Mitochondrial dysfunction in some triple-negative breast cancer cell lines: role of mTOR pathway and therapeutic potential

Cells (BT474, MCF7, BT20, T47D, ZR751, SKBR3, MDA231, MDA468, MDA436) were obtained from the American Type Culture Collection (Manassas, VA, USA). Cells were cultured in DMEM-F12 medium (Cellgro, Mediatech Inc, Herndon, VA, USA) supplemented with 10% fetal bovine serum and 2 mM glutamine at 37°C, 5% CO₂, and 95% humidity. The experiments performed in this study do not required Institute Ethics Board approval because only commercially available cell lines were used.

N,N,N’,N’-tetramethyl-p-phenylenediamine dihydrochloride (TMPD), β-actin antibody, oligomycin, rotenone, RU486, ascorbate, succinic acid, digitonin, cyanide 3-chlorophenylhydrazone (CCCP), antimycin A and 4-hydroxytamoxifen (4-HOT) were acquired from Sigma (St Louis, MO, USA). Rapamycin was from AG Scientific, Inc (San Diego, CA, USA); (³H) 2-deoxyglucose was acquired from Amersham Pharmacia Biotech (Piscataway, NJ, USA). Phospho-Akt, Phospho-AMPK, total AMPK, Glut-4, hexokinase I, GPX-1, p70S6K (total and phosphorylated form) and α/β tubulin antibodies were from Cell Signaling Technology, Inc (Danvers, MA, USA). Cytochrome c, Akt total, SCO2, Glut-1, hexokinase II antibodies were purchased from Santa Cruz Biotechnology, Inc (Santa Cruz, CA, USA). MitoProfile® Total OXPHOS Human WB Antibody Cocktail was obtained from Mitosciences (Eugene, OR, USA); 7-AAD, MitoSOX™ Red, MitoTracker™ Green and rhodamine-123 were purchased from Invitrogen/Molecular Probes (Carlsbad, CA, USA). p70S6K and MigR1 plasmids were from Addgene (Cambridge, MA, USA) and shRNA P70S6K were from Thermo Scientific/Dharmacon (Lafayette, CO, USA) and were transfected to the cells according to the method provided by the companies.

Oxygen consumption, cellular glucose uptake, lactate production and ATP content were measured as an indication of mitochondrial dysfunction, as previously described [8],[27]. To analyze the mitochondrial respiratory complex activity in breast cancer cells with different receptor status we used the oxytherm system equipped with the Clark electrode. Briefly, basal oxygen consumption was first determined in intact cells (1 × 10^6 cells) resuspended in respiratory buffer containing 20 mM HEPES pH 7.4, 10 mM MgCl2 and 250 mM sucrose, followed by addition of rotenone (500 nM, complex I inhibitor), at 6 minutes, digitonin (15 mg/ml) at 9 minutes, succinate (10 mM, complex II substrate) at 13 minutes, antimycin A (5 mM, complex III inhibitor) at 19 minutes, and TMPD + ascorbic acid (0.5 mM and 2 mM respectively, complex IV substrates) at 22 minutes. In the presence of rotenone and succinate (digitonin was also added to make the biological membranes permeable to succinate), the oxygen consumption rate would represent the maximal activity of complex II-> CoQ- > III-represent the maximal activity of complex> IV. In the presence of antimycin A and ascorbic acid/TMPD, the oxygen consumption rate would represent the uncoupled respiration through complex IV. The overall oxygen consumption at complex IV reflects the combined electron transport activities from both complex I- > CoQ- > III- > IV and complex II- > CoQ- > III- > IV.

Simultaneous multiparameter metabolic analysis of cell populations in culture was performed in the Seahorse XF24 extracellular flux analyzer (Seahorse Bioscience, Billerica, MA, USA) as described previously [28]. Briefly, breast cancer cells were seeded on XF24 V7 multi-well plates (50,000 cells per well for TNBC cells and 20,000 cells per well for ER-positive cells) and were pre-incubated overnight at 37°C in 5% CO₂. The following day, the culture medium was replaced with assay media (unbuffered DMEM/F12 supplemented with 2 mM glutamine, pH 7.4) 1 hour before the assay and for the duration of the experiment. Mitochondrial complex inhibitors (10 ng/ml oligomycin, 4 μM CCCP, 1 μM rotenone) are preloaded in the injection ports. After establishing the baseline oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) readings, mitochondrial complex inhibitors (oligomycin, CCCP, rotenone) were injected consequently and after a short period of mixing, oxygen consumption rate OCR and ECAR measurements were made using photodetectors with specific excitation and emission wavelengths of oxygen (532/650 nm) and protons (470/530 nm; 28). Experiments were run in quadruplicate. The data were normalized by cell number.

