The Emperor of All Maladies

by Siddhartha Mukherjee

  Varmus and Bishop, working with Deborah Spector and Dominique Stehelin, probed more cells, and again the src gene appeared in them: in duck cells, quail cells, and geese cells. Closely related homologues of the src gene were strewn all over the bird kingdom; each time Varmus’s team looked up or down an evolutionary branch, they found some variant of src staring back. Soon, the UCSF group was racing through multiple species to look for homologues of src. They found src in the cells of pheasants, turkeys, mice, rabbits, and fish. Cells from a newborn emu at the Sacramento zoo had src. So did sheep and cows. Most important, so did human cells. “Src,” Varmus wrote in a letter in 1976, “. . . is everywhere.”

  But the src gene that existed in normal cells was not identical to the viral src. When Hidesaburo Hanafusa, a Japanese virologist at Rockefeller University in New York, compared the viral src gene to the normal cellular src gene, he found a crucial difference in the genetic code between the two forms of src. Viral src carried mutations that dramatically affected its function. Viral src protein, as Erikson had found in Colorado, was a disturbed, hyperactive kinase that relentlessly tagged proteins with phosphate groups and thus provided a perpetually blaring “on” signal for cell division. Cellular src protein possessed the same kinase activity, but it was far less hyperactive; in contrast to viral src, it was tightly regulated—turned “on” and turned “off”—during cell division. The viral src protein, in contrast, was a permanently activated switch—“an automaton,” as Erikson described it—that had turned the cell into a dividing machine. Viral src—the cancer-causing gene—was cellular src on overdrive.

  A theory began to convulse out of these results, a theory so magnificent and powerful that it would explain decades of disparate observations in a single swoop: perhaps src, the precursor to the cancer-causing gene, was endogenous to the cell. Perhaps viral src had evolved out of cellular src. Retrovirologists had long believed that the virus had introduced an activated src into normal cells to transform them into malignant cells. But the src gene had not originated in the virus. It had originated from a precursor gene that existed in a cell—in all cells. Cancer biology’s decades-long hunt had started with a chicken and ended, metaphorically, in the egg—in a progenitor gene present in all human cells.

  Rous’s sarcoma virus, then, was the product of an incredible evolutionary accident. Retroviruses, Temin had shown, shuttle constantly out of the cell’s genome: RNA to DNA to RNA. During this cycling, they can pick up pieces of the cell’s genes and carry them, like barnacles, from one cell to another. Rous’s sarcoma virus had likely picked up an activated src gene from a cancer cell and carried it in the viral genome, creating more cancer. The virus, in effect, was no more than an accidental courier for a gene that had originated in a cancer cell—a parasite parasitized by cancer. Rous had been wrong—but spectacularly wrong. Viruses did cause cancer, but they did so, typically, by tampering with genes that originate in cells.

  Science is often described as an iterative and cumulative process, a puzzle solved piece by piece, with each piece contributing a few hazy pixels of a much larger picture. But the arrival of a truly powerful new theory in science often feels far from iterative. Rather than explain one observation or phenomenon in a single, pixelated step, an entire field of observations suddenly seems to crystallize into a perfect whole. The effect is almost like watching a puzzle solve itself.

  Varmus and Bishop’s experiments had precisely such a crystallizing, zippering effect on cancer genetics. The crucial implication of the Varmus and Bishop experiment was that a precursor of a cancer-causing gene—the “proto-oncogene,” as Bishop and Varmus called it—was a normal cellular gene. Mutations induced by chemicals or X-rays caused cancer not by “inserting” foreign genes into cells, but by activating such endogenous proto-oncogenes.

  “Nature,” Rous wrote in 1966, “sometimes seems possessed of a sardonic humor.” And the final lesson of Rous sarcoma virus had been its most sardonic by far. For nearly six decades, the Rous virus had seduced biologists—Spiegelman most sadly among them—down a false path. Yet the false path had ultimately circled back to the right destination—from viral src toward cellular src and to the notion of internal proto-oncogenes sitting omnipresently in the normal cell’s genome.

  In Lewis Carroll’s poem, when the hunters finally capture the deceptive Snark, it reveals itself not to be a foreign beast, but one of the human hunters sent to trap it. And so it had turned out with cancer. Cancer genes came from within the human genome. Indeed the Greeks had been peculiarly prescient yet again in their use of the term oncos. Cancer was intrinsically “loaded” in our genome, awaiting activation. We were destined to carry this fatal burden in our genes—our own genetic “oncos.”

  Varmus and Bishop were awarded the Nobel Prize for their discovery of the cellular origin of retroviral oncogenes in 1989. At the banquet in Stockholm, Varmus, recalling his former life as a student of literature, read lines from the epic poem Beowulf, recapitulating the slaying of the dragon in that story: “We have not slain our enemy, the cancer cell, or figuratively torn the limbs from his body,” Varmus said. “In our adventures, we have only seen our monster more clearly and described his scales and fangs in new ways—ways that reveal a cancer cell to be, like Grendel, a distorted version of our normal selves.”

  * The term oncogene had been coined earlier by two NCI scientists, Robert Huebner and George Todaro, in 1969, although on scant evidence.

