Mitochondrial DNA mutations are currently investigated as modifying factors impinging on tumor growth and aggressiveness, having been found in virtually all cancer types and most commonly affecting genes encoding mitochondrial complex I (CI) subunits. However, it is still unclear whether they exert a pro- or anti-tumorigenic effect. We here analyzed the impact of three homoplasmic mtDNA mutations (m.3460G>A/MT-ND1, m.3571insC/MT-ND1 and m.3243A>G/MT-TL1) on osteosarcoma progression, chosen since they induce different degrees of oxidative phosphorylation impairment. In fact, the m.3460G>A/MT-ND1 mutation caused only a reduction in CI activity, whereas the m.3571insC/MT-ND1 and the m.3243A>G/MT-TL1 mutations induced a severe structural and functional CI alteration. As a consequence, this severe CI dysfunction determined an energetic defect associated with a compensatory increase in glycolytic metabolism and AMP-activated protein kinase activation. Osteosarcoma cells carrying such marked CI impairment displayed a reduced tumorigenic potential both in vitro and in vivo, when compared with cells with mild CI dysfunction, suggesting that mtDNA mutations may display diverse impact on tumorigenic potential depending on the type and severity of the resulting oxidative phosphorylation dysfunction. The modulation of tumor growth was independent from reactive oxygen species production but correlated with hypoxia-inducible factor 1α stabilization, indicating that structural and functional integrity of CI and oxidative phosphorylation are required for hypoxic adaptation and tumor progression.
Respiratory chain complex I (CI) is the first component and the rate-limiting step of mitochondrial oxidative phosphorylation (OXPHOS). This enzyme couples the oxidation of NADH, produced through the tricarboxylic acid (TCA) cycle, to proton translocation generating mitochondrial membrane potential (Δψm). Hence, CI has a pivotal role in the control of energetic metabolism and the maintenance of cellular redox status, by regulating the NAD+/NADH ratio and reactive oxygen species (ROS) production (1). Isolated CI deficiencies, due to the occurrence of mitochondrial DNA (mtDNA) and nuclear DNA mutations, are the most frequent cause of mitochondrial diseases (2,3). In addition, mitochondrial dysfunction, particularly caused by CI mtDNA mutations, has been identified as an important contributor to the pathogenesis of complex disorders such as diabetes, obesity and cancer (4–6). A prominent role for CI in cancer progression is currently emerging, since several somatic mtDNA mutations have been reported in a variety of tumors. In particular, mtDNA mutations mapping in CI subunits or in genes that impact on CI assembly and function are genetic hallmarks of oncocytomas (7,8), tumors which usually do not metastasize and display a benign phenotype. It has been shown that CI-affecting mutations and the consequent mitochondrial dysfunction may modify tumorigenic potential, although the molecular mechanisms involved remain unclear. Some mtDNA mutations confer a growth advantage to cancer cells and/or enhance the metastatic potential in a ROS-dependent manner (9–12), whereas severe CI dysfunction, caused by homoplasmic truncative mtDNA mutations, induced tumor growth arrest via activation of apoptosis (13). Moreover, we reported that the lack of functional CI, due to the presence of high loads of the m.3571insC/MT-ND1 mutation, was associated with TCA cycle metabolites imbalance, chronic hypoxia-inducible factor 1α (HIF1α) destabilization and hence tumor growth arrest, due to cancer cells inability to acquire a typical advantageous Warburg profile (14,15). In this frame, the evaluation of the effects of mtDNA mutations on tumor progression must take into account not only the mutation type and load, but also the degree and nature of respiratory chain dysfunction, which may modulate changes in the metabolic status/adaptation and growth of cancer cells. OXPHOS impairment induces a cellular energetic crisis, due to the reduced ATP production, an alteration of redox homeostasis, an imbalance of TCA cycle metabolites and O2 consumption, which in turn can trigger specific sensing systems. It has been suggested that the OXPHOS defect caused by the heteroplasmic m.12418insA/MT-ND5 mutation increases ROS production and activates the PI3K/Akt proliferation pathway, leading to a higher tumorigenic potential (16). This pathway opposes the AMP-activated kinase (AMPK) signaling, the major player in the regulation of cellular energy homeostasis, which controls both glycolysis and mitochondrial oxidative metabolism, by inhibiting biosynthetic pathways and cell proliferation (17,18). On the other hand, O2 availability and the accumulation of TCA cycle intermediates, such as α-ketoglutarate (αKG) and succinate (SA), allosterically regulate the activity of prolyl-hydroxylases (PHDs), determining the stabilization of HIF1α. This transcription factor promotes the metabolic switch toward aerobic glycolysis or Warburg effect, promoting neo-vascularization, tumor progression and cell invasion (19). Since opposite effects have been attributed to CI deficit with respect to tumorigenesis, we reasoned that diverse degrees of OXPHOS defects may differently regulate the activation of cellular sensing systems, and thus modulate the metabolic adaptation and cell growth.
