Dysfunction of mitochondrial complex I and the proteasome: interactions between two biochemical deficits in a cellular model of Parkinson's disease


Address correspondence and reprint requests to Günter U. Höglinger, INSERM U 289, Experimental Neurology and Therapeutics, Hôpital de la Salpêtrière, 47, Boulevard de l'Hôpital, 75651 Paris Cedex 13, France. Tel. +331 42 26 22 02, Fax: +331 44 24 36 58, E-mail:gunter@ccr.jussieu.fr


Two biochemical deficits have been described in the substantia nigra in Parkinson's disease, decreased activity of mitochondrial complex I and reduced proteasomal activity. We analysed interactions between these deficits in primary mesencephalic cultures. Proteasome inhibitors (epoxomicin, MG132) exacerbated the toxicity of complex I inhibitors [rotenone, 1-methyl-4-phenylpyridinium (MPP+)] and of the toxic dopamine analogue 6-hydroxydopamine, but not of inhibitors of mitochondrial complex II–V or excitotoxins [N-methyl-d-aspartate (NMDA), kainate]. Rotenone and MPP+ increased free radicals and reduced proteasomal activity via adenosine triphosphate (ATP) depletion. 6-hydroxydopamine also increased free radicals, but did not affect ATP levels and increased proteasomal activity, presumably in response to oxidative damage. Proteasome inhibition potentiated the toxicity of rotenone, MPP+ and 6-hydroxydopamine at concentrations at which they increased free radical levels ≥ 40% above baseline, exceeding the cellular capacity to detoxify oxidized proteins reduced by proteasome inhibition, and also exacerbated ATP depletion caused by complex I inhibition. Consistently, both free radical scavenging and stimulation of ATP production by glucose supplementation protected against the synergistic toxicity. In summary, proteasome inhibition increases neuronal vulnerability to normally subtoxic levels of free radicals and amplifies energy depletion following complex I inhibition.

Abbreviations used

adenosine triphosphate






potassium cyanide


least significant difference test








3-nitroproprionic acid




polyacrylamide gel electrophoresis


phosphate-buffered saline


Parkinson's disease


peptidylglutamyl-peptide hydrolase


reactive oxygen species




tyrosine hydroxylase


ubiquitin-proteasome system

In degenerating nigral dopaminergic neurones in Parkinson's disease (PD), damaged proteins accumulate (Floor and Wetzel 1998) and aggregate as Lewy bodies (Munch et al. 2000). The bulk of damaged intracellular proteins are degraded by the ubiquitin-proteasome system (UPS; Sherman and Goldberg 2001). It comprises a set of enzymes that attach poly-ubiquitin chains to proteins, signalling them for degradation by the proteasome, a multicatalytic protease complex.

This system may be functionally impaired in inherited forms of PD:

Although only a minority of PD cases are inherited, they still indicate that a defect in the UPS can mimic the neurodegeneration seen in sporadic PD. Dysfunction of the UPS may therefore also contribute to sporadic PD. Indeed, reduced proteasomal activity has been reported in nigral (McNaught and Jenner 2001; McNaught et al. 2003), but not striatal or cortical (Furukawa et al. 2002) post-mortem tissue from sporadic PD patients, although technical concerns about these data have been voiced (Vigouroux et al. 2003) and they warrant further confirmation.

A second biochemical deficit found in the substantia nigra of sporadic PD patients is reduced activity of complex I of the mitochondrial respiratory chain (Schapira et al. 1989, 1990). Complex I inhibition by 1-methyl-4-phenylpyridinium (MPP+) causes selective degeneration of nigral neurones in humans, non-human primates and mice, providing a valuable model of PD (Przedborski and Jackson-Lewis 1998). Recently, chronic treatment of rats with the lipophilic complex I inhibitor rotenone has been presented as novel model of parkinsonism, leading to degeneration of nigral neurones and cytoplasmic aggregation of ubiquitinated proteins (Betarbet et al. 2000; Höglinger et al. 2003). The association of complex I and proteasome inhibition in sporadic PD prompted us to investigate the functional interactions of the two biochemical deficits in a cellular model of PD.

Materials and methods

Primary mesencephalic cell cultures

Animals were treated in accordance with the Guide for the Care and Use of Laboratory Animals (National Research Council 1996) and the European Communities Council Directive 86/609/EEC. Dopaminergic neurones from the ventral mesencephalon of embryonic day 15.5 Wistar rat embryos (CERJ, Le Genest St Isle, France) were dissociated mechanically, plated onto polyethylenimine-pre-coated 24-well culture plates, grown in N5 medium (Kawamoto and Barrett 1986) supplemented with 5 mm glucose, 5% horse serum and 2.5% fetal calf serum to favour cell attachment (Michel and Agid 1996). After 3 days, fetal calf serum was reduced to 0.5% to limit astrocyte proliferation. All experiments were initiated at day 5 in N5 medium containing 1% horse serum and 0.1% fetal calf serum to limit possible interference by serum components.


