Use of the NADH-Quinone Oxidoreductase ( NDI1 ) Gene of Saccharomyces cerevisiae as a Possible Cure for Complex I Defects in Human Cells*

The Ndi1 enzyme of Saccharomyces cerevisiae is a single subunit rotenone-insensitive NADH-quinone oxidoreductase that is located on the matrix side of the inner mitochondrial membrane. We have shown previously that the NDI1 gene can be functionally expressed in Chinese hamster cells (Seo, B. B., Kitajima-Ihara, T., Chan, E. K., Scheffler, I. E., Matsuno-Yagi, A., and Yagi, T. (1998) Proc. Natl. Acad. Sci. U. S. A. 95, 9167–9171) and human embryonal kidney 293 (HEK 293) cells (Seo, B. B., Matsuno-Yagi, A., and Yagi, T. (1999) Biochim. Biochem. Acta 1412, 56–65) and that the Ndi1 protein is capable of compensating respiratory deficiencies caused by defects in the host NADH-quinone oxidoreductase (com-plex I). To extend the potential use of this enzyme to repair complex I deficiencies in vivo , we constructed a recombinant adeno-associated virus vector carrying the NDI1 gene (rAAV-NDI1). With rAAV-NDI1 as

Mitochondrial proton-translocating NADH-quinone (Q) 1 ox-idoreductase (complex I) of bovine heart is composed of at least 43 distinct subunits (1,2). Of these subunits, seven are encoded by mtDNA. Those are designated ND1, -2, -3, -4, -4L, -5, and -6 (3,4). It has been shown that defects of complex I are involved in many human mitochondrial diseases (5,6). Unfortunately, the cause of complex I dysfunction, namely mutations in the seven mtDNA-encoded subunits of complex I, are difficult to correct (7). Repair of mutations in subunit genes coded for by nDNA is also challenging, and success has been limited (8). Nevertheless, a number of potential therapies can be envisioned (9). Thus, for example, the intramitochondrial NAD/ NADH ratio could be adjusted by oxidizing NADH with a number of membrane-permeable dyes. When attempted in vitro, however, the reduced forms of these dyes undergo oxidation by molecular oxygen, producing reactive oxygen species (ROS), which are toxic. The other potential alternative that appears to be promising is to introduce the yeast-type single subunit NADH-Q oxidoreductase (Ndi1) into complex I-deficient mammalian mitochondria. In addition to its potential to correct the symptoms associated with complex I dysfunction, the use of yeast Ndi1 has an additional advantage since electron transfer with mammalian complex I is believed to be a one-electron reaction that involves FMN and a number of ironsulfur clusters (10,11); the Ndi1 bears FAD as cofactors and appears to be a two-electron reaction enzyme (12). Thus, the mammalian complex I generates ubiquinone radicals as intermediate products that have been demonstrated to react with oxygen to produce ROS that are considered to be one of the causes of aging and cell death (13). Introduction of the twoelectron enzyme, Ndi1, should eliminate the complications associated with free radicals generated by the mammalian complex I. Therefore, our goal is to develop the incorporation of the yeast Ndi1 enzyme into human mitochondria as a potential remedy for complex I defects. As described previously (14), the yeast mitochondrial protein Ndi1, which is encoded by nDNA, is a versatile enzyme because the Ndi1 enzyme expressed in Escherichia coli functions as a member of the respiratory chain in the prokaryotic host cells. Furthermore, we have shown that Ndi1 can be functionally expressed in the Chinese hamster mutant cell line (CCL16B2) lacking the MWFE subunit of complex I (8,15) as well as in the human embryonal kidney 293 (HEK 293) cells (16).
However, the transfection procedures employed (calcium phosphate precipitation and lipofection methods using the pHook-2 vector) will have limitations when applied in vivo. First, transfection is either very difficult or practically impossible to achieve in many of human cell lines. Second, mitochon-drial defects often occur in nonproliferating or slowly proliferating tissues. Therefore, successful therapy will require that NDI1 can be expressed in differentiated and non-dividing cells.
