n = 3–6).

Stabilized lysosomes and enhanced autophagy in HTH2O2 cells

The observed perinuclear localization of Lamp1 and cholesterol in HT22H2O2 cells prompted us to analyze the lysosomal compartment in more detail. Western blot analysis for Lamp1 showed an increase of this protein in HT22H2O2 cells (Fig. 2c). Lamp1 and Lamp2 are the most abundant lysosomal membrane proteins which together with other highly glycosylated proteins form a carbohydrate-covered coat over the lumenal surface of the lysosomal membrane. The observed increase in Lamp1 could indicate a higher stability of this protein (Kundra and Kornfeld 1999). Furthermore, we observed a dramatic increase in the protein amount of the antiapoptotic stress-inducible chaperone HSP70 in HT22H2O2 cells compared with HT22WT cells (Fig. 2d). HSP70 is also known to prevent the disruption of lysosomes by H2O2 (Nylandsted et al. 2004). Notably, HSP70 is a protein that is localized in cholesterol- and sphingomyelin-enriched membrane subdomains (lipid rafts), and thus its expression is strongly influenced by the modulation of sphingomyelin or cholesterol levels (Ledesma et al. 2003).

Perinuclear enrichment of Lamp1-positive vacuoles could indicate enhanced autophagy, an important cellular response to starvation or stress conditions to maintain cell viability, where cellular components are degraded inside lysosomes to remove damaged cellular components and/or to provide nutrients (for review, see (Mizushima and Klionsky 2007). Alternatively, accumulation of Lamp1-positive vacuoles could also point to a block in autophagy because of maturation deficits of lysosomes. To distinguish between these two possibilities, we tracked the expression of the autophagosome marker LC3II. HT22H2O2 cells showed a higher expression of LC3I and LC3II than HT22WT cells (Fig. 2e). As the observed increase in LC3II might either indicate enhanced formation of autophagosomes or reduced lysosomal degradation of LC3II, we analyzed the LC3II turnover in HT22WT and HT22H2O2 cells. We therefore measured the degradation of LC3II in each cell type during a 2 h incubation period in the presence of the protease inhibitor E64d (Fig. 2e). We observed a higher autophagic activity (indicated by a higher amount of degraded LC3II during the 2 h time frame) in HT22H2O2 cells (237.4 ± 38.9%, p ≤ 0.05, paired t-test, n = 6) compared with HT22WT cells (considered as 100%). A higher autophagic activity in HT22H2O2 cells may help to adapt to chronic oxidative stress and thus support cell survival.

Sphingomyelin levels are increased in HT22H2O2 cells

The regulation of the cholesterol and sphingolipid metabolism is tightly linked. As it is known from sphingolipid storage diseases, increased cellular cholesterol alters sphingolipid trafficking such that sphingolipids accumulate in the endosomal/lysosomal pathway (Reagan et al. 2000). Excess sphingomyelin, in turn, disrupts normal cholesterol trafficking to the cell periphery and cholesterol efflux (Nagao et al. 2007). Because of the observed alterations in cholesterol metabolism in HT22H2O2 cells we expected to see changes also in the cellular sphingomyelin content. Therefore, we determined the sphingomyelin level and the activity of SMase, which degrades sphingomyelin to ceramide, a known pro-apoptotic compound. HT22H2O2 cells showed a reduced activity of nSMase and an increased level of sphingomyelin (Fig. 3a and b) which is consistent with an interdependence of cholesterol and sphingolipid metabolism. The reduced activity of nSMase correlated with decreased ceramide levels (Fig. 3c). A decreased intracellular production of ceramide may result in less cellular apoptosis.

Figure 3.

 Activity of neutral sphingomyelinase and ceramide levels are down-regulated and sphingomyelin levels are increased in HT22H2O2 cells. Activity of neutral sphingomyelinase (a) and the amount of sphingomyelin (b) in total lysates of HT22WT and HT22H2O2 cells were measured using a commercially available kit. Values are expressed as mean ± SEM (*p ≤ 0.05 and ***p ≤ 0.001) from four experiments carried out in triplicate for (a) and from six experiments carried out in duplicate for (b). Ceramide content was analyzed by TLC. Values are expressed as mean ± SEM (*p ≤ 0.05, n = 4).

