We performed a comprehensive analysis of sequential metabolic changes that occur in a panel of cell populations with loss of cell-cycle arrest or increased tumorigenic potential. These analyses gave us new insight in the relation between oncogenic transformation and the development of the typical Warburg phenotype. The direct relatedness of the different cell pools – all pools were derived from the same primary MEFs ground state cell population – gave us a unique opportunity to do comparative profiling and highlight the early-late order of adaptive metabolic changes in energy metabolism that occur during solid tumor development.
Two genetically independent immortalization procedures had only minor effects on cell size, morphology, and carbohydrate metabolism, and yielded metabolic phenotypes with strong similarity to the original Prim-MEF cells. In contrast, H-RasV12/E1A transformation caused a prominent change from flattened to spherical appearance and smaller volume (see also ). Similar changes have been reported before for other transformed cell types [23, 24].
One of the most conspicuous observations in our study was that the metabolic phenotype initially switched to aerobic, OXPHOS-based, immediately after transformation. The augmented OXPHOS rate in Ras-LP cells was associated with increased total cellular NAD(H) levels, an altered mitochondrial NAD+/NADH setpoint, increased oxygen consumption and increased superoxide production. We think that OXPHOS induction in early transformed cells may be a general concept that precedes the well-documented Warburg effect or shift to aerobic glycolysis. The initial switch to aerobic metabolism has also been reported after oncogene-mediated transformation of human mesenchymal stem cells , human epithelial cells [26, 27] and human breast cancer cells . At this stage it is, however difficult to draw direct comparison between our findings in mouse cells and findings in the different human systems, as panels with related human cells lines comparable to our panel do not exist or were generated via use of increasing numbers of oncogenes . Future work, involving analyses of metabolic fate specification in panels with human cells from mesenchymal, epithelial, or progenitor-stem cell origin, will therefore be needed to verify the prediction of high OXPHOS activity in early transformed cells.
Importantly, our results indicate that environmental factors can exert strong influence on viability of tumor cells during the early phase of the transformation process. Anchorage-independent growth capacity of Ras-LP cells increased under low oxygen conditions, presumably by a direct suppressive effect on the high intrinsic and "unbalanced" mitochondrial activity or simply by better coupling of glycolytic-mitochondrial pathways. For better explanation and discrimination between these possibilities a detailed comparison of relative mitochondrial and glycolytic pathway activity in cells in colonies formed under ambient oxygen or in hypoxia would be imperative. Unfortunately, such pathway analysis on relatively small numbers of cells in soft agar in situ is currently technically not possible, so we have to leave this analysis for future work. If we may extrapolate these findings to the in vivo situation, we would predict that nutrient levels in the environment, including oxygen levels, may determine the fate of early dysplastic cells in tissues. In the light of our results, we predict that low oxygen or hypoxia would enhance their capacity to grow as tumors, as was already proposed in the model of Gatenby and Gillies .
The later trajectory of oncogenic transformation displayed by our cell panel involved a gradual shift towards higher anaerobic glycolysis rates over 20 passages in vitro and passage through immune-deficient mice, accompanied with clear gain in survival ability. We demonstrate here that the pools of cells that progress towards a tumorigenic phenotype go through a distinct series of adaptational changes. Although this metabolic rewiring may also be controlled by specific mutations in regulatory genes that become dominant in the entire pool, we consider it more likely that cell-intrinsic metabolic signaling mechanisms, acting within the windows set by the culture conditions (i.e. nutrient and O2 supply) used in our experiments, drive the gradual shift from OXPHOS to glycolysis. We draw this conclusion because we observed a similar stepwise drift in metabolic phenotype in a parallel series of cell populations that were independently derived from MEFs from mice with a combined defect in phosphotransfer enzymes CK-B and AK1 . At this point it is important to note that this entire transition trajectory occurred in continuous presence of 21% environmental oxygen and under constant medium conditions. Strikingly, H-RasV12/E1A expression only marginally affected protein expression levels of glycolytic, TCA and OXPHOS enzymes at all stages after transformation, implying that altered flux rates dominate and underlie the observed metabolic effects. The shift in mitochondrial NAD+/NADH ratio that we observed is in support of this model, as this ratio sets the flux rate of key metabolic reactions.
