The interorganellar interaction between distinct human mitochondria with deletion mutant mtDNA from a patient with mitochondrial disease and with HeLa mtDNA.

For the examination of possible intermitochondrial interaction of human mitochondria from different cells, cybrids were constructed by introducing HeLa mitochondria into cells with respiration-deficient (ρ−) mitochondria. Respiration deficiency was due to the predominance of mutant mtDNA with a 5,196-base pair deletion including five tRNA genes (ΔmtDNA5196). The HeLa mtDNA and ΔmtDNA5196 encoded chloramphenicol-resistant (CAPr) and chloramphenicol-sensitive (CAPs) 16 S rRNA, respectively. The first evidence for the interaction was that polypeptides exclusively encoded by ΔmtDNA5196 were translated on the introduction of HeLa mitochondria, suggesting supplementation of the missing tRNAs by ρ− mitochondria from HeLa mitochondria. Second, the exchange of mitochondrial rRNAs was observed; even in the presence of CAP, CAPs ΔmtDNA5196-specific polypeptides as well as those encoded by CAPr HeLa mtDNA were translated in the cybrids. These phenomena can be explained assuming that the translation in ρ− mitochondria was restored by tRNAs and CAPr 16 S rRNA supplied from HeLa mitochondria, unambiguously indicating interorganellar interaction. These observations introduce a new concept of the dynamics of the mitochondrial genetic system and help in understanding the relationship among mtDNA mutations and expression of human mitochondrial diseases and aging.

Mammalian cells have been proposed to possess hundreds of independent mitochondria, each containing several mitochondrial DNA (mtDNA) molecules (1)(2)(3). Variation in the number and morphology of mitochondria in different cell types (4) and even in the same cell type (5) has been suggested. The interaction and fusion between mitochondria in yeast and plant cells have received support from morphological findings (6) and from the molecular and genetic evidence of mtDNA recombination (7,8). In mammalian species, however, the coexistence of mitochondria and mtDNA from different individuals is completely inhibited by their strictly maternal inheritance (9, 10); spermderived mtDNA was selectively and completely eliminated from fertilized mouse eggs, suggesting that the mtDNA popu-lation of mammalian individuals is derived exclusively from eggs (10) and is extremely homoplasmic throughout an individual (11). Therefore, it seems reasonable to suppose that mammalian mtDNAs consequently have lost the ability to recombine with each other. In fact, we previously showed the absence of extensive mtDNA recombination in mammalian cells, even when polymorphic mtDNA molecules from different individuals and from different species were mixed within a cell by the fusion of somatic cells (12).
On the other hand, metabolic cooperation of the mammalian mitochondrial genetic system was suggested by the translational complementation of mitochondrial rRNA observed in heteroplasmic cells with chloramphenicol-sensitive (CAP s ) 1 and -resistant (CAP r ) mtDNA (13). Our previous studies provided biochemical evidence for intermitochondrial interactions through translational complementation and competition of mitochondrial tRNAs using cybrids with mutant and wild type mtDNA from patients with mitochondrial diseases (14,15) and also provided molecular evidence for the rapid penetration of mtDNA and/or its products into mitochondria of mtDNA-less ( 0 ) HeLa cells (16). Based on these observations, we proposed that the mitochondria and mitochondrial genetic system in living mammalian cells have lost their individuality and thus function as a single dynamic cellular unit (16).
Recently, the absence of intermitochondrial interaction was proposed, particularly when mitochondria containing different pathogenic mutant mtDNAs derived from different patients coexisted in single cells (17). One explanation for the apparent discrepancy between the observations of the presence (13)(14)(15)(16) and absence (17) of mitochondrial interactions might be that the occurrence of interactions of mutant and wild type mtDNA molecules could be limited only when they coexist in the mitochondria from the time of the mutation event, but not when once they have been separated into different mitochondria by random segregation, or originate in distinct mitochondria of different cells (17). If this is so, intermitochondrial cooperation followed by their fusion would not occur in mammalian cells.