Treatment of HT22H2O2 cells with SMase partially reverses the H2O2-resistant phenotype while treatment of HT22WT cells with an nSMase inhibitor renders cells more resistant to H2O2

Accumulation of cholesterol and sphingomyelin have been described as rather detrimental for cell survival, e.g. in lipid storage diseases (Liao et al. 2007). However, the observed accumulation of cholesterol and sphingomyelin in the stress-resistant HT22H2O2 cells suggests that a higher cholesterol and sphingomyelin amount could contribute to improved protection toward a second hit by an oxidative stressor. Based on the observed up-regulation of sphingomyelin and down-regulation of nSMase activity in HT22H2O2 cells we hypothesized that treatment of HT22H2O2 cells with SMase might reverse the resistant phenotype and render cells more sensitive to H2O2. To prove this hypothesis we treated HT22H2O2 cells for 24 h with 25 mU/mL SMase which significantly reduced the sphingomyelin content (Fig. 4c). Stainings with LysoTracker Red, anti-Lamp1 antibody and filipin showed an overall loss of lysosomal labeling after SMase treatment, whereby especially the HT22H2O2-typical large perinuclear vacuoles, that are clearly visible in untreated HT22H2O2 cells, were lost (Fig. 4a). In line with the cytochemical data, western blot analysis showed a SMase-induced reduction in Lamp1 expression and the appearance of a faster migrating proteolytic cleavage fragment of Lamp1, which indicates a reduced stability of Lamp1 (Fig. 4b). Furthermore, we observed reduced expression of HSP70, which is involved in the stabilization of lysosomes (Fig. 4b). In addition, western blot analysis of protein extracts from SMase-treated HT22H2O2 cells revealed increased levels of rab7 and LRP (Fig. 4b), which may contribute to lysosomal maturation (Harrison et al. 2003), and the release of lysosomal enzymes into the cytosol, respectively (Ji et al. 2006). Moreover, HT22H2O2 cells were also more sensitive toward H2O2. MTT assays showed enhanced H2O2-induced cell death of HT22H2O2 cells after SMase treatment compared with untreated cells (Fig. 4d). In summary, after SMase treatment, the morphology, various biochemical features and the stress sensitivity of HT22H2O2 cells partially resemble those of HT22WT cells.

Figure 4.

 Treatment of HT22H2O2 cells with SMase partially reverses the H2O2-resistant phenotype. (a) HT22H2O2 cells were treated for 24 h with 25 mU/mL SMase in serum-free medium and subsequently incubated for 60 min with LysoTracker Red to label lysosomes. Staining with Lamp1 antibody and filipin was performed after fixation of SMase-treated cells with 4%p-formaldehyde. The white arrows point to large perinuclear vesicular structures; bar, 20 μm. (B) Western blot analysis of extracts of SMase-treated HT22H2O2 cells showed a deglycosylation of Lamp1 (indicated by an arrow), an increase of rab7 and LRP levels and a decrease of HSP70 protein compared with non-treated HT22H2O2 cells. For each protein one representative blot is shown (n = 3). (c) Treatment of HT22H2O2 cells with SMase reduced the cellular sphingomyelin content, that was measured in two separate experiments in triplicates (mean ± SEM, *p ≤ 0.05). (d) Treatment of HT22H2O2 cells with SMase significantly enhanced the toxic effects of H2O2 (2 mM for 22 h). Values are expressed as mean ± SEM from four independent MTT assays with at least four values for each treatment condition per experiment (*p ≤ 0.05).

