OPA1 mutations associated with dominant optic atrophy impair oxidative phosphorylation and mitochondrial fusion

Abstract

Dominant optic atrophy (DOA) is characterized by retinal ganglion cell degeneration leading to optic neuropathy. A subset of DOA is caused by mutations in the OPA1 gene, encoding for a dynamin-related GTPase required for mitochondrial fusion. The functional consequences of OPA1 mutations in DOA patients are still poorly understood. This study investigated the effect of five different OPA1 pathogenic mutations on the energetic efficiency and mitochondrial network dynamics of skin fibroblasts from patients. Although DOA fibroblasts maintained their ATP levels and grew in galactose medium, i.e. under forced oxidative metabolism, a significant impairment in mitochondrial ATP synthesis driven by complex I substrates was found. Furthermore, balloon-like structures in the mitochondrial reticulum were observed in galactose medium and mitochondrial fusion was completely inhibited in about 50% of DOA fibroblasts, but not in control cells. Respiratory complex assembly and the expression level of complex I subunits were similar in control and DOA fibroblasts. Co-immunoprecipitation experiments revealed that OPA1 directly interacts with subunits of complexes I, II and III, but not IV and with apoptosis inducing factor. The results disclose a novel link between OPA1, apoptosis inducing factor and the respiratory complexes that may shed some light on the pathogenic mechanism of DOA.

OPA1, DOA, oxidative phosphorylation, mitochondrial fusion, AIF

Introduction

Mitochondrial hereditary optic neuropathies, the most common being Leber's hereditary optic neuropathy, are characterized by a selective degeneration of retinal ganglion cells (RGCs), leading to optic atrophy and blindness (Carelli et al., 2004). The Kjer-type dominant optic atrophy (DOA) (Kjer, 1959) has recently been added to the list of mitochondrial optic neuropathies. A subset of DOA cases has been associated with heterozygous mutations in the OPA1 gene located on chromosome 3q28-q29 (Alexander et al., 2000; Delettre et al., 2000). This gene encodes a dynamin-related GTPase, which is targeted to mitochondria by an N-terminus import sequence motif and is anchored to the inner membrane facing the intermembrane space (Olichon et al., 2002). The OPA1 protein structure includes a transmembrane domain, a GTPase domain, a middle domain and C-terminal coiled-coil domain proposed to be a GTPase effector domain (see http://lbbma.univ-angers.fr). Eight OPA1 mRNA isoforms resulting from alternative splicing have been described. These are variably expressed in different tissues, with the highest levels being found in retina, brain, testis, heart and muscle (Delettre et al., 2001). Furthermore, the different isoforms seem to be related to distinct conserved OPA1 functions (Olichon et al., 2007a).

To investigate OPA1 function at the cellular level, gene expression was downregulated by RNA interference in HeLa cells. This led to fragmentation of the mitochondrial network, loss of mitochondrial membrane potential and drastic disorganization of the cristae. Furthermore, spontaneous release of cytochrome c and caspase-dependent activation of the apoptotic cascade were observed in this experimental model (Olichon et al., 2003). A quantitative mitochondrial fusion assay was used to demonstrate that OPA1 protein downregulation caused a complete inhibition of mitochondrial fusion (Lee et al., 2004). Therefore, it has been proposed that OPA1, like its yeast homologue Mgm1p (Sesaki et al., 2003), is involved in the regulation of mitochondrial dynamics and the maintenance of structural integrity of the cristae (Griparic et al., 2004; Chen et al., 2005). A recent study emphasized the role of OPA1 in cristae remodelling as a function independent from mitochondrial fusion, directly involved in regulating the apoptotic process through the control of cytochrome c redistribution (Frezza et al., 2006).

Over 90 OPA1 gene mutations have been described to date (Olichon et al., 2006). These mutations may be substitutions, deletions or insertions and they are spread throughout the coding region of the gene, most of them being localized to the GTPase and the GED domains. The c.2708delTTAG frameshift-inducing microdeletion in the GED domain seems to be the most common in patients of European descent (see http://lbbma.univ-angers.fr).

The functional consequences of OPA1 mutations in DOA patients have been poorly investigated. One recent study using phosphorus MR spectroscopy to assess calf muscle oxidative metabolism in patients carrying the c.2708delTTAG microdeletion, showed a subclinical defect in oxidative phosphorylation (Lodi et al., 2004). This observation was reinforced by biochemical investigations in fibroblasts carrying the severe OPA1 R445H mutation, where an oxidative phosphorylation defect was detected together with a significantly reduced rate of ATP synthesis compared to controls (Amati-Bonneau et al., 2005). The R445H mutation, identified in multiple families, is associated with a complex clinical phenotype characterized by DOA, sensorineural deafness and in a few cases adjunctive myopathic features including ptosis and progressive external ophthalmoplegia (Payne et al., 2004). Recently, the R445H mutation, as well as other mis-sense mutations at specific sites involving the GTPase domain of OPA1, have all been linked to mitochondrial DNA (mtDNA) instability and accumulation of multiple deletions in skeletal muscle (Amati-Bonneau et al., 2007; Hudson et al., 2007). Furthermore, another study reported a slight decrease in the number of mtDNA copies per cell in leucocytes from DOA patients with OPA1 mutations (Kim et al., 2005). Thus, OPA1 also seems to be involved in mtDNA stability and possibly maintenance.

The present study undertook a comprehensive investigation of skin fibroblasts derived from DOA patients carrying the common c.2708delTTAG microdeletion and four other pathogenic mutations in the OPA1 gene, all of which are predicted to truncate the protein product. The mitochondrial network morphology and ultrastructure, mitochondrial fusion and oxidative phosphorylation were significantly affected in DOA fibroblasts. Furthermore, we show that OPA1 physically interacts with different respiratory complex subunits and with apoptosis inducing factor (AIF), suggesting a possible mechanism for DOA pathophysiology.