In this study, using enucleation and cell fusion techniques, we created an environment suitable for determining whether intermitochondrial cooperation occurs even between distinct organelles derived from different cells. First, we isolated respiration-deficient Ϫ cells containing more than 95% of diseaserelated mutant mtDNA with a 5,196-base pair deletion (⌬mtDNA 5196 ), which have no mitochondrial translation activity because of the deletion of five tRNA genes. Then, mitochondria of CAP r HeLa cells were introduced into the Ϫ cells, and the translation activity of the mitochondria was examined. The results showed that even in the presence of CAP, the fusion proteins exclusively encoded by ⌬mtDNA 5196 were translated with the help of the tRNAs and CAP r rRNA supplied from the imported HeLa mitochondria, suggesting unambiguously the occurrence of intermitochondrial cooperation between distinct organelles of Ϫ mitochondria and HeLa mitochondria.

MATERIALS AND METHODS
Cells and Cell Culture-The human skin fibroblast line 197 with 21% ⌬mtDNA 5196 was isolated from a patient with chronic progressive external ophthalmoplegia, a subgroup of mitochondrial encephalomyopathies. The 0 HeLa cells and CAP r HeLa cells used are resistant to 20 M 6-thioguanine and 2 mM ouabain. These cell lines were grown in normal medium (RPMI 1640, 0.1 mg/ml pyruvate, 50 g/ml uridine, 10% fetal bovine serum).
Intercellular Transfer of the HeLa Nuclear Genome-Intercellular transfer of HeLa nuclei to 197 fibroblasts was achieved by fusion of the fibroblasts with 0 HeLa cells using polyethylene glycol 1500 as described previously (18) with slight modifications. Briefly, the fusion mixtures were cultivated in selective medium (RPMI 1640, 0.1 mg/ml pyruvate, 50 g/ml uridine, 10% fetal bovine serum, hypoxanthine/aminopterin/thymidine (HAT; Sigma), 2 mM ouabain), and on day 10 after fusion colonies grown in the selection medium were cloned by the cylinder method. Nuclear hybrids were then cultivated in normal medium.
Intercellular Transfer of the HeLa Mitochondrial Genome-Intercellular transfer of HeLa mitochondria was carried out as described previously (18) by fusion of enucleated CAP r HeLa cells with the nuclear hybrid clone H5 containing only ⌬mtDNA 5196 , and cybrid clones were isolated in selection medium (DM170 ϩ HAT). Briefly, CAP r HeLa cells grown on round glass discs were enucleated by centrifugation (23,000 ϫ g, at 34°C for 10 min) in the presence of cytochalasin B (Sigma; 10 g/ml). The resulting cytoplasts were mixed with H5 cells, and fusion was carried out in the presence of 50% (w/v) polyethylene glycol 1500 (Boehringer Mannheim). The fusion mixture was cultivated in selection medium without glucose, pyruvate, and uridine (DM170, Kyokuto Kagaku, Tokyo) supplemented with HAT (9). On day 14 -20 after fusion, cybrid clones grown in selective medium were harvested and cloned by the cylinder method. Cybrids were cultivated in DM170 medium without HAT.
Southern Blot Analyses of mtDNA-For determination of the content of ⌬mtDNA 5196 , total DNA (2 g) extracted from 2 ϫ 10 5 cells was digested with a single cut restriction enzyme, PvuII, and the restriction fragments were separated by 0.6% agarose gel electrophoresis. After blotting onto a Nytran membrane, the DNA fragments were hybridized with [␣-32 P]dATP-labeled HeLa mtDNA. The contents of HeLa mtDNA in cybrid clones were determined from the HeLa mtDNA-specific polymorphisms of HaeIII restriction patterns, which can distinguish HeLa mtDNA from other human mtDNAs (19). The total DNA (2 g) extracted from 2 ϫ 10 5 cells was digested with HaeIII, and the fragments were separated by 2% agarose gel electrophoresis. After blotting onto a Nytran membrane, the DNA fragments were hybridized with [␣-32 P]dATP-labeled polymerase chain reaction products amplified using two primers (15,756 -15,775 and 15,916 -15,897). The 1,284-base pair fragment was cleaved into two fragments of 711 and 573 base pairs due to one HaeIII site gain in HeLa mtDNA (19). The membrane was washed and exposed to an imaging plate (Fuji Photo Film, Tokyo), and the radioactivity of the fragments was measured with a BAS2000 instrument (Fuji Photo Film).
