In most retinoblastoma tumors, Dryja suspected, the two deletions in the two copies of the Rb gene would lie in different parts of the gene. Since mutations occur randomly, the chance of both mutations lying in precisely the same region of the gene is a little akin to rolling double sixes in dice that have one hundred faces. Typically, one of the deletions would “hit” the front end of the gene, while the other deletion might hit the back end (in both cases, the functional consequences would be the same—inactivating Rb). The two “hits” in most tumors would thus be asymmetric—affecting two different parts of the gene on the two chromosomes.

  But even hundred-headed dice, rolled many times, can yield double sixes. Rarely, Dryja knew, one might encounter a tumor in which both hits had deleted exactly the same part of the gene on the two sister chromosomes. In that case, that piece of chromosome would be completely missing from the cell. And if Dryja could find a method to identify a completely missing piece of chromosome thirteen in a retinoblastoma tumor cell, he would instantly land on the Rb gene. It was the simplest of strategies: to hunt the gene with absent function, Dryja would look for absence in structure.

  To identify such a missing piece, Dryja needed structural mileposts along chromosome thirteen—small pieces of DNA called probes, which were aligned along the length of the chromosome. He could use these DNA probes in a variant of the same “sticking” reaction that Varmus and Bishop had used in the 1970s: if the piece of DNA existed in the tumor cell, it would stick; if the piece did not exist, the probe would not stick, identifying the missing piece in the cell. Dryja had assembled a series of such probes. But more than probes, he needed a resource that he uniquely possessed: an enormous bank of frozen tumors. The chances of finding a shared deletion in the Rb gene in both chromosomes were slim, so he would need to test a vast sample set to find one.

  This, then, was his crucial advantage over the vast professional labs in Toronto and Houston. Laboratory scientists rarely venture outside the lab to find human samples. Dryja, a clinician, had a freezer full of them. “I stored the tumors obsessively,” he said with the childlike delight of a collector. “I put news out among patients and doctors that I was looking for retinoblastoma cases. Every time someone saw a case, they would say, ‘Get that guy Dryja.’ I would then drive or fly or even walk to pick up the samples and bring them here. I even got to know the patients by name. Since the disease ran in families, I would call them at home to see if there was a brother or sister or cousin with retinoblastoma. Sometimes, I would know [about a tumor] even before doctors knew.”

  Week after week, Dryja extracted the chromosomes from tumors and ran his probe set against the chromosomes. If the probes bound, they usually made a signal on a gel; if a probe was fully missing, the signal was blank. One morning, having run another dozen tumors, Dryja came to the lab and held up the blot against the window and ran his eyes left to right, lane after lane automatically, like a pianist reading a score. In one tumor, he saw a blank space. One of his probes—H3-8, he had called it—was deleted in both chromosomes in that tumor. He felt the brief hot rush of ecstasy, which then tipped into queasiness. “It was at that moment that I had the feeling that we had a gene in our hands. I had landed on retinoblastoma.”

  Dryja had found a piece of DNA missing in tumor cells. Now he needed to find the corresponding piece present in normal cells, thus isolating the Rb gene. Perilously close to the end, Dryja was like an acrobat at the final stretch of his rope. His one-room lab was taut with tension, stretched to its limit. He had inadequate skills in isolating genes and limited resources. To isolate the gene, he would need help, so he took another lunge. He had heard that researchers in the Weinberg lab were also hunting for the retinoblastoma gene. Dryja’s choices were stark: he could either team up with Weinberg, or he could try to isolate the gene alone and lose the race altogether.

  The scientist in Weinberg’s lab trying to isolate Rb was Steve Friend. A jovial, medically trained molecular geneticist with a quick wit and an easy manner, Friend had casually mentioned his interest in Rb to Dryja at a meeting. Unlike Dryja, working with his growing stash of tumor samples, Friend had been building a collection of normal cells—cells in which the Rb gene was completely intact. Friend’s approach had been to find genes that were present in normal retinal cells, then to try to identify ones that were abnormal in retinoblastoma tumors—working backward toward Dryja.