Material and Methods

Materials

3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), tert butylhydroperoxide (t-BH), ATP and ATP monitoring kit were purchased from Sigma (Milan, Italy). Hoechst-33342 was from Calbiochem (La Jolla, CA, USA) and Mitotracker Red from Molecular Probes (Eugene, OR, USA). The antibodies used were as follows: anti-OPA1 (BD Bioscience Pharmingen, Milan, Italy), anti-mitofusin-1 (Abnova, Taipei, Taiwan), anti-DRP1 (Santa Cruz, CA, USA), anti-β-tubulin (Sigma), anti-AIF (Santa Cruz), anti-porin, anti-NDUFA9, anti-ND6 and anti-NDUFB6 subunits of complex I, anti-FP subunit of complex II, anti-core II subunit of complex III and anti-subunit I of complex IV (Molecular Probes). Secondary antibodies were from Jackson ImmunoResearch Europe Ltd (Soham, Cambridgeshire, UK).

OPA1 sequence analysis of DOA patients

For the OPA1 mutation analysis genomic DNA was amplified by PCR with sense and anti-sense primers specific for the 30 previously described exons (Pesch et al., 2001). PCR products were purified by ExoSAP treatment (Amersham, Munich, Germany) and sequenced employing BigDye Terminator chemistry (Applied Biosystems, Weiterstadt, Germany). Sequencing products were separated on an ABI 3100 DNA Sequencer and the data were assessed manually and using alignment tools. Total RNA was isolated from whole blood using the PAXgene Blood RNA system (Quiagen, Hilden, Germany). Single-stranded cDNA was synthesized applying the SuperScript First-Strand Synthesis System (Invitrogen, Carlsbad, CA) and overlapping fragments of the OPA1 cDNA were amplified by PCR prior to direct DNA sequencing as described previously (Schimpf et al., 2006).

Cells and culture conditions

Skin fibroblasts were derived, following informed consent, from five healthy donors, five DOA patients from two unrelated families with the c.2708delTTAG deletion, two patients with the c.2819-2A>C mutation and one patient with each of the other mutations (see the Results section).

Fibroblasts were grown in DMEM medium supplemented with 10% foetal bovine serum (FBS), 2 mM l-glutamine and antibiotics. The mtDNA-lacking 143B206 Rho0 cells and the osteosarcoma-derived 143B.TK-parental cells (a kind gift from Martinuzzi, King and Attardi) were grown in the same DMEM medium containing 50 μg/ml uridine or 0.1 mg/ml bromodeoxyuridine, respectively. For the experiments, fibroblasts were grown in DMEM glucose medium or DMEM glucose-free medium containing 5 mM galactose, 5 mM pyruvate (DMEM galactose medium).

Western blotting and immunoprecipitation

10⁶ cells were resuspended in 0.1 ml of RIPA buffer (50 mM Tris–Cl pH 7.6, 150 mM NaCl, 1% NP-40, 1% NaDOC, 0.1% SDS, 5 mM EDTA, 100 μl/ml of protease inhibitor cocktail), sonicated and centrifuged at 10 000 g. The protein content of the supernatant (cell lysates) was determined according to Bradford (Bradford, 1976). Mitochondria were isolated from 4 × 10⁶ fibroblasts by standard differential centrifugation. Proteins (30 μg) were separated by 8% SDS–PAGE and transferred onto a nitrocellulose membrane (Bio-Rad, Hertfordshire, UK). Primary antibodies were visualized using horseradish peroxidase-conjugated secondary antibodies. Signals were detected using ECL (Amersham). Densitometry was performed with a Fluo-2 MAX Multimager system (Bio-Rad).

For immunoprecipitation experiments, cell lysates (200 μg) were incubated in RIPA buffer and mixed with anti-OPA1 or anti-AIF antibodies at 4°C for 2 h. Protein A-Sepharose (Sigma) was then added. Precipitates were used for western blot analysis.

Blue native electrophoresis

Isolated mitochondria were solubilized using 0.5% lauryl maltoside (Sigma), and proteins were loaded on a linear 5–12% gradient polyacrylamide gel (Nijtmans et al., 2002).

Cell viability and ATP assays

Cell viability was determined with the MTT assay (Ghelli et al., 2003). ATP was measured by using the luciferin/luciferase assay (Zanna et al., 2005).

Mitochondrial ATP synthesis

Mitochondrial ATP synthesis driven by complex I and II substrates was determined in digitonin-permeabilized fibroblasts, exactly as previously reported (Bonora et al., 2006). ATP synthesis through glycerol 3-phosphate dehydrogenase (mGPD) was determined in the presence of 2 μg/ml rotenone and 25 mM glycerol 3-phosphate. Citrate synthase activity was measured as described in (Srere, 1969).

Nuclear and mitochondrial morphology

Nuclear and mitochondrial morphology was assessed after cell staining with 2 μg/ml Hoechst and 10 nM Mitotracker for 30 min at 37°C. Fluorescence was visualized with a digital imaging system using an inverted epifluorescence microscope with 63X/1.4 oil objective (Diaphot, Nikon, Japan). Images were captured with a back-illuminated Photometrics Cascade CCD camera system (Roper Scientific, Tucson, AZ, USA) and Metamorph acquisition/analysis software (Universal Imaging Corp., Downingtown, PA, USA).

Analysis and quantification of mitochondrial fusion

Cells were transfected with 1 μg of mitochondrial matrix-target photoactivable green fluorescent protein (mt-PAGFP) (Karbowski et al., 2004) and incubated in DMEM-glucose or DMEM-galactose medium for 24 h. Images were captured with an LSM 510 confocal microscope (Carl Zeiss MicroImaging, Inc.) using a 63 × 1.4 NA Apochromat objective. The excitation and photoactivation wavelengths for mito-PAGFP were 488 nm and 413 nm, respectively. Regions of interest were selected and series of z-sections from the top to the cell bottom with intervals between the sections set to 0.5–0.75 μm were irradiated with 413-nm light. Images were captured immediately after photoactivation (0 min) and after 30 and 60 min. Mitochondrial fusion was calculated by obtaining the ratio between the number of pixels in the fluorescent area and the total cell number of pixels using the Volocity program. Data were exported to Microsoft Excel and converted into graphs.

