Dear Editor,
The question we address here is what drives cancer to grow in an accelerated fashion as it evolves. Various proposals have been made regarding the possible drivers of cancer growth such as driver mutations and autonomous growth signaling. While these are clearly relevant, they rely too much on specific types of genomic mutations or molecular abnormalities by chance across different cancer types, which makes the probability for cancer to occur/progress significantly lower than what we have witnessed about the current cancer occurrence rates worldwide, hence making them less probable to be the ultimate drivers of cancer growth (Loeb, 1998).
We present here a model for the (accelerated) growth of a cancer based on the discovered gene-expression patterns derived from genome-scale transcriptomic data of seven solid carcinoma types, namely breast, kidney, liver, lung, ovary, pancreatic, and stomach cancers. Our data analysis clearly indicates that as a cancer advances, (i) its percentage of cells in the G0 phase of the cell cycle tends to become increasingly lower, indicating accelerated cell proliferation; (ii) when the hypoxia level goes up, the activity level of oxidative phosphorylation as the main energy (ATP) producer goes down and that of glycolysis goes up, which triggers cancer cells to accelerate the uptake of glucose from the blood circulation to make up for the lost efficiency in energy production, needed for them to stay viable; (iii) this switch in energy metabolisms leads to accelerated cell proliferation and further increased hypoxia, forming a vicious cycle of (accelerated) growth of cancer; (iv) this cycle breaks down when the new angiogenesis takes place triggered by the high hypoxia level, which decreases the hypoxia level and switches back to oxidative phosphorylation as the main energy producer and continues until the cells become too hypoxic again; and (v) the cellular hypoxia level goes up and down ‘periodically’ that coincides with the increasing cancer mass and new angiogenesis, respectively, while its overall trend is going up.
Hypoxia has long been linked to cancer growth, including its link to the activation of glycolysis and angiogenesis, de-acidification of cancer cells, tumor promotion, and the emergency DNA repair mechanism. The relationship between hypoxia and the (gradual) switch in energy metabolisms from oxidative phosphorylation to glycolysis has been widely studied and well established (Harris, 2002). However, there have not been any published studies aimed to explain how hypoxia drives the growth and accelerated growth of a cancer, to the best of our knowledge.
Glycolysis has been a focus of cancer study since 1920s when Warburg (1956) discovered that glycolysis is active in cancer cells even in the presence of oxygen. The current understanding about this is that aerobic glycolysis provides a powerful way to generate carbon skeletons for biosynthesis of amino acids and lipids needed for rapid cell proliferation, and it is substantially faster in ATP generation than oxidative phosphorylation when glucose is abundant (DeBerardinis et al., 2007; Koppenol et al., 2011).
The key contribution made in this study is that our model represents the first one for explaining what possibly drives cancer growth/accelerated growth and how, which is supported by genome-scale transcriptomic data of seven solid cancer types. A key internal driver, that we found, is that the reduced energy efficiency (caused by hypoxia and/or possibly other factors) triggers increased uptake and accumulation of glucose, leading to cell proliferation. Our model does not require genomic mutations but it allows the growing process of a cancer to positively select mutations that may give the cells a competitive edge, such as multiplications of oncogenes and deleterious mutations of tumor suppressor genes.
Previous studies have suggested that cancer tends to grow increasingly faster as it evolves (Hanahan and Weinberg, 2011). To check whether this is purely due to the exponential growth with a stable cell-division rate or compounded with accelerated cell division, we have examined whether the cell division of a cancer is accelerated or not as the hypoxia level increases. We first estimated the relative cell cycle rate and its change versus cancer stages by checking the percentage of cells in a cell population (e.g. a tumor sample) in the G0 phase, using a marker gene VRK1 for phase transition from G0 to G1 (Valbuena et al., 2008), as well as the cyclin genes that are involved in the subsequent phase transitions of cell cycle. From Figure 1A, we can clearly see overall up-going trends of the expression levels of these cell cycle genes with high statistical significance as a cancer evolves for each cancer type, indicating increased cell cycle rates in all the seven cancers.