  For Dryja, the complementarity of the two approaches was obvious. He had identified a missing piece of DNA in tumors. Could Friend and Weinberg now pull the intact, full-length gene out of normal cells? They outlined a potential collaboration between the two labs. One morning in 1985, Dryja took his probe, H3-8, and virtually ran across the Longfellow Bridge (by now, the central highway of oncogenesis), carrying it by hand to Friend’s bench at the Whitehead.

  It took Friend a quick experiment to test Dryja’s probe. Using the DNA “sticking” reaction again, Friend trapped and isolated the normal cellular gene that stuck to the H3-8 probe. The isolated gene “lived” on chromosome thirteen, as predicted. When Dryja further tested the candidate gene through his bank of tumor samples, he found precisely what Knudson had hypothesized more than a decade earlier: all retinoblastoma cells contained inactivations in both copies of the gene—two hits—while normal cells contained two normal copies of the gene. The candidate gene that Friend had isolated was indisputably Rb.

  In October 1986, Friend, Weinberg, and Dryja published their findings in Nature. The article marked the perfect complement to Weinberg’s ras paper, the yin to its yang—the isolation of an activated proto-oncogene (ras) and the identification of the anti-oncogene (Rb). “Fifteen years ago,” Weinberg wrote, “Knudson provided a theoretical basis for retinoblastoma tumorigenesis by suggesting that minimally two genetic events are required to trigger tumor development.” Weinberg noted, “We have isolated [a human gene] apparently representing one of this class of genes”—a tumor suppressor.

  What Rb does in normal cells is still an unfolding puzzle. Its name, as it turns out, is quite a misnomer. Rb, retinoblastoma, is not just mutated in rare eye tumors in children. When scientists tested the gene isolated by Dryja, Friend, and Weinberg in other cancers in the early nineties, they found it widely mutated in lung, bone, esophageal, breast, and bladder cancers in adults. Like ras, it is expressed in nearly every dividing cell. And it is inactivated in a whole host of malignancies. Calling it retinoblastoma thus vastly underestimates the influence, depth, and prowess of this gene.

  The retinoblastoma gene encodes a protein, also named Rb, with a deep molecular “pocket.” Its chief function is to bind to several other proteins and keep them tightly sealed in that pocket, preventing them from activating cell division. When the cell decides to divide, it tags Rb with a phosphate group, a molecular signal that inactivates the gene and thus forces the protein to release its partners. Rb thus acts as a gatekeeper for cell division, opening a series of key molecular floodgates each time cell division is activated and closing them sharply when the cell division is completed. Mutations in Rb inactivate this function. The cancer cell perceives its gates as perpetually open and is unable to stop dividing.

  The cloning of ras and retinoblastoma—oncogene and anti-oncogene—was a transformative moment in cancer genetics. In the decade between 1983 and 1993, a horde of other oncogenes and anti-oncogenes (tumor suppressor genes) were swiftly identified in human cancers: myc, neu, fos, ret, akt (all oncogenes), and p53, VHL, APC (all tumor suppressors). Retroviruses, the accidental carriers of oncogenes, faded far into the distance. Varmus and Bishop’s theory—that oncogenes were activated cellular genes—was recognized to be widely true for many forms of cancer. And the two-hit hypothesis—that tumor suppressors were genes that needed to be inactivated in both chromosomes—was also found to be widely applicable in cancer. A rather general conceptual framework for carcinogenesis was slowly becoming apparent. The cancer cell was a broken, deranged machine. Oncogenes were its jammed accelerators an
d inactivated tumor suppressors its missing brakes.*

  In the late 1980s, yet another line of research, resurrected from the past, yielded a further bounty of cancer-linked genes. Ever since de Gouvêa’s report of the Brazilian family with eye tumors in 1872, geneticists had uncovered several other families that appeared to carry cancer in their genes. The stories of these families bore a familiar, tragic trope: cancer haunted them generation upon generation, appearing and reappearing in parents, children, and grandchildren. Two features stood out in these family histories. First, geneticists recognized that the spectrum of cancers in every family was limited and often stereotypical: colon and ovarian cancer threading through one family; breast and ovarian through another; sarcomas, leukemias, and gliomas through a third. And second, similar patterns often reappeared in different families, thereby suggesting a common genetic syndrome. In Lynch syndrome (first described by an astute oncologist, Henry Lynch, in a Nebraskan family), colon, ovarian, stomach, and biliary cancer recurred generation upon generation. In Li-Fraumeni syndrome, there were recurrent bone and visceral sarcomas, leukemias, and brain tumors.

