George T. Chen, Mary Lee, Kehui Wang, Robert A. Edwards, John S. Lowengrub, Marian L. Waterman

(This is an expanded abstract for a poster presented at the 2014 AACR Conference)

Colon cancer is the third leading cause of cancer deaths in males and second in females globally. Though improved detection and treatment have gradually decreased the incidence of cases in the United States, approximately 51,000 Americans die from colon cancer annually. Although hereditary risk factors for colon cancer exist, the majority of cases are sporadic, and strongly influenced by environmental factors such as diet and drug use.  At the cellular level, a key mutation in colon stem cells targets APC – a critical protein in the “beta-catenin destruction complex.” The destruction complex phosphorylates beta-catenin for degradation by the proteosome.  Physiologic, normal Wnt signaling inhibits the destruction, allowing beta-catenin to translocate to the nucleus and activate transcription of genes associated with growth and development. This pathway is referred to as the canonical or beta-catenin dependent Wnt signaling pathway. When APC is mutated, the destruction complex cannot form, and thus the canonical Wnt signaling pathway is constitutively active whether or not Wnt ligands are present.

In most aerobic organisms, the oxidative phosphorylation metabolic pathway is the predominant mechanism through which cellular energy is produced. This process, compared to anaerobic metabolic pathways, is much more efficient at producing ATP; however, it is not the metabolic pathway favored by colon cancer. Despite an abundance of environmental oxygen and the induction of angiogenesis by a tumor, the predominant metabolic pathway utilized is glycolysis. This apparent paradox is also known as Warburg metabolism.

Previous work in our lab and others has demonstrated a role for Wnt signaling in regulating Warburg metabolism. In mouse xenograft tumors grown from SW480 and SW620 colon cancer cells, inhibition of beta-catenin dependent Wnt signaling leads to a shift in the metabolism of the tumor, where glycolysis markers decrease and oxidative phosphorylation becomes more dominant.  Metabolism was measured by fluorescence lifetime imaging microscopy (FLIM), which utilizes the autofluorescent properties of molecules and proteins to quantify their fluorescent decay rates. In our studies we used NADH FLIM signatures of glycolysis and oxidative phosphorylation as well as conventional measures such as lactate production.  For all measures, inhibition of Wnt signaling in colon xenograft tumors reduced lactate levels, and reduced the NADH signature for glycolysis.  Taken together we observe a strong link between glycolysis and Wnt signaling.

We also determined that transcription of the pyruvate dehydrogenase kinase-1 gene (PDK1), is regulated by Wnt signaling. PDK1 protein normally inactivates mitochondrial pyruvate dehydrogenase (PDH) through phosphorylation in order to increase the conversion of pyruvate to lactate in the cytosol. In addition, decreasing Wnt signaling activity also reduced tumor vessel density and tumor volume.  PDK1 rescue expression restored the vessel network and tumor perfusion. We concluded that Wnt signaling directs Warburg metabolism in colon cancer via regulation of a key regulator of glycolysis.

When staining for phosphorylated PDH in the xenograft tumors as a measure of active PDK, we discovered a unique spotted pattern of active PDK1 within the tumor. This pattern was revealed as discrete regions of increased phospho-PDH at regular intervals throughout the xenograft tumor. The pattern was accentuated when Wnt signaling was reduced and it was completely abolished with PDK1 expression. This regular spotted pattern was also seen when the Wnt target gene LEF1 was assessed for expression, suggesting that Wnt signaling and other signaling molecules may be responsible for establishing the spotted pattern. We hypothesized that this patterning could be modeled as a Turing pattern. Beta-catenin dependent Wnt signaling and its associated inhibitors have been previously characterized to form Turing patterns, or reaction-diffusion systems, in multiple developmental biology systems.

Our collaborative group has developed a system of reaction-diffusion equations that describes the formation of these spots in relation to varying concentrations of Wnt signaling, Wnt inhibitors, and nutrients. Included in the model are equations for glycolytic cells, oxidative cells, Wnt signaling activity, Wnt inhibitors, PDK activity, lactate, HIF, and a general nutrient term. Wnt activity and Wnt inhibitor equations are based on the Gierer-Meinhardt activator-inhibitor model. We modeled Wnt activity and HIF so that they directly drive PDK activity, which in turn drives lactate production, and HIF stabilization. The tumor cells switch metabolic regimes based on PDK activity level (high activity implies a tendency towards glycolysis, and low activity tends toward oxidative phosphorylation).

Despite containing virtually identical cells, as SW480 and SW620 have extremely limited differentiation potential, a number of key differences can be observed among the cells within the xenograft. In addition to spotted pattern of high glycolysis surrounded by low glycolytic activity (as measured by phospho-PDH), a gradient of the pattern exists such that the strongest, most dense manifestation of the pattern exists in the outer region of the tumor encompassing the border, gradually decreasing in frequency of pattern and intensity of staining towards the interior of the tumor. We hypothesize that the gradient of glycolysis is mirrored in markers for Wnt signaling. Indeed, even though both SW480 and SW620 cells contain high levels of beta-catenin in every cell, immunohistological staining shows heterogeneity in beta-catenin localization in regions throughout the xenograft tumors.

Since the invading border of the tumor contains the highest levels of glycolysis, we have begun to characterize the roles that the microenvironment surrounding the tumor might play in influencing tumor metabolism. Despite the constitutively active Wnt signaling pathway, a number of factors in the environment, including “non-canonical” or “beta-catenin independent” Wnt ligands are still capable of modulating the beta-catenin dependent Wnt pathway as well as triggering non-canonical modes of Wnt signaling. We also hypothesize that specific extracellular matrix proteins secreted by tumor cells and normal fibroblasts moderate Wnt signaling, and its ability to influence the development of vasculature in xenograft tumors to show high vessel density near high glycolytic regions.

Overall, our collaborative group has defined a new function for Wnt signaling in colon cancer. Using novel imaging techniques and mathematical modeling, we have demonstrated that beta-catenin dependent Wnt signaling regulates expression of PDK1 to drive glycolysis in xenograft tumors. Furthermore, this increased glycolysis exists in a regular Turing pattern throughout the tumor. Our mathematical models will allow us to predict changes to tumor metabolism and behavior in response to modulation of Wnt signaling or external stimuli. Greater understanding of the relation between Wnt signaling and metabolism may lead to new therapeutic targets for colon cancer.