Gene regulatory systems

In 1957, Conrad Waddington presented his well-known epigenetic landscape of cel-lular differentiation, pictured in figure 2.3 [23]. The landscape is a metaphor of gene regulation-mediated differentiation of the cells of an organism, in which a ball rol-ling down a grooved hill represents the cell as it maturates. At the top of the hill, the ball represents is a pluripotent stem cell, but as it rolls down the landscape, it ends up in a groove (developmental lineage) and further sub-grooves (sub-lineages),

Kuva 2.3: Waddington’s epigenetic landscape. The ball rolling down the grooved hill repre-sents a differentiating cell. The valleys at the root of the hill correspond to distinct cell types. [23]

finally stopping at the root of the hill in a valley representing a specific, fully ma-ture cell type. With the word epigenetic, Waddington refers to all gene-regulatory activity defining the phenotype as opposed to the actual genetic information, or DNA-sequence, which is assumed to be stay unaltered during differentiation. Later, epigenetics has been defined more strictly as heritable modifications of a chromoso-me not affecting the DNA sequence. For convenience, however, we will stick here to Waddington’s broader definition of the term. [24]

Besides being an illustrative metaphor of the complex system giving rise to multiple, distinct cell types in spite of identical hereditary material, Waddinton’s landscape model does not defy exact formulation. One of the first models of epige-nesis, filling the causative gap between DNA and the diverse range of distinct cell types it enables, was proposed in 1967 by Stuart Kauffman. Introducing random Boolean networks (RBN) as discrete, dynamical models of gene regulatory networks (GRN), Kauffman suggested that different cell types are attractors, or stable states, of the network. The concept of an attractor is central in the theory of dynamical systems. In a dynamical gene network it corresponds to a gene expression state to which the system tends to even if slightly disturbed. A sufficiently large perturba-tion, however, may cause the system to drift to another attractor. In Waddington’s landscape, the attractors are the valleys and an attractor-switching perturbation would be one which pushes the ball over the hill separating two valleys. [26]

Perturbations of the GRN may be due to impulses from the environment as well as the inherent stochasticity of inter- and intracellular regulation. During the process of differentiation, randomness plays an important role in defining cellular fate [25].

However, once the cell is terminally differentiated, robustness to disturbances is de-sirable to maintain the current state, or gene expression profile. Therefore, GRNs

Regulatory gene Non-regulatory gene Induction

Repression

E D A

B

C

Kuva 2.4: A gene regulatory network. Genes A and B induce other genes, or increase their expression. C, on the other hand, is a repressor: it decreases the expression of genes, including itself. By self-repression, C regulates its own expression level. A regulates its own expression level, too, but indirectly by inducing a gene (B) which induces its repressor (C).

Genes A, B and C form the regulatory core of this GRN while D and E merely express the effects of their regulators.

wired such that their attractors are sufficiently stable enjoy a selective advantage. Es-sential for any self-regulative system are negative feedback loops. Negative feedback ensures that changes in the state of the system are moderated by adversary effects.

Gene regulatory-wise, this means that GRNs must contain self-repression. [28]

Figure 2.4 shows a small GRN. Three of its five genes have regulatory functions:

they induce or repress other genes (or themselves). The regulatory genes form the core of the GRN as changes in their expression affects the expression state of the whole system. Non-regulatory genes have no effect on the expression profile outside their own expression. In a real-life GRN, the regulatory core composes of TFs and other genes with regulatory effects. Two types of control circuits are presented in figure 2.4: direct and indirect self-repression. Both contribute to the stability of the expression profile. [1]

Any change in the wiring of a GRN may change its attractor. Biologically, a mutation affecting any regulatory connection of the GRN is such a change. As the system changes, so does its epigenetic landscape. Thus, cancer can be seen as a re-wired GRN which leads to new attractors being formed, corresponding to gene expression states enabling limitless replication. Figure 2.5 illustrates the case. The normal maturation process is halted by a new attractor which retains the replicative potential of the cell. In leukemia, for instance, such an attractor-state is the cause of blast crisis. This model, thus, places tumorigenesis in the context of cellular differentiation, only with an altered epigenetic landscape. [28]

So far, modeling a genome-wide GRN has been a daunting task. Uncovering all the regulatory connections and quantifying their degree of inhibition or repression, not to mention their combinatorial effects, requires a great amount of biological research.

Stem cell

Blast cell

Mature cell

A

Cancerous blast

B

Kuva 2.5: Cellular maturation. In cancer, normal differentiation (A) is halted by rewiring the gene regulatory network, forming a new, cancerous attractor in the gene expression space (B).

Therefore, systemic properties of whole-genome regulatory models have been studied with GRNs making very strong simplifying assumptions, such as binary expression states and random regulatory effects in RBNs. Another approach is to restrict the GRN to a small sub-network and attempt to model its regulatory connections in detail using differential equations. This, of course, enables analyzing only a fraction of the functionality of the entire system. [27]

2.7 DNA microarrays