Cross-linking ATP synthase complexes in vivo eliminates mitochondrial cristae

Mitochondria are ubiquitous and essential organelles in eukaryotic organisms. They synthesise a variety of cellular metabolites, house the reactions of the tricarboxylic acid cycle and produce the bulk of cellular ATP. Mitochondria are enclosed by a double membrane that surrounds a dense, protein-filled matrix. The enzymes of the electron transport chain reside within the inner membrane, the surface area of which is increased by a series of membrane folds termed cristae. Cristae are thought to facilitate ATP production by enabling more electron transport chain complexes to be packed into the mitochondrion. The electron transport chain pumps protons from the matrix during respiration, establishing a proton-motive force across the inner membrane. The proton-motive force drives protons back into the matrix through F₁F_O-ATPase, which harnesses proton flow to catalyse the energetically unfavourable conversion of ADP and inorganic phosphate to ATP (Capaldi and Aggeler, 2002).

Recently it has been shown that yeast mitochondrial F₁F_O-ATPase synthase (mtATPase) complexes can be recovered as dimers from detergent extracts of mitochondria using blue native gel electrophoresis (Arnold et al., 1998; Schägger and Pfeiffer, 2000). However, dimeric mtATPase could not be isolated from mitochondria lacking either of the nonessential mtATPase subunits e or g, suggesting a role for these subunits in mtATPase dimer formation and/or their stabilisation (Arnold et al., 1998). Subsequent work has shown that small amounts of dimeric mtATPase can be recovered from digitonin extracts of mitochondria from yeast cells lacking subunit g (Brunner et al., 2002; Paumard et al., 2002b). Dimeric mtATPase has also been recovered from detergent extracts of bovine heart mitochondria (Schägger and Pfeiffer, 2000), but not from bacteria or chloroplast membranes, suggesting F₁F_O-ATPase dimers may be specific to mitochondria (Schägger, 2002). Independent evidence for dimeric mtATPase complexes in mitochondrial membranes has been provided by the generation of disulfide bridges between two b subunits (Spannagel et al., 1998) and cross-links between two i subunits (Paumard et al., 2002a) of neighbouring mtATPase complexes. Most recently an oligomeric arrangement for mtATPases has been proposed following the isolation of apparent higher order mtATPase structures containing more than two mtATPase complexes from mitochondrial lysates using blue native gel electrophoresis (Paumard et al., 2002b).

The physiological relevance of mtATPase dimerisation and/or oligomerisation remains unclear. Dimerisation has been thought to confer stability to the enzyme (Arnold et al., 1998; Schägger, 2001), or promote efficient regulation of mtATPase activity by the inhibitor protein (IF₁/Inh1p) (Cabezon et al., 2000; Dienhart et al., 2002). Altered morphology of the inner mitochondrial membrane in cells lacking either subunit e or g has been reported, suggesting a role for mtATPase dimerisation, or even oligomerisation, in cristae formation (Paumard et al., 2002b). While the notion of mtATPase dimerisation has become generally accepted, the existence of mtATPase dimers in vivo has yet to be shown. However, the possibility that mtATPase dimers are created in vitro by the very methods used to recover them has been refuted (Arselin et al., 2003; Schägger, 2001; Schägger, 2002).

Recently we found that respiratory growth of yeast cells having mtATPase incorporating a subunit γ-DsRed fusion was impaired relative to control cells (Gavin et al., 2002). Furthermore, native gel electrophoresis revealed the presence of mtATPase oligomers in mitochondrial lysates prepared under conditions where normally only mtATPase monomers are recovered. By contrast, expression of an equivalent subunit γ-GFP fusion did not impair growth of cells, or result in mtATPase oligomerisation (Gavin et al., 2002; Prescott et al., 2003). DsRed is known to form tetramers in vitro and in living cells (Baird et al., 2000), a property that can cause oligomerisation of proteins to which it is fused (Gavin et al., 2002; Lauf et al., 2001; Mizuno et al., 2001). Thus, we concluded that strong interactions between the DsRed moieties of individual mtATPase complexes incorporating the γ-DsRed fusion led to stable associations between multiple adjacent mtATPase complexes. Co-expression of DsRed not fused to another protein (free DsRed) that was targeted to mitochondria partially reversed the slower growth phenotype of cells, presumably because the free DsRed competed with interaction sites on the subunit γ-DsRed fusion proteins, thereby releasing some mtATPase complexes from their oligomeric arrangement (Gavin et al., 2002).