We hence investigated the functional consequences of homoplasmic mtDNA mutations that displayed different effects on CI assembly and function, and correlated the degree of OXPHOS impairment with the tumorigenic potential of isogenic osteosarcoma cells. We demonstrated that the m.3571insC/MT-ND1 and the m.3243A>G/MT-TL1 mutations, but not the m.3460G>A/MT-ND6, induced a severe CI defect leading to a profound energetic crisis, followed by AMPK activation, alteration of αKG/SA balance and chronic destabilization of HIF1α. Taken together, energetic failure and inability of hypoxic adaptation resulted in lower tumorigenic potential of severe CI mutants, indicating that different mtDNA mutations modify tumor progression depending on the degree of respiratory CI impairment.
mtDNA mutations differently affect CI function and cellular bioenergetic competence
In order to discriminate the role of different degrees of CI dysfunction on tumorigenic potential, we used a panel of cybrid cell lines derived from osteosarcoma 143B.TK− harboring the missense m.3460G>A/MT-ND1 (RJ206 cells), the tRNA m.3243A>G/MT-TL1 (RN164 cells) and the frameshift m.3571insC/MT-ND1 mutation (HXTC1 cells).
The first two are well-established pathogenic mutations for the most common mitochondrial diseases, Leber's Hereditary Optic Neuropathy (LHON) (20) and Mitochondrial Encephalomyopathy, Lactic Acidosis, Stroke-like episodes (MELAS) (21), respectively. The MELAS mutation m.3243A>G/MT-TL1 maps within the tRNALeu(UUR) and therefore affects overall mitochondrial protein synthesis, in particular of those mtDNA encoded subunits rich in Leu, such as the ND6 of CI (22). The homoplasmic m.3571insC/MT-ND1 frameshift mutation is likely the most frequent genetic lesion in oncocytic tumors where it has been shown to induce CI disassembly (23–25).
In order to assess the overall bioenergetic competence of these cell clones, we determined the effect of the three mutations on cell viability, challenging the control (CC cells with wild-type mtDNA) and mutant cell lines in a glucose-free medium supplemented with galactose. This condition forces cells to use the oxidative phosphorylation for ATP production and enables to highlight a mitochondrial respiratory dysfunction. All mutant cell lines were unable to grow under these conditions unlike CC cells (Fig. 1A and Supplementary Material, Fig. S1). Moreover, CC cells in high glucose medium displayed a typical oxygen consumption rate (OCR) profile, being inhibited by oligomycin, greatly stimulated above basal levels by uncoupler FCCP and inhibited by rotenone (Fig. 1B). Conversely, RJ206 were characterized by a lower basal respiration (OCR), which was still sensitive to oligomycin, but could not be stimulated by FCCP to higher levels than basal OCR, indicating a markedly low respiratory capacity. HXTC1 and RN164 exhibited a nearly absent basal OCR that was neither stimulated by FCCP nor sensitive to respiratory chain inhibitors (Fig. 1B), revealing a dramatic mitochondrial dysfunction. CI in-gel activity (CI-IGA) analysis showed a band corresponding to functional CI in both CC and RJ206. Such activity was markedly reduced in RN164 and nearly undetectable in HXTC1 (Fig. 1C). These data were confirmed by spectrophotometric measurement of CI activity (NADH:DB:DCIP oxidoreductase), which was absent in HXTC1 and RN164 and reduced by 50% in RJ206 (Fig. 1D). This finding was strengthened by the observation that the ATP synthesis driven by CI substrates was undetectable in HXTC1 and RN164, whereas it was decreased by about 60% in RJ206 compared with CC (Fig. 1E). It was not surprising that ATP synthesis driven by Complex II (CII) substrates was compromised exclusively in RN164, indicating the presence of an overall respiratory defect downstream of CI (Fig. 1E). Lastly, the mitochondrial membrane potential (Δψm) was maintained in the presence of oligomycin in both RJ206 and CC, suggesting that the residual CI activity was sufficient to sustain the electrochemical gradient (Fig. 1F). On the contrary, oligomycin induced a rapid and significant decrease in Δψm in HXTC1 and RN164, unveiling the major contribution of the ATPase-mediated ATP hydrolysis in the generation of Δψm in these cells (Fig. 1F).
Overall, these findings clearly demonstrated that these mtDNA mutations differently affect CI function and, in turn, cell bioenergetic competence. In particular, the m.3571insC/MT-ND1 and the m.3243A>G/MT-TL1 mutations led to a dramatic OXPHOS impairment, whereas the m.3460G>A/MT-ND1 caused only a mild, although significant, decrease in CI activity.