Recently, adeno-associated virus (AAV) expression systems have been developed for the expression of genes in nonproliferating cells (17,18). AAV vectors have been used to deliver a number of different genes into a variety of target tissues both in vitro and in vivo, thus demonstrating the significant potential of this virus in the treatment of human diseases (19 -23). In contrast to the more commonly used retroviral and adenoviral vectors, AAV is a non-pathogenic human parvovirus, does not elicit antibodies against itself, and has high possibility for long term expression of transgenes (24). We have used the NDI1recombinant AAV vectors and expressed the NDI1 gene in growth-arrested human cells.

EXPERIMENTAL PROCEDURES
Proviral Plasmid Construction and Packaging-An rAAV proviral plasmid, pCB-NDI1, designed to express the full-length NDI1 gene from the cytomegalovirus/␤-actin hybrid (CB) promoter, was constructed as follows. An rAAV-CB-hAAT construct 2 was digested with NotI and EcoRI leaving the rAAV backbone and the CB promoter with NotI-and EcoRI-compatible ends. The NDI1 gene was then released from the pPCRScript Amp SK(ϩ) vector with NotI and EcoRI, and this fragment was ligated into the rAAV backbone to produce the pAAV-CB-NDI1 plasmid. The rAAV proviral plasmid, pAAV-CB-NDI1, was then packaged into rAAV virions by double-transfection of human embryonic kidney 293 cells and purified by iodixanol step-gradient centrifugation followed by heparin sulfate column chromatography as described previously (25).
Cell Culture, NDI1 Gene Transfection, and Infection-The 293 cells of human embryonal kidney (HEK 293) and the 143B cells of human osteosarcoma were grown in DMEM supplemented with 10% fetal calf serum, 25 mM glucose, and 50 g/ml gentamicin. Cells were maintained at 37°C in a 5% CO 2 atmosphere. The HEK 293 cells (1 ϫ 10 5 ) in 1 ml of DMEM containing 25 mM glucose and 10% horse serum were cotransfected with 8 -10 g of pAAV-CB-NDI1 plasmid by a calcium-phosphate precipitation method (15). The human 143B cells (1 ϫ 10 5 ) in 1 ml of DMEM containing 25 mM glucose and 10% fetal bovine serum were infected with 9 ϫ 10 8 infectious units of rAAV-NDI1 virions. The transduced cells were isolated by screening in the DMEM ϩ 10% fetal calf serum ϩ 0.1 M rotenone in the presence of 5 mM galactose and were grown in the same media.
Viability Measurements-Cells were plated at 1 ϫ 10 5 cells/ml onto 6-well plates (10 5 cells/well) and were harvested and counted every 24 h using a hemocytometer. Viability was determined using trypan blue exclusion. Cell number was counted in triplicate.
Cell Cycle Arrest (Synchronization)-Cells were mitotically arrested by treatment with aphidicolin which is an inhibitor of the DNA polymerases ␣ and ␦ (26,27). The HEK 293 and human 143B cells were cultured by seeding triplicate 6-well plates at approximately 1 ϫ 10 6 cells per well (1-2 ϫ 10 6 cells in 1 ml of medium) and allowed to adhere overnight. The following day, HEK 293 and 143B cells were placed in a medium containing aphidicolin at 5 g/ml for 24 h or at 10 g/ml for 17 h, respectively. On day 2, drug-treated cells and untreated controls were washed twice in ice-cold phosphate-buffered saline, resuspended at 2-3 ϫ 10 5 cells/ml aliquoted into 1-ml samples, and subsequently kept on ice.
Determination of DNA Content-To a 1-ml sample of cell suspension, 2 ml of ice-cold methanol was added dropwise while gently mixing and then incubated on ice for at least 30 min. At this point, samples were either stained immediately with propidium iodide or stored at 4°C for up to 5 days before staining. For staining, samples were centrifuged at 300 ϫ g for 5 min, and the supernatant was discarded. The cell pellets were suspended in 500 l of 0.1 mg/ml propidium iodide (or 20 g/ml) stain solution. Subsequently, 500 l of DNase-free RNase A was added to a final concentration of 100 units/ml. Samples were again vortexed, incubated in the dark at room temperature for 30 min, and then stored on ice in the dark for up to 1 h before they were subjected to flow cytometric analysis on a FACScan (Becton Dickinson, San Jose, CA). For each sample, 10,000 events were counted with the FL2 parameters. Data were collected using the data acquisition program CELLQuest (Becton Dickinson). Histograms were analyzed with the WinMDI program (Joseph Trotter, The Scripps Research Institute). Each experiment was repeated at least three times.