Hence, increasing the amount of sphingomyelin may protect cells toward oxidative stress. To increase the sphingomyelin level, we blocked degradation of sphingomyelin by nSMase using the nSMase inhibitor 3-O-Me-SM (Caruso et al. 2005). 3-O-Me-SM treatment of HT22WT cells (20 μM for 24 h) decreased the activity of nSMase as expected (Fig. 5c) and increased the number of large vacuoles that were positively labeled by LysoTracker Red, Lamp1 and filipin (Fig. 5a), which is consistent with the previously reported stabilization of lysosomes by sphingomyelin (Caruso et al. 2005). Furthermore, 3-O-Me-SM treatment reduced the protein amount of LRP and rab7 and increased Lamp1. Thus, 3-O-Me-SM-treated HT22WT cells were remarkably similar to stress-resistant HT22H2O2 cells apart from the observed increase in HSP70 in HT22H2O2 cells that we did not observe after 3-O-Me-SM treatment (Fig. 5b). Finally, treatment of HT22WT cells with 3-O-Me-SM increased resistance toward H2O2 (Fig. 5d) which supports a protective role of sphingomyelin toward oxidative stress.

Figure 5.

 Treatment of HT22WT cells with an nSMase inhibitor renders cells more resistant to H2O2. (a) HT22WT cells were treated for 24 h with the nSMase inhibitor 3-O-Me-SM (20 μM) in serum-free medium and subsequently incubated for 60 min with LysoTracker Red to label lysosomes. Staining with Lamp1 antibody and filipin was performed after fixation of 3-O-Me-SM-treated cells with 4%p-formaldehyde; bar, 20 μm. (b) Western blot analysis of extracts of 3-O-Me-SM-treated HT22WT cells showed an increase in Lamp1, a decrease of rab7 and LRP and no changes for HSP70 protein compared with non-treated HT22WT cells. For each protein one representative western blot is shown (n = 3). (c) Treatment of HT22WT cells with 3-O-Me-SM blocked nSMase activity significantly (mean ± SEM from three independent experiments performed in triplicate; ***p ≤ 0.001). (d) Treatment of HT22WT cells with 3-O-Me-SM protected against H2O2-induced cell death (500 μM H2O2 for 22 h). Values are expressed as mean ± SEM from five independent MTT assays with eight values for each treatment condition per experiment (*p ≤ 0.05).

HT22Glu cells mimic HT22H2O2 cells

HT22H2O2 cells have been shown to be cross-resistant to glutamate (Schafer et al. 2004). In HT22 cells, glutamate elicits oxidative stress via its inhibitory action on a cysteine–glutamate antiporter, which results in the depletion of intracellular cysteine and hence the antioxidant molecule glutathione (Maher and Davis 1996). The observed cross-resistance led us to assume that treatment with glutamate could induce a similar adaptive response to chronic stress than H2O2 and prompted us to analyze a second cell line that was obtained after prolonged exposure to glutamate and is cross-resistant to H2O2 (Schafer et al. 2004). Staining of HT22Glu cells with filipin and anti-Lamp1 antibody showed a similar perinuclear accumulation of cholesterol and lysosomes as in HT22H2O2 cells (Fig. S2a). In addition, similar to HT22H2O2 cells, western blot analysis of HT22Glu cells shows an increase of Lamp1 and reduced levels of rab7 and LRP protein (Fig. S2b and c). We also measured the levels of sphingomyelin, ceramide, and the activity of SMase in glutamate resistant cells and obtained similar results as for the H2O2 resistant cell line, namely an increase in sphingomyelin, a reduction in the amount of ceramide and reduced nSMase activity (Fig. S2d–f). In summary, we found that both resistant cell lines show similar characteristics although they were generated independently and by incubation with different oxidative stressors. These data suggest that the observed changes in the lipid profile and endosomal/lysosomal pathways may be general features of adaptation of neuronal cells to chronic oxidative stress.

Comparison of hippocampus and cerebellum as areas with differential vulnerabilities to oxidative stress