Adaptive changes in mitochondrial function, with or without further genetic programming, may provide growth- and survival advantage by conferring resistance to apoptotic cell death . Cells with electron flow through the ETC and enhanced glycolytic flux may perform best, as these are properties associated with reduced ROS production and increased ROS scavenging ability, respectively. ROS induced damage is a well-known factor in cell viability control . Indeed, in the H-RasV12/E1A cells mitochondria are the primary ROS source, as we observed a direct relationship between oxygen consumption and superoxide production. Indirect support for this relationship was also provided by our finding on glucose-galactose shift effects in TBX2-MEF cells.
We would like to emphasize that mechanisms other than those based on mitochondria-derived ROS effects could also explain the apparent coupling between tumor progression and increased glycolytic rate. Firstly, the metabolic infrastructure of cells could have changed due to Ras-induced alterations in cell morphology and cytoarchitecture. Cell surface changes, cell clustering, attachment and movement are processes that are coupled to actin cytoskeleton rearrangements. Bereiter-Hahn and co-workers  showed that changes in energy metabolism may result from rearrangements of the F-actin network, and conversely, that energy demand of motile actin-based structures is supplied predominantly by ATP derived from glycolysis. At the same time, mitochondrial dynamics are determined by a complex interplay between the inner membrane-matrix complex and cytoskeletal elements outside, possibly via F-actin and dynamin-related protein DRP1 . Upregulation of glycolysis or alteration of mitochondrial properties may thus be directly coupled to Ras-induced cytoskeletal remodeling, and fulfill new needs for development of anchorage independent growth, or other essential prerequisites for tumorigenic progression.
As another speculative possibility, there may also be a connection between cell cycle progression and the balance between glycolytic or oxidative metabolism. H-RasV12 induces cell cycle progression via enhanced G1 to S phase transition , and also the E1A oncoprotein induces S-phase via induction of p300/CBP and c-Myc . Recently, it was reported that slowly growing yeast cells in a low glucose environment exhibit a periodic metabolic cycle that alternates between glycolysis and respiration, where the cell division cycle is constrained to the reductive non-respiratory phase of this cycle, with DNA replication taking place only during the glycolytic phase [36, 37]. If extrapolation of the yeast model to the situation in mammalian cells is valid, our findings may suggest that glycolysis is the default metabolic pathway adapted by Ras-transformed cells with enhanced proliferation capacity.
Finally, our findings can also be explained by early alterations in NAD+/NADH setpoint, acting as driver for subsequent changes in tumorigenic conversion. We consider this an interesting possibility because we observed changes in set point level of NAD(H) (NAD+ plus NADH) metabolites and upregulation of NAMPT protein levels as parallel immediate early effects of H-RasV12/E1A transformation in our RAS-LP-HP-TUM series. Van der Veer et al.  reported that overexpression of NAMPT results in increased life span and resistance to oxidative stress. Our data show that total cellular NAD(H) levels increase upon cellular transformation and may relate to the simultaneous upregulation of the NAD-salvaging enzyme NAMPT at the protein level. Evidence is now accumulating that NAD+ levels, and/or redox state of the NAD+/NADH couple in cytosol, nucleus and mitochondria are key factors in transcriptional regulation [39, 40], modulation of growth-differentiation decisions in dividing cells , and survival control . Moreover, the need for NAD+ or NADH as essential co-factors in regulation of metabolic flux rates is already long known. While we regard our work as a first proof-of-concept it is clear that more study into these regulatory circuits – including those in human cells of different tissue origin – is necessary to draw definite conclusions about the control over early and late events in the Warburg transition.