Cells were stained 60 minutes with 60 nM MitoTracker Green to measure the mitochondrial mass, with 3 μM Mitosox Red or 100 ng/ml dihydroethidium (DHE) to detect superoxide, or with 1 μM rhodamine-123 to evaluate the mitochondrial transmembrane potential, and with annexin-V-FITC (BD Biosciences Pharmingen, San Diego, CA, USA) and PI to analysis cell death according to the manufacturer’s protocol. Analysis was performed using a FACScan flow cytometer (Becton Dickinson Bioscience, San Jose, CA, USA) with the CellQuest software.

To quantify cell proliferation, we used the (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt) (MTS) assay according to the manufacturer’s protocol (Promega Corporation, Madison, WI, USA). Each independent experiment was carried out three times.

A glutathione assay kit (Cayman Chemical, Ann Arbor, MI, USA) was used to measure cellular glutathione; NADH/NAD + and NADPH/NADP + quantification kits (BioVision, Mountain View, CA) were used to determine NADH/NAD + and NADPH/NADP + ratios according to the manufacturer’s protocols.

RNA was isolated from breast cancer cells using the RNeasy kit and reverse-transcribed using the Omniscript RT kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. The amplification primers for p70S6K were 5′-GTCGACAGCCCAGATGACTCAACTCTC-3′ (forward) and 5′-TCATAGATTCATACGCAGGTGCTCTGGC-3′ (reverse), for PDK1 were 5′- CCGCTCTCCATGAAGCAGTT- 3′ (forward) and 5′- TTGCCGCAGAAACATAAATGAG-3′ (reverse) and for β-actin were 5′-GCATCGTCACCAACTGGGAC-3′ (forward) and 5′-ACCTGGCCGTCAGGCAGCTC-3′ (reverse). cDNA was amplified using standard PCR conditions.

The quantification of western blots and PCR was performed using Image J software and can be found in Additional file 1.

We first evaluated the mitochondrial bioenergetic profiles of TNBC cells by measuring their oxygen consumption and glycolysis rates in comparison with other types of breast cancer cells. The nine cell lines used in this study and their ER, PR, and HER2 expression status are shown in Figure 1A (left panel). A Seahorse extracellular flux analyzer XF24 was used to measure the ECAR and OCR of each cell line, as the indicators of lactic acid production during glycolysis and mitochondrial respiration during OXPHOS, respectively. As shown in Figure 1A (right panel), all TNBC cell lines and the ER/PR double-negative, HER2-positive SKBR3 cells (group I) exhibited higher ECAR and lower OCR compared with the receptor-positive breast cancer cells (group II). Quantitative analysis revealed that TNBC cells had lower OXPHOS and higher glycolysis, as they exhibited significantly lower OCR/ECAR ratios than other breast cancer cells (Figure 1B). Using traditional biochemical analyses, we further validated that TNBC cells exhibited lower mitochondrial respiration and a compensatory increase in glycolysis, as evidenced by a decrease in oxygen consumption and an increase in glucose uptake and lactate production (Additional file 2: Figure S1A-C). To further test if TNBC cells might be more dependent on glycolysis for cell proliferation and viability, breast cancer cells were cultured in medium containing various concentrations of glucose or with only galactose, which generates ATP mainly through mitochondrial metabolism and can only support the cells with competent mitochondrial function. As shown in Additional file 2: Figure S1D, when TNBC cells were grown in medium containing high glucose (3.15 g/L) they exhibited rapid proliferation and the cell numbers increased 4- to 8-fold by day 4. The cell proliferation rate became slower when glucose was reduced to 1.0 g/L. The cells failed to proliferate and even showed a decrease in cell number when the culture medium contained only galactose (Additional file 2: Figure S1D, F). In contrast, ER-positive cells were able to maintain cell viability and showed moderate proliferation with low-glucose or in galactose-only medium (Additional file 2: Figure S1E, F). Furthermore, the high dependency of TNBC cells on glucose was further demonstrated in clonogenic assays (Additional file 3: Figure S2). TNBC cells exhibited much lower ability to form colonies in low-glucose or galactose medium. Analysis of cellular ATP showed that all subtypes of breast cancer cells had similar ATP content (Figure 1C), suggesting that the upregulation of glycolysis in TNBC cells might be sufficient to compensate for the decreased ATP generation in their mitochondria. These data together further confirm that TNBC cells with low mitochondrial respiration were highly dependent on glucose, and that the presence of high glucose provide the key metabolic substrate for generation of ATP and metabolic intermediates for lipid and nucleic acid synthesis to support the robust growth of TNBC cells.