Stock solutions of 1 mm epoxomicin, 25 mm MG-132, 10 mmclasto-lactacystin β-lactone (all Calbiochem, Fontenay sous Bois, France), 100 mm antimycin A and 10 mm oligomycin (all Sigma-RBI-Aldrich, St Quentin Fallavier, France) in dimethylsulphoxide were stored at −20°C. Solutions of 100 mm rotenone in dimethylsulphoxide and 10 mm MPP+, 100 mm 6-hydroxydopamine hydrobromide (6-OHDA), 50 mm 3-nitroproprionic acid (3-NP), 100 mm potassium cyanide (KCN; all Sigma), 100 mmN-methyl-d-aspartate (NMDA) and 100 mm kainic acid (Research Biochemicals International, Natic, MA, USA) in H2O were freshly prepared for each experiment. The chemicals were further pre-diluted in N5 medium prior to addition to the cultures.


Neurones were identified immunocytochemically using the pan-neuronal marker MAP2 or the dopaminergic marker tyrosine hydroxylase (TH). Cultures were fixed with 4% formaldehyde, incubated with a rabbit anti-TH (Peel Freez, Paris, France, 1/500) or mouse anti-MAP2 antibody (Chemicon, Hofheim, Germany, 1/200; 12 h, 4°C), then with a secondary biotinylated anti-mouse or anti-rabbit IgG (Vectastain, Burlingame, CA, USA; 1/200) and a peroxidase-conjugated-avidin complex (Vectastain), or directly with a Cy3-conjugated anti-rabbit IgG (Jackson ImmunoResearch, West Grove, PA, USA; 1/500) or an Alexa Fluor 488-conjugated antimouse IgG (Molecular Probes, Interchim, Montilucon, France; 1/500). Nuclei were visualized with the fluorescent DNA stain Hoechst 33342.

ATP levels

Cells were washed with cold phosphate-buffered saline (PBS), scraped off the culture wells with 100 µL H2O and stored at − 80°C. Adenosine trihphosphate (ATP) levels in 10 µL of the lysates were quantified in a tube luminometer using the Vialight HS kit (Bio Whittaker, Verviers, Belgium), which utilizes luciferase to catalyse the formation of light from ATP and luciferin.

Complex I activity

Adult rat brain fragments were homogenized mechanically in 10 mm Tris–HCl pH 7.2, 225 mm mannitol, 75 mm saccharose and 0.1 mm EDTA and centrifuged (600 g, 4°C, 20 min) to obtain the post-nuclear supernatant. Complex I activity was measured spectrophotometrically at 37°C during 3 min by the rate of NADH oxidation at 340 nm in an assay medium containing 40 µg protein of post-nuclear supernantants in 1 mL 25 mm phosphate buffer pH 7.5, 2.5 mg/mL bovine serum albumin, 100 µm decylubiquinone and 200 µm NADH. Reactions were performed in the absence and the presence of 2 µm rotenone, and the rotenone-sensitive activity was attributed to complex I.

Quantification of reactive oxygen species

Cultures were exposed to neurotoxins for 5.5 h before 50 µm dihydrorhodamine-123 (Molecular Probes, Montluçon, France) was added for 30 min to allow its oxidation to rhodamine-123 by reactive oxygen species (ROS). Then, cells were washed three times and rhodamine-123 fluorescence (excitation 488 nm, emission 515 nm) was quantified by computer-assisted image analysis (Fluostar, Imstar, Paris, France) on at least five individual neuronal profiles in 10 randomly distributed fields per experimental condition.

Proteasomal peptidase activities

Cells were washed, scraped off and stored at −80°C in 25 mm Tris–HCl pH 7.5. Supernatants were collected after sonication and centrifugation (20 000 g, 4°C, 20 min). Degradation of 25 µm Suc-Leu-Leu-Val-Tyr-7-amido-4-methylcoumarin [chymotrypsin (CT)- like activity], 40 µm N-t-Boc-Leu-Ser-Thr-Arg-7-amino-4-methylcoumarin [trypsin (T)-like activity], and 150 µm N-Cbz-Leu-Leu-Glu-β-naphthylamide [peptidylglutamyl-peptide hydrolase (PGPH) activity; substrates from Sigma] was monitored for 20 min using 20 µg protein in 200 µL Tris–HCl at 37°C by spectrofluorimetry (excitation/emission wavelengths: 350/440 nm for aminomethylcoumarin, 333/410 nm for β-naphtylamine), as described (Bulteau et al. 2001). Peptidase activities were calculated from the slopes of the linear curves in relation to standards generated using pure fluorescent products. Samples were analysed three times in the absence, and once in the presence of the proteasome inhibitor MG-132 (20 µm). The MG-132-sensitive activity was attributed to the proteasome. We verified the effective contribution of the proteasome to the MG-132-sensitive activity by enrichment of the proteasomes in crude rat mesencephalic protein extracts (n = 8) by centrifugation (160 000 g, 4°C, 4 h), as described by Vigouroux et al. (2003).