Quantification of mtDNA

DNA was isolated using QIAmp DNA Blood Mini kit (Qiagen, Valencia, CA). The cell content of mtDNA was determined by double-competitive polymerase chain reaction (Cossarizza et al., 2003).

Transmission electron microscopy

Fibroblasts grown in DMEM-glucose or DMEM-galactose medium were fixed with 4% paraformaldehyde. Thin sections were examined on a Jeol 1200 EXII TEM.

Statistical analysis

The data were analysed using the Student's t-test, unless otherwise indicated. Only values of P < 0.05 were considered significant.

Results

Energetic competence of fibroblasts with OPA1 mutations

We studied fibroblasts derived from 10 patients carrying OPA1 mutations. The localization of these mutations in the OPA1 gene is reported in Fig. 1A. We also studied fibroblasts from five healthy donors. Five patients carried the frameshift c.2708delTTAG microdeletion in exon 27, whereas another patient had the previously reported c.1705 + 1G>T mutation in the splice donor sequence of intron 17, which activated a cryptic splice site (Schimpf et al., 2006). The other four patients harboured three new mutations, inducing either splice defects (c.1516 + 1G>C in intron 15, one patient; c.2819-2A>C in intron 27, two patients) or a translational frame shift (c.1346_1347insC in exon 14, one patient). All the mutations are predicted to generate a truncated OPA1 protein as confirmed by cDNA analysis with total RNA isolated from patients’ blood. All mutations were heterozygous and strictly co-segregated with individuals affected by optic atrophy in families with autosomal dominant inheritance.

Fig. 1

Expression level of OPA1, mitofusin-1 and DRP1 proteins in control and DOA fibroblasts. (A) Map of the OPA1 gene and localization of the mutations analysed in this study. (B) Representative western blot of OPA1, MFN1, DRP1, β-tubulin and porin, carried out in fibroblast lysates obtained from one control and from patients carrying the indicated OPA1 mutations. OPA1, MFN1 and DRP1 bands were normalized to β-tubulin (C) and porin (D) band density. Results are means ± SEM of four controls and eight DOA fibroblasts lysates obtained from three independent experiments and resolved in at least three blots.

Figure 1B shows the expression level of OPA1, mitofusin-1 (MFN1) and dynamin-related protein-1 (DRP1), the main proteins involved in the processes of mitochondrial fusion and fission, determined in fibroblast lysates by western blotting. We detected two bands corresponding to OPA1, slightly above and below 90 kDa, likely corresponding to the 92 kDa and 86 kDa OPA1 isoforms previously reported in heart and brain. This pattern is similar to that shown by others using homemade polyclonal antibodies in mitochondria isolated from other tissues or in other cell types (Olichon et al., 2002). Two bands were also apparent for mitofusin-1. β-Tubulin and the mitochondrial protein porin were used as controls for quantification. Normalization relative to both β-tubulin and porin showed a significant reduction of OPA1 protein in DOA fibroblasts compared to controls (Fig. 1C and D), confirming that haploinsufficiency results from mutations predicted to generate a truncated protein. Conversely, mitofusin-1 levels were slightly upregulated in DOA fibroblasts, this being more evident when normalized to β-tubulin compared to porin. DRP1 on the other hand remained unchanged in both cases (Fig. 1C and D).

The growth capability and ATP content of control and DOA fibroblasts bearing OPA1 mutations were not significantly different in glucose medium (Fig. 2A and B). To further characterize the energetic competence of DOA cells, we assessed their viability during incubation in a glucose-free medium containing galactose. Under these conditions, cells are forced to rely predominantly on oxidative phosphorylation for ATP synthesis, given the low efficiency of this carbon source to feed the glycolytic pathway (Robinson et al., 1992). Growth in galactose medium was significantly slower than in glucose medium for both DOA and control cells. Again, in galactose medium no significant difference was observed between controls and DOA patients. Under both growth conditions, the number of DOA fibroblasts was higher at all times compared to controls although not significantly different and within the range of variation previously experienced by similar cell growth experiments in our laboratory (Figure 2A and C). Furthermore, the ATP levels of both control and DOA fibroblasts increased after 24 h of incubation in galactose medium compared to glucose medium, as a consequence of the shift from glycolytic to oxidative metabolism. However, this ATP increase was significantly higher in controls compared to patient fibroblasts at 24 h, this difference being negligible after 48 and 72 h of incubation (Fig. 2D). Therefore, DOA fibroblasts initially showed a lower capacity to produce ATP when forced to utilize solely the respiratory chain for their energy production. However, at longer incubation times their ATP levels were comparable to those displayed by control cells and they were viable and able to grow in galactose medium.

Fig. 2

Cell viability and ATP content of control and DOA fibroblasts. (A) Controls and DOA fibroblasts were incubated in DMEM-glucose medium for the times indicated. The MTT absorbance value at time zero was considered the 100% value of the number of viable cells. Each data point represents the mean ± SEM of values obtained from two controls and four DOA fibroblasts. Each cell line was tested at least in triplicate. (B) The ATP content of five controls and eight DOA fibroblasts in DMEM-glucose medium was determined in at least three different preparations for each cell line. (C) Controls and DOA fibroblasts were incubated in DMEM-galactose medium for the times indicated and cell viability determined. The 100% value is the MTT absorbance value determined in DMEM-glucose (time = 0). Each data point represents the means ± SEM of values obtained from two controls and four DOA fibroblasts, each tested in three to five independent experiments. (D) ATP levels of controls and DOA fibroblasts were determined at different times of incubation in galactose medium and expressed as nanomoles per milligram protein. Data are means ± SEM of values obtained from three controls and seven DOA fibroblasts, each tested in three to five independent experiments. Asterisk denotes a value significantly different from that of controls at the same time (P < 0.05).

