Device Allows Comparison of Prognostic Gene Clusters

WASHINGTON—Stanford University researchers have developed a device for gene comparison that could lead to a more exact way to categorize cancer tumors and assess the survival chances of patients. In initial experiments with breast cancer patients, they detected specific gene combinations in tumors and found that certain combinations indicated either a good or bad prognosis.

“We find that we can classify tumors into subsets,” said David Botstein, PhD, professor and chair of the Department of Genetics, Stanford University School of Medicine.

Dr. Botstein described the new approach during a congressional briefing on the importance of the computer to sequencing the human genome and eventually using the information it contains to benefit patients.

“These two communities [genome mapping and computing], which have revolutionized the way humans think about biology and how they calculate, are converging,” he said.

The briefing was one of a regular series held by the Congressional Biomedical Research Caucus, a group of House members interested in encouraging and promoting such research.

“The important thing is that DNA is an information carrier,” Dr. Botstein said. “It holds the information in a code of nucleotide letters, and the cell decodes this into a protein for each gene.”

In computer terms, the human genome contains about 750 megabytes of data. “You all have laptops that can remember it,” he added. “However, before you become impressed with the computer industry, a cell that is 1 micron cubed has a nucleus that not only remembers it, but uses all this information. So computers have a long way to go to compete with biology.”

Dr. Botstein noted that when the human genome is 100% sequenced (the two competing groups doing the work still have some gaps to fill), scientists will only have a string of letters from a “book” that must be deciphered into meaningful “words,” which represent genes. “The sequence of the human genome, in some nontrivial way, is absolutely useless as it now stands,” he said.

To determine human genes and their functions, researchers compare human genes with known genes of some other organism.

“Sequence comparison essentially consists of looking for close similarities in the genomes of two different organisms and selecting the ones from each genome that most closely match one another,” Dr. Botstein said.

Yeast and eukaryocytes share many genes with humans, and the functions of a number of these genes are known, he said. Work is complete on several organisms, and the mouse genome is now in draft form, he added.

“It’s not enough to have the human sequence,” Dr. Botstein explained. “You must have these model organisms’ sequences as well. When you have the complete sequences, you can do a comparison of one whole genome against another—all the instructions for making a worm against all the instructions for making a yeast.”

Matching human genetic materials against those known in other species will quicken the process of sorting the human genome sequences into specific genes. “By sequence comparison, we can tell which of the human proteins does what,” Dr. Botstein said. “This is extremely difficult and cannot be done without a computer.”

Genes transcribe their code for making proteins into messenger RNA (mRNA), which carries the message to cells, where proteins are manufactured. “When you know which messenger RNAs are in a cell, then you know which proteins are being made, and when you know that, then you know the nature of that cell,” he said.

At Stanford, Patrick O. Brown, MD, PhD, a Howard Hughes Medical Institute investigator collaborating with Dr. Botstein, has invented a machine that makes thousands of mRNA copies and puts them on a single slide. “So on a single slide, we can have the entire genome of an organism,” Dr. Botstein said.

A gene expressed in a cell will make many copies of mRNA. “And if we have two cells and compare them, we can tell which gene is expressed in one cell type compared to another cell type,” he said. “And because we have this little gadget [invented by Dr. Brown], we can do this for thousands of genes. If we are talking about the yeast genome, that’s 6,000 genes. If we’re talking about the worm genome, that’s 20,000 genes. And the human genome is, who knows—60,000?”

In one experiment, the Stanford team looked at breast cancer samples from a Norwegian study. In that trial, 20 women with late-stage disease had surgery, followed by chemotherapy and then a second surgery 16 weeks after the first to remove remaining tumor. Biopsies samples were preserved at both operations.

Dr. Botstein and his colleagues analyzed the entire tumor and its heterogeneous cell types from each biopsy. “We boiled it down to 1,500 genes that changed by more than threefold on three different assays,” he said.

The “really striking thing,” he said, is that with very few exceptions, “the before and after samples are more similar to each other than they are to anything else. So the tumor is stable and orderly. It’s not a mess. In all these genes, the level of expression is more similar in the metastasis and the primary tumor than the primary tumor or the metastasis is to anything else.”

What that means, he said, is that “when a cell migrates to a lymph node, it creates—presumably by signaling different parts of the body—an environment much like that of the primary tumor.”

The team also found it could associate certain clusters of genes with different kinds of tumor characteristics, such as the presence of extensive T cells or B cells.

“It turns out that if you go back into the records and see what happened to these women. In one subset, nothing happened; these women were still with us after 100 months,” Dr. Botstein said. Another subset, however, had extremely poor survival. Such findings make the potential for using the technique to categorize tumors quite promising, he added.