Analysis of protein oxidation

Cells were lysed in 20 mm Tris–HCl with 150 mm NaCl, 2 mm EDTA pH 8, 1% triton, 10% glycerol, 2% complete mini-protease inhibitor cocktail (Roche, Mannheim, Germany), 1 mm sodium orthovanadate, 2 mm sodium pyrophosphate and 50 mm sodium fluoride. Five micrograms of protein were incubated for 15 min with 2,4,-dinitrophenylhydrazin to allow its reaction with carbonyl groups in protein side-chains, separated by polyacrylamide gel electrophoresis (PAGE), blotted onto a nitrocellulose membrane, incubated with a rabbit-antidinitrophenyl antibody (OxyBlot kit, Intergen, New York, NY, USA), and revealed using ECL (Amersham). OxyBlot experiments were done in duplicate.

Statistical analysis

Data are expressed as the mean ± SEM from at least three different experiments. p-values were calculated by one- or two-way anova, as appropriate, followed by post-hoc least significant difference (LSD) test.


Cell death induced by proteasome inhibitors in mesencephalic cultures

Epoxomicin, MG-132 and clasto-lactacystin β-lactone inhibited CT-like proteasomal activity in protein extracts from mesencephalic cultures in a concentration-dependent manner (Figs 1a–c), but only epoxomicin and MG-132 markedly inhibited the three proteasomal proteolytic activities in the neuronal cultures, when present for 24 h in the medium (Figs 1a–c). Clasto-lactacystin β-lactone is rapidly hydrolysed in aqueous buffers (Dick et al. 1996) and may not penetrate sufficiently into mesencephalic cells, possibly explaining its weak effects. Consistently, only epoxomicin and MG-132 resulted in significant neurotoxicity (Figs 1a′–c′) and were used for further experiments. They induced only a mild loss (< 20%) of neurones in general (MAP2-ir) and dopaminergic neurones in particular (TH-ir) after 24 h even at the highest concentrations used, but a severe and concentration-dependent cell loss after 48 h, equally affecting MAP2-ir and TH-ir neurones. Only a few cells had condensed chromatin in control conditions and after 24 h of proteasome inhibition, but after 48 h in presence of high concentrations of epoxomicin or MG-132, all remaining cells had condensed chromatin, suggesting that death occurred by apoptosis (Figs 1d–f).

Figure 1.

Proteasome inhibitors reduce proteolytic activity and induce cell death in mesencephalic cultures. Effects of the proteasome inhibitors epoxomicin (a), MG-132 (b) and clasto-lactacystin β-lactone (c) on the chymotrypsin (CT)-like (•), trypsin (T)-like (○) and PGPH (▴) activity in mesencephalic cultures (24 h exposure) and in protein extracts from cultures (▵ 10 min exposure). Effects of epoxomicin (a′), MG-132 (b′), and clasto-lactacystin β-lactone (c′) on survival of the total neuronal population (MAP2-ir, •, ○) and of dopaminergic neurones (TH-ir, ▴, ▵) after 24 h (▴, •) and 48 h (▵, ○) in the presence of the inhibitor. Photomicrographs illustrate MAP2-ir (green) and TH-ir (red) neurones in control cultures (d) and after exposure to 1 µm epoxomicin for 24 h (e) or 48 h (f). DNA-staining with Hoechst 33342 (blue) showed few condensed nuclei (arrows) suggestive of apoptosis in control cultures and after 24 h. In contrast, all nuclei remaining after 48 h of treatment with epoxomicin were condensed. Insets at lower magnification show that TH-ir neurites are dystrophic by 24 h, preceding cell death. Scale bar 30 µm, inset 90 µm.

Proteasome inhibition exacerbates the neurotoxicity of complex I inhibitors

The lipophilic complex I inhibitor rotenone induced the death of both TH-ir neurones and MAP2-ir neurones to the same degree at all concentrations tested (Figs 2a and b). In contrast, the hydrophilic complex I inhibitor MPP+ killed TH-ir neurones specifically within a concentration-range of 0.3–10 µm. At higher concentrations, the general neuronal population (MAP2-ir) began to degenerate as well (Figs 2c and d). A total of 100 nm epoxomicin, a concentration that did not induce cell death at 24 h (Fig. 1a′), significantly increased the toxicity of both complex I inhibitors (Figs 2a–d). Specifically, epoxomicin led to significant cell death at 24 h when cultures were exposed to normally non-toxic concentrations of rotenone (Figs 2a and b), and significantly increased cell death caused by partially toxic concentrations of MPP+ (Figs 2c and d). This synergy affected TH-ir (Figs 2a and c) and MAP2-ir cells (Figs 2b and d) equally. At a fixed concentration of rotenone or MPP+, loss of TH-ir (Figs 2a′ and c′) and MAP2-ir neurones (Figs 2b′ and d′) increased as a function of the epoxomicin concentration. Figure 2(e) illustrates the synergistic toxicity of epoxomicin with rotenone and MPP+ in cultures immunostained for TH.

