Horizontal transfer of whole mitochondria restores tumorigenic potential in mitochondrial DNA-deficient cancer cells

1Mitochondria are vital organelles of eukaryotic cells responsible for energy production and other key biochemical functions. Many human diseases, including cancer, are characterised by mutations in mitochondrial DNA (mtDNA). This is linked to altered mitochondrial bioenergetics, such that cancer cells are more glycolytic than their non-malignant counterparts, as postulated almost a century ago by Warburg. Altered energy metabolism is now regarded as one of the hallmarks of cancer. In this context, the mitochondrial genome has been reported to play a role in tumorigenesis and in metastatic cancer.

2Although many cancer cells are biased towards the glycolytic metabolism, they also need oxidative phosphorylation (OXPHOS) for their ‘pathophysiological’ requirements. Recent research has demonstrated a new concept of cancer metabolism, emphasising the importance of OXPHOS in the tumour environment beyond its role in bioenergetics. A novel paradigm has emerged, according to which respiration is important for cancer cell proliferation, and also for tumour formation, progression and metastasis. This is probably linked to metabolic re-modelling. There are reports that some cancer cells retain OXPHOS capacity and have no obvious respiratory defects. Further, inhibiting glycolysis may restore higher rates of OXPHOS in neoplastic cells. Hence, depriving cancer cells of their capacity to respire may preclude them from forming tumours.

3We recently reported on the importance of respiration in tumour formation and progression. We showed that cancer cells without mtDNA (ρ0 cells) form tumours after a considerable delay compared to their parental counterparts. Tumour progression was associated with mtDNA acquisition from the host, resulting in respiration recovery. While these findings point to a new phenomenon of horizontal transfer of mtDNA between mammalian cells in vivo, direct evidence for the role of mitochondrial respiration in tumour formation as well as understanding the mode of mtDNA acquisition is lacking.

4Here we provide a link between efficient tumour formation and recovery of mitochondrial respiration, and show that mtDNA acquisition occurs via trafficking of whole mitochondria.

5We have shown that B16ρ0 cells injected subcutaneously into C57BL/6J mice formed syngeneic tumours with a 2–3-week delay compared to B16 cells, and next generation sequencing (NGS) indicated the host origin of mtDNA. Since NGS would not detect heteroplasmy of less than about 3%, a much more sensitive single cell/digital droplet (sc/dd) PCR method was used in the current study to document that the homoplasmic polymorphism at the tRNAArg locus of mtDNA of cell lines isolated from tumours grown subcutaneously from B16ρ0 cells (B16ρ0SC cells) is of the host origin. The assay is able to detect heteroplasmy down to 0.5%, demonstrating with very high confidence that the mtDNA in B16ρ0SC cells is of host origin, and that original B16 polymorphism is either completely absent or present below the detection limit of 0.5% of mtDNA (Figure 1A; see also Appendix 1—figure 1 for validation of sc/ddPCR).

6We next analysed the properties of B16, B16ρ0 and B16ρ0SC cells, as well as a B16ρ0CTC sub-line derived from circulating tumor cells and a B16ρ0SCL sub-line derived from lung metastases. Appendix 1–figure 2A documents confocal microscopy analysis of mtDNA in mitochondria, showing that B16ρ0 cells lack mtDNA, whereas mtDNA appears homogeneously distributed in mitochondria in all other sub-lines. Super-resolution stimulated emission depletion (STED) microscopy exerted similar levels and distribution of mtDNA nucleoids in B16, B16ρ0SC, B16ρ0CTC and B16ρ0SCL cells, and no nucleoids in B16ρ0 cells, also showing largely unchanged levels of Tom20 and low level of TFAM (Figure 1B; see also Appendix 1—figure 2E). Native blue gel electrophoresis (NBGE) revealed that B16ρ0 cells do not contain the supercomplex/respirasome (formed by CI, CIII and CIV), but contain low amounts of sub-CV (Figure 1C). CII was found to be fully assembled in all sub-lines, which is reasonable considering that all four subunits in CII are encoded by nuclear DNA (nDNA) (Figure 1C). To test if cells replicate their mtDNA, we established the mitochondrial chromatin immunoprecipitation (mitoChIP) assay. This showed a high level of DNA polymerase-γ1 (POLG1) binding to the D-LOOP region of mtDNA in all cells except B16ρ0 cells (Figure 1D). B16ρ0SC, B16ρ0CTC and B16ρ0SCL cells showed similar respiration to B16 cells, but no respiration was observed with B16ρ0 cells (Figure 1E,F). Accordingly, B16ρ0 cells produced more lactate (Figure 1G) and less ATP (Figure 1H). B16ρ0 cells also had lower succinate dehydrogenase (SDH) (Figure 1I) and succinate quinone reductase (SQR) (Figure 1J) activity, as well as lower citrate synthase (CS) activity (Figure 1K). Finally, we observed higher glucose uptake in B16ρ0 and B16ρ0SC cells, and lower uptake in B16ρ0SCL cells (Figure 1L). Collectively, these results document that mitochondrial function is already fully restored in cells derived from the primary tumour.