Mitochondrial network of DOA fibroblasts

Considering that OPA1 function is thought to be relevant in mitochondrial network organization, we investigated the mitochondrial morphology of DOA fibroblasts bearing the OPA1 microdeletion in glucose and after 24 h incubation in galactose medium. After loading with Mitotracker Red and examination by fluorescence microscopy, fibroblasts could be scored into three categories on the basis of mitochondrial morphology. Class I cells showed a typical filamentous network, class II cells showed filamentous mitochondria containing balloon-like structures and class III cells showed complete fragmentation resulting in only isolated mitochondrial balloons. Representative images of these three classes of mitochondrial morphology are shown in Fig. 3A–C.

Fig. 3

Mitochondrial morphology. Control and DOA fibroblasts with the OPA1 microdeletion were incubated in DMEM-glucose or in DMEM-galactose medium for 24 h, and then loaded with Mitotracker Red and Hoechst, as described in Materials and Methods section. Representative images of the three mitochondrial morphologies of DOA fibroblasts incubated in DMEM-galactose medium are shown: (A) class I; (B) class II; (C) class III. The inset at the right of each panel shows an enlarged detail of the mitochondrial network. The inset at the left shows the corresponding nuclear morphology. (D, E) Bar graphs show quantification of the three categories by blind test. Fifty-one control and 56 DOA fibroblasts were counted in glucose medium; 40 control and 44 DOA fibroblasts in galactose medium. Fibroblasts analysed were from three controls and three DOA patients. Asterisk denotes values significantly different from controls (P < 0.01).

In glucose medium, both control and DOA fibroblasts exhibited a similar percentage of class I and class II cells (Fig. 3D). After 24 h incubation in galactose medium, control fibroblasts maintained the same distribution of the three cell classes. Conversely, in DOA fibroblasts the percentage of class I cells decreased, whereas that of class II and III consistently increased, these values becoming significantly different from controls (Fig. 3E). Fragmentation of the mitochondrial network has frequently been reported in cells undergoing apoptosis (Youle and Karbowski, 2005). However, the nuclear morphology, as evaluated after Hoechst staining, in all three classes of mitochondrial network organization never displayed chromatin condensation (insets of Fig. 3A–C), clearly indicating that despite the different degrees of mitochondrial fragmentation these cells did not spontaneously undergo apoptosis.

Representative images of the mitochondrial morphology of control and DOA fibroblasts with the other mutations in DMEM-glucose and DMEM-galactose medium are shown in Fig. 4. Also in fibroblasts bearing these mutations, the percentage of class II and III consistently increased in galactose medium. It is therefore apparent that although DOA fibroblasts are still able to maintain ATP levels and growth capability in galactose medium, a large percentage of cells undergo a drastic remodelling of the mitochondrial network when forced to utilize oxidative phosphorylation for ATP production.

Fig. 4

Mitochondrial morphology of DOA fibroblasts with other OPA1 mutations. DOA fibroblasts with the indicated OPA1 mutations were incubated in DMEM-glucose or in DMEM-galactose medium for 24 h, and then loaded with Mitotracker Red as described in Materials and Methods section. Representatives out of eight similar images are shown for each condition.

Major disruption of the overall organization of the cristae has been reported in cells where OPA1 expression has been downregulated by RNA interference (Olichon et al., 2003). Ultrastructural investigation by transmission electron microscopy was therefore carried out in fibroblast mitochondria under the two different growth conditions. In glucose medium, a remarkable decrease in the number and organization of cristae was evident in ∼50% of DOA fibroblasts with the OPA1 microdeletion compared with controls (Supplemental Fig. 1A and C). In the large majority of images control fibroblasts had a regular pattern of parallel organization of cristae along the short diameter (Supplemental Fig. 1A). Growth in galactose medium occasionally altered cristae organization also in controls compared to glucose medium, and further affected the DOA fibroblast mitochondria. In fact, aberrant elongation of isolated cristae running along the longitudinal diameter of the organelle was frequently observed, dramatically contrasting with the parallel organization essentially maintained in control mitochondria (Supplemental Fig. 1B and D).

Mitochondrial fusion in DOA fibroblasts with c.2708delTTAG microdeletion

The OPA1 protein has been implicated in the mitochondrial fusion process (Olichon et al., 2003; Lee et al., 2004; Chen et al., 2005). A quantitative in vivo mitochondrial fusion assay, based on the measurement of the dilution rate of mito-PAGFP (Karbowski et al., 2004), was used to evaluate whether this process was altered in DOA fibroblasts bearing the OPA1 microdeletion. Small regions of cells transfected with mito-PAGFP were photoactivated and the decrease in fluorescence over time was determined by confocal microscopy. A representative image of one DOA fibroblast in glucose medium at 0, 30 and 60 min after photoactivation is shown in Fig. 5A. In glucose medium, the decrease in fluorescence intensity within the activated regions of DOA fibroblasts was significantly faster than in controls (Fig. 5C). After 24 h incubation in galactose medium, DOA fibroblasts presented both filamentous and balloon-like patterns of mitochondrial network. An example of the latter is shown in Fig. 5B. No class III cells with fragmented mitochondrial network were detected, possibly because these fibroblasts may be less susceptible to successful transfection with mito-PAGFP. The overall rate of mitochondrial fusion in galactose medium was similar in control and DOA cells (Fig. 5D). However, once we scored the DOA fibroblasts analysed in Fig. 5D according to mitochondrial morphology, about 40% of them belonged to class I and 60% to class II (Fig. 5E). When the decrease of mito-PAGFP fluorescence was re-calculated in the two different DOA cell populations (class I versus class II), it was apparent that mitochondrial fusion was almost abolished in class II cells, whereas in class I cells it remained faster than controls, similar to what was observed in glucose medium (Fig. 5F).