Figure 2.

Synergistic toxicity of complex I and proteasome inhibition. Proteasome inhibition by 100 nm epoxomicin (Epox) increased the concentration-dependent loss of TH-ir (a and c) and MAP2-ir neurones (b and d) induced by a 24-h treatment with the complex I inhibitors rotenone (a and b) and MPP+ (c and d). The potentiation of rotenone (a′ and b′) and MPP+ (c′ and d′) toxicity for both TH-ir (a′ and c′) and MAP2-ir neurones (b′ and d′) was a function of the epoxomicin concentration. (e) Photomicrographs show representative fields of TH-ir neurones in control cultures (Con) or cultures treated for 24 h with 100 nm epoxomicin (Epox), 30 nm rotenone (Rot), 30 nm rotenone + 100 nm epoxomicin (Rot + Epox), 3 µm MPP+ (MPP+), or 3 µm MPP+ + 100 nm epoxomicin (MPP+ + Epox). Statistics: (a and b) two-way anova, effect for rotenone p < 0.001, epoxomicin p < 0.001, interaction p < 0.01; post-hoc LSD test *p < 0.01 100 nm versus 0 nm epoxomicin. (c and d) Two-way anova, effect for MPP+p < 0.001, epoxomicin p < 0.001, interaction p < 0.05; post-hoc LSD test *p < 0.05 100 nm versus 0 nm epoxomicin. (a′–c′) anova followed by post hoc LSD-test *p < 0.05 versus 0 nm epoxomicin.

Proteasome inhibition does not exacerbate all toxin-induced neuronal cell death

To determine whether neurotoxins acting by other mechanisms are also potentiated by proteasome inhibition, we examined the effect of 100 nm epoxomicin on the toxicity of the following substances: inhibitors of mitochondrial complex II (3-NP), III (antimycin A), IV (KCN) and V (oligomycin), excitotoxins (NMDA, kainic acid) and the dopamine analogue 6-OHDA. Figure 3 shows their dose–response curves for MAP2-ir neuronal death at 24 h in absence and presence of epoxomicin. Only neuronal death induced by 6-OHDA was potentiated by epoxomicin. TH-ir neurones were no more sensitive than MAP2-ir neurones (not shown). Synergistic toxicity of proteasome inhibition with rotenone, MPP+ and 6-OHDA was confirmed with MG-132 (Table 1).

Figure 3.

Effect of proteasome inhibition on neuronal cell death induced by neurotoxins other than complex I inhibitors. Survival of MAP2-ir cells in mesencephalic primary cultures was quantified after coexposure for 24 h to either 0 nm (○) or 100 nm epoxomicin (•) and inhibitors of the mitochondrial respiratory chain [complex II: 3-NP (a), complex III: antimycin A (b), complex IV: KCN (c), complex V: oligomycin (d)], excitotoxins [NMDA (e), kainic acid (f)], or the toxic dopamine analogue 6-OHDA (g). Synergy was only observed with 6-OHDA. Two-way anova, effects for 6-OHDA, epoxomicin and interaction p < 0.001; post-hoc LSD test *p < 0.001 100 nm versus 0 nm epoxomicin.

Table 1.  MG-132 exacerbates the toxicity of rotenone, MPP+ and 6-OHDA
 No MG-1321 µm MG-132
  1. TH-ir cell survival in mesencephalic cultures is presented as percentage of controls after a 24-h exposure to 30 nm rotenone, 3 µm MPP+ or 30 µm 6-OHDA either in the absence or presence of 1 µm MG-132. Control values were 100.0 ± 3.3%. *p < 0.05, anova followed by post-hoc LSD-test.

Rotenone90.9 ± 3.5%,6.6 ± 1.4%*
MPP+44.4 ± 3.1%,22.1 ± 2.5%*
6-OHDA65.9 ± 14.0%44.4 ± 3.4%*

A threshold level of toxin-induced ROS production is required for synergistic toxicity with proteasome inhibitors

Rotenone, MPP+ and 6-OHDA induce ATP depletion, ROS production or both. To determine whether these mechanisms contribute to the synergism with proteasome inhibitors, we quantified ROS production and ATP depletion after 6 h exposure to rotenone, MPP+, 3-NP, antimycin A, KCN, oligomycin and 6-OHDA, and in parallel cultures, the survival of MAP2-ir neurones after 24 h (Fig. 4). In all situations in which all neurones died within 24 h, ATP levels decreased more than 80% within the first 6 h. When residual ATP levels remained above 20%, there was no or little neuronal loss, suggesting that 20% of normal ATP levels was the minimum compatible with neuronal survival. Rotenone, MPP+ and 6-OHDA increased ROS to ≥ 40% over control levels already at low concentrations that were subtoxic or only moderately toxic. In contrast, 3-NP, antimycin A, KCN and oligomycin, that showed no synergy with proteasome inhibitors, did not increase ROS production over baseline levels at subtoxic concentrations, but only at high concentrations that cause ≥ 80% ATP depletion and complete neuronal death. These data suggest that a threshold of a 40% increase in ROS at subtoxic or partially toxic conditions predisposed cells to synergistic toxicity with proteasome inhibitors.