7We next analysed cells for their mtDNA levels and expression of selected transcripts. Appendix 1—figure 2B shows no mtDNA in B16ρ0 cells, while mtDNA was present at similar levels in other sub-lines. No mtDNA-encoded transcripts were present in B16ρ0 cells. Their levels were low in B16ρ0SC cells, while higher levels were seen for most transcripts in B16ρ0CTC and in B16ρ0SCL cells. Transcripts of the assembly factor SCAFI were present at similar levels in all sub-lines, but TFAM transcripts were low in B16ρ0 cells relative to the other cells. Transcripts of nDNA genes coding for subunits of respiratory complexes were found to be present in all cells, with some being lower in B16ρ0 cells. WB revealed that most proteins investigated were relatively abundant in all sub-lines with LC3AII levels lower in B16ρ0 cells, indicating stalled autophagy (Appendix 1—figure 2C). Interestingly, although present in B16ρ0 cells, many nDNA-encoded mitochondrial proteins, including subunits of RCs, were unstable in these cells, as evidenced using cycloheximide treatment (Appendix 1—figure 2D). Exceptions were SDHA and ATPβ; a plausible reason is the absence of binding partners encoded by mtDNA, which could render the ‘unassembled nDNA-encoded’ subunits unstable. In summary, these results indicate that, in contrast to 4T1 cells, the respiratory function of B16 sublines is already fully recovered at the primary tumour stage.

8Given the rapid recovery of respiratory function in the B16 model, we next tested tumour-forming capacity of B16 and B16ρ0 cells, and of the sub-lines B16ρ0SC, B16ρ0CTC and B16ρ0SCL, derived from various tumour stages as described above. We found that all sub-lines formed tumours without delay except for B16ρ0 cells with >2 week delay (Figure 2A). Sectioning of tumours derived from all five sub-lines revealed melanomas with necrotic cells away from blood vessels (Figure 2B). Tumours derived from the sub-lines were assessed by NBGE, identifying full assembly of all complexes as well as the respirasome, with somewhat higher levels of the respirasome in tumours derived from B16ρ0 cells (Figure 2C). WB revealed similar levels of expression of most proteins tested in the five types of tumours (Figure 2D). No obvious differences in the expression of genes coding for transcripts of individual subunits of mitochondrial complexes encoded by mtDNA and nDNA were observed (Figure 2E). Finally, we tested respiration of tumour tissues derived from the sub-lines. Figure 2F reveals increased CI-dependent respiration and the maximum electron transfer capacity (ETS) in tumours derived from B16ρ0 and from B16ρ0SCL cells, and no change in CII-dependent respiration. Normal liver tissue from the same animals used as internal control showed no changes in respiration. The restoration of the tumorigenic potential fully correlates with the observed recovery of respiration.

