By David N. Leff

Cancers come mainly in two physical forms, liquid and solid. And clinically, many malignant tissues and organs can be divided into those indispensable to life, such as brain, colon or heart, and nonessential ones that in extremis can be dispensed with. Among the latter are breast, ovary and prostate.

Perceiving this dichotomy, molecular biologist Ira Pastan at the National Cancer Institute tested a new way to screen cells of the prostate gland for candidate diagnostic and therapeutic genes. His point of departure was lymphoma, an essentially liquid cancer of B cells in the blood.

"The FDA," Pastan told BioWorld Today, "has recommended for approval a monoclonal antibody against CD20, for therapy of lymphomas. That antibody recognized the differentiation antigen not just in normal B cells but also in B cell lymphomas. It proved the principle," he pointed out, "that if one could identify differentiation antigens on cancers arising in nonessential tissues, they could be therapeutically useful.

"It's relatively easy to make antibodies to B cells," he reflected, but "we do not have any comparable antibodies to breast, prostate, ovary — other tissues in which cancers commonly arise, because it's technically difficult to get the isolated cells; there are many technical barriers.

"So we decided," he recalled, "that we'd like to search for such antigens — for targets of immunotherapy — thinking if we could find new antigens for prostate cells, particularly cell-surface antigens, they might also be present in prostate cancers, and we could make antibodies to them.

"But how do you go about searching for these things?" Pastan asked himself and his coworkers.

The answer they found appears in the current issue of the Proceedings of the National Academy of Sciences (PNAS), dated Jan. 6, 1998. Their report bears the title: "Discovery of three genes specifically expressed in human prostate by expressed sequence tag database analysis."

The team chose prostate because, Pastan explained, "Prostate cancer is such a deadly disease, with really no adequate therapy, in principle. We believed there should be such antigens present in its cells, and that we had to develop a method of finding them. Since we couldn't use conventional ways of isolating cells and making monoclonal antibodies, we'd use EST [expressed sequence tag] databases."

Aiming At Therapeutic Prostate Antibodies

"We hoped then," he went on, "to be able to make antibodies to the appropriate antigens, and see if the antibody alone, or with any conjugate on it — toxins, radioactivity, drugs — might be useful therapy. So that was, and is, our goal."

ESTs are small gene fragments, typically 400 to 600 base pairs long, which can be used as hooks for fishing out entire sequences from complementary-DNA libraries. There are now well over 1 million ESTs in public databases, believed to represent over half of all human genes.

"Many investigators," Pastan said, "and even the National Cancer Institute, support the generation of ESTs from various tissues, to identify new genes not previously identified. For instance, NCI has a project for outpaying, supporting and contracting with investigators who make RNA from different kinds of normal and cancerous prostate tissues. The RNA is then used to make ESTs, which can be sequenced."

To begin with, the team generated ESTs from all the library databases containing the words "Homo sapiens" and "prost." This initial search involved 1,137,304 EST entries in 907 cDNA libraries, 539 of them human, of which 16 were from the prostate.

Then they matched the prostate ESTs against all the human ones, looking for strong sequence similarities. Next, they divided the resulting hits into two lists, prostate and nonprostate ESTs. The first grouped the prostate hits; the second measured their specificity among all organ systems.

Proof Of Principle: Zeroing In On PSA

Further refinements led them at last to bingo: the prostate-specific antigen, PSA. (PSA is now a standard clinical marker for prognosis, diagnosis and monitoring of prostate cancer. See BioWorld Today, April 21, 1995, p. 1.)

"Looking for known prostate genes," Pastan recounted, "we found 274 EST hits for PSA, 115 of them with zero hits in other organs. The remaining 159 had EST hits in lung and breast. "

Besides PSA, Pastan's chart lists 33 other prostate gene classes. "As you go down the list," he observed, "you have only 8 to 12 ESTs in each. Fifteen of them reflect the rare genes that no one has yet described, expressed less abundantly than PSA."

Three of these 15 newfound cDNA's, he said, are prostate-specific, so they "could be useful in the targeted therapy of prostate cancer."

So far, he observed, "there's not enough information in the sequences to figure out gene function. That will come only when we get the full-length sequence; maybe that will give a clue. Otherwise, we may have to do even more." He added, "That's what we're doing now."

Pastan thinks his report in PNAS may well be describing an innovation in database analysis. "We actually were the first to show it can be done," he allowed, "and how to do it. In science, of course, there often are several ways to do it."

He concluded that "the procedure can easily be applied to discover other genes specifically expressed in other organs or tumors." Suiting his actions to those words, he and his co-authors "have done some work on breast, but we're not at the stage of publishing it yet."

Pastan directs the laboratory of molecular biology at the National Cancer Institute, in Bethesda, Md. *