  †Art Levinson, in Mike Bishop’s lab at UCSF, also discovered this phosphorylating activity; we will return to Levinson’s discovery in later pages.

  The Wind in the Trees

  The fine, fine wind that takes its course through the chaos of the world

  Like a fine, an exquisite chisel, a wedge-blade inserted . . .

  —D. H. Lawrence

  The developments of the summer of 1976 drastically reorganized the universe of cancer biology, returning genes, again, to its center. Harold Varmus and Michael Bishop’s proto-oncogene theory provided the first cogent and comprehensive theory of carcinogenesis. The theory explained how radiation, soot, and cigarette smoke, diverse and seemingly unrelated insults, could all initiate cancer—by mutating and thus activating precursor oncogenes within the cell. The theory made sense of Bruce Ames’s peculiar correlation between carcinogens and mutagens: chemicals that cause mutations in DNA produce cancers because they alter cellular proto-oncogenes. The theory clarified why the same kind of cancer might arise in smokers and nonsmokers, albeit at different rates: both smokers and nonsmokers have the same proto-oncogenes in their cells, but smokers develop cancer at a higher rate because carcinogens in tobacco increase the mutation rate of these genes.

  But what did human cancer genes look like? Tumor virologists had found src in viruses and then in cells, but surely other endogenous proto-oncogenes were strewn about in the human cellular genome.

  Genetics has two distinct ways to “see” genes. The first is structural: genes can be envisioned as physical structures—pieces of DNA lined up along chromosomes, just as Morgan and Flemming had first envisioned them. The second is functional: genes can be imagined, à la Mendel, as the inheritance of traits that move from one generation to the next. In the decade between 1970 and 1980, cancer genetics would begin to “see” cancer-causing genes in these two lights. Each distinct vision would enhance the mechanistic understanding of carcinogenesis, bringing the field closer and closer to an understanding of the core molecular aberration in human cancers.

  Structure—anatomy—came first. In 1973, as Varmus and Bishop were launching their initial studies on src, a hematologist in Chicago, Janet Rowley, saw a human cancer gene in a physical form. Rowley’s specialty was studying the staining patterns of chromosomes in cells in order to locate chromosomal abnormalities in cancer cells. Chromosome staining, the technique she had perfected, is as much an art as a science. It is also an oddly anachronistic art, like painting with tempera in an age of digital prints. At a time when cancer ge
netics was zooming off to explore the world of RNA, tumor viruses, and oncogenes, Rowley was intent on dragging the discipline back to its roots—to Boveri’s and Flemming’s chromosomes dyed in blue. Piling anachronism upon anachronism, the cancer she had chosen to study was chronic myelogenous leukemia (CML)—Bennett’s infamous “suppuration of blood.”

  Rowley’s study was built on prior work by a duo of pathologists from Philadelphia who had also studied CML. In the late 1950s, Peter Nowell and David Hungerford had found an unusual chromosomal pattern in this form of leukemia: the cancer cells bore one consistently shortened chromosome. Human cells have forty-six chromosomes—twenty-three matched pairs—one inherited from each parent. In CML cells, Nowell found that one copy of the twenty-second chromosome had its head lopped off. Nowell called the abnormality the Philadelphia chromosome after the place of its discovery. But Nowell and Hungerford could not understand where the decapitated chromosome had come from, or where its missing “head” had gone.

  Rowley, following this study, began to trace the headless chromosome in her CML cells. By laying out exquisitely stained photographs of CML chromosomes enlarged thousands of times—she typically spread them on her dining table and then leaned into the pictures, hunting for the missing pieces of the infamous Philadelphia chromosome—Rowley found a pattern. The missing head of chromosome twenty-two had attached itself elsewhere—to the tip of chromosome nine. And a piece of chromosome nine had conversely attached itself to chromosome twenty-two. This genetic event was termed a translocation—the flip-flop transposition of two pieces of chromosomes.

  Rowley examined case after case of CML patients. In every single case, she found this same translocation in the cells. Chromosomal abnormalities in cancer cells had been known since the days of von Hansemann and Boveri. But Rowley’s results argued a much more profound point. Cancer was not disorganized chromosomal chaos. It was organized chromosomal chaos: specific and identical mutations existed in particular forms of cancer.

  Chromosomal translocations can create new genes called chimeras by fusing two genes formerly located on two different chromosomes—the “head” of chromosome nine, say, fused with the “tail” of a gene in chromosome thirteen. The CML translocation, Rowley postulated, had created such a chimera. Rowley did not know the identity or function of this new chimeric monster. But she had demonstrated that a novel, unique genetic alteration—later found to be an oncogene—could exist in a human cancer cell, revealing itself purely by virtue of an aberrant chromosome structure.

  In Houston, Alfred Knudson, a Caltech-trained geneticist, also “saw” a human cancer-causing gene in the early 1970s, although in yet another distinct sense.