9We have shown for B16 cells (Figures 1 and 2, Appendix 1—figure 2) and previously for 4T1 cells that tumour formation from their respective ρ0 variants correlates with the acquisition of mtDNA and recovery of respiration. However, direct evidence for the requirement of respiration for tumour formation has been missing. We therefore prepared B16 and B16ρ0 cells with suppressed levels of NDUFV1 (the catalytic subunit of CI) or SDHC (ubiquinone-binding CII subunit essential for its SQR activity) by RNA interference (RNAi) using two different shRNAs for each protein. Figure 3A shows that NDUFV1 shRNA#2 and SDHC shRNA#2 were rather efficient in knocking down the respective proteins. Figure 3A also reveals that NDUFV1 knock-down (KD) cells exhibit lower level of CI subunits (NDUFV1 and NDUFS3) while CII subunits (SDHA and SDHC) were unaffected. Conversely, SDHC KD cells showed low levels of CII subunits and also decreased levels of CI subunits. We used cells stably transfected with NDUFV1 shRNA#2 or SDHC shRNA#2 in subsequent experiments. Both NDUFV1 KD and SDHC KD cells proliferated at a slower rate than parental cells (Figure 3B). NDUFV1 KD as well as SDHC KD cells showed lower routine respiration and lower ETS (Figure 3C). NDUFV1 KD cells respired less via CI, while CII-dependent respiration was largely unaffected; SDHC KD cells not only showed much lower CII-dependent respiration, but also significantly suppressed CI-dependent respiration (Figure 3D).

10We next grafted B16, B16ρ0 and B16 cells, as well as their derived NDUFV1 KD and SDHC KD cells into C57BL/6J mice. Figure 3E shows that tumours started to grow from B16ρ0 cells shortly after day 20 post-grafting of B16ρ0 cells in all mice. 4 out of 6 mice grafted with B16ρ0 SDHC KD cells formed tumours with delays of 15 to 40 days compared to B16ρ0 cells. Only one out of 5 mice of the B16ρ0 NDUFV1 KD group formed a tumour, with a lag of about 40 days compared to B16ρ0 cells. A similar pattern was observed for B16 cells and their NDUFV1 KD and SDHC KD variants, though the lag for these sub-lines was considerably shorter (Figure 3F). On pathological examination of the different tumours, relatively subtle differences in morphology were observed (Figure 3G). B16 melanoma cells form tumours in control mice (NS) showed prominent nuclear pleomorphism. They appeared more aggressive than cells from ρ0 tumours. Tumours derived from both B16 NDUFV1 KD and B16 SDHC KD cells showed better histological differentiation associated with eosinophilic cytoplasm, vesicular nuclei and frequent deposits of melanin pigments, indicating the KD cells are less aggressive than their parental counterpart. There were no obvious differences in morphology of the 3 types of ρ0 melanoma cells. Both B16ρ0 NDUFV1 KD and B16ρ0 SDHC KD tumours showed focal areas of nuclear pleomorphism and frequent mitotic features.

11These data clearly point to respiration recovery as essential for driving efficient tumour formation, since suppression of respiration completely deregulated this process. Interestingly, while with CII suppression the majority of mice formed tumours (albeit with additional lag time), most mice grafted with CI-compromised cells failed to form tumours within 100 days. In summary, we show that respiration is important for efficient tumour formation, which is consistent with recent reports.

12We have previously documented the host origin of mtDNA in cancer cells isolated from primary tumours derived from 4T1ρ0 and B16ρ0 cells based on NGS analysis, which we confirmed here for B16ρ0SC cells using the more sensitive sc/dd PCR assay (Figure 1A). The only plausible explanation for this phenomenon is the transfer of mtDNA from host cells to tumour cells with compromised mtDNA. The question of how mtDNA moves between cells has not been addressed. An attractive scenario is that whole mitochondria with their payload of mtDNA are transferred, but alternative explanations such as cell fusion have also been suggested. Mesenchymal stem cells (MSCs) have been proposed as donors of mitochondria due to their high levels of Miro-1, which is the adaptor protein responsible for mitochondrial association with the microtubule mobility complex. We performed co-culture of MSCs isolated from C57BL/6J mouse bone marrow with either B16 or B16ρ0 cells. Prior to co-culture, the mitochondria of the potential donor (MSC or B16) cells were labelled with mitochondrial dark red fluorophore (MitoDR). Carboxyfluorescein succinimidyl ester (CFSE) that binds covalently to amino acid residues inside cells was used for staining the potential recipient (B16 or B16ρ0) cells. Following co-culture, cells were assessed by confocal microscopy to determine the extent of mitochondrial transfer to recipient cells. Representative images and quantitative evaluation (Figure 4A, Appendix 1—figure 3A,B) reveal accumulation of MitoDR stained mitochondria in CFSE stained ρ0 cells, as a result of their transfer from MitoDR stained MSCs. In similar experiments, pairs of either CFSE+B16 / MitoDR+MSC or CFSE+B16ρ0/MitoDR+B16 cells did not show any evidence of MitoDR stained mitochondria in CFSE positive cells, thereby excluding dye diffusion. Collectively, this indicates that transfer of mitochondria between cells in a regulated process.

