Cancer progression as a learning process

Abstract

Drug resistance and metastasis-the major complications in cancer-both entail adaptation of cancer cells to stress, whether a drug or a lethal new environment. Intriguingly, these adaptive processes share similar features that cannot be explained by a pure Darwinian scheme, including dormancy, increased heterogeneity, and stress-induced plasticity. Here, we propose that learning theory offers a framework to explain these features and may shed light on these two intricate processes. In this framework, learning is performed at the single-cell level, by stress-driven exploratory trial-and-error. Such a process is not contingent on pre-existing pathways but on a random search for a state that diminishes the stress. We review underlying mechanisms that may support this search, and show by using a learning model that such exploratory learning is feasible in a high-dimensional system as the cell. At the population level, we view the tissue as a network of exploring agents that communicate, restraining cancer formation in health. In this view, disease results from the breakdown of homeostasis between cellular exploratory drive and tissue homeostasis.

Keywords: Cancer systems biology; Evolutionary theories.

Graphical abstract

Figure 1

The fingerprints of exploratory adaptation in drug resistance and metastasis (A) Dormancy: stressed cancer cells often enter a non-proliferative state that might be crucial for them to contrive adaptive states. (B) Induced heterogeneity: despite the selective pressure of the treatment or the secondary organ, resistant cells and metastases exhibit high heterogeneity. This is concordant with exploratory adaptation as it yields multiple solutions to the same problem. (C) Stress-induced stemness: the transition to a stem cell state is enhanced when cancer cells are exposed to stress such as a new environment. This transition provides cells with transient high plasticity that enables them to search for a new adaptive state. Created with BioRender.com

Figure 2

Molecular mechanisms supporting trial-and-error learning (A) Chromatin remodeling: (top) Stress induced by the secondary organ or the drug can reshape the chromatin landscape, making it more permissive. (bottom) This endows the cells with higher plasticity to explore alternative states by lowering the barriers of transition between them. (B) Regulatory network plasticity: (left) Regulatory networks are dynamic and transcription factors can even take contradictory roles depending on context. Arrows represent activating interactions and caps represent inhibitory interactions. (right) Intrinsically disordered proteins can confer plasticity by alternating between conformations that enable them to interact with different proteins. Created with BioRender.com.

Figure 3

Modeling exploratory adaptation (A) A low-dimensional example of exploratory adaptation is chemotaxis, where bacteria move toward a higher gradient of an attractant (blue gradient) by a biased random walk. (B) When exposed to stress, a system will search for a new state that relaxes the stress through a trial-and-error exploration. When the system reaches such a state, exploration stops. The time that takes to reach this state is the convergence time (T_conv). (C) Convergence time increases dramatically with the number of dimensions. In the case of a genetic network, each gene is a dimension. This makes it hard for large networks performing exploratory adaptation to converge. Nevertheless, networks that harbor outgoing hubs—nodes with disproportionately high number of outgoing connections—can converge in a plausible time. (D) Cells exposed to stress perform exploratory adaptation in which the connections between the genes are modified by a random walk. The strength of the random walk is dictated by the stress $S\left(y-y^{\ast }\right)$, where $\left(y-y^{\ast }\right)$ is the distance between the low dimensional state of the cell, y, and the constraint $y^{\ast }$. Outgoing hubs (big purple nodes) enable exploratory adaptation to converge. Created with BioRender.com