Analysis of Mitochondrial Translation Products-Mitochondrial translation products were labeled with [ 35 S]methionine as described previously (18) with slight modifications. Briefly, semiconfluent 2 ϫ 10 6 cells in a dish were incubated in methionine-free medium containing 2% bovine serum for 30 min at 37°C. Then, the cells were labeled with [ 35 S]methionine for 2 h in the presence of 0.2 mg/ml emetine. The mitochondrial fraction was obtained by homogenization in 0.25 M sucrose, 1 mM EGTA, and 10 mM HEPES-NaOH, pH 7.4, followed by differential centrifugation. Proteins in the mitochondrial fraction (30 mg) were separated by SDS-polyacrylamide gradient gel (12-20%) electrophoresis. The dried gel was exposed to an imaging plate (Fuji Photo Film) for 6 h, and the labeled polypeptides were located with a bioimaging analyzer, Fujix BAS 2000 (Fuji Photo Film).
Cytochrome c Oxidase (COX) Electron Micrographs-Cells grown on coverslips were fixed in 2% glutaraldehyde, 0.05 M phosphate buffer, pH 7.4, for 15 min and stained for COX by the procedure of Seligman et al. (20) with slight modifications (21) to detect the COX activities of individual mitochondria at the ultrastructural level. The cells were then postfixed in OsO 4 for 15 min and embedded in epoxy resin.

Isolation of Ϫ Nuclear Hybrid Clones Containing Predomi-
nantly ⌬mtDNA 5196 -To determine whether there is intermitochondrial cooperation between distinct organelles derived from different cells, we first isolated Ϫ nuclear hybrid cells, which show no mitochondrial translation or oxidative phosphorylation activities due to predominance of patient-derived ⌬mtDNA 5196 with a deletion of five tRNA genes (14). Previously, cybrid clones were isolated by fusion of 0 HeLa cells with enucleated fibroblasts from a chronic progressive external ophthalmoplegia patient with ⌬mtDNA 5196 in selective medium without pyruvate and uridine to remove 0 HeLa cells (14). In this selective medium, however, Ϫ cybrids containing ⌬mtDNA 5196 predominantly must also be removed because of their overall respiration deficiency. Therefore, we tried to isolate Ϫ nuclear hybrid clones by fusing 0 HeLa cells with the fibroblasts followed by HAT selection, which does not have any selective pressure upon growth of Ϫ cells but can remove 0 HeLa cells.
No mitochondrial translation was observed in H5 cells with only ⌬mtDNA 5196 , suggesting that they are Ϫ cells (Fig. 1). On the other hand, quantitative estimation of mtDNA translation products showed that the overall translation capacity of H8 cells with 59% ⌬mtDNA 5196 was comparable to that of H9 cells without ⌬mtDNA 5196 . Moreover, because of the 5,196-base pair long deletion with a breakpoint between 8,563 and 13,788, ⌬mtDNA 5196 newly acquired a unique fusion gene that encoded two fusion proteins, ATP8/ND5 (FA) and ATP6/ND5 (FB) (14), which were exclusively translated in H8 cells (Fig. 1). These features indicate that polypeptides encoded by ⌬mtDNA 5196 were translated using the tRNAs supplied from the endogenous wild type mtDNA of the patient, suggesting intramitochondrial cooperation. Oua ϩ HAT 0 H4 Oua ϩ HAT 0 H5 Oua ϩ HAT Ͼ95 H6 Oua ϩ HAT 0 H7 Oua ϩ HAT 0 H8 Oua ϩ HAT 41 H9 Oua ϩ HAT 0 H10 Oua ϩ HAT 72 H11 Oua ϩ HAT 0 H12 Oua ϩ HAT 79 a Fibroblast line derived from a chronic progressive external ophthalmoplegia patient.
b No mtDNA molecules were detected even by polymerase chain reaction techniques.

Analysis of Intermitochondrial Cooperation between Distinct
Organelles Derived from Different Cells-For determination of whether intermitochondrial cooperation occurs even between distinct organelles from different cell types, the chance of interaction between Ϫ mitochondria from H5 cells and ϩ mitochondria from HeLa cells was created by fusion of enucleated CAP r HeLa cells with H5 cells. All cybrid clones isolated in selective HAT medium without pyruvate and uridine contained both ⌬mtDNA 5196 and HeLa mtDNA (Table II).