In the present study we sought to further exploit the tetrameric associations of subunit γ-DsRed fusion proteins by using DsRed as an `in vivo cross-linker' to specifically examine the association of mtATPase complexes within the membrane and the consequences of cross-linked mtATPase oligomers for respiring cells. The key finding of this study was that mitochondrial cristae were eliminated in cells expressing the subunit γ-DsRed fusion, indicating that the correct arrangement of mtATPase complexes within the inner mitochondrial membrane directly affects cristae formation. Recovery of mtATPase tetramers from mitochondrial lysates suggests that structured associations of mtATPase complexes can be trapped in respiring cells.

Saccharomyces cerevisiae strains used in this study are described in Table 1. YRD15 (MATα, his3, ura3, leu2,[rho⁺]) was the parental strain (Straffon et al., 1998). Strains lacking endogenous subunit γ and expressing γ-DsRed and γ-GFP fusions have been described elsewhere (Gavin et al., 2002; Prescott et al., 2003). The chromosomal ATP3 gene encoding the γ subunit was modified in strain γ-27-DsRed to express the DsRed protein (Clontech, Palo Alto, CA) fused to the C-terminus of subunit γ via a 27 amino acid linker. Strain γ-27-GFP expresses a yeast enhanced variant of GFP (YEGFP3) (Cormack et al., 1997) linked by an identical 27 amino acid linker to the C-terminus of subunit γ. Growth media used were as described by Boyle et al. (Boyle et al., 1999) and supplemented with uracil, histidine and leucine as required. Yeast cultures were grown aerobically in liquid medium (SaccE) containing 2% (v/v) ethanol at 28°C. All experiments used cells harvested in mid-logarithmic growth phase. The petite (rho^–/rho⁰) frequency of cells was determined following overnight growth in liquid medium containing 2% glucose, followed by spreading aliquots on solid medium containing 2% ethanol and 0.1% glucose. Small petite cells unable to utilise ethanol for respiratory growth were scored and expressed as a percentage of the total population.

Mitochondria were prepared as described previously (Arnold et al., 1998) and stored at –70°C. ATPase assays were performed on mitochondrial membranes as in Bateson et al. (Bateson et al., 1999).

Clear Native-PAGE was performed as described previously (Gavin et al., 2002; Schägger et al., 1994). Mitochondrial detergent extracts were centrifuged (100,000 g, 20 minutes) and supernatants loaded onto polyacrylamide gels (gradient 3-13%). Gels were imaged for fluorescence with a ProXPRESS multi-wavelength imager (PerkinElmer Life Sciences, Rowville, Australia) using appropriate filters for excitation (GFP, 480±20 nm; DsRed 540±25 nm) and emission (GFP 535±20 nm; DsRed 590±35 nm). Following imaging, gels were analysed in situ for ATPase activity (Yoshida et al., 1975), or bands excised for second dimension SDS-PAGE (Gavin et al., 2002). PVDF membranes (Pall Gelman Laboratory, Sydney, Australia) were used for western blotting. Polyclonal antibodies against subunits e and g were kindly provided by Jean Velours (Institut de Biochimie et Génétique Cellulaires du CNRS, Université Victor Segalen, Bordeaux, France). Membranes were incubated with alkaline-phosphate conjugated antibodies (Amrad-Pharmacia, Melbourne, Australia) and visualised using a Storm Phosphoimager (Molecular Dynamics Australia, Melbourne) following the addition of chemifluorescent Vistra substrate (Amersham Pharmacia Biotech, Sydney, Australia) as described previously (Bateson et al., 1999).

Yeast strains were sampled during mid-logarithmic growth in SaccE for fluorescence microscopy. A 5 μl aliquot was sealed between a microscope slide and coverslip using nail polish and examined using an Olympus BX60 fluorescence microscope (Melbourne, Australia) equipped with a 1.35 NA oil immersion lens (Olympus 100X; UPlanapo). Images were captured using a DAGE CCD-72 ETX camera and MCID acquisition software (Imaging Research, Ontario, Canada).