Severe CI dysfunction alters cellular glucose metabolism and induces AMPK activation
Fast proliferating cells do not only utilize glucose as a carbon source for cell growth, but they also rely on L-glutamine (26). Hence, to assess the contribution of glutaminolysis on cell growth capacity, all clones were incubated in a medium without L-glutamine supplemented with 25 mm of glucose (QFM), or in a medium lacking glucose and supplemented with 2 mm L-glutamine (GFM). Despite the cytostatic effect induced upon L-glutamine deprivation, all clones were able to survive under this condition, whereas they all rapidly died without glucose, indicating that glycolysis, rather than glutaminolysis, mainly contributes to cell proliferation in culture (Fig. 2A and Supplementary Material, Fig. S1). We next analyzed the impact of different CI dysfunctions on the cells ability to use glucose as a carbon source. Severely impaired HXTC1 and RN164 cells displayed an increased glucose uptake and consumption, when compared with CC and RJ206 cells (Fig. 2B and C). Moreover, lactate production and release in the culture medium was significantly higher in severe mutants (Fig. 2D). Taken together, these results indicate the prevalence of a compensatory glycolytic metabolism in cells with severe OXPHOS defects. Conversely, no difference was found between RJ206 and CC, indicating that the mild CI dysfunction caused by the m.3460G>A mutation does not alter glucose metabolism (Fig. 2B–D).
To assess the overall energetic status, we evaluated the levels of adenine nucleotides, which are known to be sensitive indicators of low energy conditions and to trigger specific signaling cascades in order to maintain the energetic homeostasis for cell survival (27). Hence, we measured the AMP, ADP and ATP levels in the presence of either high or low glucose, the latter being a condition similar to physiological and representing a mild energetic stress for cultured cells. Surprisingly, under high glucose conditions, the ATP content was significantly decreased to comparable levels in all mutant clones (Fig. 3A), whereas AMP and ADP levels were not affected by the mitochondrial dysfunction (Supplementary Material, Fig. S2A and B). It is likely that the up-regulated glycolysis of severe mutants may produce ATP levels similar to those found in mild mutants, thus allowing cell survival in vitro, but not fully compensating the mitochondrial deficit and the consequent energetic imbalance. To investigate whether this energetic impairment triggers cellular energetic sensor AMPK, which is known to be activated by high levels of AMP and inhibited by ATP (28), we evaluated its phosphorylation by western blot analysis. In the presence of high glucose, a modest increase in the levels of phosphorylated Thr172 AMPK was detectable in cells bearing a severe CI defect, compared with CC and RJ206 cells (Fig. 3B), suggesting its potential role in the maintenance of basal ATP levels.
However, upon a mild metabolic stress, such as incubation in low glucose, a slight but significant increase in total ATP levels was observed in CC and RJ206, but not in severe mutants (Fig. 3C). Conversely, a striking increase in AMP concentration was detected in severe mutants after incubation in low glucose, but not in CC and RJ206, confirming the presence of an energetic failure in HXTC1 and RN164 (Supplementary Material, Fig. S2A and B). Furthermore, under mild metabolic stress conditions, the increase in AMP content induced a dramatic activation of AMPK in severely affected mutants, whereas CC and RJ206 displayed only a modest activation of the kinase (Fig. 3D and E and Supplementary Material, Fig. S2C). To infer whether a mild metabolic stress is able to switch on the fatty acid oxidation (FAO) as a compensatory mechanism of ATP supply, we analyzed the status of acetyl-CoA carboxylase (ACC), a regulator of fatty acid metabolism directly phosphorylated and inhibited by active AMPK (29). Interestingly, AMPK activation was reflected in the levels of ACC phosphorylation in all cell lines and particularly in severe mutants (Fig. 3D–F). Collectively, these data indicate that the energetic failure caused by a severe functional CI defect induces the activation of the energetic sensor AMPK that may likely stimulate glycolysis and FAO in severe mutants, allowing the preservation of a certain level of ATP. However, we cannot exclude that ACC phosphorylation reflects only the state of energetic stress without actually engaging functional FAO, since severe mutants are unable to grow and proliferate under nutrients paucity.
Severe mitochondrial CI dysfunction induces HIF1α destabilization and down-modulates tumor progression
In order to assess the impact of mtDNA mutations differently affecting OXPHOS efficiency on the tumorigenic potential, we determined the ability of clones to form colonies in vitro and xenografts in vivo. Anchorage-independent growth assay showed that the number of colonies produced by RJ206 in soft-agar was comparable with that of CC, whereas HXTC1 and RN164 formed significantly fewer and smaller colonies (Fig. 4A and B). Moreover, after the injection of clones in nude mice, the tumor growth rate and volume at the time of sacrifice were similar in CC and RJ206, whereas they were significantly reduced in HXTC1 and RN164 (Fig. 4C). At sacrifice, mice injected with controls and cells bearing a mild CI defect developed masses of 1.96 ± 0.57 cm3 (14/15 positive animals) and 2.47 ± 0.50 cm3 (15/15 positive animals), respectively. Conversely, cells carrying a severe CI deficit formed tumors of 0.21 ± 0.13 cm3 (RN164, 5/15 positive animals) and 0.07 ± 0.05 cm3 (HXTC1, 3/10 positive animals). The few available xenografts derived from RN164 and HXTC1 resulted to be revertants for the mitochondrial pathogenic mutation (Supplementary Material, Fig. S3) and were not carried forward in subsequent experiments, although reinforcing the finding that at least a partial OXPHOS recovery is necessary to allow tumor growth. Overall, in terms of in vivo tumorigenicity, the two mutations that severely affect CI activity clustered together against the mild mutants and controls, supporting the previously proposed association between the lack of functional CI and the inability of cancer cells to proliferate (14,15).
