Background
Otto Warburg observed that many cancers lose their capacity for mitochondrial respiration, limiting ATP production to anaerobic glycolytic pathways [
1]. The phenomenon is particularly prevalent in aggressive malignancies, most of which are also hypoxic. Hypoxia induces a stochastic imbalance between the number of reduced mitochondrial species
vs. available oxygen, resulting in increased reactive oxygen species (ROS) whose toxicity can lead to apoptotic cell death. One mitochondrial adaptation to increased ROS is over-expression of uncoupling protein 2 (UCP2) which has been reported in a number of human cancer cell lines [
2‐
4]. Horimoto et al. [
3] demonstrated UCP2 over-expression in most of the 120 colon cancer lines tested, the extent correlating with the degree of tumor aggressiveness. Increased UCP2 expression was also associated with reduction in ATP production in malignant oxyphilic thyroid tumors [
2], and in mouse leukemia and human lymphoma cell lines [
4]. It is reasonable to ask whether one can take advantage of the ability of UCP2 to disrupt regulation of ATP generation in cancer as a therapeutic method.
One mechanism by which aggressive cancers that over-express UCP2 might inhibit glycolytic ATP is the so-called Randle cycle. In normal cells, the Randle cycle has been proposed as a mechanism by which increased availability of fatty acids and ketone bodies from lipolysis inhibit glycolysis in order to maintain stable production of ATP. Products of lipolysis, according to this mechanism, supply acetyl-CoA for the TCA cycle whose intermediates inhibit glycolysis [
5]. We propose that aggressive cancers, under conditions of ketosis, employ a Randle-like cycle to inhibit glycolytic ATP generation, but, in contrast to normal cells, cannot supply compensatory mitochondrial ATP due to uncoupling.
We have cultured three human primary fibroblast lines and seven human cancer cell lines in media of varying glucose and ketone body content to compare cell growth rates, ATP concentration, and expression of UCP2. Restriction of cancer cell lines to colon and breast was meant to parallel the kinds of patients we expect to be eligible for our ongoing clinical trial of very low carbohydrate diets in advanced cancer[
6]. (The trial is registered with ClinicalTrials.gov; Registry No. NCT00444054; with information available at
http://clinicaltrials.gov/ct2/show/NCT00444054).
Discussion
The uncoupling proteins (UCP's) constitute a structurally related family with variable ability to uncouple mitochondrial production of ATP [
7][
8,
9]. Uncoupling reduces the proton gradient across the inner mitochondrial membrane. UCP1, first discovered in rodents, is present in brown fat in infants but its abundance decreases strikingly in most people with age, along with the reduction of brown fat tissue. UCP2 has about 58% structural homology with UCP1. While UCP2 concentration is very low or absent in most normal tissues it has been found to be more highly expressed in aging vs. young fibroblasts [
10], in pancreatic islets in type 2 diabetes mellitus[
11], and in macrophages under inflammatory stimulation [
7,
12]. In all these cases its presence has been postulated to represent an adaptive reaction to increased reactive oxygen species (ROS). Controversy exists whether UCP2 causes significant mitochondrial uncoupling or ROS inhibition [
13,
14]. It is instructive, however, to note that tissues with the highest UCP2 expression, as above, represent those which generate high ROS; and that ROS mitigation could be explained by mitochondrial uncoupling, an effect observed only at high UCP2 concentrations. Cancers are among the tissues reported to express UCP2 in much higher concentrations than normal tissues [
2‐
4]. Our proposal starts with the assumption that UCP2 over-expression is an adaptive response which limits ROS in cancer lines by causing uncoupling. The proposal asks whether this assumption is borne out with consistent results under specific experimental conditions.
An increase in UCP2 has been reported to be associated with reduction of respiratory ATP production in several cancer lines [
2,
4] that have preserved TCA cycle enzymes. Horimoto et al [
3] showed that UCP2 over-expression is present in
most colon cancer lines, the degree of over-expression correlating with the histologic grade. It was proposed that UCP2 over-expression was an adaptive mechanism that reduced ROS-induced apoptosis among the cancer cells.