In these cybrids, the fusion proteins (FA and FB) can be used as specific markers of the translation of ⌬mtDNA 5196 . On the other hand, since ND3 of HeLa cells, ND3Ј, migrated slightly faster in SDS-polyacrylamide gel than the corresponding ND3 of other human cells (13,17,22) because of a single amino acid exchange (Asp to Asn) (23), ND3Ј could be used as a specific marker of the translation of HeLa mtDNA.
First, as a negative control of intermitochondrial interaction, we examined mitochondrial translation in a simple mixture of H5 and HeLa cells with 53% ⌬mtDNA 5196 . In this case, as Ϫ mitochondria in H5 cells and ϩ mitochondria in HeLa cells were completely separated and confined to each type of cell, cooperation of the Ϫ and ϩ mitochondria could not occur, and thus their translation should be limited to mitochondria in HeLa cells. As expected, the results showed that ND3Ј but not fusion proteins was translated and that the total amount of [ 35 S]methionine incorporation corresponded exactly to the amount of HeLa mtDNA (Figs. 2 and 3).
Then mitochondrial translation activity was examined by [ 35 S]methionine labeling using a cybrid clone CH5-2 with 53% ⌬mtDNA 5196 ( Fig. 2A), in which the Ϫ and ϩ mitochondria coexisted by cell fusion. Figs. 2B and 3 show that the total amount of [ 35 S]methionine incorporation into mtDNA-encoded polypeptides in CH5-2 cells was comparable to that in HeLa cells. Furthermore, the fusion proteins FA and FB exclusively encoded on ⌬mtDNA 5196 were observed in the CH5-2 cells (Fig.  2B). Thus, in contrast to the case in the simple mixture, ⌬mtDNA 5196 -encoded polypeptides were translated using the tRNAs supplied from the imported HeLa mitochondria in CH5-2 cells, showing unambiguously the occurrence of intermitochondrial cooperation between distinct organelles of the Ϫ and ϩ mitochondria derived from H5 cells and HeLa cells, respectively.
Effects of CAP on the Mitochondrial Translation Activity in CH5-2 Cells-By use of CH5-2 cells, the occurrence of intermitochondrial cooperation could be confirmed by examining the effects of CAP on the translations of both CAP s ⌬mtDNA 5196 and CAP r HeLa mtDNA. If intermitochondrial cooperation oc-curred, the fusion proteins encoded by CAP s ⌬mtDNA 5196 should be translated by the use of CAP r 16 S rRNA transcribed from CAP r HeLa mtDNA.
First, we carried out a negative control experiment by testing the effect of CAP on translation of mitochondria in a simple mixture of 53% H5 and 47% HeLa cells, in which no intermitochondrial interaction occurred. In the presence of CAP, translation in CAP r HeLa mitochondria was slightly inhibited (Fig.  2B). The amount of mitochondrial translation in the mixture of H5 and HeLa cells corresponded exactly to that expected from the proportion of HeLa cells in the mixture (Fig. 2B), and no fusion proteins were translated (Fig. 2B), reflecting the absence of cooperation.
If there is no interaction between mitochondria from different cells, the mitochondrial translation profiles in CH5-2 cells with 47% HeLa mtDNA must correspond to those in the simple parental cell mixture with 47% HeLa mtDNA. However, the translation of the fusion proteins was observed in CH5-2 cells even in the presence of CAP (Fig. 2B). Furthermore, the amount of translation of ND3Ј exclusively encoded by CAP r HeLa mtDNA, as well as those of CAP s ⌬mtDNA 5196 -specific fusion proteins, was reduced simultaneously to much lower levels than those in the simple parental cell mixture with 47% HeLa mtDNA (Figs. 2B and 3).
The translation of a small amount of fusion proteins in the presence of CAP represents the occurrence of intermitochondrial cooperation, since CAP r 16 S rRNA transcribed from the imported CAP r HeLa mtDNA should be necessary for translation of the fusion protein in the presence of CAP, suggesting rRNA complementation. The progressive inhibition of the translation of CAP r HeLa mtDNA-encoded polypeptides including ND3Ј in CH5-2 cells by CAP might be explained by  supposing that the CAP s allele is dominant over CAP r in mammalian cells. This phenomenon could not occur when mitochondria with CAP s ⌬mtDNA 5196 and with CAP r HeLa mtDNA did not cooperate with each other as in the simple mixture of mitochondria prepared from the cells (Figs. 2 and 3).