Transmission electron microscopy (TEM) was essentially performed as in O'Malley et al. (O'Malley et al., 2001). Cells were fixed using a phosphate-buffered fixative (2% paraformaldehyde, 2% glutaraldehyde in 0.1 M phosphate buffer pH 6.8) for 5 minutes. After centrifugation (1000 g, 10 minutes), pellets were resuspended in a second fixative (1% paraformaldehyde, 1% glutaraldehyde in 0.1 M phosphate buffer, pH 6.8) and incubated for 30 minutes. Cells were washed three times with PBS before being permeabilised with 1% sodium metaperiodate (20 minutes). A further series of washes with PBS was followed by further fixation in 0.5% potassium permanganate (20 minutes). Cells were washed with distilled water, stained with 2% uranyl acetate in the dark (15 minutes), and then washed thoroughly with distilled water. Cells were dehydrated by immersion in a series of alcohol concentrations (10-100%) before being embedded in Spurr's resin (ProSciTech, Kirwan, Australia). Sections (90 nm) were stained for 10 minutes with lead citrate and photographed with a Jeol 200CX transmission electron microscope (Brooksvale, Australia) at 100 kV.

Yeast cells were prepared for immunogold labelling as follows. An equal volume of 4% paraformaldehyde in 50 mM Hepes (Sigma, Sydney, Australia) buffer with 150 mM NaCl₂, 4 mM MgCl₂, 4 mM CaCl₂ and 2 mM KCl pH 6.8 was added to a suspension of yeast cells in culture medium and left for 10 minutes. The cells were pelleted (1000 g for 3 minutes), resuspended in fresh 4% paraformaldehyde in Hepes buffer and fixed at room temperature for a further 60 minutes. The cells were then washed in Hepes buffer followed by 0.1 M phosphate buffer pH 7.4. The cells were then suspended in 10% gelatin dissolved in 0.1 M phosphate buffer and cooled to 4°C. The gelatin blocks were then cut into small pieces (1 mm × 1 mm) that were placed in 15% Polyvinyl-Pyrrolidone (Sigma, Sydney, Australia), 1.7 M sucrose in 0.1 M phosphate buffer overnight at 4°C before being frozen by immersion in liquid nitrogen. The pieces were then freeze substituted in methanol for 44 hours at –90°C and infiltrated with Lowicryl HM20 resin (Polysciences, Warrington, PA) at –50°C for 8 hours and finally cured at –50°C with UV light for 65 hours using a Leica EM AFS. Ultrathin sections were cut using a Leica Ultracut UCT ultramicrotome and mounted on Formvar coated nickel grids. Grids were incubated overnight at room temperature with monoclonal (mouse) antibody RH51 raised against mtATPase subunit α (Grasso et al., 1991) and used in conjunction with anti-mouse IgG conjugated to 10 nm gold (AURION, Wageningen, The Netherlands) as per manufacturer's instructions.

Yeast strains γ-27-DsRed and γ-27-GFP have been described previously (Gavin et al., 2002; Prescott et al., 2003) (see also Table 1). In brief, the DsRed or GFP protein was expressed fused to the C terminus of subunit γ by a 27 amino acid linker, in yeast cells lacking endogenous subunit γ. These cells grow on a nonfermentable substrate (ethanol), indicating the assembly of functional mtATPase complexes.

Strain γ-27-DsRed showed a marked increase in generation time relative to both γ-27-GFP and the parental strain YRD15 (Gavin et al., 2002) (see also Table 1). The association of mtATPase complexes recovered from mitochondrial lysates prepared from these strains was compared using clear native polyacrylamide gel electrophoresis (CN-PAGE). Mitochondria were solubilised using dodecyl β-maltoside or digitonin and detergent extracts separated under native conditions on a 3-13% acrylamide gradient. Following electrophoresis, gels were analysed to detect DsRed or GFP fluorescence followed by in situ assay of ATPase activity (Fig. 1A and 1B, respectively).