We suggest that ketone bodies, if metabolized in cancers of this type, can inhibit glycoloysis and resultant ATP generation without compensatory mitochondrial respiratory ATP production, thereby inhibiting cell growth.
The Randle cycle was proposed more than 30 years ago as a mechanism whereby products of lipolysis (fatty acids and ketone bodies) would inhibit glycolysis in normal cells to maintain stable ATP production [
5]. According to this mechanism, abundant acetyl CoA causes feedback inhibition of pyruvate dehydrogenase and increases the level of TCA intermediates, specifically citrate, to inhibit phosphofructokinase, both processes reducing the rate of glycolysis. The relevance to normal metabolism has been challenged by Wolfe et al [
15,
16]. These investigators showed convincingly that under most physiologic conditions carbohydrate-stimulated insulin secretion increases intracellular glucose availability. Increased intracellular glucose concentrations drive glycolysis, so that carbohydrate, via the dominating effect of insulin, becomes the regulator of lipid metabolism rather the other way around. Wolfe
et al. pointed out that the Randle cycle has only been persuasively demonstrated in
in vitro preparations under conditions of euglycemic, hyperinsulinemic clamp, in which intracellular glucose availability remains high and fixed. Under these circumstances, in rat myocardium preparations, for example, variable fatty acid availability does indeed regulate the rate of glycolysis.
Interestingly, high fixed intracellular glucose availability is a likely condition
in vivo for most aggressive cancer cells whose membranes express high affinity
insulin-independent glucose 1 transporters (GLUT1) [
17]. The Km for GLUT1, at 18 mg/dl, predicts maximum fixed glucose transport at ordinary physiologic glucose concentrations of 60–100 mg/dl. This contrasts with normal human tissues (e.g. myocytes, adipocytes, endothelial cells) which, capable of metabolizing lipids and carbohydrates, express insulin-dependent GLUT4. Our hypothesis suggests that metabolism of ketone bodies in cells over-expressing GLUT1 and UCP2 may provide the opportunity for acetyl COA and citrate to inhibit glycolysis, just as in the euglycemic-hyperinsulinemic Randle cycle. However, UCP2 over-expression differs from the Randle model in that uncoupling prevents full mitochondrial compensation of ATP production. We consider this an "inefficient" Randle cycle.
Evidence for such an inefficient Randle cycle is found in the seven cancer lines of our investigation which demonstrated parallel reductions in ATP concentration and cell growth, the greatest inhibition found in CaCO2 and SW48 lines, and the least in MCF7 (Figure
3B). In addition, UCP2 staining correlates inversely with ATP production in all cells lines (Figure
4,
5). The three control fibroblast cell lines demonstrated neither growth inhibition nor reduced ATP production, and showed the least quantitative staining for UCP2 among all cell lines measured. The findings extend the observations of previous investigators and point to a specific kind of mitochondrial abnormality, UCP2 over-expression that under conditions of ketosis is capable of reducing the growth rate of certain cancers. We believe that the findings are also capable of reconciling apparently contradictory results in previous animal models. Tisdale
et al. reported MAC16 tumor growth inhibition in mice on ketogenic diets [
18] but failure of ketosis to inhibit Walker carcinosarcoma [
19]. The latter tumor lacks the succinyl-CoA ketotransferase required to metabolize ketone bodies. Metabolism of ketone bodies is required by the present hypothesis to enable inhibition of glycolysis via the TCA cycle, and therefore absence of this enzyme would predictably fail to inhibit cancer growth.