Neurodegenerative disorders are characterized by selective vulnerability and protective capacity of specific brain areas to oxidative stress. In the presented in vitro models we observed differences between oxidative stress-sensitive cells and cells that adapted to chronic oxidative stress which may be relevant also under in vivo conditions. To analyze whether our observations in vitro could be at least in part transferred to the in vivo situation in the brain, we compared adult mouse hippocampus and cerebellum as examples for especially oxidative stress-sensitive and relatively stress-resistant brain areas, respectively (Cardozo-Pelaez et al. 2000). We performed immunohistochemical Lamp1-staining on brain sections of 4-month-old mice that revealed an exceptionally strong labeling of cerebellar Purkinje cells compared with the pyramidal cells of the hippocampal CA3 region (Fig. 6a). This is in accordance with the significantly higher amount of Lamp1 protein in cerebellar compared with hippocampal homogenates as revealed by western blot analysis (Fig. 6b). In addition, we analyzed the expression of the major lysosomal enzyme cathepsin L. We found higher levels of the proform and of the mature cathepsin L form in the cerebellum than in the hippocampus (Fig. 6c). Together, these data suggest that Purkinje cells show a higher lysosomal activity than hippocampal neurons. An increased lysosomal activity may contribute to higher oxidative stress resistance and is consistent with findings describing that lysosomal protease inhibitors induce signs of AD-pathology in rodent slices (Bi et al. 1999).

Figure 6.

 Comparison of hippocampus and cerebellum as examples for relative vulnerable and stress-resistant brain areas, respectively. (a) Immunohistochemical analysis of Lamp1 expression in cryosections through the hippocampal CA3 region and the cerebellum; bar, 50 μm. (b–d) Brain homogenates of 4-month-old C57/BL6 mice were analyzed by western blot for Lamp1 (b), cathepsin L (c), and rab7 (d) expression. (e) The amount of sphingomyelin in these brain homogenates was determined using a commercially available kit. Values shown are mean ± SD (n = 3; *p ≤ 0.05).

In addition, we showed a reduced amount of rab7 protein in cerebellar compared with hippocampal homogenates, which mirrors the situation in both stress-resistant cell lines compared with the stress-sensitive parental cells (Fig. 6d). Furthermore, we analyzed the sphingomyelin levels of hippocampal and cerebellar homogenates and found a higher sphingomyelin content in the cerebellum than in the hippocampus (Fig. 6e). This is in line with the reported lower level of nSMase in cerebellar tissue of adult rats compared with hippocampus (Alessenko et al. 2004) and mimics data from the stress-resistant cell lines.


Here, we investigated the characteristic features of oxidative stress-resistant neuronal cells and found that cells adapted to chronic oxidative stress show major changes in the lipid composition of their membranes. As lipids are known to modulate membrane traffic along endocytic routes, we also investigated the endocytic/lysosomal compartments of these cells, and found stabilized lysosomes and signs of increased autophagy together with an increase in the anti-apoptotic molecule HSP70 in stress-resistant cells. The concept of a lipid-modulated variability in oxidative stress resistance was supported by results from experiments in which incubation with SMase rendered stress-resistant cells more vulnerable to oxidative stress, whereas treatment of vulnerable cells with a SMase inhibitor increased oxidative stress resistance. Finally, we found that the biochemical features detected in stress-resistant and vulnerable cells closely resembled those found in brain areas with differential vulnerability.

Lipid profiling of HT22WT and HT22H2O2 cells showed an increase in total cholesterol and cholesterol precursors in HT22H2O2 cells compared with HT22WT cells. Among the cholesterol precursors, desmosterol levels showed the highest increase. Desmosterol is a substrate for seladin1, for which we measured reduced mRNA and protein levels in HT22H2O2 cells. That is consistent with the observed strong accumulation of desmosterol and confirms a recently published study which described a decline of seladin1 expression upon exposure to chronic oxidative stress (Kuehnle et al. 2008). Interestingly, in this report, Kuehnle et al. found protective effects of both seladin1 over-expression and seladin1 ablation, depending on the duration of the oxidative stress. Lower seladin1 levels conferred resistance under conditions of chronic oxidative stress which is consistent with our data.

The reduced levels of rab7 and ABCA1 in stress-resistant cells may contribute to the observed cholesterol accumulation. Rab7 over-expression has been shown to correct lipid trafficking in Niemann–Pick C (NPC) cells and to reduce the amount of intracellularly stored cholesterol (Choudhury et al. 2002). ABCA1 is responsible for cholesterol efflux, and over-expression of ABCA 1 in hepatocytes resulted in compensatory up-regulation of genes that raise hepatic cholesterol, including HMG-CoA reductase and LRP, to correct for the enhanced ABCA1 mediated cholesterol efflux (Basso et al. 2003). In stress-resistant cells, we found the opposite situation. LRP and HMG-CoA reductase levels were reduced, which probably helps to keep the accumulation of cholesterol in stress-resistant cells at a level that allows cell survival.