Two bands (a and b) having intense DsRed fluorescence were detected in extracts from γ-27-DsRed mitochondria solubilised using dodecyl β-maltoside (Fig. 1A, lane 1). On the basis of the reported stoichiometry for DsRed oligomers (Baird et al., 2000), the band of least mobility (a) in γ-27-DsRed lysate was considered to represent a tetrameric mtATPase arrangement and the band of higher mobility (b) was considered to represent trimers (see also Gavin et al., 2002). Under identical extraction conditions, mtATPase from γ-27-GFP mitochondria was recovered predominately in a single band (d) assumed to be the monomeric form (Fig. 1A, lane 4). Mitochondrial lysate prepared from the parental strain YRD15 contained no fluorescent bands (Fig. 1A, lane 6) as expected, but the presence of monomeric mtATPase was visible following in situ ATPase assay (Fig. 1B, lane 6). The mtATPase complexes in all lanes displayed in situ ATPase activity (Fig. 1B) that was sensitive to oligomycin (not shown), confirming they contained only fully assembled mtATPase complexes and not combinations of mtATPase and F₁. Solubilisation of mitochondria with dodecyl β-maltoside normally leads to recovery of predominately monomeric mtATPase, showing that the strong association between DsRed moieties responsible for cross-linking mtATPase complexes in cells expressing γ-DsRed is maintained under conditions of dodecyl β-maltoside extraction.

The expression of free DsRed targeted to the mitochondrial matrix (strain γ-27-DsRed-mtDsRed; Table 1) led to the recovery of four fluorescent bands (a-d) with an increase in the amount of trimer (b) relative to the tetramer (a) on native gels, as well as the recovery of mtATPase dimers (c) and monomers (d) (Fig. 1A, lane 2). The change in the oligomeric association of mtATPase was accompanied by a decrease in the generation time of cells (Gavin et al., 2002) (see also Table 1). The oligomeric rearrangements arising from the expression of free DsRed do not occur during the detergent extraction of mitochondria, as free DsRed was shown to be lost with the soluble contents of the matrix during the procedure used to isolate mitochondrial membranes (Gavin et al., 2002). Furthermore, the addition of purified DsRed to γ-27-DsRed mitochondria during the extraction procedure did not alter the amount of tetramer or trimer recovered (not shown). These results strongly suggest that the formation of the mtATPase oligomers recovered must have occurred in vivo. Indeed, as DsRed fluorescence only matures on the association of multiple DsRed proteins (Sacchetti et al., 2002), the DsRed moiety associated with mtATPase must interact at some point with DsRed proteins from neighbouring complexes in order to be visible in whole cells (discussed further below).

Second dimension SDS-PAGE of oligomeric mtATPase bands (a and b) recovered from γ-27-DsRed (Fig. 1A, lane 1) showed the same profile of mtATPase subunits as detected in monomeric mtATPase band (d) from strain γ-27-GFP (Fig. 1A, lane 4), with no additional bands of significant intensity evident (Fig. 1C, compare lanes 1a and 4d). Thus, bands with decreased mobility on CN-PAGE (Fig. 1A, lanes 1 and 2) represent higher order associations of mtATPase complexes and not associations or aggregation of other mitochondrial proteins, such as components of the electron transport chain, with monomeric or dimeric mtATPase.

Solubilisation of mitochondrial membranes with digitonin allows the recovery of mtATPase dimers (Schägger, 2001). Lysates (4g digitonin/g protein) of γ-27-GFP mitochondria contain fluorescent mtATPase in a predominately dimeric state (Fig. 1A, lane 3) that shows ATPase activity when assayed in situ (Fig. 1B, lane 3). YRD15 mitochondrial lysates contain an enzymatically active mtATPase dimer that is not fluorescent (compare lane 5 of Fig. 1A with lane 5 of Fig. 1B). However, mtATPase could not be detected in digitonin extracts of γ-27-DsRed mitochondria prepared under similar conditions (not shown). Increasing the digitonin/protein ratio (10g/g) and decreasing the speed of centrifugation (from 100,000 g to 11,000 g) after solubilisation enabled the recovery of a weakly fluorescent band having a lower mobility than tetrameric mtATPase (Fig. 1A, lane 7). The relative recovery compared with other mtATPase bands (Fig. 1A, lanes 1 and 2) was poor. The in situ ATPase assay made visible a faint lead deposit at the position of this band (not shown), identifying it as containing mtATPase; however, this was of insufficient intensity to be captured by the imaging methods employed here. Nevertheless, second dimension SDS-PAGE followed by silver staining showed a similar profile of mtATPase subunits within this band (Fig. 1C, lane 7) to that present in mtATPase recovered from γ-27-DsRed and γ-27-GFP mitochondria (see Fig. 1C, lanes 1 a-b, 2a-d, 3c and 4d).

