Heterogeneity of the human metabolic network in human tumors

Late last month, the paper Heterogeneity of tumor-induced gene expression changes in the human metabolic network was published in Nature Biotechnology. In 1924, Otto Heinrich Warburg theorized that tumor cells might not derive their energy source via the usual aerobic respiratory pathway but adopt the anaerobic pathway instead. Since then, this theory has gained much support, especially for explaining the survival of tumor cells in the dense hypoxic cores of neoplasias. Regardless however, cancer metabolism has been largely ignored -- until recently. In this paper, the group of Hu, J. et al. sought to look (at the individual and systems level) for similarities or difference in the expression profiles of metabolic genes in multiple tumor types.

In brief summary, the authors analyzed a compendium of gene expression profiles collected over the past decade from microarray studies of 22 different tumor types. They used only data from the most comprehensive microarray platform to-date (HG U133 Plus 2.0) in order to capture expression profiles from the most human genes possible. Furthermore, they used data derived only from studies which used tissue from the biopsies of primary tumors.

They reported several interesting findings including:

1. The identification of several key pathways in the glycolytic cycle which are overall upregulated or downregulated across all or most of the tumor types.
2. A significant amount of heterogeneity in gene expression patterns across tumor types
3. The rewiring of the gene expression program for several key respiratory pathways

Overall, they statistically showed clear examples of metabolic pathways/genes that clearly change expression patterns when comparing tumor to normal tissue. Of these examples, some appear to be conserved across different tumor types, some appear to be dependent on cancer type. But overall, they show that the metabolic network of human tumors do go haywire and really does warrant further attention by the cancer research community.

Relation to the Warburg effect?

The authors do not make clear how their results support or refute the claim of the Warburg effect. They show that the metabolic network changes going from a normal to a tumor tissue but how these changes, together, play out is still unclear. The goal of network biology is to put biological context onto the wiring details of a biological system but the functional understanding appears to be sparse.

One hypothesis could stem from the authors' finding that, at the individual gene level, the expression program of TCA cycle components appears to have mostly changed in colon cancers. This example might hold the clues to suggest that perhaps this pathway imbalance somehow "forces" the cell to adapt anaerobic respiration. Further biochemical studies must be undertaken to address the biological question surrounding this mechanism.

Unavoidable bias.

The authors were unable to perform paired-analyses (paired based on tumor origin in terms of the patient) at the single gene level due small sample size. Furthermore, it is unclear how heterogeneous the biopsy samples from each tissue type were and the age of the patients. These are clearly unavoidable pitfalls of the data but it emphases the need for more numbers in order to better understand what is happening. There are statistical methods to attempt to "correct" for these biases but the gold-standard is usually to correct for these biases at the experimental, data-collecting stage.

Comparison to yeast.

Interestingly, the yeast metabolic network also changes during growth, depending on the availability of glucose - shifting from aerobic respiration during the early phase of growth to anaerobic during the late phase (hence the eventual production of ethanol). Since yeast are much easier to work with in the lab, it would be interesting to compare the yeast versus human cancer phenomenons. The only simple (yet major) caveat to such comparison would be that yeast were "built" within their biology to readily switch respiration pathways. The underlying difficulty with human tumors are the multitude of various mutations that must first occur. These mutations vary widely from tumor to tumor and it is currently tough to differentiate the driver from passenger mutations. This might make it difficult to make generalized conclusions about such comparisons.

Final remarks.

The paper definitely reveals insightful findings, although it is not clear how much "WOW factor" there is to the general conclusion that the metabolic network is rewired in cancer. It definitely provides some clues for where scientists can and should focus for future work.  But, as with most computational work, much experimental work will be required to better understand the biology.