Other Analytical Procedures-Measurements of respiratory chain activities were performed using digitonin-permeabilized cells as reported previously (15). Immunofluorescence (28) was done using anti-Ndi1 antibody and a mitochondria-directed fluorescent probe, Mito Tracker Red (Molecular Probes). Protein concentration was determined by the bicinchoninic acid method (Pierce). SDS-polyacrylamide gel electrophoresis was carried out by the modified method of Laemmli (29). Any variations from the procedures and other details are described in the figure legends.

Transduction of Human Cells with the Yeast NDI1 Gene
Using rAAV-As described in our previous papers (8,15,16), we have successfully introduced the yeast Ndi1 enzyme into mitochondria of the Chinese hamster cells and HEK 293 cells. The expressed protein was functional and was able to restore the respiratory activities of those cells that were defective in complex I. The method of gene transfer used, however, involved use of chemicals and, thus, had some limitations. For example, the transfer efficiency was generally low. Also, by using this technique we were unable to transfect other human cells such as the osteosarcoma 143B cells. In order for this technique to be developed into a possible gene replacement therapy, it must work not only on dividing cells but also with non-dividing or growth-arrested cells and, ultimately, tissues. Thus, it was essential to develop another method of transfection. For this purpose, we turned to recombinant technology and constructed a recombinant AAV proviral plasmid containing a full-length NDI1 gene. This recombinant proviral plasmid, rAAV-NDI1, was then used to transfect HEK 293 cells. Immunofluorescence microscopic analyses using antibodies directed against the yeast Ndi1 polypeptide and a fluorescent probe selective for mitochondria indicated that the expressed Ndi1 was predominantly localized in the mitochondria of the 293 cells (data not shown). rAAV-NDI1 virions were then produced from the rAAV proviral plasmid as detailed under "Experimental Procedures" and were used to infect the human 143B cells. The efficiency of transduction as observed in immunofluorescent staining ranged from 50 to 90%. The expression of the Ndi1 protein was retained after 4 months of continuous culture.
Effect of the Ndi1 Expression on the Electron Transfer Activity- Fig. 1 shows the respiratory activities of the nontransduced 143B cells (control) and the NDI1-transduced 143B cells (rAAV-NDI1). In the case of the nontransduced cells, oxygen consumption in the presence of the respiratory substrates malate/glutamate was inhibited by addition of rotenone, a complex I inhibitor, but enhanced by the following addition of succinate. By contrast, the respiratory activities of the NDI1transduced cells in the presence of glutamate/malate was insensitive to rotenone but sensitive to flavone, an inhibitor of Ndi1. Furthermore, antimycin A inhibited this respiration completely. These results indicate that the expressed Ndi1 enzyme functions as an upstream member of the respiratory chain of the cell. HEK 293 cells transfected with the rAAV-NDI1 vector showed similar results (data not shown) which were basically the same as those obtained with the pHookmediated transfection (16).
Cell Cycle Arrest-To ascertain our ability to transfect nondividing cells using the rAAV-NDI1 virions, it was necessary to establish conditions for arresting cell growth. We used aphidicolin that blocks nuclear DNA replication without interfering with mitochondrial DNA synthesis (30). HEK 293 cells were treated with aphidicolin, and the DNA content was examined with a flow cytometer. As shown in Fig. 2, cells incubated with 5 g/ml aphidicolin for 24 h were primarily in the S phase, whereas the control cells, without the inhibitor, exhibited a normal distribution of growth phases. The same results were obtained with 143B cells (data not shown). To ascertain further the extent to which growth was arrested, we monitored the number of cells for up to 3 days after the aphidicolin treatment ( Fig. 3). At the end of 24 h of incubation (labeled 0 h), the number of cells remained the same when aphidicolin was present but doubled if no inhibitor was added. Furthermore, after aphidicolin was removed from the media, the treated cells remained viable for a few days without showing appreciable increase in the cell number. Thus it is clear that both HEK 293 and 143B cells stopped dividing when they were treated with aphidicolin.