Figure 4.

ATP levels, ROS production and neuronal survival in mesencephalic cultures treated with respiratory chain inhibitors and 6-OHDA. Cultures were exposed to rotenone (a), MPP+ (b), 3-NP (c), antimycin A (d), KCN (e), oligomycin (f), or 6-OHDA (g) and analysed after 6 h for intracellular ROS production (○) and ATP-levels (□), or after 24 h for MAP2-ir cell survival (▴). An 80% decrease in ATP levels (dashed lines) was always followed by complete neuronal cell loss. Concentrations at which toxicity could be amplified by proteasome inhibition are marked with arrows. These concentrations always increased ROS production ≥ 40% over baseline levels (dotted lines).

Rotenone and MPP+ reduce proteasomal activities in the absence of cell death via ATP depletion, not via ROS production

We hypothesized that complex I inhibitors affect proteasomal activity, and could explain their synergy with proteasome inhibitors. Indeed, all three proteasomal peptidase activities significantly decreased in cultures grown for 6 h in presence of 30 nm rotenone or 30 µm MPP+(Figs 5a and b). We excluded MAP2-ir cell loss or chromatin condensation in these conditions, demonstrating that loss of proteasomal activity was not due to cell death. We excluded a direct inhibition of the proteasome by rotenone and MPP+ by measuring the CT-like-activity in protein extracts exposed for 10 min to these molecules (control 100 ±1.7%; 30 nm rotenone 94.2 ± 4.0%; 30 µm MPP+ 101.1 ±0.7%). To exclude the possibility that non-proteasomal MG132-sensitive CT-like activity was lost in the presence of the complex I inhibitors, we enriched mesencephalic protein extracts in proteasomes by ultracentifugation (Vigouroux et al. 2003) and showed that only 12.8% of the MG-132-sensitive CT-like activity was not attributable to the proteasome. Rotenone and MPP+, however, reduced the MG-132-sensitive CT-like activity more than could be explained by this non-proteasomal activity (Figs 5a and b). Together these data suggest that complex I inhibition, indeed results in reduced proteasomal activities in living cells.

Figure 5.

Rotenone, MPP+ and 6-OHDA differentially affect proteasomal activity. Chymotrypsin (CT)-like, trypsin (T)-like, and peptidylglutamyl-peptide hydrolase (PGPH) activity in primary mesencephalic cultures after a 6-h-exposure to 30 nm rotenone (a), 30 µm MPP+ (b) and 30 µm 6-OHDA (c) either alone or in combination with the energy substrate glucose (50 mm, Gluc) or the radical scavenger NAC (1 mm). The effect of these treatments on ATP levels (a′ and b′) and ROS production (c′) analysed in parallel cultures. Statistics: anova followed by post-hoc LSD test: *p < 0.05 versus corresponding controls in (a–c′); #p < 0.05 versus rotenone alone in (a and a′) versus MPP+ alone in (b and b′) versus 6-OHDA alone in (c and c′).

To determine whether ROS production or ATP depletion were responsible for this decrease, we analysed the effects of the radical scavenger NAC and the energy substrate glucose. One mm NAC fully prevented the rise in ROS production by the complex I inhibitors (control 100.0 ±3.0%; 30 nm rotenone 151.6 ± 4.6%; rotenone + NAC 97.3 ± 3.8%; 30 µm MPP+ 176.4 ± 5.6%; MPP+ + NAC 93.4 ± 4.8%), but did not prevent the loss of proteasomal activity (Figs 5a and b). In contrast, raising the glucose levels in the culture medium from 250 µm (normal) to 50 mm prevented the decrease in proteasomal activity with both complex I inhibitors (Figs 5a and b). In parallel, glucose partially restored the ATP loss induced by rotenone (Fig. 5a′) and MPP+ (Fig. 5b′). These data suggest that the decrease in proteasomal activities after complex I inhibition resulted from ATP depletion, not from ROS production.

To confirm this conclusion, the effect of 6-OHDA on proteasomal activity was monitored, as 6-OHDA did not significantly decrease cellular ATP levels in our conditions (Fig. 4g). In contrast to rotenone and MPP+, a 6-h exposure to 6-OHDA increased rather than decreased the three proteasomal proteolytic activities (Fig. 5c). This increase was transient and CT-like activity decreased again to 113.9 ± 3.3% of control (ns) after 12 h of 6-OHDA treatment. In parallel with its effect on proteasomal activity, 6-OHDA increased ROS production (Fig. 5c′). The increase in both proteasomal activity and ROS production was prevented by the radical scavenger NAC (1 mm), but not by 50 mm glucose (Fig. 5c′).