In summary, the above observations show that mtATPase exists in `cross-linked' oligomeric arrangements in γ-27-DsRed cells during periods of respiratory growth.

Yeast cells were sampled from cultures (grown on the nonfermentable carbon source ethanol) during mid-logarithmic growth phase and subjected to examination by fluorescence microscopy. During respiratory growth, mitochondria in yeast are typically arranged in a branched, tubular network distributed around the periphery of the cell (Egner et al., 2002; Hermann and Shaw, 1998). Fluorescence from free DsRed targeted to the mitochondrial matrix of YRD15 (Fig. 2A) and fluorescence detected in γ-27-GFP cells highlight this normal tubular network (Fig. 2B). However, in γ-27-DsRed cells the filamentous reticulum was fragmented and replaced by punctate ball and ring-like structures, often aggregated to one side of the cell (Fig. 2C,D). Branched tubular networks as seen in control cells were never observed in γ-27-DsRed cells. The mitochondrial morphology of γ-27-DsRed, γ-27-DsRed-mtDsRed and those of control strains YRD15 and γ-27-GFP are quantified in Table 2.

GFP not fused to another protein was targeted to the mitochondrial matrix of γ-27-DsRed cells (strain γ-27-DsRed-mtGFP) to establish whether the restricted DsRed fluorescence (Fig. 2C,D) represented gross changes in mitochondrial morphology or specific localisation of fluorescent mtATPase oligomers within the tubular network. GFP fluorescence in γ-27-DsRed-mtGFP cells (Fig. 2E) highlighted the same abnormal mitochondrial morphology as indicated by DsRed fluorescence (Fig. 2F), with no filamentous network visible, indicating cross-linked mtATPase tetramers responsible for alterations to the entire mitochondrial reticulum.

When viewed by bright-field or differential interference contrast microscopy, γ-27-DsRed cells appeared slightly larger than control cells with considerable heterogeneity in size among the population (Fig. 2C,D). Larger cells often looked `bloated', with an increase in the size and/or number of vacuoles visualised by CellTracker Blue CMAC (Molecular Probes) staining (not shown) and an accumulation of vesicle-like bodies (Fig. 2C,D). Some of these bloated cells were observed to burst during microscopic observation, releasing their fluorescent internal components (Fig. 3A-C). In addition, γ-27-DsRed cells displayed altered budding during cell division (Fig. 3A), with as many as seven buds observed attached to a single mother cell. Indeed, examination of the samples taken from cultures during mid-logarithmic growth show that 24.9% and 27.6% of cells from YRD15 and γ-27-GFP, respectively, were budding. Of these budding cells, none from YRD15 contained additional buds, while 4.2% of budding cells from γ-27-GFP contained one additional bud. By contrast, 23.9% of γ-27-DsRed cells observed were undergoing cell division, with 49.7% of these containing two or more buds (Fig. 3D). The accumulation of additional buds attached to mother cells during division presumably contributes to the longer generation time of γ-27-DsRed cells. A decrease in the percentage of dividing cells with multiple buds (35.2%) accompanied the decrease in generation time (Table 1) and mtATPase oligomerisation (Fig. 1A, lane 2) in γ-27-DsRed-mtDsRed cells (Fig. 3D). An increased percentage of petite (rho^–/rho⁰) cells was observed in γ-27-DsRed cultures, suggesting that mitochondrial DNA transmission to daughter cells during mitosis is impaired. For cells grown in a nonfermentable substrate (ethanol) the petite frequency of YRD15, γ-27-GFP and γ-27-DsRed cells was 8.3%, 23.4% and 42.4%, respectively.

We examined mitochondrial ultrastructure using transmission electron microscopy (TEM). TEM sections from control strains YRD15 and γ-27-GFP exhibited normal mitochondria with numerous cristae evident (Fig. 4A,B). However, TEM performed on sections of γ-27-DsRed cells did not show any normal mitochondria. Instead, large ring or long tubular structures were present, composed of multiple concentric layers of membrane (Fig. 4C-E), but having no cristae. These unusual structures are consistent with the punctate ball and larger ring-like mitochondria visualised in γ-DsRed cells using fluorescence microscopy (Fig. 2).

