Stochastic gene expression is the driving force of cancer

I agree with the analysis made by Kunihiko Kaneko, in the article recently published in your journal [1]. Dr. Kaneko points out the role of cell-cell interactions in cellular differentiation, and the importance of their disruption in the origin of cancers. He especially highlights that a decrease in stochastic variability of gene expression might be caused by the cellular interactions established during cell differentiation, and inversely that disruption of cell-cell interactions could be the primary cause of cancer accompanied by a global increase of stochastic gene expression. I reached similar conclusions in an article also published in your journal in 2005, although from a slightly different perspective [2]. Indeed, from a series of experimental data, theoretical analyses and computer simulation works, it has been known for some time that normal cells can be canalized along differentiation pathways by a process combining stochastic expression of their differentiation features and selection (or stabilization) of the adequate differentiation features by cell-cell interactions between neighbouring cells. Thus, differentiation is associated with a global decrease in the stochastic variability of gene expression mediated by signalling between cells and epigenetic modifications of chromatin [3]. Recently, strong experimental evidence has been obtained supporting the key points of this model. Stem cells exhibit highly stochastic gene expression profiles [4] and it is demonstrated that cellcell interactions stabilize stochastic gene expression in embryonic and adult cells [5, 6]. Based on this model and on a series of studies concerning carcinogenesis itself, I proposed that when cell-cell interactions are disrupted, destabilization of gene expression in differentiated cells (corresponding to dedifferentiation) or failure of canalization of progenitor cells that are intrinsically unstable (corresponding to non-differentiation) generates cells that exhibit confused and unstable gene expression profiles. In those cases, stochasticity in gene expression is no longer ‘controlled’, and differentiation no longer stabilized or canalized by the microenvironment. I further suggested that genetic and epigenetic instabilities in cancer cells could be the consequence of this unstabilized stochastic gene expression. Indeed, stochastic expression of DNA repair and maintenance genes in cancer cells could prevent efficient DNA repair leading to the accumulation of genetic alterations. In this perspective, those genetic alterations are not the primary cause of cancer but a consequence of the destabilization of gene expression following the disruption of cell-cell interactions, subsequently acting as ‘accelerators’ necessary to reach full progression of cancer. Dr. Kaneko’s modelling works serve to confirm my own analysis. Indeed, he explains that, ‘because the normal developmental course is a result of cell-cell interaction in the theory presented (. . .), the appearance of cancertype cells is also expected to depend on cell-cell interaction’. Moreover, he highlights that ‘genetic instability in cancer cells would not be a cause but a result of phenotypic instability’. This theoretical support for the conclusions I drew from biological observations emphasizes the need to re-evaluate the way in which cancer cells appear. Cancer has to be understood as originating from a global imbalance in gene expression profiles in many cells, linked to disruption of cell-cell interactions generating dedifferentiation or non-differentiation. If it is not stabilized by micro-environmental cues, the stochastic nature of gene expression may generate the necessary variable gene expression pattern, and constitute the primary driving force in cancer. This perspective is the only one able to take into account the fact that (i) alteration of cell-cell interactions is able alone to induce carcinogenesis [7], (ii) restoration of cellular interactions with ‘normal’ cells normalizes cancer cells in spite of genetic alterations [8], (iii) no specific genetic alterations can be unequivocally associated with particular forms of cancer, and (iv) loss of differentiation, understood here as global increase of stochastic variability in gene expression, is the fundamental hallmark of cancer cells. Finally, it is important to point out a crucial aspect ofmodels considering disruption of tissue equilibrium as the primary origin of cancer [1–3]. In these models, cell differentiation and cell proliferation are not a matter of induction but stabilization and inhibition, respectively. Undifferentiated (or dedifferentiated) cells are intrinsically unstable and proliferation is their default state [9]. This ‘instability postulate’ [10] disagrees fundamentally with classical theories of DOI 10.1002/bies.201100092