Because the mitochondrial electron transport chain is a major site of ROS production, we used the mitochondrial redox fluorophore MitoSOX™ Red to measure mitochondrial ROS and found that TNBC cells had higher ROS compared to other breast cancer cells (Additional file 4: Figure S3A, left panel), suggesting an increase in electron leakage from the respiratory chain to form ROS within the mitochondria. The specificity of superoxide detection was confirmed using DHE, and the results were similar with those obtained by MitoSOX™ Red analysis (Additional file 4: Figure S3A, right panel). Consistent with the possible impairment in electron flow through the mitochondrial respiratory chain, the TNBC cells exhibited a significant increase in NADH/NAD + ratio (Figure 1D), likely due to less oxidation of NADH to NAD + by the compromised complex I activity and more conversion of NAD + to NADH during active glycolysis. We have also measured NADPH/NADP + ratio in breast cancer cells, and showed that TNBC cells had a slightly lower NADPH/NADP + ratio compared with the receptor-positive cells (Additional file 4: Figure S3B), which might reflect more consumption of NADPH to counteract higher ROS in TNBC cells. There was also a decrease in cellular glutathione (GSH) level (Additional file 4: Figure S3C) and an increase in glutathione peroxidase 1 expression (GPX-1) (Additional file 4: Figure S3D), reflecting cellular response to increased ROS generation. In addition, TNBC cells exhibited an increase in mitochondria mass (Additional file 4: Figure S3E), suggesting a compensatory response to the impaired respiratory function. The mitochondrial transmembrane potential (Δψm) was higher in TNBC cells as indicated by a higher Rhodamine 123 signal (Additional file 4: Figure S3F), suggesting that the integrity of the mitochondrial inner membranes was maintained.

To further investigate the mitochondrial alterations in TNBC cells, we used oligomycin in conjunction with oxygen consumption analysis to evaluate the mitochondrial OXPHOS coupling efficiency (MCE) [29]. Oligomycin suppresses OXPHOS activity by inhibiting ATP synthase. As shown in Figure 1E, although TNBC cells (MDA-468) exhibited substantially lower basal oxygen consumption compared to the triple-positive breast cancer cells (BT474), the presence of oligomycin (10 ng/ml) caused a similar decrease in oxygen consumption (approximately 50%) in both cell lines, suggesting that TNBC cells had similar mitochondrial coupling efficiency compared to other breast cancer cells. Similar results were observed in other cell lines (Figure 1F), indicating no significant impairment in oxygen consumption/ATP synthesis coupling in TNBC.

We then compared the spare respiratory capacity (SRC) in TNBC cells and other breast cancer cells. SRC was calculated as the difference between the basal mitochondrial O₂ consumption and the maximal mitochondrial O₂ consumption when the respiratory chain was uncoupled by CCCP [30]. In the presence of CCCP, the respiratory chain activity is independent from ATP synthesis at complex V (F₁F₀ ATPase). Addition of CCCP (4 μM) to BT474 and MCF7 cells resulted in a dramatic increase in mitochondrial respiration (OCR), suggesting that these cells exhibit a high reserve of mitochondrial respiratory capacity (Figure 1E, F). In contrast, TNBC cells seemed to have much lower respiratory reserve capacity, as they exhibited only a limited increase in OCR when CCCP was added (Figure 1E, F). As CCCP acts as a protonophore across all membranes, acidifying the cytosolic compartment and disrupting the function various cellular compartments, this could affect substrate supply and thus affect the uncoupled rate. For this reason CCCP might have different oxidation rates in different breast cancer cells and thus potentially contribute to the differences in the maximum respiratory rates observed to some degree.