ROS-mediated signaling has been proposed to modulate the tumorigenic and metastatic potential of cancer cells harboring mtDNA mutations (10,16). It is possible that a CI mutation may allow the main ROS-generating site to over-produce radicals by preventing an efficient electron transfer. The measurement of superoxide anion production with the dye MitoSOX™ by flow cytometry and fluorescent microscopy did not reveal any difference among all cell clones (Fig. 4D and E). However, significantly higher H2O2 levels were found in RJ206 compared with the other cells (Fig. 4F), confirming the previously reported existence of a basal oxidative stress in LHON mutants (30). Noticeably, no differences in H2O2 levels were found between controls and HXTC1 and RN164 (Fig. 4F). These data indicate that different degrees of CI dysfunction may induce a different basal oxidative stress. However, it is likely that severe mutations inducing CI disassembly, such as those harbored by HXTC1 and RN164, block the entrance of electrons into the main ROS-producing respiratory complex and thus prevent the generation of ROS. Conversely, the m.3460G>A mutation, which reduces only CI activity without affecting its structural integrity, may enhance ROS production (30,31), thus stimulating cell growth. Despite the higher H2O2 levels found in RJ206 cells in comparison to controls, they both still displayed similar tumor growth rate and volume in vivo, suggesting that the increased oxidative stress might contribute to tumor progression, but that this stimulus is not sufficient to further enhance the tumorigenic potential of highly aggressive osteosarcoma cells.
We have recently demonstrated that CI function is necessary for the induction of the Warburg effect and CI disassembly is associated with an imbalance in the αKG/SA ratio and with HIF1α destabilization (14,15). Interestingly, cells carrying the severe CI defect, namely HXTC1 and RN164, displayed a strong imbalance in the αKG/SA ratio compared with CC and RJ206 (Fig. 5A and Supplementary Material, Table S1). Moreover, the alteration in the αKG/SA ratio correlated with reduced levels of HIF1α protein in CI-defective cells (Fig. 5B), likely due to the continuous feeding of the PHDs reaction by αKG. These findings emphasize that a severe CI dysfunction is tightly associated with chronic HIF1α destabilization and lack of hypoxic adaptation, which in turn may be responsible for the arrest of tumor growth (14,15).
Since the crosstalk between AMPK and HIF1α is still controversial, we attempted to elucidate whether active AMPK regulates HIF1α stabilization or vice versa in our models. The activation of AMPK induced by AICAR in CC and RJ206 resulted in no modifications in HIF1α status or alterations in the propensity to form colonies in soft agar (Fig. 5C–E). On the other hand, HIF1α chronic stabilization after treatment with DMOG in HXTC1 and RN164 results ineffective on AMPK and its target ACC phosphorylation, while it increased the tumorigenic potential of CI-defective cells (Fig. 5C, F and G). Taken together, these results suggest that in osteosarcoma-derived cybrids carrying a severe mitochondrial defect, HIF1α is the major player in the regulation of tumor progression.