13We next conducted a co-culture experiment, in which MSCs were isolated from transgenic C57BL/6Nsu9-DsRed2 mice with red fluorescent mitochondria in somatic cells. These cells were cultured with ρ0 cells transfected with a plasmid coding for nuclear targeted blue fluorescence protein (nBFP) and plasma membrane green fluorescence protein (pmGFP). Appendix 1—figure 3C documents transfer of DsRed mitochondria from MSCs into the recipient ρ0 cells via an intracellular bridge.

14To investigate the origin of mtDNA and the manner of its transfer between host and cancer cells in vivo, we used transgenic C57BL/6Nsu9-DsRed2 mice with red fluorescent mitochondria in somatic cells. B16ρ0 cells transfected with a plasmid coding for nuclear-targeted blue fluorescent protein (nBFP) were injected subcutaneously into C57BL/6Nsu9-DsRed2 mice. Several days later, mice were sacrificed, the pre-tumour lesion excised, and single cell suspension sorted for double-positive (DP) cells with both red and blue fluorescence that on average were found with the frequency of 0.23 ± 0.18 in the BFP-positive population. Immediately after sorting, B16ρ0 DP cells were plated and inspected by confocal microscopy within about 12 hr before red fluorescence in mitochondria has been lost. Figure 4B shows an image of a B16ρ0 DP cell prepared from a day 11 pre-tumour lesion, identifying mouse stromal cells as a source of mitochondria that moved into a grafted B16ρ0 cell. The sorted DP cells were established as a sub-line. B16, B16ρ0, and B16ρ0 DP cells were subjected to transmission electron microscopy and STED microscopy to show the presence of fully formed mitochondrial cristae (Figure 4C) and mtDNA nucleoids in DP cells (Figure 4D). DP cells also showed strong binding of POLG1 to the D-LOOP region of mtDNA (Figure 4E) as well as recovery of respiration (Figure 4F), and showed a propensity to form tumours without a lag phase (Figure 4G). These results document that mtDNA is transferred from stromal cells to B16ρ0 cells within intact mitochondria, resulting in the restoration of respiration and in efficient tumour formation.

15Horizontal gene transfer is a process that until recently had not been known in mammals, but has been described for lower eukaryotes, affecting their phenotype. In mammals, the mitochondrial gene transfer has been inferred in a 10,000 year-old canine transmissible venereal tumour (CTVT). These predictions have now been confirmed, and a recent study showed at least 5 mtDNA transfers in CTVT within the last 1–2 thousand years. Mitochondrial transfer was reported in vitro with functional consequences, as well as in mice with endogenously injected MSCs, but these results could also be explained by the association of membrane-bound particles or exosomes containing mitochondria with the damaged cells rather than functional mitochondria proliferating/dividing inside cells. A recent publication from our laboratory showed, for the first time, horizontal transfer of mitochondrial genes between mammalian cells in vivo, based on acquisition of host mtDNA by ρ0 tumour cells and on the presence of host mtDNA markers in cell lines derived from these tumours. Several subsequently published papers also provide evidence for horizontal transfer of mitochondria under (patho)physiological conditions, pointing to OXPHOS as a factor complicating cancer therapy. Hence, our data, along with those of other groups, point to mitochondrial transfer as a highly dynamic field of research, with important implications for the conceptual understanding of cancer.