The cellular localization of accumulated cholesterol probably determines if cholesterol exerts either protective or toxic effects. Obviously, HT22H2O2 cells are more resistant to oxidative stress despite their high intracellular cholesterol content. Cholesterol storage in lysosomes is possibly a means of HT22H2O2 cells to deal with increased cholesterol levels which could otherwise lead to dangerously rigidified plasma membranes. If cholesterol accumulation exceeds a certain cell type specific level, cell degeneration could be induced. This hypothesis is supported by observations in NPC1-deficient mice, a model for NPC disease. There, among other accumulating lipids a massive intracellular cholesterol accumulation was observed as early as 3 weeks of age although neurodegeneration and clinical symptoms were detected much later and the amount of free cholesterol in different neuronal cell types and the extent of neurodegeneration did not directly correlate (Treiber-Held et al. 2003).

Furthermore, our lipid analysis revealed higher levels of oxysterols, especially of 27-OH-cholesterol, which has been found to increase cellular sphingomyelin levels by stimulating sphingomyelin synthesis and inhibiting nSMase activity (Zhou et al. 2004). In line with these reports we found increased sphingomyelin levels and decreased nSMase activity in HT22H2O2 cells compared with HT22WT cells. Reduced nSMase activity might protect cells against apoptotic stimuli because of the diminished generation of apoptosis-inducing ceramide, the degradation product of sphingomyelin and an important apoptosis mediator (Komatsu et al. 2001). Previous studies have shown that oxidative stress can induce ceramide generation from sphingomyelin via nSMase activation and lead to accumulation of cholesterol esters which consequently results in cell death (Goldkorn et al. 1998; Cutler et al. 2002). In contrast, GSH, which is increased in oxidative stress-resistant cells (Bose Girigoswami et al. 2005), has been shown to block nSMase activation (Rutkute et al. 2007). Cutler et al. (2004) showed that sphingomyelin levels decrease in aging mice in an AD-vulnerable area, the middle frontal gyrus, but not in the relatively spared cerebellum, and observed a correlation between these changes with enhanced oxidative stress shown by the increased levels of lipid peroxidation products. In addition, the authors detected similar changes in AD-patient brains, a reduction of sphingomyelin in the middle frontal gyrus, that was characterized by extensive AD-pathology (Aβ plaques and neurofibrillary tangles), but not in the cerebellum, where only diffuse Aβ deposits and few tangles were found. Interestingly, Alessenko et al. (2004) showed that in adult rat brains, nSMase activity is higher in the vulnerable hippocampal area than in the less sensitive cerebellum. Higher nSMase activity correlated with higher ceramide levels in the hippocampus compared with the cerebellum (Alessenko et al. 2004). We found lower sphingomyelin levels in the adult mouse hippocampus than in the cerebellum. This is in line with the higher hippocampal nSMase activity observed by Alessenko et al. Moreover, filipin staining indicated lower cholesterol levels in the plasma membranes of HT22H2O2 and HT22Glu cells compared with the plasma membranes of HT22WT cells although the total cholesterol content of the stress-resistant cells was higher than that of HT22WT cells because of its intracellular accumulation in lysosomes. Similar to our cell culture data, Chochina et al. (2001) measured lower cholesterol levels in synaptic plasma membranes of the cerebellum than in plasma membranes of the hippocampus. In addition, they observed that the cerebellum was more resistant toward amyloid β peptide-induced changes of the fluidity of synaptic plasma membranes than the hippocampus. In summary, the observed changes in the lipid profiles of HT22WT versus HT22H2O2 cells and of hippocampus versus cerebellum most likely contribute to their specific level of stress resistance. This hypothesis is supported by our data from MTT assays that showed reduced survival of HT22H2O2 cells under oxidative stress conditions after treatment with SMase.