MtDNA mutations, in particular those in CI-encoding genes, have been recognized as important modifiers of tumor growth and progression. We previously reported that missense mutations, most likely exerting mild functional effects, are frequently found in different types of cancer, whereas mtDNA mutations leading to severe mitochondrial deficit are significantly underrepresented, suggesting that most of them may be subjected to a negative selection (7,32). In this context, it is noteworthy that with the advent of ultra-deep next-generation sequencing techniques, a high heterogeneity in terms of mitochondrial genotypes has been revealed both in non-cancer and cancer cells (33,34). The discovery of the occurrence of a plethora of differently heteroplasmic mtDNA mutations implicates that the overall effect on tumor progression, or even the response to therapeutic treatments, may result from the combination of several variants. These findings pose the issue regarding which selective pressures operate in vivo to allow for the accumulation or the purification of mtDNA mutations. Nonetheless, it has to be taken into account that most mtDNA mutations are subjected to a threshold effect to display a pathological phenotype. The rationale for this study derives from our and other groups reports in which highly damaging homoplasmic disruptive mtDNA mutations were frequently found in oncocytomas, a particular type of epithelial tumors which usually present a low-proliferating and indolent behavior (23,24,32). Moreover, damaging mtDNA mutations are frequently identified in cancer tissues and some functional studies suggested their possible positive contribution to tumor progression (35–37). Thus, it is still a matter of discussion whether, when and how mtDNA mutations exert their modifying effects during tumor progression (7,8,38). In order to clarify these issues, we modeled different mitochondrial deficiencies by placing three mtDNA mutations within the same nuclear background. We here demonstrate that the mild CI functional impairment found in cells harboring the LHON pathogenic m.3460G>A/MT-ND1 mutation is not sufficient to influence tumor progression, likely since the nuclear hits within the osteosarcoma nucleus are preponderant in determining aggressiveness. Conversely, the oncocytoma-associated m.3571insC/MT-ND1 and MELAS pathogenic m.3243A>G/MT-TL1 mutations, which induce a severe CI defect, inhibit tumor growth, overcoming the nuclear oncogenic hits. These and our previous data clearly indicate the multiple nature of mtDNA mutations that can play neutral, anti- or a pro-tumorigenic effects, depending on the degree of the induced bioenergetic defects. In this regard, we demonstrated that the severe m.3571insC/MT-ND1 and m.3243A>G/MT-TL1 mutations induce an imbalance of the αKG/SA ratio, probably via NADH accumulation and allosteric regulation of TCA cycle, which, in turn, leads to chronic HIF1α destabilization even during hypoxia (pseudonormoxia). Under this condition, cells are not able to acquire a HIF1α-dependent Warburg profile and may hardly proceed to malignancy (14,15). In fact, allotopic CI complementation restores a physiological αKG/SA ratio and allows HIF1α stabilization in vivo, ensuring a glycolytic shift and recovering tumorigenic potential (15). On the contrary, the m.3460G>A/MT-ND1 mutation slightly reduces CI activity and still allows NADH consumption. In this case, TCA cycle is not relented and cancer cells harboring this mutation are able to adapt to hypoxia and to form tumors. Interestingly, MELAS mutations leading to CI disassembly have quite frequently been detected as homoplasmic somatic in oncocytomas (24,39–41), whereas LHON mutations never occur in association with this benign, senescent-like type of tumors.
It has also been proposed that HIF1α can be regulated by ROS, which are mainly generated by mitochondrial CI and complex III (CIII) (42–45). However, this matter is still controversial, since it has also been shown that HIF1α can be stabilized during hypoxia in cells lacking mtDNA and hence defective for CI and CIII (46). Some studies reported that CI dysfunction, caused by mtDNA mutations, enhanced tumorigenic and metastatic potential in a ROS-dependent manner, by PI3K/Akt/PKC pathway activation and HIF1α over-expression (10,13,16,47). Interestingly, in our models we found that only cells carrying a mild CI functional alteration displayed increased H2O2 levels, whereas CI defective cells were not subjected to chronic oxidative stress. It is reasonable to hypothesize that the lack of the main ROS production site, namely CI, suppresses oxidative stress, whereas missense mutations that only reduce CI activity, such as the m.3460G>A/MT-ND1, may enhance the generation of ROS due to the electron flow slowdown (48). Hence, the contribution of ROS to HIF1α stabilization and tumor progression cannot be excluded for mtDNA mutations inducing mild functional alterations of CI, although such signaling seems to be insufficient to further promote the tumorigenic potential of highly aggressive osteosarcoma.
Lastly, transcriptional and post-translational regulation of HIF1α may also depend on other pathways, in particular on the LKB1/AMPK/mTORC1 axis, although the molecular mechanism is far from being elucidated. It is well known that the AMPK complex is activated by increased AMP/ATP ratio, acting as a metabolic checkpoint that redirects cells from anabolic to catabolic metabolism (49). Interestingly, we here report that CI-deficient tumor cells display an elevated AMP/ATP ratio and AMPK activation that, although to some extent observable under high glucose condition, were particularly highlighted under mild metabolic stress. AMPK has long been suspected to regulate tumor progression, since it is placed downstream and upstream of two well-known tumor suppressors, respectively, LKB1 and TSC2 (50–53). However, only recently its role as negative regulator of tumor progression has been proven (54). Faubert et al. (54) proposed a molecular mechanism in which AMPK functions as the coordinator of glycolytic and oxidative metabolism in proliferating cells by negatively regulating HIF1α transcription and stabilization and by triggering the Warburg effect even in normoxia via TCA cycle metabolites alteration. Such pseudohypoxic condition has been demonstrated in hereditary leiomyomatosis renal cell carcinoma, familial paraganglioma and familial pheochromocytoma patients, in which imbalance of TCA intermediates, namely fumarate and SA, affects PHD activity leading to a chronic stabilization of HIF1α (55,56). Moreover, it has been shown in fumarate hydratase-deficient cells that AMPK can influence the PHDs activity and HIF1α stabilization by regulating cytosolic availability of iron, one of the essential cofactors of the hydroxylation reaction (57). In this complex crosstalk between HIF1α and AMPK, the findings that energetic impairment and AMPK activation occur in cell models bearing a severe CI defect and constitutive HIF1α destabilization lead us to hypothesize that AMPK may negatively regulate HIF1α stabilization and may concur to the inhibition of tumor growth. However, when AMPK is pharmacologically activated by AICAR, HIF1α status and the propensity to form colonies in vitro are similar to untreated cells. Conversely, colony formation is stimulated when HIF1α is chronically stabilized in severe mitochondrial mutants, revealing its major contribution to tumor progression.
