Developmental Networks

Most eukaryotic development begins with a fertilized egg, which undergoes a series of cell divisions to generate a multicellular organism in which the diverse cell types function in harmony. Central processes in development include creating distinctions between cells and producing coordination among different cells so that they function as units. In plants, both processes have been shown to rely heavily on cell-to-cell communication and activation and/or repression of subsets of genes. Signals that pass between cells frequently result in the activation of signaling pathways within cells. These signals can result in modifications of transcription factors, which bind to specific sites on the DNA causing the activation or repression of genes. The products of the activated genes are then synthesized. Among them are transcription factors that bind other DNA sequences, activating or repressing a new set of genes. The term transcriptional networks is used for the sequential activity of transcription factors that activate or repress the expression of other transcription factors responsible for developmental regulation. An important feature of developmental transcriptional networks is that at each stage some of the induced or repressed gene products are enzymes, receptors, or other molecules whose activity ultimately results in differentiated functions that define specific cell types.

The creation of different cell types begins with asymmetric cell divisions: when a cell divides to give two daughters that go on to perform different functions. One way to achieve an asymmetric division is to partition the cellular contents at the time of cell division in an unequal manner. This gives each of the two daughters a different complement of regulators. For example, a transcription factor can be partitioned so that most of its molecules go to one daughter. This leads to the activation of a different set of genes in each of the daughter cells, which, in turn, results in different cellular states. Alternatively, the two daughters of a cell division can initially be identical in their cellular contents, but their immediate environments may differ. Usually, the environmental difference is in the types of cells that contact each of the daughters. These surrounding cells send signals that cause each of the daughters to respond by activating different sets of genes. These two processes are not mutually exclusive; for the same two daughter cells, there can be differences in partitioning of cellular contents and differences in the signals received from neighboring cells. Communication between cells is then crucial to ensure that acquisition of cell identities is coordinated at the level of the entire organ so that the organ will be able to perform the higher-order functions necessary to maintain the integrity of the organism (for example, water and nutrient absorption by the root).

Thus, signals act to coordinate all of the major stages of development from asymmetric cell division to organ functioning. Signaling during development can be short range (e.g. a ligand secreted from one cell binds a receptor on a neighboring cell), midrange (e.g. a secreted ligand diffuses through several layers of cells before activating a receptor on its target cell), or long range (e.g. a hormone is secreted from a cell within one tissue and passes from one organ to another until it activates a response in a distant cell). Frequently, these different types of signals are intertwined, forming a signaling network that is integrated by the cell to produce developmental fate decisions.

A key challenge in developmental biology is understanding the interplay and regulation of the signaling and transcriptional networks and how they modulate cellular function. There is increasing awareness that a systems approach integrating global datasets can lead to important insights about biological networks (Levesque and Benfey, 2004). The questions then are the following. What types of data will be essential to model these networks? What modeling approaches best represent these networks? What are the salient features of these networks? How robust are these networks, and how do the properties of network structure confer biological function?

While signaling and transcription are equally important for development, high-throughput techniques for identifying the nodes and links in transcriptional networks have matured more rapidly. Specifically, the critical datasets include global expression profiles at cell type-specific resolution and direct binding data for identification of targets of transcription factors expressed in the cells of interest. Pioneering work in yeast (Saccharomyces cerevisiae) has shown that transcriptional networks can be deduced when these two datasets are combined (Lee et al., 2002; Harbison et al., 2004). For plants, the simplifying aspects of development in an organ such as the root make it highly tractable for the application of these approaches.

The Arabidopsis (Arabidopsis thaliana) root develops continuously from four sets of stem cells in its tip. These stem cells (or initials) divide asymmetrically to regenerate themselves and produce a daughter cell, which in turn divides asymmetrically to generate the first cells of each of the root lineages (Benfey and Scheres, 2000). Because plant cells don't move, these cell lineages are constrained in cell files. Thus, in the root, each stage of development is found in a specific set of cells along the longitudinal axis, with the youngest cells in each file being closest to the initials. The developmental processes that lead to elaboration of the plant root reduce what is normally a four-dimensional problem (three spatial dimensions and time) to a two-dimensional problem (two spatial dimensions). Developmental age can be read from the anatomical position. Moreover, this means that time during development is discretized: each cell in the file represents a discrete stage of development in that lineage. The other simplifying aspect of root development is that, at least for the four outer layers of cells, the root can be viewed as a radially symmetric cylinder. These simplifying aspects have allowed an initial determination of global expression patterns in individual cell types in the root (Birnbaum et al., 2003).

Once the appropriate datasets are available, the challenge will be to model the changes in the networks as stem cells divide asymmetrically and their progeny differentiate. Although we view the steps in the developmental pathway as discretized, a major challenge will be to determine if there are incremental changes in the network along the lineage or if major transitions occur in network architecture as cells mature.