16Our recent findings provoke a number of questions, such as whether respiration is essential for tumour formation, and what is the mode of mtDNA acquisition. To address these queries, B16 metastatic melanoma cells without mtDNA and with compromised respiratory function were used. We show that B16ρ0 cells do not form tumours unless they acquire mtDNA and that severe suppression of either CI- or CII-dependent respiration leads to impaired ability to form tumours, directly linking mitochondrial respiration and tumor growth. The reason(s) for the apparently greater importance of CI-dependent respiration in tumour formation (Figure 3E,F) remain to be explored, particularly when the role of SDH in metabolic re-modelling is taken into consideration. From a mechanistic point of view, CI is essential for respirasome assembly that is needed for CI-dependent respiration. Another indication of the importance of CI is the finding that tumours derived from B16ρ0 cells feature much higher CI-dependent respiration (Figure 2F).

17We document here that the primary tumour-derived B16ρ0SC cells feature a fully assembled respirasome and complete recovery of respiration, while corresponding 4T1ρ0SC cells show only partial (20–25%) recovery of respiration. Consistent with this, 4T1ρ0SC cells formed tumours with a longer delay than parental 4T1 cells, while no delay was observed for B16ρ0SC cells. This is likely related to the higher requirement of B16ρ0 cells for respiration recovery to form tumours. This ‘threshold’, apparently lower for 4T1ρ0 cells, may be due to higher routine respiration of B16 cells of some 100 pmol O2/s/106 cells (Figure 1E) when compared to less than 20 pmol O2/s/106 cells for 4T1 cells. Additionally, the different genetic changes in B16 melanoma and 4T1 breast carcinoma cells could underpin altered respiration recovery. Whether there is a link to a requirement for ‘threshold’ recovery of respiration for cancer cells to initiate tumour formation has yet to be determined. We propose the term ‘OXPHOS addiction’ to describe the requirement for mitochondrial respiration across the landscape of tumours, and will investigate this in more detail in future research.

18A key question has been the mode of movement of mtDNA between cells. Both selective transfer of mitochondria and cell fusion have been proposed in the past. Using C57BL/6Nsu9-DsRed2 mice with red fluorescent mitochondria in somatic cells, we now provide evidence for acquisition of mtDNA by the trafficking of whole mitochondria from host donor cells to ρ0 cells both in vivo (Figure 4B), resulting in long-lasting respiration recovery and, consequently, efficient tumour formation (Figure 4E–G). The transient nature of dsRed expression in recipient B16ρ0 cell mitochondria in the in vivo model sheds additional light on the mechanism of mitochondrial transfer. When dsRed-containing mitochondria are selectively transferred from the host into the recipient cells, dsRed cannot be replenished in the donated mitochondria by de novo synthesis (not being encoded in the recipient’s nuclear genome), and the red fluorescence in donated mitochondria is quickly lost. On the other hand, if cell fusion were responsible for the observed acquisition of mitochondria from the donor cells, the donor’s nuclear material would also be transferred, and mitochondrial dsRed expression would be maintained. As this is not the case, the selective transfer of whole, intact mitochondria remains the only possible explanation of our experimental data. The next step will be to investigate the mechanism of mitochondrial trafficking between cells in vivo, with tunnelling nanotubes being a plausible mode of intercellular transfer of the organelles.

19We conclude that recovery of respiration in tumor cells with damaged mtDNA is essential for efficient tumour formation and that this is accomplished by the intercellular transfer of whole mitochondria. Our findings are consistent with the emerging notion of the essential role of respiration in cancer cell proliferation and tumor progression, and of a role for mitochondrial transfer in maintaining the bioenergetics balance. From translational angle, recent studies on horizontal mitochondrial transfer indicate two tantalising, novel approaches to cancer therapy: targeting mitochondrial respiration and blocking transfer of mitochondria from stromal cells to cancer cells. While we have started exploring the first approach, the other approach remains untested, with recent papers pointing to its plausibility. Finally, to the best of our knowledge, this paper is the first report to show lasting functional consequences of a well-documented mitochondrial transfer event.