Analysis of COX Activity in Individual Mitochondria in Single CH5-2 Cells
Using COX Electron Microscopy-The observed complementation of mitochondrial tRNA and rRNA excludes the possibility that ⌬mtDNA 5196 and HeLa mtDNA remained separated in respective mitochondria. Accordingly, the ⌬mtDNA 5196 and HeLa mtDNA in cells of the cybrid clone CH5-2 should mix homogeneously throughout the mitochondria. This prediction can be tested by electron microscopic analysis of COX activity in mitochondria using CH5-2 cells with 53% ⌬mtDNA 5196 , since this technique clearly identifies COX activity of individual mitochondria in single cells.
No individual mitochondria with predominantly ⌬mtDNA 5196 in H5 cells showed COX activity, whereas most mitochondria in HeLa cells showed clear COX activity (Fig. 4). Therefore, the COX activity of individual mitochondria could be used as a probe to identify mitochondria with ⌬mtDNA 5196 and with HeLa mtDNA. If they do not interact with each other and remain unfused in CH5-2 cells after the import of HeLa mitochondria, they can be distinguished unambiguously by COX electron microscopy, and the ratio of COX-negative to COXpositive mitochondria in a cell should be proportional to the ratio of ⌬mtDNA 5196 to HeLa mtDNA (53:47).
COX activity of individual mitochondria in single CH5-2 cells was analyzed immediately after clonal isolation to minimize intercellular variations in the ⌬mtDNA 5196 contents of different cells in the clone. Fig. 4 shows that most mitochondria in CH5-2 cells were COX-positive, suggesting that mitochondria from H5 cells and from HeLa cells did not remain segregated, but interacted with each other in the CH5-2 cells. This uniform distribution of COX-positive mitochondria within single CH5-2 cells containing 53% ⌬mtDNA 5196 can be explained by rapid diffusion of ⌬mtDNA 5196 , HeLa mtDNA, and their products throughout the mitochondria.
Accordingly, all observations in this study support the idea that intermitochondrial interaction occurs even between distinct organelles with ⌬mtDNA 5196 from a patient with mitochondrial disease and wild type mtDNA of HeLa cells. DISCUSSION In this study, to provide direct evidence for the presence of interaction between distinct mitochondria originating in different cells, we created using cell fusion technique the chance of their interaction by isolating CH5-2 cells containing both Ϫ Closed and open arrowheads denote ND3Ј and fusion proteins, which are exclusively encoded by HeLa mtDNA and ⌬mtDNA 5196 , respectively. Note that HeLa mtDNA-specific ND3Ј could be distinguished from ND3 only when SDS-polyacrylamide electrophoresis was carried out in the absence of urea. On the other hand, two fusion proteins (⌬FA and ⌬FB) could be separated only in the presence of urea (14). In this study, as SDS-polyacrylamide gel was carried out in the absence of urea to distinguish ND3Ј from ND3, fusion proteins comigrated as indicated by the open arrowheads. First, we demonstrated complementation of mitochondrial tRNAs between mitochondria originating from different cells. Polypeptides encoded by ⌬mtDNA 5196 , which has a deletion of five tRNA genes, were translated when exogenous normal mitochondria imported from HeLa cells coexisted within CH5-2 cells (Fig. 2). This suggests the mitochondrial fusion and subsequent diffusion of the tRNAs transcribed from the imported HeLa mitochondria into the host Ϫ mitochondria with ⌬mtDNA 5196 .
Second, we observed complementation of mitochondrial rR-NAs. Even in the presence of CAP, fusion proteins exclusively encoded by CAP s ⌬mtDNA 5196 were translated in the CH5-2 cells. Furthermore, the mitochondrial translations of all polypeptides including the CAP r HeLa mtDNA-specific polypeptide ND3Ј in CH5-2 cells were reduced simultaneously to much lower levels than those in a simple mixture of the parental cells with an equivalent amount of HeLa mtDNA (47%) (Figs. 2B and 3). These results could be explained by supposing that intermitochondrial interaction, i.e. complementation of mitochondrial 16 S rRNAs, occurs between mitochondria with CAP s and CAP r mtDNA derived from different cells and that the CAP s allele is dominant over the CAP r allele.