A cancer growth model based on gene expression analysis. (A) Expression-level changes of cell-cycle genes versus stages 0 (normal), 1, 2, 3, 4 for breast, kidney, liver, lung, ovary, pancreatic, stomach cancer (in this order). The y-axis represents the relative expression level (normalized and log-transformed) of cell cycle genes VRK1, CCND1, CCNE1, CCNA2 and CCNB2 averaged over the available samples for each cancer group (Supplementary Table S2). (B) The expression level of cell-cycle genes versus HIF1A level, with a higher HIF1A level indicating a higher hypoxia level. (C) A switch from oxidative phosphorylation (OxPhos) to glycolysis. (D) The expression levels of the glucose transporter gene, GLUT (blue), lactate transporter gene, MCT (red) and an estimated inhibition level (green) of the expression of OxPhos based on the expression of PDK1 (i.e. xy = −1 with x being the slope of the green line and y being the slope of the linear approximation of the PDK1 across stages) versus the HIF1A level. CC represents the Pearson correlation coefficients, along with the P-value. Detailed expression data are shown in Supplementary Figure S1. (E) The expression levels of PDK1 (green) and VRK (red) versus the expression level of heme oxygenase-1 (HMOX1), a responder to oxygen with a higher HMOX1 level representing a higher oxygen level. (F) The expression level of VEGF versus the HIF1A level, with a higher VEGF level reflecting a higher angiogenesis level. Each blue diamond represents one patient's data of the two relevant genes, and each black line represents the best linear regression model for the underlying data. (G) A driver model for accelerated cell division.
We have also checked the change of these expression levels versus the change of the hypoxia level. Here we used the expression level of the hypoxia marker gene HIF1A to reflect the level of hypoxia. All the seven cancers show an increasing trend of the expression level of VRK1 as the hypoxia level increases, indicating accelerated cell division as the hypoxia level goes up (Figure 1B).
A natural question is what drives cancer to accelerate its cell division as it becomes more hypoxic. To address this question, we have examined how hypoxia affects the activity levels of oxidative phosphorylation and glycolysis using the gene-expression levels of two rate-limiting enzymes, PFKL and HK1, in the glycolysis pathway and the expression levels of two inhibitory factors PDK1 and PDK3 of oxidative phosphorylation as a way to measure the expression level of the two pathways. Figure 1C shows down-going and up-going trends of the expression levels of the oxidative phosphorylation and glycolysis, respectively, when the cell becomes more hypoxic. It is known that oxidative phosphorylation (plus glycolysis and TCA cycle) is at least 18-fold more efficient than glycolysis in their glucose-to-ATP conversion. Hence, a key to answering the above question lies in understanding of how cancer cells make up for the lost energy production efficiency as well as the total lost energy production when the activity level of oxidative phosphorylation goes down. The ‘make up’ is essential for the survival of a cell since a human cell typically requires 1 × 107 to 1 × 108 ATPs at any time for executing its basic cellular functions. Hence, we hypothesize that cancer cells make up for the lost energy efficiency and total energy production through increasing the glycolysis-based energy production, which requires accelerated accumulation of glucose.
To check whether this hypothesis holds, we have examined publicly available transcriptomic data of seven cancers. Our data analysis revealed that the expression levels of glucose transporters (GLUTs) and the lactate transporter genes (MCTs) all have an up-going trend as the cancer becomes more hypoxic (Figure 1D), indicating an increasing need for glucose as well as an increasing quantity of lactate produced, which is also consistent with the imaging data that cancer cell may have 10–100-fold increase in glucose than their healthy counterparts. These data provide direct evidence supporting the above hypothesis. While the cells need to uptake more glucose to produce sufficient level of ATPs needed for them to remain viable, the accumulated cellular glucose will trigger cell division (Singh et al., 1999; Vander Heiden et al., 2009).