  Using powerful molecular genetic techniques, cancer geneticists in the 1980s and 1990s could clone and identify some of these cancer-linked genes. Many of these familial cancer genes, like Rb, were tumor suppressors (although occasional oncogenes were also found). Most such syndromes were fleetingly rare. But occasionally geneticists identified cancer-predisposing gene alterations that were quite frequently represented in the population. Perhaps the most striking among these, first suggested by the geneticist Mary Claire-King and then definitively cloned by Mark Skolnick’s team at the pharma company Myriad Genetics, was BRCA-1, a gene that strongly predisposes humans to breast and ovarian cancer. BRCA-1 (to which we will return in later pages) can be found in up to 1 percent of women in selected populations, making it one of the most common cancer-linked genes found in humans.

  By the early 1990s, the discoveries of cancer biology had thus traversed the gap between the chicken tumors of Peyton Rous and real human cancers. But purists still complained. The crusty specter of Robert Koch still haunted the genetic theory of cancer. Koch had postulated that for an agent to be identified as the “cause” of a disease, it must (1) be present in the diseased organism, (2) be capable of being isolated from the diseased organism, and (3) re-create the disease in a secondary host when transferred from the diseased organism. Oncogenes had met the first two criteria. They had been found to be present in cancer cells and they had been isolated from cancer cells. But no one had shown that a cancer gene, in and of itself, could create a bona fide tumor in an animal.

  In the mid-1980s, a series of remarkable experiments allowed cancer geneticists to meet Koch’s final criteria. In 1984, biologists working on stem cells had invented a new technology that allowed them to introduce exogenous genes into early mouse embryos, then create a living mouse out of those modified embryos. This allowed them to produce “transgenic mice,” mice in which one or more genes were artificially and permanently modified. Cancer geneticists seized this opportunity. Among the first such genes to be engineered into a mouse was c-myc, an oncogene discovered in lymphoma cells.

  Using transgenic mouse technology, Philip Leder’s team at Harvard altered the c-myc gene in mice, but with a twist: cleverly, they ensured that only breast tissue in the mouse would overexpress the gene. (Myc could not be activated in all cells. If myc was permanently activated in the embryo, the embryo turned into a ball of overproliferating cells, then involuted and died through unknown mechanisms. The only way to activate myc in a living mouse was to restrict the activation to only a subset of cells. Since Leder’s lab was studying breast cancer, he chose breast cells.) Colloquially, Leder called his mouse the OncoMouse. In 1988, he successfully applied for a patent on the OncoMouse, making it the first animal patented in history.

  Leder expected his transgenic mice to explode with cancer, but to his surprise, the oncomice sprouted rather mousy cancers. Even though an aggressive oncogene had been stitched into their chromosomes, the mice developed small, unilateral breast cancers, and not until late in life. Even more surprisingly, Leder’s mice typically developed cancers only after pregnancy, suggesting that environmental influences, such as hormones, were strictly required to achieve full transformation of breast cells. “The active myc gene does not appear to be sufficient for the development of these tumors,” Leder wrote. “If that were the case, we would have expected the uniform development of tumor masses involving the entire bilateral [breast] glands of all five tumor-bearing animals. Rather, our results suggest at least two additional requirements. One of these is likely to be a further transforming event. . . . The other seems to be a hormonal environment related to pregnancy that is only suggested by these initial studies.”

  To test the roles of other oncogenes and environmental stimuli, Leder created a second OncoMouse, in which two activated proto-oncogenes, ras and myc, were engineered into the chromosome and expressed in breast cells. Multiple tumors sprouted up in the breast glands of these mice in months. The requirement for the hormonal milieu of pregnancy was partially ameliorated. Still, only a few distinct clones of cancer sprouted out of the ras-myc mice. Millions of breast cells in each mouse possessed activated ras and myc. Yet, of those millions of cells, each endowed with the most potent oncogenes, only a few dozen turned into real, living tumors.

