Introduction

The goals of NCT are: 1) To help facilitate the use of gene expression profiling, proteomics and bioinformatics to the understanding of how cells respond to stress; 2) to create a public database relating environmental stress to biological responses; 3) to collect information relating environmental exposure to disease; 4) to develop an improved paradigm for use of computational mathematics for understanding response to environmental stress; and 5) to identify biomarkers of disease or exposure to enhance environmental health.

To help define the path towards fulfilling its goals, a series of workshops are being held through which the research community can learn about NCT and provide input to NCT. The first workshop in this series entitled "Functional Genomics and Environmental Health" was held at Massachusetts Institute of Technology (MIT) in Cambridge, MA and was co-organized by Ben van Houten (NIEHS) and Leona Samson (Harvard School of Public Health).

In his opening remarks, Samuel Wilson (NIEHS) spelled out the goals of this meeting. The scientific community is excited about the potential of techniques such as DNA microarray expression profiling to advance knowledge of basic biology and to improve human health and medical practice. The fast pace and continually changing new field of toxicogenomics presents many opportunities and challenges to the research community; however, the future of the field is not well defined at present. We are now faced with the challenge of lending definition to the field and understanding where the field of toxicogenomics is going. One endpoint of this workshop, and the NCT program itself, is to evaluate "problem definition" in understanding the cellular response to stress. In addition, this workshop will begin to define how toxicogenomics might change the practice of toxicology research.

The workshop included scientific presentations by Richard A. Young (Whitehead Institute), George Church (Harvard Medical School), Edwin Clark (Millennium Pharmaceuticals), and Leona Samson, all of whom have extensive experience in applying DNA microarray technology to problems of basic biology. The scientific presentations were followed by a roundtable discussion session. This document summarizes the main points of each presentation and of the discussion session.

Transcriptional Networks in Yeast and Human Macrophages

Young described several types of microarrays currently in use that are distinguished from one another by their method of preparation and their source. Commercial microarrays are produced primarily by Affymetrix, Rosetta, Agilent and Corning. These arrays are prepared either by in situ oligonucleotide synthesis of probes or by spotting probes onto a glass slide surface. When compared directly, the commercially available arrays appear to be of similar quality. In contrast, Young indicated that "homemade" glass slide arrays, prepared in many facilities where microarray research is being carried out, are subject to large variations in quality and can be of much lower quality than an equivalent commercially-prepared array.

The methods for analyzing microarray data are evolving rapidly. Scatter plots are commonly used at present to estimate error in microarray data. The scatter plot compares data obtained when a single sample is differentially labeled with two dyes. It is customary to discount data points whose error is 2-fold or higher using such an analysis. Young indicated that scatter plots are no longer the method of choice for estimating error in microarray data. A method was recently developed in which an error model is used to determine an error probability boundary. Error models generate a p-value for each data point. Young emphasized that this method is far superior to more commonly used methods, and he encouraged researchers to begin to use it more widely.

Young indicated that database management is a key component of microarray studies. One database system called "Resolver" has been developed by Rosetta Inpharmatics. This system is available from Rosetta, but for many microarray researchers it may be prohibitively expensive.

Transcriptional networks, especially in yeast and eukaryotic organisms, are a major research interest of Young's. The mechanism of transcription is complex. Other complex processes are also involved in transcriptional networks such as the following: the interaction between transcription factors and chromatin; modulation of chromatin structure; coordination and recruitment of transcription cofactors; and regulatory mechanisms that co-regulate groups of genes. Transcriptional networks for metabolic pathways have been elucidated in great detail over the past several decades. The understanding of these pathways is used extensively in drug development. Thus, one rationale for studying transcriptional networks is to enhance and expedite the drug discovery process.

Young and his colleagues have concluded that simple microarray approaches can not be used to deduce transcriptional networks. For example, they studied the expression of 6218 yeast genes in cells exposed to hydrogen peroxide. Expression profiles showed that approximately 30% of the yeast genome was involved in a short-term transient response to hydrogen peroxide. Attempts were made to discern which of these responses were linked in a significant manner using iterative simulations and neural network theory, but these efforts were not successful.

To better understand transcriptional networks, Young developed an approach called genome-wide location analysis (GWLA). The first step in GWLA is to crosslink a crude cell extract using a reagent that forms DNA-protein and protein/protein crosslinks. The DNA is sheared into short fragments, differentially labeled, and then screened with a selectiveprobe such as a monoclonal antibody. Two differentially labeled DNA pools result; one pool is enriched for a bound protein and one is not enriched. These DNA pools are hybridized to an array containing all intergenic regions of the yeast genome. The magnitude of the signal is compared in enriched and unenriched DNA; a potential network of protein-DNA interactions can then be deduced.

Genome-wide location analysis was applied to the transcriptional networks of the verywell studied yeast transcription factor Gal4. The goal of this GWLA experiment was to identify genes that are bound and upregulated by Gal4 in the presence of galactose. Ten genes were identified; 7 of the genes are known targets of Gal4 and 3 genes were notpreviously known to interact with Gal4. The new Gal4 targets are FUR4, PCL10 and MTH1, which are genes involved in sugar metabolism in yeast. In addition, the algorithm AlignACE identified a putative Gal4 binding motif in the upstream region of these genes.

Young emphasized his view that GWLA is a powerful tool for understanding transcriptional networks. He hopes to apply this technique systematically to up to 200 transcriptional activators and repressors in yeast, and he eventually hopes to use GWLA to analyze gene regulation in higher eucaryotic cells. In the future, GWLA may help annotate the function of many genes and improve our understanding of complex pathways such as genome replication and repair.

Young has also used microarray analysis to understand the response of human macrophages to infectious agents. Macrophages in culture wereexposed to bacteria and harvested after different periods of exposure. Extracts of exposed macrophages were analyzed using a high density Affymetrix array for expression of 2500 human genes. Cluster analysis revealed that exposure to 8 infectious agents, including strains of Escherichia coli, Salmonella, Staphylococcus, Mycobacterium tuberculosis (TB) and others, induced a similar group of genes. Analysis with an array for 68 human cytokine and chemokine genes revealed that about a fourth of these genes were induced in a similar manner after exposure to different bacteria. Further analysis indicated that bacterial lipopolysaccharide and heat shock proteins were likely to be inducing agents contributing to this pattern of macrophage gene expression. Careful study identified a unique characteristic of the response to TB. Macrophages fail to induce interleukin-12 after exposure to TB or after combined exposure to TB and E. coli. However, cells exposed only to E. coli induce interleukin-12 expression. The results indicate that IL-12 therapy might be useful to enhance recovery from TB infection; importantly, this therapy might be useful in cases involving a drug-resistant strain of TB.