Fig. 5

Visualization and quantification of mitochondrial fusion in control and DOA fibroblasts with the OPA1 mutation c.2708delTTAG. Control and DOA fibroblasts with the OPA1 microdeletion were transfected with mito-PAGFP and observed using a confocal microscope. Representative images of one DOA fibroblast incubated in DMEM-glucose (A) or after 24 h incubation in DMEM-galactose medium (B) at 0, 30 and 60 min are shown. The values of mito-PAGFP fluorescence after 30 and 60 min were expressed as per cent of the fluorescence value at 0 min (100% value) in control and DOA fibroblasts incubated in DMEM-glucose (C) or DMEM-galactose (D). Data are the means ± SEM of control (n = 45) and DOA (n = 50) fibroblasts analysed. Fibroblasts were from two controls and two DOA patients. Asterisk denotes values significantly different from controls (P < 0.01). (E) The images of DOA fibroblasts incubated in DMEM-galactose medium analysed in panel D were scored into class I and class II categories. (F) The values of mito-PAGFP fluorescence of 50 DOA fibroblasts incubated in DMEM-galactose were grouped according to the mitochondrial morphology (class I versus class II). Asterisk denotes values significantly different between class I and class II cells (P < 0.01).

Mitochondrial ATP synthesis in fibroblasts with OPA1 mutations

The lower capacity to produce ATP and the induction of a fragmented mitochondrial network under conditions of forced oxidative metabolism in galactose medium, in addition to the abnormal mitochondrial ultrastructure, prompted us to investigate the energetic efficiency of mitochondria in more detail. We measured the rate of ATP synthesis in digitonin-permeabilized cells, after normalization for the citrate synthase activity, an indicator widely used to evaluate mitochondrial mass. The values of citrate synthase activity of control and DOA fibroblasts were 132 ± 57 and 122 ± 50 nmol/min/mg, respectively in glucose medium and 124 ± 42 and 148 ± 61 nmol/min/mg, respectively in galactose medium. As reported in Fig. 6A, in glucose medium the rate of ATP synthesis driven by the complex I substrates (pyruvate and malate) was significantly decreased in fibroblasts with OPA1 mutations compared to controls. To evaluate the ATP synthesis excluding complex I, we used substrates transferring reducing equivalents into the electron transport chain to ubiquinone through complex II (succinate) or through the mitochondrial glycerol 3-phosphate dehydrogenase (mGPD). A tendency towards reduction, without reaching a statistical significance, was observed for complex II, whereas values similar to controls were obtained for mGPD (Fig. 6A).

Fig. 6

Mitochondrial ATP synthesis in control and DOA fibroblasts. Fibroblasts incubated for 24 h in DMEM-glucose (A) or DMEM-galactose (B) were treated with 50 μg/ml digitonin and the rate of ATP synthesis driven by 5 mM pyruvate plus 5 mM malate (complex I) or 10 mM succinate plus 4 μM rotenone (complex II) or 25 mM glycerol 3-phosphate plus 4 μM rotenone (mGPD) was subsequently determined. The rate of ATP synthesis was normalized for citrate synthase (CS) activity. Data (mean ± SEM) were obtained in fibroblasts from five controls and seven DOA patients. The experiment was performed at least in duplicate. Asterisk denotes values significantly different from controls (P < 0.05).

After 24 h of incubation in galactose medium, in control fibroblasts the rates of mitochondrial ATP synthesis driven through complex I and mGPD, but not through complex II were increased compared to those in glucose medium, without reaching statistical significance (Fig. 6A and B). Conversely, in DOA fibroblasts ATP synthesis through all complexes was similar to that in glucose medium. Thus, in galactose medium ATP synthesis driven by complex I and also by mGPD was significantly different between DOA and control fibroblasts (Fig. 6B).

Overall, these results reveal a clearly defective oxidative phosphorylation prevalently through complex I, as a common pathological feature characterizing all the OPA1 mutations investigated in this study.

Analysis of mtDNA content and respiratory chain complexes in DOA fibroblasts

A reduced mtDNA copy number has been reported in leucocytes from DOA patients bearing different OPA1 mutations (Kim et al., 2005), thus providing a potential link between OPA1 mutations and the respiratory chain defect. However, determination of the mtDNA content by quantitative real-time PCR showed an increase in mtDNA copy number per cell in DOA fibroblasts compared with controls, which was significant in DMEM-glucose medium, but not in galactose medium (Fig. 7A).

Fig. 7

Determination of mtDNA copy number and expression level of complex I subunits in control and DOA fibroblasts. (A) Fibroblasts from controls and DOA patients were incubated in DMEM-glucose medium or for 24 h in DMEM-galactose medium and mtDNA copies/cell calculated for each sample from the ratio between mtDNA copies and nDNA copies multiplied by two. Data were obtained in fibroblasts derived from four controls and six DOA patients, each tested in at least two experiments run in triplicate. Comparative box plots are shown: the top and bottom of the box indicate the 25th and 75th percentiles, the middle line indicates the median and the black square shows the mean numbers measured in each group. The whiskers indicate the 5th and 95th percentiles. Asterisk denotes values significantly different from controls (P < 0.05) determined by the Mann–Whitney Rank Sum Test. (B) The expression level of NDUFA9 (39 kDa), ND6 (20 kDa) and NDUFB6 (17 kDa) subunits of complex I and of porin was determined in mitochondria isolated from control and DOA fibroblasts by western blot. A representative out of two similar blots is shown. (C) Each band was normalized to porin band density. Results are means ± SEM (n = 4 controls; n = 12 DOA).

Impairment of oxidative phosphorylation could reflect a defective assembly or a decreased number of respiratory complexes. To investigate respiratory complex assembly, mitochondria were isolated from fibroblasts derived from one control and three DOA patients and respiratory complexes were resolved by blue native electrophoresis (Nijtmans et al., 2002). Bands corresponding to complexes I, II, III, IV and V were consistently well resolved. No significant difference was observed between control and DOA patients (Supplemental Fig. 2). Furthermore, the expression level of NDUFA9, ND6 and NDUFB6 subunits of complex I was also determined in the mitochondrial fractions (Fig. 7B). Normalization to porin showed that the levels of these complex I subunits were slightly increased in DOA fibroblasts, this effect being greater, but not significant, for the ND6 subunit (Fig. 7C). Similarly, the level of the complex II FP subunit was not affected (data not shown).