  Even so, this was a landmark experiment: cancer had artificially been created in an animal. “Cancer genetics,” as the geneticist Cliff Tabin recalls, “had crossed a new frontier. It was not dealing with just genes and pathways and artificial lumps in the lab, but a real growing tumor in an animal.” Peyton Rous’s long squabble with the discipline—that cancer had never been produced in a living organism by altering a defined set of cellular genes—was finally laid to its long-overdue rest.

  * In fact, the “normal” cells that Weinberg had used were not exactly normal. They were already growth-adapted, such that a single activated oncogene could tip them into transformed growth. Truly “normal” cells, Weinberg would later discover, require several genes to become transformed.

  †In fact, ras, like src, had also been discovered earlier in a cancer-causing virus—again underscoring the striking capacity of these viruses to reveal the mechanisms of endogenous oncogenes.

  * The Laskerites had largely been disbanded in the aftermath of the 1971 National Cancer Act. Mary Lasker was still involved in science policy, although with nowhere near the force and visceral energy that she had summoned in the sixties.

  * Although cancer is not universally caused by viruses, certain viruses cause particular cancers, such as the human papilloma virus (HPV), which causes cervical cancer. When the mechanism driving this cancer was deciphered in the 1990s, HPV turned out to inactivate Rb’s and p53’s signal—underscoring the importance of endogenous genes in even virally induced cancers.

  The Hallmarks of Cancer

  I do not wish to achieve immortality through my works. I wish to achieve immortality by not dying.

  —Woody Allen

  Scurrying about in its cage in the vivarium atop Harvard Medical School, Philip Leder’s OncoMouse bore large implications on small haunches. The mouse embodied the maturity of cancer genetics: scientists had created real, living tumors (not just abstract, etiolated foci in petri dishes) by artificially manipulating two genes, ras and myc, in an animal. Yet Leder’s experiment raised further questions about the genesis of cancer. Cancer is not merely a lump in the body; it is a disease that migrates, evolves, invades organs, destroys tissues, and resists drugs. Activating even two potent proto-oncogenes had not recapitulated the full syndrome of cancer in every cell of the mouse. Cancer genetics had illuminated much about the genesis of cancer, but much, evidently, remained to be understood.

  If two oncogenes were insufficient to create cancers, then how many activated proto-oncogenes and inactivated tumor suppressors were required? What were the genetic steps needed to con
vert a normal cell into a cancer cell? For human cancers, these questions could not be answered experimentally. One could not, after all, proactively “create” a human cancer and follow the activation and inactivation of genes. But the questions could be answered retrospectively. In 1988, using human specimens, a physician-scientist named Bert Vogelstein at Johns Hopkins Medical School in Baltimore set out to describe the number of genetic changes required to initiate cancer. The query, in various incarnations, would preoccupy Vogelstein for nearly two decades.

  Vogelstein was inspired by the observations made by George Papanicolaou and Oscar Auerbach in the 1950s. Both Papanicolaou and Auerbach, working on different cancers, had noted that cancer did not arise directly out of a normal cell. Instead, cancer often slouched toward its birth, undergoing discrete, transitional stages between the fully normal and the frankly malignant cell. Decades before cervical cancer evolved into its fiercely invasive incarnation, whorls of noninvasive premalignant cells could be observed in the tissue, beginning their first steps in the grisly march toward cancer. (Identifying and eradicating this premalignant stage before the cancer spreads is the basis for the Pap smear.) Similarly, Auerbach had noted, premalignant cells were seen in smokers’ lungs long before lung cancer appeared. Colon cancer in humans also underwent graded and discrete changes in its progression, from a noninvasive premalignant lesion called an adenoma to the highly invasive terminal stage called an invasive carcinoma.

  Vogelstein chose to study this progression in colon cancer. He collected samples from patients representing each of the stages of colon cancer. He then assembled a series of four human cancer genes—oncogenes and tumor suppressors—and assessed each stage of cancer in his samples for activations and inactivations of these four genes.*