Functional Expression of NDI1 in Non-dividing Cells-After growth arrest was confirmed, the 143B cells were infected with the rAAV-NDI1 particles. The success of the transduction and the subsequent expression of Ndi1 enzyme was shown using immunofluorescence as illustrated in Fig. 4. This is in contrast to our earlier attempts using the pHook vector which gave rise to no transfection. The overall transduction efficiency was typically in the range of 50 -80% (see below for further discussion). The degree of expression, however, seems to vary as is evident in Fig. 4 where cells overexpressing the protein appear brighter (green) and those expressing less are fainter. Results with growth-arrested HEK 293 cells were similar (data not shown). Thus, the expression levels of Ndi1 in non-dividing cells were at least equivalent to those in actively proliferating cells. Although these data confirmed successful transduction and expression in the growth-arrested cells, it was still necessary to make certain that the expressed Ndi1 was functionally active and did not harm cell metabolism. To confirm this point, we carried out the experiment illustrated in Fig. 5. First, 143B cells were incubated with aphidicolin to arrest the cell cycle. When this was complete, cells were infected with the rAAV- NDI1 virions. After 24 h, cells were placed in a medium containing the complex I inhibitor, rotenone, and provided with galactose, instead of glucose, as the carbon source. Under these conditions, glycolysis is too slow to sustain the cell, and energy must be provided by oxidative phosphorylation (31); therefore, non-transduced cells are expected to die because the respiration is blocked by rotenone at the level of complex I. The data show that the cells infected with rAAV-NDI1 survived under the non-glycolytic conditions. In the case of control cells that did not receive the NDI1 gene, the number of viable cells under the same conditions was reduced to Ͻ10% within a day and almost zero by the end of the 2nd day. It is interesting to note that the sustained viability of the transduced cells in the presence of rotenone and galactose was exactly the same as that observed with the non-transduced cells kept in the glucose medium after the growth arrest (Fig. 3B). These data clearly demonstrate that the Ndi1 enzyme is able to support respiration in the growth-arrested cells and that the enzyme does not harm the cell in other ways.
Effects of Complex I Inhibitors on Cell Growth- Fig. 6 illustrates the effects of complex I inhibitors on cell growth of non-transduced and NDI1-transduced 143B cells. In the upper panels, the medium contained galactose as the carbon source with or without rotenone (left) or MPP ϩ (right) as the complex I inhibitor. As explained for the experiment in Fig. 5, cells depend on oxidative phosphorylation for the energy source in this medium (non-glycolytic conditions), and inhibition of the electron transfer at the level of complex I leads to cell death within 2-3 days. This is clearly seen for the non-transduced control cells. On the contrary, the NDI1-transduced cells were able to grow in the presence of rotenone almost to the same extent as that of control cells cultured without rotenone. Addition of MPP ϩ seems to have some retarding effect when compared with rotenone on the growth of the transduced cells. This could be due to possible secondary actions of MPP ϩ (see, for FIG. 4. Immunofluorescence of non-dividing 143B cells infected with the rAAV-NDI1 particles. Cells were first treated with aphidicolin as in Fig. 3, and growth arrest was achieved. Cells were then transduced using rAAV-NDI1. The transduced and the nontransduced control cells were double-labeled with affinity-purified rabbit antibody to Ndi1 (fluorescein isothiocyanate, green) and a mitochondria-selective fluorescent probe (Mito Tracker, red).

FIG. 5. Survival of NDI1-transduced cells in the presence of complex I inhibitor.