Proteasome inhibition and complex I inhibition interact on the level of protein oxidation and ATP depletion

Complex I inhibition increased ROS production (Fig. 6a) and protein oxidation (Fig. 6b). Proteasome inhibition did not increase basal or rotenone-induced ROS production (Fig. 6a), but still led to an increase in oxidized proteins (Fig. 6b), suggesting that oxidized proteins accumulated because of reduced proteolysis rather than increased oxidation. Thus, complex I and proteasome inhibition contribute independently to the accumulation of oxidized proteins by increasing oxidation and reducing proteolysis, respectively.

Figure 6.

Interactions between complex I and proteasome inhibition on ROS production and ATP depletion, and their relevance for synergistic toxicity. (a) Intracellular ROS production in mesencephalic cultures under control conditions (Con) or after a 6-h exposure to 1 µm epoxomicin (E1µm) or to 30 nm rotenone (R30nm) either alone or in combination with 1 µm epoxomicin, 1 mm NAC or 50 mm glucose (Gluc); dotted line: 40% increase responsible for synergy with proteasome inhibitors. anova followed by post-hoc LSD test: *p < 0.001 versus control. (b) Protein oxidation in cultures analyzed by OxyBlot under control conditions or after 24 h exposure to 10 nm and 30 nm rotenone or 10 nm and 100 nm epoxomicin. (c) ATP-levels in cultures analyzed after a 6-h exposure to 0, 10, 30, or 100 nm rotenone in combination with 0, 100, or 1000 nm epoxomicin; dotted line: 20% threshold critical for neuronal survival. anova followed by post-hoc LSD test: *p < 0.05 versus 0 nm epoxomicin. (c′) Raising glucose concentrations from 250 µm (normal) to 50 mm increases ATP levels in cultures exposed to 30 nm or 100 nm rotenone or a combination of 30 nm rotenone + 1 µm epoxomicin; dotted line: 20% threshold. anova followed by post-hoc LSD test: *p < 0.001 Gluc 50 nm versus Gluc 250 µm (d) 50 mm glucose, but not 1 mm NAC, protects MAP2-ir cells against 100 nm rotenone, suggesting cell death results from energy depletion. anova followed by post-hoc LSD test: *p < 0.001 versus control; #p < 0.001 versus R100nm (d′) Both 1 mm NAC and 50 mm glucose protect against synergistic toxicity of 30 nm rotenone and 1 µm epoxomicin suggesting cell death results from both ROS attacks and ATP depletion. anova followed by post-hoc LSD test: *p < 0.001 versus control; #p < 0.001 versus R30nm + E1µm.

Complex I inhibition with rotenone for 6 h resulted in a concentration-dependent decrease in ATP levels (Fig. 6c). Proteasome inhibition, however, did not affect ATP levels (Fig. 6c). This is consistent with the absence of an effect of proteasome inhibitors on the activity of complex I in rat brain homogenates (control 186.2 ± 8.7 nmol/min/mg protein; 10 µm epoxomicin 185.5 ± 19.8 nmol/min/mg; 10 µm MG132 169.6 ± 1.9). However, epoxomicin significantly increased the ATP loss induced by 10 and 30 nm rotenone (Fig. 6c), in the absence of MAP2-ir cell loss and chromatin condensation, demonstrating that the synergistic effect of complex I and proteasome inhibition on ATP depletion results from a functional interaction preceding cell death.

Proteasome inhibition renders mesencephalic neurones vulnerable to normally subtoxic ROS levels induced by complex I inhibitors

To assess the relative contribution of ROS production and ATP depletion to cell death, we first determined that increasing glucose levels in the culture medium to 50 mm did not prevent the rotenone-induced ROS increase (Fig. 6a), but significantly increased ATP levels in cultures exposed to rotenone or to a combination of rotenone and epoxomicin (Fig. 6c′). 1 mm NAC, in turn, did not affect rotenone-induced ATP depletion (control 100.0 ± 5.9%; 30 nm rotenone 22.4 ± 6.7%; 30 nm rotenone + NAC 23.8 ± 3.4%), but completely prevented the rotenone-induced ROS increase (Fig. 6a).

When cultures were exposed to a high concentration of rotenone (100 nm), leading to complete cell death, glucose fully prevented cell death, whereas NAC had no effect, demonstrating that cell death in this situation was due to energy depletion (Fig. 6d).

When cultures were exposed to a lower concentration of rotenone (30 nm) or to 1 µm epoxomicin, neither toxin induced significant cell death alone, but their combination led to complete cell death (Fig. 6d′). Both NAC and glucose significantly prevented cell death in this situation, demonstrating, that for this synergistic toxicity ROS production and ATP depletion were equally important (Fig. 6d′). A schematic representation of these interactions is shown in Fig. 7.