Our results underline the importance of mitochondrial energy production and hypoxic adaptation in tumor progression, highlighting the link between mtDNA mutations, mitochondrial dysfunction and tumorigenesis. Altogether, our data suggest that mtDNA mutations may be classified into two functional groups with respect to their oncojanus effect on tumorigenesis (Fig. 6); (i) neutral [(16) and present study] and/or pro-tumorigenic mutations (9,10,12,16) that permit and/or promote tumor growth and (ii) mutations that hamper tumorigenesis. Pro-tumorigenic mtDNA mutations have been shown to increase cellular proliferation by elevating oxidative stress, pyruvate and lactate levels and by inducing PI3K/Akt pathway activation (9,10,12,16). On the other hand, anti-tumorigenic mtDNA mutations have been demonstrated to negatively modify tumor growth by hindering cellular hypoxic adaptation (14), promoting apoptotic cell death (13) and introducing severe energetic defects, leading to AMPK phosphorylation, as here reported.
The complexity of mitochondrial metabolism and the peculiar mitochondrial genetics suggest that the mechanisms linking mtDNA mutations and tumorigenesis are numerous and many still need to be identified and understood. Our study highlights how the biochemical outcomes of mtDNA mutations must be carefully taken into account when drawing mtDNA genotype–phenotype correlations, especially when mtDNA mutations are used as diagnostic or prognostic markers in oncology.
MATERIALS AND METHODS
Cell cultures and growth conditions
Human osteosarcoma 143B.TK−-derived cybrid cell lines were generated and mutation loads were assessed as previously described (58–61). In particular, CC indicates a pool of wild-type cybrid clones, whose mitochondria derived from the same control fibroblasts, characterized by similar growth rate and OXPHOS efficiency. Cybrids were grown in DMEM high-glucose containing 10% fetal bovine serum (FBS), 2 mm L-glutamine, 100 units/ml penicillin and 100 μg/ml streptomycin (HG) at 37°C in an incubator with humidified atmosphere at 5% CO2. To induce a mild metabolic stress, cells were incubated in a DMEM without glucose supplemented with 10% FBS, 2 mm L-glutamine, 100 units/ml penicillin and 100 μg/ml streptomycin (LG).
Cell viability measurement
Cybrids were seeded in 24-well plates (3 × 104 cells/well) in high glucose medium, after 24 h cells were washed twice in PBS and incubated in different media. Galactose medium was composed by glucose-free DMEM supplemented with 5 mm galactose, 5 mm Na-pyruvate, 2 mm L-glutamine, 100 units/ml penicillin and 100 μg/ml streptomycin, and 10% FBS. Glucose-free medium (GFM) was glucose-free DMEM supplemented 5 mm Na-pyruvate, 2 mm L-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin and 10% FBS. Glutamine-free medium (QFM) was composed of DMEM-high glucose medium with 100 units/ml penicillin, 100 μg/ml streptomycin and 10% FBS. Cell viability was determined using the colorimetric sulforhodamine B assay, as previously detailed (62).
Oxygen consumption rate
OCR in adherent cells was measured with an XF24 Extracellular Flux Analyzer (Seahorse Bioscience, Billerica, MA, USA), as previously described (63). Briefly, cells were seeded in XF24 cell culture microplates (Seahorse Bioscience) at 3 × 104 cells/well in 200 μl of high glucose medium and incubated at 37°C in 5% CO2 for 24 h. Assays were initiated by replacing the growth medium in each well with 670 μl of unbuffered DMEM-high glucose pre-warmed at 37°C. Cells were incubated at 37°C for 30 min to allow temperature and pH equilibration. After an OCR baseline measurement, 70 μl of oligomycin, FCCP, rotenone and antimycin A were sequentially added to each well to reach final concentrations of 1 μg/ml oligomycin, 0.2 μm (CC and RJ206 cells) or 0.1 μm (HXTC1 and RN164 cells) FCCP, and 1μm rotenone and antimycin A. Data are expressed as picomoles of O2 per minute per 3 × 104 cells. At the end of each experiment, the medium was removed and cells were incubated with 0.1 μm calcein-AM in 200 μl/well of Hank's balanced salt solution supplemented with 10 mm Hepes and 1.6 μm cyclosporine H (CsH) for 30 min, at 37°C. Calcein-labeled adherent cells were observed with the Olympus IX71/IX51 inverted microscope (excitation and emission 495 and 515 nm, respectively, exposure time 100 ms, 6% illumination intensity) using a 10×, 1.3 NA oil immersion objective (Olympus).