Immunoelectron microscopy positively identified these abnormal structures as mitochondria. A monoclonal antibody directed against the mtATPase α subunit was found to localise to the mitochondrial cristae of both YRD15 and γ-27-GFP cells (Fig. 5A,B), and the inner rings of the abnormal structures in γ-27-DsRed sections (Fig. 5C-E), identifying them as the inner mitochondrial membrane. Gold particles were absent from the outer membrane of YRD15 and γ-27-GFP, as well as the outermost membrane of γ-27-DsRed mitochondria. Thus, cross-linking of mtATPase complexes into oligomers prevented the formation and/or stabilisation of normal mitochondrial cristae. To our knowledge, a similar abnormal mitochondrial phenotype involving multiple layers of inner mitochondrial membrane, an absence of cristae and an enlarged `onion-like', appearance has only been described once before in yeast, for cells lacking expression of either mtATPase subunit e or g (Paumard et al., 2002b; Soubannier et al., 2002). We confirmed that TEM sections (Fig. 4G,H) of A20N cells (YRD15 lacking expression of subunit g) showed aberrant structures similar to those in γ-27-DsRed cells.

Given the similarity between the abnormal mitochondrial phenotypes displayed by cells expressing γ-27-DsRed and cells lacking expression of subunit e or g, mitochondrial membranes were examined for the presence of subunits e and g using immunoblot analysis. Both subunits were detected in mitochondrial membranes of γ-27-DsRed cells at levels equivalent to those found in mitochondria from strains YRD15 and γ-27-GFP (Fig. 6). Given that subunits e and g require interactions with other F_O subunits to be maintained within the membrane (Arnold et al., 1998; Arnold et al., 2002; Paumard et al., 2002b; Soubannier et al., 1999), we concluded that these two subunits must be associated with mtATPase oligomers in γ-27-DsRed mitochondria at normal levels, although it is unknown if their arrangement within mtATPase complexes is distorted. The alterations to mitochondrial morphology and the absence of cristae we observe in γ-27-DsRed cells must therefore directly result from the oligomeric arrangement of mtATPase complexes, and not the loss of one or both of subunits e and g. The corollary is that normal arrangement of mtATPase complexes within the inner membrane is required for the generation of mitochondrial cristae.

Free DsRed targeted to the mitochondrial matrix of γ-27-DsRed cells reduced the tetrameric associations of mtATPase within the membrane (Fig. 1A, compare lanes 1 and 2). TEM sections from γ-27-DsRed-mtDsRed cells showed a partial restoration of normal mitochondrial morphology, with the presence of cristae in some mitochondria being evident (Fig. 7); this was presumably due to the relative decrease of mtATPase tetramers within the inner mitochondrial membrane.

Subunit γ is a dynamic rotary component of mtATPase (Noji et al., 1997), yet is able to tolerate GFP attached to its C-terminus (Prescott et al., 2003) or being cross-linked to subunit α at its C-terminus (Gumbiowski et al., 2001) without significant effect on enzyme function. Nevertheless, loss of cristae may reflect the function and central location of the γ subunit within F₁. Therefore, subunit b, a peripheral subunit of the stator stalk, was tagged with DsRed to confirm that alterations to mitochondrial cristae resulted from the arrangement of mtATPase within the membrane and not specifically from interference with the function of subunit γ and the activity of the enzyme. Yeast strain b-25-DsRed expresses subunit b with DsRed fused to its C-terminus via a 25 amino acid linker in the absence of endogenous subunit b. These cells grow on the nonfermentable substrate ethanol with a generation time of 11.38±0.13 hours (Table 1), slower than YRD-15 (8.67±0.14 hours), but faster than strain γ-27-DsRed (15.77±0.50 hours). Dodecyl β-maltoside extracts of b-25-DsRed mitochondria exhibited a similar mtATPase profile following CN-PAGE (Fig. 8A) to that of γ-27-DsRed-mtDsRed (Fig. 1A, lane 2), with tetramer, trimer, dimer and monomer species recovered. TEM sections of b-25-DsRed cells showed mitochondria with aspects of both normal and abnormal mitochondrial morphology (Fig. 8B). GFP fused to the C-terminus of subunit b in identical fashion has no effect on generation time, mtATPase association or mitochondrial morphology (not shown). Thus, b-25-DsRed cells show that oligomerisation of mtATPase mediated by DsRed cross-linking is not specific to the γ subunit, and reinforces the contention that normal arrangements of mtATPase complexes within the inner membrane are required for the generation of mitochondrial cristae.