OPA1 directly interacts with respiratory complexes and with AIF

To explore the hypothesis that the OPA1 protein could influence oxidative phosphorylation through a physical interaction with one or more of the respiratory complexes, cellular lysates from controls and DOA fibroblasts bearing four different OPA1 mutations were immuno-precipitated with anti-OPA1 antibody and immunoblotted with antibodies specific to individual structural subunits of the respiratory complexes. Figure 8A shows that the 39-kDa (NDUFA9) subunit of complex I, the 70-kDa (FP) subunit of complex II and the core II subunit of complex III co-precipitated with OPA1, whereas subunit I of complex IV was not detected. It is worth noting that similar results were obtained by co-immunoprecipitation experiments using both cellular lysates or mitochondrial fractions (data not shown).

Fig. 8

Interaction of OPA1 with mitochondrial proteins. (A) Cellular lysates obtained from control and DOA fibroblasts with the indicated OPA1 mutations were immunoprecipitated with anti-OPA1 antibody and immunoblotted with antibodies against the NDUFA9 subunit of complex I, the FP subunit of complex II, the core II subunits of complex III, the subunit I of complex IV, AIF and OPA1. The reaction mixture prior to immunoprecipitation was also analysed by immunoblotting using anti-OPA1 antibody (input). In each blot, immunoreactivity between lysates and protein-A Sepharose in the absence of antibodies was determined (no-Ab). (B) The same lysates tested in panel (A) were immunoprecipitated with antibody against anti-AIF and revealed with the same antibodies used in panel (A). (C) Cellular lysates obtained from 143B206 Rho0 and 143B.TK⁻ cells were immunoprecipitated with anti-AIF or anti-OPA1 antibodies and immunoblotted with antibodies against OPA1, AIF and the NDUFA9 subunit of complex I. The reaction mixture prior to immunoprecipitation was analysed by immunoblotting using anti-OPA1, AIF, NDUFA9 and porin antibodies (input). Blots are representative of three similar experiments.

Given that AIF has been shown to participate in complex I assembly and maintenance (Vahsen et al., 2004), the same nitrocellulose membranes were also immunoblotted with anti-AIF antibody revealing a high affinity interaction between OPA1 and AIF. Similar results were obtained when cellular lysates were immunoprecipitated with anti-AIF antibody (Fig. 8B). Furthermore, when we tested the same nitrocellulose membranes with antibodies against actin, tubulin and DRP1 we did not detect any signal (data not shown), suggesting that the interaction among OPA1, AIF and respiratory complexes I, II and III was specific.

We then used anti-OPA1 and anti-AIF antibodies for immunoprecipitation of cellular lysates obtained from the mtDNA-lacking 143B206 Rho0 cells and the original parental osteosarcoma-derived 143B.TK⁻ cells (King and Attardi, 1989). The 143B206 Rho0 cells completely lack the mitochondrial genome and consequently do not have a functional respiratory chain. It has been shown that due to the absence of the mtDNA-encoded ND subunits the complex I is only partially assembled (Ugalde et al., 2004). Consistently with this notion, Fig. 8C (input) shows that 143B206 Rho0 cells expressed drastically reduced levels of the nuclear encoded NDUFA9 subunit of complex I compared to the parental 143B.TK⁻ cells, as previously reported (Ugalde et al., 2004). Accordingly, NDUFA9 subunit was not detected in the anti-OPA1 and anti-AIF immunoprecipitates. Conversely, under these conditions of partially assembled complex I, the cross-immunoreactivity between OPA1 and AIF was maintained in 143B206 Rho0 cells, similar to the parental 143B.TK⁻ cells (Fig. 8C).

Sensitivity of DOA fibroblasts to a pro-oxidant agent

OPA1 mutations have been shown to sensitize fibroblasts to apoptotic cell death induced by staurosporine (Olichon et al., 2007b). Figure 9A shows that incubation with the pro-oxidant agent t-BH induced a significantly more rapid loss of viability in DOA fibroblasts compared to controls. This was associated with a marked decrease of ATP content, which became significant after 16 h (Fig. 9B). Analysis of nuclear morphology showed that the number of nuclei exhibiting chromatin condensation was significantly increased in DOA fibroblasts treated with t-BH compared to controls (71 ± 5%, n = 50 in DOA and 29 ± 9%, n = 46 in controls). Representative images of nuclei before and after t-BH treatment are reported in Fig. 9C and D.

Fig. 9

Viability and ATP content of control and DOA fibroblasts treated with t-BH. Control and DOA fibroblasts were incubated with 250 μM t-BH for the times indicated and the number of viable cells (A) and the ATP content (B) were determined as described in Fig. 2. Data are expressed as % of the values measured in DMEM-glucose before addition of t-BH (time = 0). Data are means ± SEM of values obtained from two controls and two DOA fibroblasts, each tested in at least three independent experiments. Asterisk denotes a value significantly different from those of controls at the same time (P < 0.05). Representative images of Hoechst-loaded nuclei from one DOA patient before (C) and after 6 h incubation with t-BH (D).

Discussion

The present study shows that the common c.2708delTTAG OPA1 microdeletion and four other pathogenic mutations, all predicted to generate a truncated protein, cause a significant reduction in the level of the OPA1 protein in fibroblasts. This would support the hypothesis that the pathological mechanism involves haploinsufficiency. Reduced OPA1 expression was associated with a significant impairment of oxidative phosphorylation, mostly mediated at the level of complex I. When cells were forced to use oxidative metabolism in galactose medium, ATP levels failed to increase as in controls, but cells remained viable for several days. Furthermore, growth in galactose medium strongly promoted an extensive mitochondrial network remodelling and inhibited mitochondrial fusion in a consistent subset of cells. We also documented that OPA1 mutant fibroblasts were more prone to cell death than controls after an exogenous oxidative stress. Finally, we showed for the first time that OPA1 interacts with some respiratory complexes and AIF.