Figure 7.

Schema representing the interactions of complex I and proteasome inhibitors described in this study. Pathways relevant for cell death are indicated by solid arrows; pathways activated, but without major contribution to cell death by dashed arrows. (a) Complex I inhibition increases oxidized protein levels via ROS production and decreases proteasomal activity via ATP depletion. Toxicity of complex I inhibitors at high concentration results from energy depletion, because restoration of ATP-levels by glucose, but not radical scavenging, protects neurones from death. (b) Complex I inhibitors at low concentrations increase ROS-production, deplete ATP-levels and impair proteasome function, but neurones resist these challenges. (c) The additional presence of a proteasome inhibitor impairs detoxification of oxidized proteins and exacerbates ATP depletion induced by low concentrations of complex I inhibitors, resulting in synergistic toxicity that can now be prevented both by restoring ATP-levels with glucose and radical scavenging with NAC.


The major new finding in this study is that reduced proteasomal activity increases the vulnerability of mesencephalic neurones in vitro to MPP+, rotenone and 6-OHDA, toxins used to model parkinsonism, but not to a variety of other neurotoxins. Given that mutations in components of the UPS have up to now been identified only in families with PD, but not other diseases, our data suggest that an increased vulnerability to such toxins as a consequence of UPS malfunction may be a basic pathomechanism in both sporadic and familial forms of PD. The second new insight is that complex I inhibition can decrease proteasomal activity, suggesting a possible mechanism by which the UPS might be impaired in sporadic PD.

We chose mesencephalic cultures as in vitro model of PD to be able to compare the vulnerability of dopaminergic and non-dopaminergic neurones. Decreased complex I (Schapira et al. 1989, 1990) and proteasomal activity (McNaught and Jenner 2001; McNaught et al. 2003), as measured in post-mortem nigral tissue in PD, are unlikely to be restricted to the degenerating dopaminergic neurones, as these comprise only a small proportion of all nigral cells. Still, our results suggest that neither a generalized complex I inhibition nor a generalized proteasome inhibition induced by exogenous agents sufficiently explains the selective dopaminergic cell death seen in PD. Of all the toxins used in this study, only MPP+, that is selectively taken up into dopaminergic neurones via the dopamine transporter (Javitch et al. 1985), was selectively toxic for dopaminergic neurones at low concentrations, either alone or in synergy with proteasome inhibitors. Striking, however, was the selectivity with which proteasome inhibition increased the toxicity of MPP+, rotenone and 6-OHDA, all of which have been associated with toxin-induced parkinsonism (Przedborski and Jackson-Lewis 1998; Betarbet et al. 2000; Blum et al. 2001; Höglinger et al. 2003).

Our data suggest that significant ROS production, observed even at non-toxic or partially toxic concentrations, is common to these three toxins and predisposes them for synergistic toxicity with proteasome inhibitors. All respiratory chain inhibitors used in this study induced ROS production. However, for 3-NP, antimycin A, KCN and oligomycin, this occurred only at concentrations high enough to cause a depletion that resulted in severe ATP depletion that in itself was toxic to the cells. This fundamental difference has already been described for rotenone and the complex III inhibitor antimycin A: both complex I or III activity must be decreased by about 70% before major ATP depletion occurs in non-synaptic brain mitochondria (Davey and Clark 1996). Also, significant ROS production requires a 70% inhibition of complex III, but only a 16% inhibition of complex I (Sipos et al. 2003). This opens a large window for complex I inhibition in which ROS production occurs in the absence of bioenergetic failure. Our data suggest that this ROS-ATP gap is unique for complex I in the mitochondrial respiratory chain and predisposes cells to synergistic toxicity with proteasome inhibitors.

Synergistic toxicity between 6-OHDA and MG-132 has already been reported (Elkon et al. 2001). 6-OHDA also inhibits complex I in isolated mitochondria (Glinka et al. 1996), but in intact cells, its toxicity does not appear to depend on ATP depletion but rather on ROS arising from deamination or auto-oxidation of 6-OHDA (Blum et al. 2001). This is consistent with our findings of 6-OHDA-induced ROS production without ATP depletion, and could explain why 6-OHDA, like rotenone and MPP+, can synergize with proteasome inhibition. The two excitotoxins studied, NMDA and kainic acid, that primarily induce an abnormal Ca+-influx and increase ROS only during the final downstream events leading to cell death (Dugan et al. 1995; Carriedo et al. 1998), showed no synergy with proteasome inhibition.