The key finding arising from the results presented here is that the correct arrangement of mtATPase complexes within the inner mitochondrial membrane is crucial for cristae formation in vivo. The novel DsRed cross-linking approach we have utilised led to the formation of mtATPase oligomers in respiring cells. Here, oligomerisation of mtATPase was mediated through subunits not normally associated with mtATPase dimerisation. We have shown the presence in cells of mtATPase oligomers formed as a result of DsRed cross-linking, although the precise pathway of formation of these cross-linked oligomers remains unclear. It may be that mtATPase complexes are normally present in an oligomeric arrangement within the inner membrane, with individual mtATPase complexes being close enough for association between DsRed moieties, thereby `trapping' the tetrameric assemblies subsequently recovered from mitochondrial lysates. Alternately, much more transient associations of mtATPase complexes within the membrane may be `captured' by virtue of the interaction between DsRed moieties when complexes come into sufficiently close proximity so as to eventually build tetramers. The high efficiency of mtATPase oligomerisation seen in γ-27-DsRed mitochondrial lysates and the relative absence of monomers and dimers suggests that stable, structured associations between mtATPase complexes exist in vivo during respiratory growth. This is the first evidence for the association of mtATPase complexes within mitochondrial membranes of intact yeast cells.

Aberrant mitochondrial morphology has been described in cells lacking either of two nonessential mtATPase subunits, e and g, shown to be required for the recovery of dimers from detergent lysates of mitochondria (Paumard et al., 2002b). An apparent mtATPase oligomer was recovered using native gels from mitochondrial lysates prepared using low digitonin/protein ratios (<1g/g). This oligomer could not be recovered in cells lacking subunit e or g. Moreover, the formation of disulfide bonds between two b subunits from separate mtATPase complexes could be shown in mitochondrial membranes prepared from such cells, suggesting the presence of dimers formed through a b-b subunit interface. A model integrating these observations was presented whereby mtATPase dimers formed via a subunit e/g interface are able to associate via a second interface involving subunit b, and are thereby incorporated into a larger oligomeric network of mtATPase complexes. From these observations, it was proposed that oligomeric mtATPase arrays were responsible for cristae formation (Paumard et al., 2002b). It was acknowledged that a relationship between subunit e and proteins within the intermembrane space known to directly affect mitochondrial morphology, such as Mgm1p (Shepard and Yaffe, 1999; Wong et al., 2003), could not be excluded.

In the present study digitonin/protein ratios normally sufficient to extract mtATPase dimers proved ineffective in releasing mtATPase complexes containing γ-27-DsRed fusions from the mitochondrial membranes. Why this should be the case is unknown. It may be that the formation of DsRed cross-linked mtATPase oligomers affects digitonin access to the membrane. Alternatively, the oligomeric networks of mtATPase stabilised by the DsRed cross-linker may become large enough to be pelleted with the insoluble membrane fraction during extraction. Increasing the digitonin concentration and decreasing the speed of centrifugation allowed the recovery of small amounts of a presumptive mtATPase oligomer larger than that of the tetramer, suggesting that larger mtATPase arrays may form in mitochondrial membranes. The exact nature and stoichiometry of mtATPase complexes within this assembly remains to be defined.

How do the associations of multiple mtATPase complexes lead to cristae formation? Oligomeric networks of proton pumps were proposed by Allen and colleagues (Allen, 1995; Allen et al., 1989) to be responsible for the formation of tubular cristae (F₁F_O-ATPase) and microtubules of the contractile vacuole complex (V₁V_O-ATPase) in Parameciummultimicronucleatum. Allen (Allen, 1995) proposed that association of proton pumps could lead to the formation of a rigid arc that protruded from the planar surface of the membrane (Fig. 9A), such that extension of the enzyme chain would impart sufficient deformation to the membrane to result in its tubulation. Paumard et al. (Paumard et al., 2002) hypothesised that disruption of oligomeric mtATPase networks in yeast impairs folding of the inner mitochondrial membrane into cristae. Here, we have promoted the formation of, or stabilised, mtATPase oligomers by DsRed cross-linking and in doing so produced essentially the same phenotype. Why should this be the case? One possibility is that the DsRed moieties anchoring mtATPase tetramers may prevent the component complexes from achieving the distortion necessary to curve out of the plane of the lipid bilayer, thereby favouring sheets of membrane lacking cristae-like deformations (Fig. 9B). The tetramer itself may be sufficiently distorted that it prevents assembly in a larger oligomeric network with other tetramers. It could be argued that cross-linking any large membrane-bound complexes would lead to abnormal deformation of the membrane and that this would not necessarily implicate such complexes in having a normal role in cristae formation. The fact that cells lacking expression of subunit e or g (Paumard et al., 2002b) exhibit a similar `loss of cristae' phenotype in the absence of DsRed cross-linking argues strongly that mtATPase tetramers affect mitochondrial morphology by altering the normal associations of mtATPase.