To further examine the mitochondrial dysfunction in TNBC cells, we measured the activity of each respiratory chain complex using specific complex inhibitors in combination with proper complex substrates in digitonin-permeabilized cells [31],[32]. As shown in Figure 2A, respiration of breast cancer cells permeabilized with digitonin (15 μg/ml) was blocked by rotenone (500 nM, complex I inhibitor), restored by addition of succinate (10 mM, complex II substrate), inhibited again by antimycin A (5 μM, complex III inhibitor). The electron flow through complex IV was resumed with addition of electron donors ascorbic acid (2 mM) and TMPD (0.5 mM), which provided electrons directly to cytochrome c, thus allowing an estimation of oxidation through complex IV. As shown in the upper panel of Figure 2A, TBNC cells (MDA231) exhibited very low rotenone-sensitive respiration (complex I) and lower succinate-driven respiration (complexes II-IV) compared to the triple-positive breast cancer cells (BT474). Furthermore, addition of ascorbic acid and TMPD to assay the function of complex IV revealed a smaller increased of oxygen consumption in MDA231 cells compared to BT474 cells (Figure 2A, upper panel). Quantitative analysis of data for each segment of the mitochondrial respiratory chain activity in four different cell lines, two triple-negative lines (MAD231 and MDA468) and two receptor-positive lines (BT474 and MCF7), showed that TNBC cells exhibited functional defects in multiple respiratory complexes, since all segments of the respiratory chain activities in TNBC cells were lower than the receptor-positive lines (Figure 2A, lower panel). We further confirmed these observations by analyzing their mitochondrial respiratory chain components by western blotting, which showed a reduction in expression of complex I and complex III proteins (Figure 2B). However, there were no significant changes in the expression of complex II, IV and V components (Figure 2B). Interestingly, the expression of SCO2, a nuclear-encoded complex IV assembly factor, was missing in BT20 (triple-negative), reduced in SKBR3 cells (ER-/PR-/HER2+) and T47D cells (ER+/PR+/HER2-), and unchanged in other cell lines including three TNBC lines (Figure 2B). The expression of cytochrome c was slightly increased in TNBC cells (Figure 2B). These results together suggest that TNBC cells exhibit lower OXPHOS due to dysfunction in several mitochondrial respiratory complexes, particularly complexes I and III.

As TNBC cells display low OXPHOS and high glycolysis, we further examined the expression of Akt, Glut-1/-4, HKII, PKM2, LDH-A, and PDH, as these molecules are known to be involved in regulation of glucose metabolism in cancer [13],[33]. Western blot analysis revealed that Akt was activated in the majority of the TNBC cell lines (BT20, MDA468, MDA436), as evidenced by an increase in Akt phosphorylation at ser473 (Figure 2C). However, although MDA231 expressed a similar level of total Akt protein, this TNBC cell line did not show detectable phosphorylated Akt for a yet unknown reason. Thus, it is possible that some of the TNBC cells could be highly glycolytic without Akt activation. The expression of HKII, Glut-1 and Glut-4 was increased in 3 TNBC cell lines (MDA468, MDA231, MDA436), while HKI expression remained unchanged (Figure 2C). Furthermore, TNBC cells also exhibited high expression of PDK1 mRNA (Figure 2D). PDK1 is known to inactivate pyruvate dehydrogenase (PDH) and thus might contribute to a lower utilization of pyruvate by the mitochondria and higher lactate production.

Because AMP-activated protein kinase (AMPK) is a metabolic sensor that responds to alterations in cellular energy levels to maintain energy balance and stimulate glycolysis in response to metabolic stress through direct phosphorylation and activation of 6-phosphofructo-2-kinase [34],[35], we used western blot analysis to examine the expression of AMPK and phospho-AMPK. TNBC cells exhibited an increase in AMPK activity, evident by elevated Thr172 phosphorylation (Figure 2E). The high phospho-AMPK levels in TNBC cells were inversely associated with low expression of p70S6K protein and its phosphorylation at Thr389 (Figure 2E). The mRNA expression of p70S6K in TNBC cells was as high as that in other breast cancer cells (Figure 2F), suggesting that the difference in p70S6K between TNBC cells and receptor-positive breast cancer cells might be due to a difference in post-transcriptional and post-translational regulation. These data are consistent with the role of AMPK in inactivation of p70S6K [36].