The third line of evidence for intermitochondrial cooperation was obtained by COX electron microscopy. The observed mitochondrial tRNA and rRNA complementation could not occur if ⌬mtDNA 5196 and HeLa mtDNA remained confined to their respective mitochondria after the introduction of HeLa mitochondria. This interpretation predicts that coexisting ⌬mtDNA 5196 and HeLa mtDNA in single CH5-2 cells should mix homogeneously throughout the mitochondria. This prediction was examined by electron microscopic analysis of COX activity in individual mitochondria within CH5-2 cells containing 53% ⌬mtDNA 5196 , and the results demonstrated that all of the individual mitochondria showed COX activity (Fig. 4).
These observations consistently suggest the coexistence and cooperation of ⌬mtDNA 5196 and HeLa mtDNA within the same organelles, supporting the occurrence of interaction between distinct organelles. Such intermitochondrial interaction could not occur if mitochondria did not fuse and exchange mtDNA and its transcripts between fused mitochondria. Therefore, the presence of intermitochondrial interaction provides evidence not only for the mitochondrial fusion but also for subsequent mixing of their contents and supports the idea we reported previously that mitochondria and mitochondrial genome function as a single dynamic cellular unit in living human cells (16). Moreover, this concept will help in understanding the relationship between mtDNA mutations and expression of human mitochondrial diseases and aging. For example, although ageassociated accumulations of various somatic deletion mutant mtDNAs and age-associated mitochondrial dysfunction were observed in mammalian tissues (3,18,24), our results show that mtDNA mutations can complement each other, and thus they do not have a serious influence on age-associated mitochondrial dysfunction.
With respect to the dominant behavior of the CAP s allele in CH5-2 cells, however, it has been proposed that the CAP s allele could not be dominant for the following reasons (13,17). First, it is reported that in bacteria CAP s ribosomes were able to move along mRNA without forming polypeptides in the presence of CAP (25). However, this simple observation does not exclude the possibility of the dominant behavior of the CAP s allele. The second reason is that both CAP s and CAP r mtDNAencoded polypeptides can be translated even in the presence of CAP (13,17). However, the amount of the translation of CAP r mtDNA-encoded polypeptide ND3Ј in the presence of CAP was reduced progressively to a much lower level than that expected from the amount of CAP r mtDNA in the cells (Fig. 2), suggesting the presence of intermitochondrial cooperation in which the CAP s allele behaves dominantly.
If the CAP s phenotype is dominant, the question arises of why a small amount of CAP s mtDNA-encoded polypeptides, fusion proteins, was translated in the presence of CAP, as observed in Fig. 2B. This question can be explained by supposing that CAP s rRNA and CAP r rRNA are distributed randomly throughout the mitochondria of cells by intermitochondrial interaction. In this case, most polysomes in CH5-2 cells would have more than one CAP s rRNA, and in these translation would be inhibited by CAP. On the other hand, a small proportion of polysomes happened to contain no CAP r rRNA, and in these both CAP s and CAP r mtDNA-encoded polypeptides were translated proportionally in the presence of CAP. Accordingly, all of the results in this study are consistent with the occurrence of intermitochondrial cooperation between organelles derived from different cells.
Recently, the absence of intermitochondrial interaction on mixing mitochondria derived from different cells by the cell fusion techniques was reported (17); interactions of mutant and wild type mtDNA molecules occur only when they have been coexisting in the mitochondria from the time of the mutation event, but not when they originate in distinct organelles in different cells. However, these possibilities are unlikely, since originally coexisting mutant and wild type mtDNA within a mitochondrion would eventually segregate stochastically during culture of the cells, if exchange of mtDNA between mitochondria did not occur (26) as in the case between cells. It is well known that coexisting wild type and mutant mtDNA (27)(28)(29) and coexisting mtDNAs of different species (30) in a cell segregate in a stochastic way. Furthermore, this study provided direct evidence for the presence of cooperation even between organelles originating from different cells. However, it is still possible that cooperation does not occur between respiration-deficient mitochondria (17), possibly because of a lack of a sufficient energy supply for mitochondrial fusion and mixing of their contents. We are now investigating this possibility by isolating cells with syn Ϫ mitochondria, which are respirationdeficient due to point mutations of the tRNA Leu(UUR) or tRNA Ile gene (15), and then creating a chance for interaction between syn Ϫ mitochondria or syn Ϫ and Ϫ mitochondria.