Reverse-Engineering Transcriptional Networks

Finding the Master Regulators

A cell may change states several times in its lifetime—from a stem cell to a specialized cell, for example, or from a normal cell to a cancerous one. Each time this happens, a veritable army of genes must be raised to do the tasks needed by the new cell type. Now, researchers have successfully used computational approaches to identify the “master regulators” that, like generals, control the transformation of benign brain cells into the malignancies that cause high grade glioma, one of the most aggressive forms of brain cancer. The computational findings were then confirmed experimentally.


This transcriptional network of high-grade glioma cells shows the two master regulators in red and other significant transcription factors in orange. Together, these transcription factors control about 80 percent of an HGG tumor’s signature. Image courtesy of Columbia University, Califano Lab. Reprinted with permission from MacMillan Publishers, Ltd.: Carro, M.S., et al., The transcriptional network for mesenchymal transformation of brain tumours, Nature 463, 318-325 (21 January 2010).The work, which was published in Nature in February 2010, demonstrates the value that can come from reverse engineering molecular interaction networks for specific cell types. Coauthor Andrea Califano, PhD, professor of bioinformatics at Columbia University and director of the Center for the Multiscale Analysis of Genetic Networks (MAGNet), hopes to apply these methods to other questions of cellular transformation and development, particularly those relevant to disease states such as cancer. “We can now ask what are the genes that control an arbitrary transformation,” he says.


For a healthy cell to become the beginnings of a high-grade glioma (HGG) tumor, it needs to express a large number of genes that otherwise would never be activated. To find the key genes that produce that altered gene expression state, Califano’s team first mapped out the regulatory logic of the most aggressive type of HGG cells using an information theory algorithm called ARACNE. The method can reconstruct regulatory networks from gene expression profiles of particular cell populations, even pruning out indirect interactions to determine which genes directly control others. Next, the researchers looked for genes in this network that were part of the tumor’s signature – those that are highly expressed in HGG cells but not in normal brain cells. A handful of transcription factors emerged that together control about 80 percent of the characteristic genes. Two in particular, STAT3 and C/EBP, appeared to hierarchically control the others, even though they are expressed at levels so small they do not appear in the signature.


Further experiments, done with brain tumor experimentalist Antonio Iavarone, MD, verified the model, showing that activating the two genes simultaneously in neural cells causes the shift to a tumor-like cell. Likewise, silencing the genes together eliminated the malignant phenotype.


While the networks in this study were built from gene expression data, the method could also work with other information, such as proteomics or chromatin structure data. “I see this work as being a prototype of the power of this type of approach, but it’s really just the beginning,” says Howard Fine, MD, chief of the Neuro-Oncology branch at the National Cancer Institute.
Fine is also hopeful that the results of this work could lead to glioma treatments. “They’ve identified one small module within this very complex signaling network that is a cancer cell,” he says. “This says to us, we might be able to translate findings from these kinds of approaches to new therapies for patients well before we can fully understand the complexity of the tumor cell.”

All submitted comments are reviewed, so it may be a few days before your comment appears on the site.

Post new comment

The content of this field is kept private and will not be shown publicly.
This question is for testing whether you are a human visitor and to prevent automated spam submissions.
Enter the characters shown in the image.