What is the mechanism underlying the synergy of proteasome inhibitors with rotenone, MPP+ and 6-OHDA? We first hypothesized that these three toxins may reduce proteasomal activity, thus aggravating the effect of proteasome inhibitors. We indeed observed an inhibition of the three proteasomal activities by rotenone and MPP+. This could reflect direct inhibition of the proteasome or be caused by oxidative damage from ROS (Friguet and Szweda 1997; Reinheckel et al. 1998; Okada et al. 1999; Keller et al. 2000) or by decreased ATP levels (Benaroudj et al. 2003). We have excluded direct proteasome inhibition by rotenone and MPP+. We also showed that the rotenone- and MPP+-induced proteasome inhibition was not mediated by ROS, as the radical scavenger NAC did not prevent the loss of proteasomal activity, although it blocked rotenone- and MPP+-induced ROS production. This conclusion is supported by the results obtained with 6-OHDA, showing that increased ROS production can also be associated with increased proteasomal activity. In contrast, dopamine (≥ 100 µm), that shares some structural similarity with 6-OHDA, has been shown to inhibit proteasomal activities in PC12 cells via ROS production (Keller et al. 2000). However, in line with our observations, several studies have also shown that moderate oxidative stress can stimulate proteasomal activity (Strack et al. 1996; Grune et al. 1998; Reinheckel et al. 1998). Because elevation of ATP levels with glucose was sufficient to prevent proteasome inhibition by rotenone and MPP+, we concluded that complex I inhibition decreased the activity of the highly energy-dependent proteasome (Benaroudj et al. 2003) via ATP depletion. However, as 6-OHDA did not decrease ATP levels and increased proteasomal activity, the synergistic toxicity of rotenone, MPP+ and 6-OHDA with proteasome inhibitors cannot be explained on this basis.

We then hypothesized that, inversely, proteasome inhibition might affect mitochondrial respiration, thus aggravating the effect of complex I inhibitors. We have excluded a direct inhibition of complex I by proteasome inhibitors. A detailed study has already excluded an effect of proteasome inhibition on complex I activity in living cells, but demonstrated that proteasome inhibition compromises the respiratory chain via cytochrome C leakage (Marshansky et al. 2001). Others have shown a decrease in mitochondrial membrane potential after proteasome inhibition (Qiu et al. 2000; Ding and Keller 2001). We investigated whether proteasome inhibition could decrease ATP levels and thereby compromise cell survival. Proteasome inhibition alone did not affect ATP production at all, but it aggravated ATP depletion induced by complex I inhibition to lethal levels, as for example with 30 nm rotenone (Figs 6c and d′). This observation clearly demonstrates that proteasome inhibition can exacerbate impaired mitochondrial energy production enough to trigger cell death. However, we did not find synergistic toxicity of proteasome inhibitors with the inhibitors of complex II–V, although they also decrease ATP production. This mechanism of synergistic ATP-depletion appears therefore not generally applicable to all respiratory chain components.

The most satisfactory explanation for the synergistic toxicity of rotenone, MPP+ and 6-OHDA with proteasome inhibitors appears to implicate events downstream of the proteasome and respiratory chain: the accumulation of oxidatively damaged proteins due to increased ROS production and decreased proteolysis (Figs 6a and b). Accumulation of damaged proteins has deleterious effects on cellular integrity and ultimately triggers apoptosis (Gabai et al. 1998). Therefore, a significant fraction of the basal energy turnover in cells is invested in the detoxification of damaged proteins, the bulk which is handled by the UPS (Sherman and Goldberg 2001). This system appears to be functionally impaired in sporadic and familial PD (McNaught et al. 2001). As shown in Fig. 7(a), cell death induced by complex I inhibition alone appeared to be exclusively a consequence of energy depletion in our system, because restoration of intracellular ATP levels, but not free radical scavenging, protected neurones from a high concentration of rotenone (100 nm). In contrast, radical scavenging did protect neurones from the synergistic cell death that results when a lower, normally subtoxic concentration of rotenone (30 nm) was combined with a proteasome inhibitor (Fig. 7c). Thus, our data suggest, that an impaired UPS renders normally subtoxic levels of ROS critical for neuronal viability, explaining the specific interaction with the toxins rotenone, MPP+ and 6-OHDA.

In summary, our study presents two major novel findings. First, we demonstrated that complex I inhibition decreases proteasomal activity, a finding that may explain the aggregation of ubiquitinated proteins in vivo after rotenone exposure (Betarbet et al. 2000; Höglinger et al. 2003), and may be relevant to the formation of Lewy bodies in sporadic PD. Secondly, we suggest a mechanism by which impaired proteasome function may increase neuronal vulnerability for a specific group of toxins, a mechanism that may be active in both sporadic and familial PD.


This work was supported by the Institut National de la Santé et de la Recherche Médicale, the Fondation de France (UB/SS 500307–2002011991), and the National Parkinson Foundation, Miami. GH was funded by the Deutsche Forschungsgemeinschaft (ho 2402/2–1), GC by the European PROTAGE program (QLKG-CT1999-02193), PPM by the Centre de Recherche Pierre Fabre, Castres, France. We thank M. Hamelin for help with the oxyblot experiments.