There are limitations in the present study. The investigation does not prove metabolic inhibition of cancer growth and ATP production due to uncoupling. Proof requires direct demonstrations of acetoacetate metabolism as well as inhibition of the rate of ATP production, not just of steady state ATP concentrations. Second, glucose and acetoacetate were both incubated at supraphysiologic concentrations to maximize detectable effects. It is not possible without further
in vivo study to predict effects under physiologic conditions. Third,
in vitro culture of primary cell lines is difficult for normal human cell types, and only fibroblasts were used as controls, so the response of other normal tissue types was not measured. Further, fibroblast lines capable of growth in cell culture are not normal; however, neither are they the equivalent to malignant cell lines and their difference from malignant lines was demonstrated in this investigation. Also, the range of tumor susceptibility to acetoacetate varies widely, plausibly due to increased levels of UCP2 expression, but other factors must be considered. For example SW48, with the greatest reduction of ATP and cell growth in acetoacetate medium, was a prominent outlier with the least UCP2 expression among the cancer cell lines (Figure
5). Finally, we did not manipulate UCP2 expression using knock-down and knock-in techniques. Single protein manipulations, while very important, cannot be expected to reliably predict phenotypic changes in light of growing awareness of complex feedback regulation of gene-protein networks. We plan to explore these manipulations in subsequent investigation, until now having focused on phenotypic behaviors of seven different cancer lines in contrast to controls.
Cancers are notorious for their ability to evade metabolic regulation [
20] and therapeutic use of metabolic inhibitors has been avoided until recently. Specifically targeted pharmacologic inhibitors such as dichloroacetate [
21] and 2-deoxyglucose [
22,
23], the latter originally proposed as a metabolic inhibitor in 1960 [
24], have shown promise in the inhibition of cancer growth, although toxicity to normal cells has been a limitation. It is worthwhile to ask whether less toxic metabolic regulation of cancer cells is possible in light of current thinking about cancer evolution, development and progression [
25,
20]. We have previously outlined a general hypothesis bearing on this [
6] and proposed, along with others [
26] that human evolutionary adaptations to starvation render normal cells tolerant of flexible nutrient sources, including ketone bodies and fatty acids.
On the other hand the initiation of a primary event transforming a normal cell into a pre-malignant phenotype is followed by a long period of subsequent adaptive cellular response to a local hypoxic, nutrient-deprived microenvironment. During this interval of abnormal cellular growth multiple adaptations eventually result in the cancer phenotype[
27]. A typical response to the hypoxic environment, for example, is expression of insulin-independent GLUT1's [
25,
28] triggered by HIF 1α [
29] insuring maximum glucose uptake for glycolytic ATP generation. Solid cancers require at least 30 to 35 generational doublings before they attain a clinically detectable size of 1–2 cm diameter (assuming average mammalian cell diameter of 10 microns) [
30] over durations of 3 to 15 years. The abundance and low cost of grain characteristic of efficient modern agriculture has made sustained ketosis an unfamiliar internal metabolic environment for many people living in developed countries [
6]. It follows that in large cohorts of individuals, sustained ketosis is not a familiar microenvironment for newly evolving abnormal cells, on the path to malignancy. Nevertheless, our normal tissues remain adapted to these microenvironments as shown by the absence of toxicity in experimental fasting of morbidly obese individuals in the 1960's[
31]. In human cancers in the developed world, and in the absence of sustained ketosis as a selective pressure over durations of 30 or more doubling times, a diversity of evolutionary responses should then be expected including accidental vulnerabilities and survival adaptations. One response that would correspond to an accidental vulnerability to ketosis would be over-expression of UCP2 with intact TCA enzymes.
Competing interests
The authors have no conflicting or competing interests to declare in the conception, evolution, or execution of this research.
Authors' contributions
EJF conceived the study, directed its primary design and coordination, performed much of the data analysis, and drafted and revised the manuscript; AM grew and counted the cell cultures, performed the ATP assays, performed substantial data analysis and participated in writing the Methods section of the manuscript; EVQ coordinated execution of the study, contributed to important aspects of the research direction, and made important revisions to the manuscript; JMS performed the UCP2 assay and wrote the methods for this aspect of the work; RDF coordinated execution of the study, contributed to important aspects of the research direction, and made important revisions to the manuscript. All authors read and approved the final manuscript.