One reason cancer is not considered a single disease but many is that every cancer cell seems to be dysfunctional in its own way. Random mutations in a cell’s DNA initiate its slide into abnormal behavior. And as additional mutations accumulate, that randomness is also thought to account for the diversity in different patients’ tumors, even when they are cancers of the same tissue. But evidence is growing that there is a method to the madness of tumor cells, making some scientists reevaluate the nature of cancer.

Analyzing tumors from dozens of tissue types, Isaac S. Kohane of the Harvard-MIT Division of Health Sciences and Technology has catalogued surprising yet familiar patterns of gene activity in cancer cells—they are the same programmed genetic instructions active during various stages of embryonic and fetal development. Entire suites of genes that drive an embryo’s early growth and the later formation of limbs and other structures in the womb normally go silent during the rest of life, but these genetic programs are switched back on in many tumor cells.

Grouping tumors according to the developmental stage their gene activity most resembles reveals predictive information about those tumors, Kohane has found. In groups of lung tumors, for instance, “malignancy and even time to death of actual patients were directly proportional to the ‘earliness’ of the gene signatures,” he says.

In his largest and latest tumor study, Kohane showed that the same holds true across different types of cancer. Comparing gene activity for nearly three dozen kinds of cancer and precancerous conditions against a timeline of 10 developmental processes, he could group seemingly disparate diseases into three categories. Among the tumors with signatures characteristic of the earliest embryonic development stages were lung adenocarcinoma, colorectal adenoma, T cell lymphomas and certain thyroid cancers. The highly aggressive cancers in this group also tend to look most undifferentiated and embryonic. The tumors with gene signatures that mirrored third-trimester and neonatal developmental gene expression patterns tend to be slower-growing types, including prostate and ovarian cancers, adrenal adenoma and liver dysplasia. A third category of tumors represented a mixed bag, in which activity matched aspects of both the other two groups.

Similarities between embryos and tumors “should be paid attention to,” says pioneering cancer researcher Lloyd J. Old, chairman of the Ludwig Institute for Cancer Research New York Branch. “The reason this is so interesting is that the idea that cancer and development are in some way linked goes way back,” he explains. The 19th-century pathologist John Beard, for example, noted the similarity between tumors and the trophoblast, a part of an early embryo that eventually becomes the placenta. “If you’ve ever seen the trophoblast invading the uterus, it invades, spreads, creates a blood supply. It also suppresses the maternal immune system,” Old says. “All of those are characteristics of cancer.”

In his own research, Old has found common genetic programs at work in tumor cells and gametes. One subject of his immunology studies are the cancer/testis (CT) antigens, a group of proteins manufactured almost exclusively by tumors and by sperm- and egg-producing germ-line cells. The specificity of CT antigens makes them ideal targets for cancer vaccines or antibody-based drugs, Old says; moreover, the activation of CT genes in tumors is telling. “These are programs that you and I used as gametes,” he explains. Seeing these primordial programs reactivated in tumors has led Old to describe cancer as a “somatic cell pregnancy.”

The fact that cancer cells switch on these normally silenced programs suggests to Old that the important characteristics of cancers are not random. “This is a fundamentally different way of thinking. A cell that mutates looks for genes that can help it flourish” and finds them in the suites of developmental genes, he says. “It’s a programmatic origin rather than a Darwinian origin” for cancer traits.

The two views of malignancy, however, do not necessarily conflict. “It’s not as if accumulating mutations are at odds with the discernible program,” remarks Robert A. Weinberg of the Massachusetts Institute of Technology, noting that the activation of developmental programs could be a downstream consequence of the mutations. Weinberg showed last year that gene activity involved in maintaining embryonic stem cell identity is a common feature of the most undifferentiated-looking and aggressive tumors. Whether that kind of evidence indicates an embryonic program driving those cancers remains to be determined, he cautions: “It’s an interesting concept, but at this stage what they talk about is highly speculative. One can ascribe all manner of human traits to cancers and speculate that it will lead one into therapeutic insights. But the devil is in the details.”

Arresting Cancer’s Development
Evidence is growing that tumor cells may grow and spread by co-opting the genetic programs normally active only during embryonic and fetal development. If true, then disrupting those programs with gene-silencing therapy could undermine a tumor’s most threatening traits. Isaac S. Kohane of the Harvard-MIT Division of Health Sciences and Technology has organized cancers into three groups that correlate with different embryonic stages. He suggests that at the very least, drugs already known to work well on one cancer in his groupings should be tested against other cancers with the same profile.

This story was originally published with the title "Primal Programs"