GNN - Genome News Network  
  Home | About | Topics
   
The Evolving Art of Arrays
The molecular portraits of gene microarrays are creating new classifications of disease. Here come the protein arrays.
  
By
Edward R. Winstead



Featured Article.

Gene microarrays are a kind of modern art. Thousands of colored squares comprise the grids of microarrays, each square representing the activity of one gene under certain circumstances. A single microarray is a mere snapshot of activity, and its patterns of red, green, and black are essentially meaningless. But when collections of microarrays are assembled and analyzed, as happened in a recent breast cancer study, the result can be new classifications of disease and an appreciation for the nuances of biology.


A 23,000-gene microarray used in the David Botstein-Patrick Brown laboratory at Stanford University School of Medicine (detail). View full

Researchers at Stanford University School of Medicine in California have recently found distinguishable differences in gene expression in a sample of 40 breast cancer tumors. Based on the activity of 8,102 genes, the researchers characterized what appear to be at least four subgroups of the disease in the sample population. The researchers are now trying to determine whether each subgroup is associated with a particular disease outcome.

"Having profiled all the tumors, we can now go back and identify certain types that are predictive of biological or clinical behavior," says Charles M. Perou, of the Stanford laboratory jointly headed by David Botstein and Patrick Brown. The group investigates a number of tumor types, including breast, prostrate, liver, and lung. "Our work is premised on the belief that there are clear differences in gene expression within tumors of a specific type," says Perou.

Two previous gene microarray studies have reported new classifications of cancers—in one case lymphoma and the other melanoma. Subjects were grouped into two categories based on a mathematical analysis of tumor gene activity. Like the breast cancer findings, the research appeared this year in Nature.


Hierarchical-clustering analysis and data display of gene-expression patterns for a set of 80 human tumor samples.
Each row represents a gene, and each column represents a tumor sample. The behavior of a gene in the experiment is represented by the color and intensity of the squares: red indicates above average expression; black indicates average expression; and green indicates below average expression. View larger

Analyzing microarray data involves sorting relatively large amounts of information, which is done using algorithms. From the initial results, the researchers identify genes of particular interest—those that are expressed at both higher and lower than normal levels. After more algorithmic analysis, the researchers arrange the data into hierarchical 'clusters' that reveal patterns among groups of genes, allowing them to begin the work of classifying tumors according to the new subsets of disease.

The researchers sampled twenty of the breast cancer tumors twice—once before and once after a 16-week course of chemotherapy. The patterns of gene activity in microarrays from the same individual were almost always more similar to each other than either was to any other. Still, among the entire sample were four distinct tumor types that no one had previously reported.

The breast tumors in the sample are basically indistinguishable during a clinical exam or under the microscope, according to Perou. Previous genetic studies also failed to reveal that certain groups of genes play an important role in tumor development. Genetic studies typically focus on a single gene or several genes, not hundreds or even thousands. "When you look at one gene at a time, you can't see relationships between genes and groups of genes," says Perou, adding: "the more samples, the finer the distinctions."

Toward personalized medicine

Preliminary data suggest that one type of tumor—those derived primarily from breast basal epithelial cells—may be associated with a very poor prognosis, according to Perou. If confirmed, this information would be critical for the treatment of this type of tumor. Indeed, gene array technologies have generated so much interest in part because they seem to promise more precise diagnoses, which might allow doctors and patients to 'personalize' the treatment. "As we begin to individualize the therapy based on the type of tumor, I suspect many treatments will prove more effective than we now realize," says Perou.

Experimental microarrays are also being used to spot gene activity associated with metastasis—the spread of tumor cells into previously unaffected tissue. For example, researchers at the Whitehead Institute for Biomedical Research and the Massachusetts Institute of Technology recently used arrays to identify genes that are more highly expressed in metastatic mouse and human melanoma cells compared with their non-metastatic counterparts.

For all the information gene microarrays provide, they reveal relatively little about proteins, the molecules that carry out most of the functions of a cell. Gene arrays detect the presence of messenger RNA, the chemical involved in translating DNA into proteins. Tracking this middle step in the production process reveals nothing about three areas of interest to researchers: protein function, the abundance of protein in a cell, and modifications to proteins after they are produced—changes that may be critical in the development of disease.

Solving technical problems

"If you really want to know what's going on in a cell, you have to look at the molecules responsible for cellular functions, rather than intermediates in the process," says Gavin MacBeath of the Center for Genomics Research at Harvard University in Cambridge, MA. His laboratory is developing microarray technologies for studying protein function and screening large numbers of protein-protein interactions. Many academic and industry researchers have tried in recent years to solve the technical problems that are delaying the development of functional protein arrays.

The technology lags behind gene arrays in part because proteins are naturally uncooperative. "DNA is very well behaved, and there are powerful ways to amplify the chemical," says MacBeath. "Any DNA will work for a common set of conditions, whereas proteins are very different from each other, and some proteins are more stable than others." Another obstacle to protein arrays is the difficulty of immobilizing proteins on slides while preserving function (often the whole point of developing the array is to study function).


Robots spot 10,000 proteins on a slide

MacBeath and Stuart L. Schreiber, also of Harvard University, have worked out some of the problems. In a recent issue of Science, the researchers describe the construction of glass slides densely arrayed with proteins for functional studies. Borrowing a technology from gene microarrays, the researchers used a high-precision robot to print more than 10,000 protein samples on a surface about half the size of a microscope slide (1,600 spots per square centimeter).

Previous studies have reported versions of a protein array, including a kind of test tube array created at the University of Washington. The Harvard team points out two characteristic features of their project: First, the method solves the problem of attaching proteins to a structure without losing function. And second, the technology is relatively simple to use and available to anyone. A purpose of the project, says MacBeath, was to make the technology easily accessible and compatible with standard instrumentation.

MacBeath's laboratory is using the array to study families of between 50 and 200 proteins in humans and yeast. The yeast proteome, he notes, includes some 6,200 proteins. "As the technology improves, we'll go after the whole thing," he says.

. . .

 
MacBeath, G. & Schreiber, S.L. Printing proteins as microarrays for high-throughput function determination. Science 289, 1760-1763 (September 8, 2000).
 
Perou, C.M. et al. Molecular portraits of human breast tumours. Nature 406, 747-752 (August 17, 2000).
 

Back to GNN Home Page