In addition to changes of the lipid composition, other cellular processes such as altered lysosomal activity and protein sorting will be influenced by lipid alterations and will also be part of the complex regulation that allows HT22H2O2 cells and less vulnerable brain areas like the cerebellum to cope with oxidative stress. The cellular lipid composition might strongly influence lysosomal activity and protein sorting and vice versa. Sphingomyelin, for example, has been reported to stabilize lysosomes and protect them against oxidant-induced damage (Caruso et al. 2005). This is consistent with our observations in stress-resistant cells, showing a higher sphingomyelin level concomitant with higher Lamp1 expression.

Lamp1 plays a critical role in autophagy, which is an important survival mechanism for mammalian cells during oxidative stress and takes place within the tight compartment of the lysosomes (for review, see Kroemer and Jaattela 2005; Kiffin et al. 2006). Lamp1 together with the other major lysosomal membrane protein Lamp2 are known as main components of the lysosomal membrane, which serves as a barrier to lysosomal hydrolases, preventing their liberation into cytoplasm and subsequent cell death (Granger et al. 1990). According to the work of Huynh et al. (2007), Lamp1 and Lamp2 have additional functions and are required for the fusion of phagosomes with lysosomes. Furthermore, rab7, a marker for late endosomes/lysosomes, has been described to play an important role during the final maturation step of autophagosomes (Jager et al. 2004). Thus, the observed decrease of rab7 in stress-resistant cells and in cerebellar tissue was surprising in light of its proposed function as a promoter of autophagosome maturation. The most likely explanation would be that down-regulation of rab7 in stress-resistant cells serves to keep the extent of autophagy below a detrimental level. Interestingly, we observed an increase in rab7 protein levels concomitant with Lamp1 degradation and a reduction in LysoTracker Red and Lamp1 staining after treatment of HT22H2O2 cells with SMase, which enhanced sensitivity of these cells toward oxidative stress. In contrast, treatment of HT22WT cells with an inhibitor of nSMase, 3-O-Me-SM, decreased rab7 levels and increased labeling with LysoTracker Red and Lamp1, which was accompanied by increased oxidative stress resistance. Although we are not able to distinguish between enhanced or reduced autophagy in tissue extracts, taking our data from the stress-resistant cells into account, the observed changes in the cerebellum correlate well with its reported stress resistance. However, further experiments are necessary to determine the contribution of autophagy to brain region specific oxidative stress resistance. Our resistant neuronal cell lines may provide a valuable tool to analyze the underlying complex mechanisms.

In conclusion, we have identified several molecules and pathways that might contribute to oxidative stress resistance (Fig. 7). The lipid composition and the lysosomal pathways of a given cell appear to be important determinants of its specific vulnerability toward oxidative stress. Modulating one of the components, as we have demonstrated for SMase, will likely influence the whole intracellular defense system and change the ability to cope with oxidative stress.

Figure 7.

 Schematic illustration of the putative effects of oxidative stress on cellular cholesterol and sphingomyelin homeostasis leading to increased stress resistance. Oxidative stress leads to increased amounts of cholesterol precursors (mainly desmosterol), of free cholesterol, and of cholesterol metabolites (mainly 27-OH-cholesterol). The increased cholesterol level is accompanied by an increased expression of HSP70 which is known to interfere with ceramide-induced apoptosis. Normal cholesterol efflux is probably impaired as indicated by the down-regulation of rab7 and ABCA. Potentially, to keep cholesterol accumulation below a detrimental level, further cholesterol biosynthesis is reduced by reduction of HMG-CoA reductase and seladin activities, and autophagic activity is enhanced (increased LC3II as autophagosomal marker). 27-OH-cholesterol is known to increase (+) sphingomyelin levels and to block (−) neutral sphingomyelinase (nSMase) activity which leads to reduced levels of ceramide. Sphingomyelin biosynthesis and degradation correlate with cholesterol metabolism. Modulation of any of these molecules (in this paper shown for sphingomyelin and nSMase activity) will regulate the function of the other components of this cycle and thus influence sensitivity toward oxidative stress.


We thank Dr C. U. Pietrzik for anti-LRP-antibody and H. Nagel for expert technical assistance. This work was supported by DFG (SFB645) and BMBF (01G0708). The initial phase of this project was supported by Alzheimer-Schwerpunkt of the DFG (Be1475/2-1/2-2).