If mtATPase oligomerisation normally affects membrane curvature, then cross-linking mtATPase complexes via DsRed must alter the dynamics of the mitochondrial membranes. Mitochondria in S. cerevisiae normally form a branched tubular reticulum distributed throughout the periphery of the cytoplasm (Egner et al., 2002; Hoffmann and Avers, 1973). The morphology and distribution of this network is determined by the energy requirements of the cell, and is maintained by opposing fission and fusion events (Karbowski and Youle, 2003; Shaw and Nunnari, 2002). The inner mitochondrial membrane is also a dynamic structure that alters conformation rapidly in response to changes in the cellular environment and growth requirements (Frey and Mannella, 2000). Large transitions in the shape of cristae have been observed in concert with fission and fusion events occurring within the inner mitochondrial membrane itself (Mannella et al., 2001). If mtATPase oligomerisation participates in the organisation of the inner mitochondrial membrane, then dynamic changes in membrane shape presumably must also involve the rearrangement/disassembly of existing oligomeric networks. Cross-linking mtATPase complexes in γ-27-DsRed cells may prevent the network from being collapsed or reorganised, thereby impairing changes to the inner mitochondrial membrane in response to altered cellular conditions. Indeed, regions of mitochondria were seen in γ-27-DsRed cells that lacked GFP fluorescence, suggesting that portions of mitochondria were excluded from mixing matrix contents with the rest of the mitochondrial network, possibly reflecting problems with mitochondrial fusion.

The loss of mitochondrial structure and/or its ability to reorganise is presumably reflected by impaired bud separation from cells and the increased frequency of petites in cell cultures. Although little published data is available on the subject, the presence of multiple buds attached to mother cells has been attributed to delays in mitochondrial inheritance (Roeder et al., 1998). It is unlikely that cross-linking of mtATPase plays a direct role in this process, but the loss of normal mitochondrial structure must influence the distribution of mitochondria to budding cells or the separation of buds from the mother cell. Mitochondrial DNA transmission to daughter cells during mitosis was impaired as supported by the increased percentage of petite (rho^–/rho⁰) cells in γ-27-DsRed cultures. Presumably the loss of normal mitochondrial structure affects mtDNA attachment to membranes and its distribution within remaining structures and/or the function of the mitochondrial DNA replication machinery that is membrane associated (Berger and Yaffe, 2000; Hanekamp et al., 2002).

The finding that mtATPase organisation within the membrane affects the formation of cristae is timely, given the revisionist view of mitochondrial ultrastructure emerging, courtesy of electron microscope tomography studies (Frey and Mannella, 2000; Mannella et al., 2001). Far from being the open baffle structures depicted in many textbooks, cristae are seen to be pleomorphic with an extensive tubular nature. The morphology of mitochondria and the shape of the inner mitochondrial membrane are known to reflect the energy requirements of the cell (Hermann and Shaw, 1998; Mannella et al., 2001). Recent work in bovine mitochondria provides further evidence that mitochondrial cristae comprise a regulated submitochondrial compartment specialised for cellular energy production (Gilkerson et al., 2003). It is intriguing to think that the enzyme responsible for satisfying much of the cell's energy demands takes an active, rather than passive, role in the modelling of the membrane in which it functions. As such, mtATPase may help regulate the morphology of the inner membrane in order to optimise its own performance as the primary producer of ATP. Indeed broader questions relating to the importance of cristae remodelling in relation to mitochondrial function and disease, and involvement in the induction of apoptosis (Karbowski and Youle, 2003; Olichon et al., 2002; Scorrano et al., 2002) are only now being addressed.

We thank Gunta Jaudzems from Monash Imaging facility for her work with electron microscopy and sample preparation. This work is funded in part by the Australian Research Council.