The 143B cells were growth-arrested by the aphidicolin treatment (10 g/ml for 17 h) followed by incubation in fresh medium for 8 h. The growth-arrested cells were then incubated in the presence or absence of 1 g/ml AAV-NDI1 for another 24 h. Both NDI1-infected cells and noninfected cells were placed in a medium containing 0.1 M rotenone and 5 mM galactose to disallow glycolysis. At 24 and 48 h after the medium change, cell viability was assessed by trypan blue exclusion, and cell number was determined by using a hemocytometer. Each measurement was done in triplicate. In upper panels (non-glycolytic conditions) 5 mM galactose was used as the carbon source to make the cell dependent on oxidative phosphorylation for the energy source. In lower panels (glycolytic conditions), the medium contained 25 mM glucose to allow the cell to grow under glycolysis. The experiments were carried out in triplicate. example, Refs. [32][33][34]. These data clearly indicate that the Ndi1 protein renders the cells resistant to complex I inhibitors by providing an alternative means to oxidize NADH. In the lower panels, the same set of experiments were carried out in a medium containing high glucose to allow glycolysis. Under these conditions, all cells, regardless of the transduction, can now grow in the presence of added inhibitors. In other words, the inhibitory effects of rotenone and MPP ϩ are only observed when NADH oxidation is taking place. DISCUSSION We have previously shown that the yeast Ndi1 protein is a versatile enzyme and can be used to supplement complex I-deficient cells (15,16). However, use of chemicals or liposomes as the gene delivery techniques, although convenient in in vitro experiments, may be restricted in clinical applications because of relatively low transfection efficiencies and other limitations. Viral vectors, on the other hand, have advantages in that they can be used to transduce quiescent cells such as hepatocytes, myocytes, and neurons and that the gene can be integrated into the DNA of the host cell so that it will be replicated (in dividing cells) and expressed indefinitely (35)(36)(37)(38). We chose recombinant AAV vector because AAV has not been implicated as the causative agent for any diseases, does not elicit antibodies against itself, and supports long term transgene expression (17,18,24).
In this paper we demonstrated that rAAV can be used to deliver the NDI1 gene into non-dividing human cells. The transduction efficiency is as high as 80% as judged by immunofluorescence. However, these numbers could be underestimates. As seen in Fig. 5, the number of viable cells remained at the same level when glucose was replaced with galactose, just as observed with the normal cells in a corresponding experiment of growth arrest but being kept in the glucose medium (Fig. 3B). This implies nearly a 100% expression level because any cells that do not have functional Ndi1 fail to survive under these conditions, and thus the number of viable cells would have declined accordingly if a fraction of the cells did not have active Ndi1. It is likely that immunofluorescent staining revealed only highly expressing cells and that lower amounts of Ndi1 may be sufficient to sustain the cells. It should be noted that, in the experiment of Fig. 5, no selection was made after the rAAV infection before the cells were tested for survival under non-glycolytic conditions. This indicates that because cell division had been prevented, the sustained number of viable cells should represent a high efficacy of transduction but was not a result of a balance between new growth of surviving cells and death of non-transduced cells. It is also noteworthy to add that the presence of "excess" Ndi1 protein does not have negative impact on cell growth. This point is evident from the results shown in Fig. 6 as well. When cells are grown glycolytically, addition of complex I inhibitors has little or no effect on cell viability.
Successful introduction of functional Ndi1 protein into mitochondria of non-proliferating cells is a significant step forward toward the clinical applications of this protein as a remedy for complex I defects. Its seemingly non-toxic nature is important especially because targeted tissues are often populated with both normal cells and cells with varying degrees of deficiency. In addition to the efficiency, longevity of gene expression is another important factor to consider. Once again AAV appears to excel here because of its ability to integrate into chromosomes. Of course, use of AAV vectors are not always without complications. Although it is expected that its immunoreactivity is conceivably low compared with other viral vectors such as adenoviruses, this must be critically tested with in vivo experiments (39). Another concern might be that the Ndi1 protein is foreign to humans, possibly leading to host immune response. There are, however, a number of encouraging reports where non-self proteins were successfully expressed, and repeated injections could be carried out in animal models (40 -42). Tissue-specific expression is another subject that will need to be addressed. Fortunately, a variety of promoters are being developed. In conclusion, it is highly likely that what was made possible with the yeast Ndi1 protein can be extended to other proteins that have to be targeted into mitochondria at the desired location with full activities.