In-gel and spectrophotometric CI activity
Isolation of mitochondria-enriched fractions was carried out as previously described (64). CI assembly and in-gel activity (CI-IGA) were determined after Blue-Native electrophoresis of isolated mitochondria as previously described (65). CI activity (NADH:DB:DCIP oxidoreductase) was also assessed in 50 mm KH2PO4 buffer (pH 7.6), 3.5 mg/ml BSA and 20 μg of mitochondrial protein in the presence of 60 μm 2,6-dichloroindophenol (DCIP, λ = 600 nm; ɛDCIP = 19.1 mm−1cm−1), 70 μm DB and 200 μm NADH, as previously described (64), after subtraction of 1 μm rotenone-insensitive activity. Data were normalized on citrate synthase activity and protein content, as previously described (63).
ATP synthesis assay
The mitochondrial ATP synthesis rate driven by CI and CII was performed in aliquots of digitonin-permeabilized cells and normalized on citrate synthase activity and protein content as previously described (63).
Mitochondrial membrane potential (Δψm) determination
Δψm was measured based on the accumulation of the fluorescent dye tetramethylrhodamine methyl ester (TMRM, Molecular Probes, Eugene, OR, USA) in the presence of 1.6 μm CsH, in resting cells and after the addition of oligomycin (Sigma-Aldrich, Milan, Italy) or FCCP (carbonylcyanide-p-trifluoromethoxyphenyl hydrazone, Sigma-Aldrich) as detailed elsewhere (66).
Glucose uptake and consumption
Glucose uptake was determined by using a fluorescent non-hydrolyzable glucose analog, 6-(N-(7-Nitrobenz-2-oxa-1,3-diazol-4-yl)amino)-6-Deoxyglucose (6-NBDG) (Life Technologies, Milan, Italy). Cells were seeded (3 × 105) onto glass-bottomed 36 mm diameter dishes (MatTek Corporation, Ashland, MA, USA) and grown in high glucose medium. After 24 h, cells were washed twice in PBS and incubated in GFM for 2 h at 37°C. Subsequently, cells were incubated in GFM supplemented with 30 µm 6-NBDG for 10 min at 37°C and visualized with a digital imaging system, using an inverted epifluorescence microscope with ×63/1.4 oil objective (Nikon Eclipse Ti-U; Nikon) at 488 nm. Images were captured with a back-illuminated Photometrics Cascade CCD camera system (Roper Scientific) and elaborated with Metamorph Acquisition/Analysis Software (Universal Imaging Corp.). Glucose consumption was determined by using the Glucose (GO) Assay Kit (Sigma-Aldrich) scaling down the manufacturer protocol to a final volume of 1 ml. Briefly, cells were seeded in 6-well plates (1 × 105 cells/well) in high glucose medium. After 48 h, cells were washed in PBS and incubated with DMEM-high glucose without phenol red. Aliquots of 100 µl of medium were taken at time 0 and after 24 h of incubation and 3 µl of samples were used to determine the glucose concentration, by the enzymatic reaction of glucose oxidase coupled to peroxidase-mediated oxidation of reduced o-Dianisidine (λ = 540 nm, 30 min, 37°C). The data obtained were normalized on cell number.
Cybrids were seeded in 6-well plates (3 × 105 cells/well) in high glucose medium. After 24 h, aliquots of medium were collected and deproteinated with 6% perchloric acid (PCA), vortexed and incubated in ice for 1 min. Samples were centrifuged at 13 000 rpm at 4°C for 2 min and the lactate concentration in supernatants was determined by measuring the NADH (λ = 340 nm; ɛ = 6.22mm−1 cm−1) production in a buffer containing 320 mm glycine, 320 mm hydrazine, 2.4 mm NAD+ and 2 units/ml L-lactic dehydrogenase (LDH, Sigma-Aldrich) after 30 min of reaction at 37°C.
Adenine nucleotides measurements
Aliquots of 1.5 × 106 cells were washed, resuspended in 1 ml of ice-cold PBS and extracted for ATP, ADP and AMP determination. Briefly, cell suspension was treated with 1 m PCA, immediately cooled on ice and centrifuged at 4°C to remove insoluble material. PCA was neutralized with potassium hydroxide and samples were centrifuged immediately before injection. The supernatant (100 µl) was injected on C18 column. Adenosine nucleotides (ATP, ADP and AMP) were extracted and detected as described by Jones (67) on a Kinetex reversed phase C18 column (250 × 4.6 mm, 5μm; Phenomenex, CA, USA), with a two pump Waters 510 system equipped with a variable volume injector. Absorbance at 260 nm was monitored by a photodiode array detector (Waters 996). ATP, ADP and AMP peaks were identified by comparison of their retention times with those of standards and confirmed by co-elution with standards. The quantification was obtained by peak area measurement compared with standard curves and data were normalized for protein content, determined by Bradford assay (68).