The above analysis revealed strong statistical correlations (i) between increasing hypoxia and reducing level of oxidative phosphorylation and the increasing level of glycolysis and (ii) between increasing hypoxia and increasing uptake of glucose. While these statistical correlations alone do not provide causality relationship directly, we believe that the most logical explanation of these observations is that hypoxia drives up and down of the activities of oxidative phosphorylation and glycolysis, respectively; this triggers cells to uptake more glucose to make up for the lost energy efficiency, which in turn triggers increased cell proliferation. Clearly the accelerated cell proliferation will further increase the level of hypoxia due to the increased cell accumulation, potentially forming a vicious cycle. A key missing piece is that the glycolysis may not generate enough ATPs needed to support the rate of cell division comparable with the accumulated glucose, which will ultimately come from new angiogenesis.
Intuitively, we would expect that cancer cells would become increasingly hypoxic as the cancer mass grows increasingly larger. However, our data analysis shows that while the general trend is true, cancer cells have a wide range of hypoxia levels at each developmental stage (Supplementary Figure S2). One most plausible explanation is that the cancer has new angiogenesis taking place triggered by the high hypoxia level, which leads to reduced hypoxia level locally and temporarily. We note that among the highly hypoxic tissue samples across the seven cancer types, there is a negative correlation between the increase of the oxygen level due to new angiogenesis (revealed by change in the level of heme oxygenase 1) and the rate of cell division as well as the inhibition level of the oxidative phosphorylation pathway in general (Figure 1E). This suggests that as a cancer evolves, the rate of cell division fluctuates resulted from the combination of the changing hypoxia level and new angiogenesis. This is also supported by Figure 1A, which shows fluctuating expression levels of the cell cycle genes as a cancer evolves for all seven cancer types, and Figure 1F, which shows the level of angiogenesis goes down as hypoxia goes down. As the hypoxia level goes down due to new angiogenesis, cancer cells will switch back to oxidative phosphorylation.
In summary, as a cancer evolves, the hypoxia level has an overall up-going trend, which drives cell division to go faster, but the cell division rate may go down from time to time due to new angiogenesis. During such a downwards period, cell division slows down and the cells prepare for material and energy needed for the next rapid growth phase. Overall the driving force of cancer growth comes from the combination of two factors: hypoxia and the miscoupling between the increased uptake of nutrients triggered by reduced energy efficiency and cell proliferation signaling induced by increased accumulation of nutrients (Figure 1G). Hypoxia might have made the further transition from a fast growing cell to cancer cells possible since it can induce a number of cancer hallmark capabilities such as activation of emergency DNA repair mechanism, which leads to increased DNA mutations, inhibition of apoptosis and activation of angiogenesis. We have noted from the public gene expression data sets (Poola et al., 2005) that precancerous tissues could be highly hypoxic (data not shown), making it possible for hypoxia to be an early key driver of cancer. Recent studies have linked a number of factors such as long-term inflammation (Eltzschig and Carmeliet, 2011) that may cause cellular hypoxia, which fits well with our model here.
We believe that this model, while coarse in nature, captures the essence of the key driving force of the accelerated cell cycle, as well as the growth patterns as a cancer progresses. The actual growth rate and pattern of a specific cancer may also depend on other factors such as the cancer location and activity levels of apoptosis and immune responses. This model will provide a useful framework for experimental studies of cancer cells, as well as for building predictive models for cell growth that will take into consideration other contributing factors.
[Supplementary material is available at Journal of Molecular Cell Biology online. We thank Dr Yanbin Yin, Profs Dave Puett and Dan DerVartanian of UGA and Prof. Luis Liu of GSU for their suggestions. This work was supported in part by the National Institutes of Health (No. 1R01GM075331), a ‘Distinguished Scholar’ grant from the Georgia Cancer Coalition, NSFC (No. 60903097) and seed funding from the University of Georgia.]
References
Author notes
These authors contributed equally to this work.