Some indications that defective oxidative phosphorylation is caused by OPA1 mutations have been previously reported. A reduction of mitochondrial ATP synthesis through the respiratory chain as a whole was shown in fibroblasts from patients suffering from DOA and sensorineural deafness patients carrying the severe mis-sense OPA1 mutation R445H (Amati-Bonneau et al., 2005), which is not predicted to generate a truncated protein. We also reported that the very same patients bearing the c.2708delTTAG microdeletion in the OPA1 gene, whose cell lines were investigated in this study, had a defective oxidative phosphorylation in skeletal muscle as detected by phosphorus MR spectroscopy (Lodi et al., 2004). The present study in fibroblasts from these same patients, pooled with others bearing four different pathogenic mutations, shows for the first time that the rate of mitochondrial ATP synthesis was significantly reduced at the level of complex I, with this phenomenon being more evident in galactose medium. The investigation of ATP synthesis driven by complex II and mGPD did not reveal any other major site of impairment in DOA fibroblasts, thus pointing to complex I as the candidate for the defective oxidative phosphorylation. The failure to increase ATP levels after 24 h of incubation in galactose medium, shown by DOA cells compared to controls, was paralleled by the significant difference in complex I and mGPD-driven ATP synthesis seen between controls and DOA fibroblasts in the same conditions. The observed upregulation of electron flux through complex I and mGPD in control cells when forced to use mitochondrial oxidative phosphorylation to sustain ATP production deserves further study, as it seems to be an interesting compensatory mechanism, and one which DOA cells failed to operate. Overall, these data suggest that a respiratory chain impairment is common to these OPA1 mutations, and complex I seems to play a major role. However, it is noteworthy that under forced oxidative metabolism in galactose medium the viability of all OPA1 mutant fibroblasts was essentially preserved, indicating that a sufficient ATP production is maintained in DOA.

Why ATP synthesis is impaired prevalently through complex I needs to be explained. Complex I is the largest complex of the respiratory chain and the first site for electron transfer and energy conservation in oxidative phosphorylation (Brandt, 2006). It is built up of 46 subunits in mammals, seven of which are encoded by mtDNA. The gene expression of nuclear and mitochondrial-encoded subunits of complex I, and their coordinated assembly in human cells are both still poorly understood processes (Ugalde et al., 2004). Furthermore, only a few proteins directly interacting with complex I and possibly involved in its assembly and stabilization have been identified, such as AIF, B17.2L, NDUFAF1 and possibly paraplegin (Atorino et al., 2003; Vahsen et al., 2004; Ogilvie et al., 2005; Vogel et al., 2005).

Kim and colleagues recently reported a small reduction of mtDNA copy number in leucocytes from DOA patients bearing different OPA1 mutations (Kim et al., 2005). Both mtDNA and OPA1 protein are anchored to the inner mitochondrial membrane and it was reasoned that the destabilization of cristae organization induced by OPA1 mutations, as was also shown by our ultrastructural findings, may disrupt the stability of mtDNA nucleoids leading to a partial depletion. This would provide a potential link between OPA1 mutations and the respiratory chain defect. Furthermore, considering that complex I is over-represented in terms of structural genes encoded by mtDNA (seven ND subunit genes), it is reasonable to speculate that partial mtDNA depletion may first and prevalently impair complex I function. However, this hypothesis was not confirmed by our evaluation of mtDNA copy number in DOA fibroblasts where, contrary to expectations, we observed a significant increase in mtDNA content. This may be interpreted as a classical compensatory mechanism for an oxidative phosphorylation defect, as commonly seen in patients with mitochondrial encephalomyopathies (DiMauro and Schon, 2003). Furthermore, we observed no decrease in the expression levels of a few complex I (NDUFA9, NDUFB6 and ND6) and complex II (FP) subunits for any of the OPA1 mutations investigated. The augmented levels of the ND6 subunit indicate an increase in mitochondrial biogenesis. The results of blue native electrophoresis further rule out the occurrence of a defective assembly of the respiratory complexes.

The immunoprecipitation experiments indicate for the first time that OPA1 and AIF physically interact, and that both proteins associate with subunits of the respiratory complexes. This interaction between OPA1 and AIF persisted even when the respiratory complexes containing mtDNA-encoded subunits were only partially assembled, in particular complex I, as shown by our experiments with the 143B206 Rho0 cells. However, from this set of experiments we did not detect differences between OPA1 mutant and control fibroblasts. These results suggest that AIF and OPA1 may have a complex role in regulating and stabilizing the respiratory chain, possibly having a major impact on complex I. Both proteins are located in the same compartment, the inner mitochondrial membrane, which also hosts the respiratory chain complexes. There is accumulating evidence that both OPA1 and AIF have multifunctional roles. OPA1 has been shown to accomplish different unrelated functions in mitochondrial network dynamics and in regulating cristae organization, cytochrome c storage and apoptosis (Olichon et al., 2003; Frezza et al., 2006), and also in mtDNA stability (Amati-Bonneau et al., 2007; Hudson et al., 2007). AIF is a flavoprotein with an NADH oxidase activity, which has a further role in stabilizing complex I and, once released, becomes a pro-apoptotic factor (Vahsen et al., 2004). We propose therefore a new scenario where the OPA1 protein may play a role in the respiratory chain by interacting with other proteins such as AIF as shown here, and also paraplegin as recently reported (Ishihara et al., 2006). Thus, OPA1 haploinsufficiency may lead to a defective oxidative phosphorylation as clearly identified in this and previous studies (Lodi et al., 2004). Furthermore, a well-established propensity of OPA1 mutant cells to undergo apoptosis, which has been documented by others (Olichon et al., 2007b) and also in this study, may well relate to the release of cytochrome c, as shown by Frezza and colleagues (2006), but also to differences in the availability of other pro-apoptotic factors, such as AIF.