Proteins from total lysates or nuclear fraction from cultured cells (80 μg) were separated by 8 or 10% SDS–PAGE and transferred onto a nitrocellulose membrane that was incubated with antibodies against AMPK (1:1000, Cell Signaling Technology, Danvers, MA, USA), AMPKThr172 (1:1000, Cell Signaling Technology), ACC (1:1000, Cell Signaling Technology), ACCSer79(1:1000, Cell Signaling Technology), ACTB (1:500, Santa Cruz Biotech, Santa Cruz, CA, USA) and HIF1α (1:500, Bethyl Laboratories, Montgomery, TX, USA). Primary antibodies were visualized using horseradish peroxidase-conjugated proper secondary antibodies (1:2000). Chemiluminescence signals were measured with Kodak Gel Logic imaging system (Kodak, Rochester, NY, USA). Densitometric analyses were performed with the Image J ver. 1.46r software (69).
Soft agar assay
Anchorage-independent cell growth was determined in 0.33% agarose with a 0.5% agarose underlayer. Cell suspensions (3 × 104 cells) were plated in semisolid medium in the absence or presence of 600 µm 5-amino-imidazole-4-carboxamide 1-β-d-ribofuranosyde (AICAR) or 1 µm dimethyloxalylglycine (DMOG) and colonies were counted after 14–20 days. Images taken at maximum magnification using Kodak Gel Logic imaging system (Kodak).
In vivo studies
Cells (3 × 106) were suspended in 0.2 ml sterile PBS and injected subcutaneously in 4- to 7-week-old athymic Crl:CD-1-Foxn1nu/nu mice (hereafter referred to as nude mice, purchased from Charles River, Italy). All mice were kept under sterile conditions. Experiments were authorized by the institutional review board of the University of Bologna and done according to Italian and European guidelines. Individually tagged virgin female mice (at least 10 per experimental group) were used. Tumor growth was assessed by measuring with a caliper and recorded every 7 days from injection; tumor volume was calculated as π[√(a*b)]3/6, where a is the maximal tumor diameter and b the tumor diameter perpendicular to a. At sacrifice, tumors were dissected and analyzed.
Quantification of mitochondrial superoxide production was performed by flow cytometry. Briefly, cell clones were incubated with 2.5 μm MitoSOX™ (Life Technologies) for 30 min at 37°C. Subsequently, cells were harvested, washed in PBS and resuspended in DMEM High Glucose without phenol red. Flow cytometric analyses were performed in triplicate for each sample with a Coulter Epics XL-MCL (Beckman Coulter) equipped with an argon ion laser at λ = 488 and 620 nm. Data were analyzed with WinMDI ver. 2.9 software (http://facs.scripps.edu/software.html). Moreover, mitochondrial superoxide production was determined by fluorescent microscopy as previously described (14).
To determine the H2O2 production, cells were incubated with 2 μm 2,7-dichlorodihydrofluorescein diacetate (H2DCFDA) (Life Technologies) as detailed elsewhere (70). Fluorescence was detected at 535 nm using a multilabel counter Victor3 (Perkin Elmer, Turku, Finland).
α-KG and SA measurements
The α-KG and SA cellular levels were measured as previously described (65).
Statistical significance was defined as a P < 0.05 with one-way analysis of variance test unless otherwise indicated.
This work was supported by Associazione Italiana Ricerca sul Cancro—AIRC grant IG8810; by Ministero dell'Istruzione, dell'Università e della Ricerca—MIUR grant FIRB Futuro in Ricerca ‘TRANSMIT’ to G.G. This work was also supported by grant DISCO TRIP from the Fondazione Umberto Veronesi, Milan, Italy, to G.G.; L.I. is supported by an annual fellowship ‘Young Investigator Programme 2012–2013’ from the Fondazione Umberto Veronesi, Milan, Italy. I.K. is supported by a triennial fellowship ‘Borromeo’ from Associazione Italiana Ricerca sul Cancro—AIRC.
We are much grateful to Professor Eric A. Schon, Columbia University, New York City, NY, USA, for the RN164 cell line, to Dr Andrea Martinuzzi, IRCCS ‘E. Medea’, Conegliano Veneto, Italy, for the generation of cybrid cell lines and to Professor Paolo Bernardi, University of Padua, Italy, for critical reading of the manuscript. We thank Dr Manuela Voltattorni for technical assistance with flow cytometry experiments.
Conflict of Interest statement. None declared.