20Cell lines were prepared and maintained as described. C57BL/6J mice were used as host animals for grafting B16 sub-lines as indicated. B16 lines formed syngeneic tumours in C57BL/6J mice indicating their authenticity. Parental and ρ0 cells with BFP nuclei were prepared by stable transfection with the pTagBFP-H2B plasmid (Evrogen) followed by clonal selection. RNAi was used to knock down NDUFV1 or SDHC subunits, using two different shRNAs (OriGene) for each protein. In brief, cells were transfected with shRNA using a standard protocol and inspected by WB for protein levels. Cells with more efficient knock-down of NDUFV1 or SDHC were used in further experiments.

21Mouse MSCs were isolated from C57BL/6J mice as described earlier. The primary cells were plated at the density of 106 cells/ml in T25 culture flasks and experiments were performed after the fourth passage. For co-culture experiments, MSCs were stained with Mito Deep Red (Ex/Em, 644/665 nm; Invitrogen) and B16ρ0 cells with the CFSE dye (Ex/Em, 492/517 nm; Invitrogen) for 15 min. The cells were then co-cultured for 24 hr and evaluated by flow cytometry (FACS Calibur) and confocal microscopy (63 x; Leica SP8). Mitochondrial transfer was primarily ascertained by a fraction of CFSE-positive cells that were also MitoDeepRed-positive. The quantitative mitochondrial transfer was primarily ascertained by pixel counts on Z-stack confocal images i.e. red pixels in cells with green background. Geometric mean intensity of MitoDeerRed fluorescence in double positive cells was additionally calculated, as a confirmatory measurement of the degree of mitochondrial transfer. Other co-culture combinations (MSCs with B16 cells, B16 cells with B16ρ0 cells) were performed in a similar fashion. Additionally, ρ0 cells were transfected with a plasmid coding for nBFP and pmGFP followed by clonal selection. Transfected ρ0 cells were seeded with MSCs prepared from C57BL/6Nsu9DsRed2 mice in glass-bottom dishes (In Vitro Scientific) at 1:1 ratio and co-cultured for 24 hr. Live cells were inspected using the inverted fluorescence microscope Delta Vision Core with laser photo-manipulation. The acquired images were deconvolved by Huygens Professional software (Scientific Volume Imaging) and processed by FiJi ImageJ software.

22C57BL/6J mice were used for most of the experiments. They were purchased from the Animal Resources Centre or produced by the animal breeding facilities of the Institute of Biotechnology and Malaghan Institute. In all cases, the mice were grafted subcutaneously with various cell lines at 5 × 105 cells per animal. Tumours were monitored by ultrasound imaging (USI) using the Vevo770 system (VisualSonics, Toronto, Canada). Transgenic mice expressing red fluorescent protein in somatic cell mitochondria (the CAG/su9-DsRed2 transgene) were generated in the Transgenic Unit of the Czech Centre for Phenogenomics, Institute of Molecular Genetics, Prague, Czech Republic, using a pronuclear injection from the construct provided by Prof. Masaru Okabe (Osaka University, Japan) and C57BL/6N mice. The stable colony of transgenic mice was housed in the animal facility of the Faculty of Science, Charles University, Prague, Czech Republic, and food and water were supplied ad libitum. The mice used for the grafting experiments were healthy 10 weeks old animals with no sign of stress or discomfort. All animal procedures and experimental protocols were approved by the Animal Welfare Committee of the Czech Academy of Sciences (Animal Ethics Number 18/2015).

23Details of the methodology are in the Supplemental information.

24Details of the methodology are in the Supplemental information.

25Details of the methodology are in the Supplemental information.

26Unless stated otherwise, data are mean values ± S.D. of at least three independent experiments. In mouse experiments, groups of 6 animals were used, unless stated otherwise. The two-tailed unpaired Student’s t test was used to assess statistical significance with p<0.05 being regarded as significant. Images are representative of three independent experiments.