To evaluate the potential role of the altered AMPK-mTOR pathway in affecting mitochondrial function in TNBC cells, we then tested if suppression of this pathway by the mTOR inhibitor rapamycin or the AMPK activator AICAR would lead to changes in mitochondrial respiration. Treatment of the receptor-positive breast cancer cells (BT474 and MCF7) with rapamycin resulted in a rapid inactivation of the mTOR downstream molecule p70S6K, evidenced by a decrease in phospho-p70S6K at Thr389, and a substantial reduction of oxygen consumption, suggesting that the mTOR/p70S6K pathway was important for maintaining mitochondrial respiration in these breast cancer cells (Figure 3A, B). We noted that the mTOR downstream target molecules p70S6K and p85S6K were both detected by the antibody used in immunoblotting (Figure 3B), as p85S6K is derived from the same gene and is identical to p70S6K except for 23 extra residues at the amino terminus [37]. The phosphorylation of p70S6K at threonine 389, however, is known to correlate with S6K activity [38]. Interestingly, the same concentration of rapamycin (0.1 μM) showed no effect on the respiration in TNBC cells (MDA468 and MDA231, Figure 3A), as these cells express very low levels of p70S6K and had low basal respiration. AICAR treatment resulted in activation of AMPK in all breast cancer cells tested, as shown by an increased in the phosphorylation of AMPK at Thr172 (Figure 3C, top panel). However, AICAR only caused a significant decrease of mitochondrial respiration in BT474 and MCF7 cells, while the respiration in TNBC cells (MDA468 and MDA231) was only slightly affected (Figure 3C , bottom panel), consistent with the observation in the experiments using rapamycin.

To further test the role of p70S6K in affecting mitochondrial respiration, we ectopically expressed p70S6K in TNBC cells (MDA468), and showed that such forced expression of p70S6K led to an increased expression of the active protein and caused a significant increase in mitochondrial respiration (Figure 3D). This was associated with a decrease in lactate production and a reduction in glucose uptake (Additional file 5: Figure S4A, B), suggesting that the MDA468 cells with forced p70S6K expression were more active in mitochondrial OXPHOS with less glycolysis. Similarly, reduction of p70S6K expression in triple-positive cells (BT474) by shRNA knockdown led to a significant decrease in respiration (Figure 3E), an increase in glucose uptake and lactate production (Additional file 5: Figure S4C, D). These data together suggest that the AMPK-mTOR-p70S6K axis play a significant role in maintaining mitochondrial respiration in breast cancer cells, and that its loss in TNBC cells led to respiration defect.

As TNBC exhibited significant alteration in AMPK and p70S6K expression and in the mitochondrial respiration, we then investigated if inhibition of ER, PR, and HER2 signaling could modulate the AMPK-p70S6K pathway and affect mitochondrial respiration in breast cancer cells. Herceptin, RU486, and 4-HOT were used to inhibit HER2, PR, and ER, respectively, in BT474 cells (triple-positive). As shown in Figure 4A, inhibition of each receptor with the corresponding compound at sub-toxic concentrations (100 nM 4-HOT, 100 nM RU486, 10 μg/ml herceptin) caused a moderate activation of AMPK and a decrease in phospho-p70S6K, while combinations of these inhibitors resulted in further activation of AMPK and more inhibition of p70S6K, with the combination of all three inhibitors having the strongest effect. Consistently, each inhibitor had a moderate effect on mitochondrial respiration (approximately 15% reduction compare to the control), and the combination of herceptin with either RU486 or 4-HOT caused a further decrease in oxygen consumption (approximately 20 to 30%). Inhibition of all three receptors by combining compounds led to a substantial loss of respiration by 40% (Figure 4B). Under these conditions, we did not observe a significant decrease in cell proliferation (measured by counting cell numbers) or any increase in cell death using annexin-V/PI double staining (data not shown). These data together suggest that suppression of ER, PR, and HER2 signaling in breast cancer cells results in activation of AMPK and suppression of p70S6K, leading to a decrease in mitochondrial respiration.

