GNN - Genome News Network  
  Home | About | Topics
The New Brain Scans
Scientists generate images of gene expression in the Alzheimer's brain
Edward R. Winstead

Featured article.

Gene chips have crossed the brain barrier. In a novel marriage of genomics and brain imaging, scientists have compared autopsied Alzheimer's brains and normal brains using gene chips and imaging software. They identified clusters of genes that were expressed differently in the normal and diseased brains. The researchers then generated images to visually represent the expression patterns for certain genes at the anatomically correct location on a cross-section of the brain.

Expression patterns of the YWHAH gene projected over the relevant neuroanatomy. View larger

The new brain images look like those created with positron emission tomography (PET) or magnetic resonance (MR) imaging. With enough of these images to cover the entire brain, researchers could in theory generate an interactive three-dimensional atlas of gene expression. Like the 'activity' maps from PET studies, the atlas could be used to learn how networks of genes function in the brain and what goes wrong during disease.

‘We're trying to map how the genome makes the brain’

"Our long-term vision is to study many normal and diseased brains to gain insights about gene-expression patterns and diseases," says Desmond J. Smith, of the University of California, Los Angeles, School of Medicine, who led the research. His team developed the new method, called voxelation.

The name voxelation comes from 'voxel,' which is the term in imaging studies for a small region ('VOlume ELement') of the brain. In voxelation, the brain is sectioned according to a standard grid, and each voxel is analyzed with microarrays to determine which genes are switched on or off. These data are analyzed further to identify similarly expressed genes.

"We're essentially trying to map the process by which the genome makes the brain," says Smith. "The idea is to use fairly sophisticated mathematical tools to try to understand how all these genes conspire together to create this wonderfully complex structure."

The grid system allows for comparisons between different specimens. Demonstrating the strategy, Smith's team analyzed tissue from a normal brain and an Alzheimer's brain. They detected consistent differences among several thousand genes between the normal and the diseased specimen. For confirmation, they repeated the process using two new specimens.

Schematic of voxelation. View larger

"When we repeated the original analysis on two new samples, we found very similar differences in gene expression," says Smith. "This gives us confidence that the differences we found in the initial analysis represent real differences and weren't simply due to chance."

At that point, when they had a set of genes whose expression changed similarly, the researchers examined the genes closely and discovered that some similarly expressed genes appeared to have regulatory elements in common. The results, which will have to be confirmed in further studies, are published in Genome Research.

"The novelty of this work is that we have applied the idea of using voxels, which is standard in imaging techniques, to studies of gene expression in the brain," says Richard M. Leahy, of University of Southern California in Los Angeles.

An electrical engineer by training, Leahy directs the Signal and Image Processing Institute at USC. He met Smith a few years ago while on sabbatical at UCLA and joined the project with the expectation that he would simply apply his expertise in imaging systems to a new biological problem.

"I was completely naïve about the microbiology side of things," he says. "I thought it would be a matter of taking the data and applying analytical tools to produce beautiful images." Instead, they faced challenges such as identifying possible causes of contamination in the data sets.

"Our job was first to remove known sources of variation and then apply other statistical and computational tools to understand the remaining variability in the data," says Leahy.

Gene chips, or microarrays, have in recent years revealed biological changes associated with diseases that are otherwise undetectable. Researchers have used microarrays to show that patients with a similar type of leukemia actually have two different forms of the disease, for example.

Profiling expression patterns in the brain, however, presents unique challenges, according to Andrew S. Peterson, of the University of California, San Francisco, who wrote a commentary accompanying the findings in Genome Research.

The first and foremost challenge comes from the overwhelming number of cell types that are present in the brain. A brain disease may involve changes in cells that make up only one percent of the total number of cells. At this level, it is difficult to separate statistically significant changes from the 'noise' of the data.

Diagram of spatial gene expression patterns for a subset of correlated genes. View larger

A common method of profiling gene expression in the brain is to dissect regions that may be important to a particular biological process or disease. But it is not always possible to know which are the relevant brain regions. A second problem is that published data do not always include information about the precise region in the brain that was analyzed.

"One of the main issues in brain studies is replication," says Peterson. "The noteworthy aspect of the voxelation study is that the authors produce a method of standardization." The results of this study—the set of differentially expressed genes in the Alzheimer's brain—need to be replicated, and Smith's group has provided the means to do so.

Standardization is the basis for developing three-dimensional atlases of gene expression. A single reference atlas of one brain would allow researchers to visualize patterns in particular regions and across regions. Researchers could ask the computer, for example, to find where in the brain a particular gene is expressed above a certain level.

"A reference atlas would be of great value," says Peterson, adding that cost remains a major impediment. "I'm sure it will happen, but I don't know when."

"These are early days still," says Smith. "We look forward to the time when high-resolution maps are available on the Web to researchers everywhere. That's when the real insights will start flowing."

See related GNN articles
»The amount of gray matter matters
»Whole-genome survey leads to a new classification of leukemia

. . .

Brown, V.M. et al. High-throughput imaging of brain gene expression. Genome Res 12, 244-254 (February 2002).
Peterson, A.S. Pixelating the brain. Genome Res 12, 217-218 (February 2002).

Back to GNN Home Page