The other set of results from this study concerns the mitochondrial morphology of fibroblasts carrying OPA1 mutations. The morphology of mitochondria from patients was similar to that of control cells in glucose medium, whereas it changed remarkably, displaying fragmented/balloon-like mitochondria, when cells were forced to use oxidative metabolism in galactose medium. Our data regarding mitochondrial morphology are in agreement with the current concept that mutations inducing haploinsufficiency do not influence the mitochondrial network in glucose medium (Olichon et al., 2006). We also provided the first quantitative evaluation of the fusion process in OPA1 mutant cells, closely associated with the mitochondrial network organization. DOA cells with balloon-like mitochondrial morphology essentially lacked fusion events, whereas unexpectedly, those with a filamentous mitochondrial network showed a significantly increased fusion rate. Our observation that mitofusin-1 levels were increased, albeit not significantly, may indicate a compensatory mechanism for OPA1 haploinsufficiency, possibly explaining the increased fusion rate in the subset of cells maintaining a filamentous mitochondrial network.

It is intriguing that when downregulated or mutated, both AIF and paraplegin have been linked to defective complex I and optic atrophy in animals and humans (Klein et al., 2002; Atorino et al., 2003). In fact, the Harlequin mouse with reduced AIF expression is characterized pathologically by an early and prevalent degeneration of RGCs leading to optic atrophy, and a defect in complex I stabilization (Ishihara et al., 2006). Similarly, paraplegin mutations in humans lead to complicated spastic paraplegia, which includes optic atrophy, and again to a defective assembly of complex I (Atorino et al., 2003). Likewise, these observations suggest that defective OPA1, AIF and paraplegin proteins share a final common pathological mechanism, which involves complex I impairment and leads to degeneration of RGCs and optic atrophy. However, most remarkably, the finding that complex I is impaired in DOA directly establishes a link with LHON, the other non-syndromic mitochondrial optic neuropathy sharing many features with DOA and caused by mtDNA point mutations affecting complex I (Carelli et al., 2004, 2007). Both diseases share a selective loss of RGCs, preferentially hitting the small, less myelinated and more energy-dependent fibres of the papillomacular bundle. The main difference distinguishing DOA and LHON is the timing of the degenerative processes, which in most cases is acute/subacute in LHON with massive and synchronized loss of RGCs, and slowly progressive with frequent stable conditions in DOA. However, the clinical end-point of optic atrophy and severe visual impairment is identical. Our study strongly suggests the existence of common steps in the pathogenic mechanism in both DOA and LHON. In particular, the link between complex I dysfunction and mitochondrial network organization is of great interest in the context of mitochondrial biogenesis in RGC somata and mitochondrial axonal transport along the optic nerve (Hollenbeck and Saxton, 2005). It is known that a skewed distribution of mitochondria characterizes the unmyelinated initial portion of the RGC axons before the lamina cribrosa, where mitochondria are abundant, compared to the postlaminar optic nerve, which is myelinated and has less mitochondria (Carelli et al., 2004, 2007). Thus, any perturbation of organellar transport in axons and mitochondrial distribution seems very relevant to the degenerative process of RGCs. The evidence that mitochondrial fusion is altered in DOA and the fact that this relates to an impairment prevalently expressed at complex I level, suggests that the complex I defect in LHON may have similar effects on mitochondrial network organization. The further and final link paralleling DOA and LHON is the evidence that in both disorders cells are predisposed to undergo apoptosis, under the appropriate triggering conditions (Danielson et al., 2002; Ghelli et al., 2003; Zanna et al., 2005).

In conclusion, our study on fibroblasts with pathogenic mutations in the OPA1 gene inducing haploinsufficiency suggests that there are different layers of interconnected consequences at the cellular level. First, we link the OPA1 mutations to a mitochondrial respiratory chain dysfunction, which is mainly mediated by complex I, possibly through a direct interaction of the OPA1 protein with the respiratory complexes and other proteins such as AIF. We are well aware that this new scenario suggested by the current results warrants an extensive further set of investigations. Second, we show a close relationship between the energetic state of the cell and the mitochondrial network organization, emphasizing the central role of OPA1 in both functions. Thus, we propose that energy production to transport and distribute mitochondria within the axonal system in association with mitochondrial network dynamics must both be crucial to maintain the efficiency of this cellular system. When this fails, a propensity to apoptotic death of RGCs may occur leading to progressive neurodegeneration and optic atrophy.

Supplementary material

Supplementary material is available at Brain online.

Acknowledgements

We thank Dr A. Antignani, National Institutes of Health, Bethesda, for technical support with confocal microscopy and Dr A. Iuso, University of Bari, for help and suggestions for BN PAGE. We also thank Dr R. Liguori and Dr V. Donadio at the Department of Neurological Sciences for kindly performing the skin biopsies. C.Z. was a recipient of a Marco Polo Fellowship from the University of Bologna. This work was partially supported by grants from Ministero della Salute, progetto di ricerca sanitaria finalizzata to V.C. and M.R., a Telethon grant GGP06233 to V.C. and a grant from the Interdisciplinary Center of Clinical Research Tübingen (IZKF) to B.W. We are deeply indebted to all DOA patients and their families for participating in this project. Funding to pay the Open Access publication charges for this article was provided by Telethon grant GGP06233.

Abbreviations:

Abbreviations:
• AIF

  apoptosis inducing factor

• DMEM

  Dulbecco's modified Eagle's medium

• mito-PAGFP

  mitochondrial matrix-targeted photoactivable green fluorescent protein

• mtDNA

  mitochondrial DNA

• MTT

  3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyl tetrazolium bromide

• RGC

  retinal ganglion cell

Author notes

*These authors contributed equally to this work.