Based on the observations that TNBC cells exhibited impaired mitochondrial function and have an active glycolytic metabolism, we hypothesized that TNBC cells might be more dependent on glycolysis and thus more sensitive to glycolytic inhibition. Two glycolytic inhibitors, 2-deoxyglucose (2-DG) and 3-bromo-2-oxopropionate-1-propyl ester (3-BrOP), were used to test this possibility. As shown in Figure S5A (Additional file 6), the 4 TNBC cell lines tested were sensitive to growth inhibition by 2-DG, with an average IC50 value of approximately 10 mM. In contrast, the triple positive breast cancer cells (BT474), which have competent mitochondrial respiratory function, were less sensitive to glycolytic inhibition by 2-DG with an IC50 value of approximately 50 mM (Additional file 6: Figure S5B). The ER-positive/HER2-negative cells exhibited intermediate sensitivity (Additional file 6: Figure S5C). Interestingly, SKBR3 cells, which are ER-/PR-/HER2+ with defective mitochondrial respiration (Figure 1), were also more sensitive to 2-DG (Additional file 6: Figure S5D). Consistently, a more potent glycolytic inhibitor 3-BrOP showed a stronger inhibitory effect on the TNBC cells and caused massive cell death in TNBC cells as revealed by annexin-V/PI double-staining (Figure 5A, B) and measurement of mitochondrial transmembrane potential (Figure 5C, D), while the respiration-competent BT474 and MCF7 cells were much less sensitive to 3-BrOP (Figure 5). Most TNBC cells and ER-/PR-/HER2+ cells (SKBR3) cells with low mitochondrial respiration were highly sensitive to 3-BrOP compared to other breast cancer cells with competent mitochondrial respiration (Figure 5B, D). Interestingly, MDA231 cells (TNBC) exhibited only modest response to 3-BrOP for yet unknown reasons.

We further tested if manipulation of P70S6K could have any influence on cell death. As shown in Figure S6 (Additional file 7), the inhibition of p70S6K by rapamycin and by shRNA rendered BT474 cells more sensitive to glycolysis inhibition (Additional file 7: Figure S6A, B). In contrast, over-expression of P70S6K in TNBC cells (MDA468) rendered cells more resistant to 3-BrOP (Additional file 7: Figure S6C). Furthermore, activation of AMPK by AICAR could also make BT474 cells more vulnerable to cell death by 3-BrOP (Additional file 7: Figure S6A). Altogether these data suggest that manipulation of p70S6K/AMPK could sensitize breast cancer cells to the cytotoxicity of glycolytic inhibitors.

We then determined the effect of 3-BrOP on ATP levels in breast cancer cells. 3-BrOP induced a more severe loss of ATP in TNBC cells and ER-/PR-/HER2+ (SKBR3) cells (group I) compared to triple-positive and ER-positive cells (Group II, Figure 6A), consistent with the high dependency of TNBC cells on glycolysis for generation of ATP. To further test the dependency of TNBC cells on glucose metabolism in comparison with the mitochondria-competent breast cancer cells, we used siRNA strategy to knock down HKII expression in MDA436 cells (TNBC) and in MCF7 cells (mitochondria-competent), and then examined the effect of such knockdown on cell viability. As shown in Figure 6B, the basal HKII expression level was much higher in MDA436 cells than in MCF7 cells, consistent with the high glycolytic activity in TNBC cells. Importantly, although siRNA was able to completely knockdown the expression of HKII in both cell lines, such a knockdown caused only moderate cell death in MCF7 cells while significantly more apoptotic cells were observed in MDA436 cells (Figure 6B, bottom panel). These results indicate that TNBC cells were more dependent on the glycolysis and that targeting key glycolytic enzyme(s) was effective in killing TNBC cells.

To evaluate the potential therapeutic selectivity of glycolytic inhibition, we compared the sensitivity of TNBC cells to 3-BrOP with immortalized human mammary epithelial cells (HMLE). As shown in Figure 6C, 3-BrOP at 50 μM caused only moderate cytotoxic effect in HMLE, whereas this concentration of 3-BrOP induced massive cell death in MDA436 cells, suggesting that inhibiton of glycolysis might selectively kill breast cancer cells with compromised mitochondrial function.