  Rowley had visualized cancer-causing genes by studying the physical structure of the cancer cell’s chromosomes. Knudson concentrated monastically on the function of a gene. Genes are units of inheritance: they shuttle properties—traits—from one generation to the next. If genes cause cancer, Knudson reasoned, then he might capture a pattern in the inheritance of cancer, much as Mendel had captured the idea of a gene by studying the inheritance of flower color or plant height in peas.

  In 1969, Knudson moved to the MD Anderson Cancer Center in Texas, where Freireich had set up a booming clinical center for childhood cancers. Knudson needed a “model” cancer, a hereditary malignancy whose underlying pattern of inheritance would reveal how cancer-causing genes worked. The natural choice was retinoblastoma, the odd, rare variant of eye cancer that de Gouvêa had identified in Brazil with its striking tendency to erupt in the same family across generations.

  Retinoblastoma is a particularly tragic form of cancer, not just because it assaults children but because it assaults the quintessential organ of childhood: the tumor grows in the eye. Afflicted children are sometimes diagnosed when the world around them begins to blur and fade. But occasionally the cancer is incidentally found in a child’s photograph when the eye, lit by a camera flash, glows eerily like a cat’s eyes in lamplight, revealing the tumor buried behind the lens. Left untreated, the tumor will crawl backward from the eye socket into the optic nerve, and then climb into the brain. The primary methods of treatment are to sear the tumor with high doses of gamma radiation or to enucleate the eye surgically, leaving behind an empty socket.

  Retinoblastoma has two distinct variants, an inherited “familial” form and a sporadic form. De Gouvêa had identified the familial form. Children who suffer from this familial or inherited form may carry strong family histories of the disease—fathers, mothers, cousins, siblings, and kindred affected—and they typically develop tumors in both eyes, as in de Gouvêa’s case from Rio. But the tumor also arises in children with no family history of the disease. Children with this sporadic form never carry a history in the family and always have a tumor in only one eye.

  This pattern of inheritance intrigued Knudson. He wondered whether he could discern a subtle difference in the development of cancer between the sporadic and the inherited versions using mathematical analyses. He performed the simplest of experiments: he grouped children with the sporadic form into one cohort and children with the familial form in a second. And sifting through old hospital records, Knudson tabulated the ages in which the disease struck the two groups, then plotted them as two curves. Intriguingly, he found that the two cohorts developed the cancers at different “velocities.” In inherited retinoblastoma, cancer onset was rapid, with diagnosis typically two to six months after birth. Sporadic retinoblastoma typically appeared two to four years after birth.

  But why did the same disease move with different velocities in different children? Knudson used the numbers and simple equations borrowed from physics and probability theory to model the development of the cancer in the two cohorts. He found that the data fit a simple model. In children with the inherited form of retinoblastoma, only one genetic change was required to develop the cancer. Children with the sporadic form required two genetic changes.

  This raised another puzzling question: why was only one genetic change needed to unleash cancer in the familial case, while two changes were needed in the sporadic form? Knudson perceived a simple, beautiful explanation. “The number two,” he recalled, “is the geneticist’s favorite number.” Every normal human cell has two copies of each chromosome and thus two copies of every gene. Every normal cell must have two normal copies of the retinoblastoma gene—Rb. To develop sporadic retinoblastoma, Knudson postulated, both copies of the gene needed to be inactivated through a mutation in each copy of the Rb gene. Hence, sporadic retinoblastoma develops at later ages because two independent mutations have to accumulate in the same cell.

  Children with the inherited form of retinoblastoma, in contrast, are born with a defective copy of Rb. In their cells, one gene copy is already defective, and only a single additional genetic mutation is needed before the cell senses the change and begins to divide. These children are thus predisposed to the cancer, and they develop cancer faster, producing the “rapid velocity” tumors that Knudson saw in his statistical charts. Knudson called this the two-hit hypothesis of cancer. For certain cancer-causing genes, two mutational “hits” were needed to provoke cell division and thus produce cancer.

  Knudson’s two-hit theory was a powerful explanation for the inheritance pattern of retinoblastoma, but at first glance it seemed at odds with the initial molecular understanding of cancer. The src gene, recall, requires a single activated copy to provoke uncontrolled cell division. Knudson’s gene required two. Why was a single mutation in src sufficient to provoke cell division, while two were required for Rb?

  The answer lies in the function of the two genes. Src activates a function in cell division. The mutation in src, as Ray Erikson and Hidesaburo Hanafusa had discovered, creates a cellular protein that is unable to extinguish its function—an insatiable, hyperactive kinase on overdrive that provokes perpetual cell division. Knudson’s gene, Rb, performs the opposite function. It suppresses cell proliferation, and it is the inactivation of such a gene (b
y virtue of two hits) that unleashes cell division. Rb, then, is a cancer suppressor gene—the functional opposite of src—an “anti-oncogene,” as Knudson called it.

  “Two classes of genes are apparently critical in the origin of the cancers of children,” he wrote. “One class, that of oncogenes, acts by virtue of abnormal or elevated activity. . . . The other class, that of anti-oncogenes [or tumor suppressors], is recessive in oncogenesis; cancer results when both normal copies have been mutated or deleted. Some persons carry one such mutation in the germline and are highly susceptible to tumor because only one somatic event is necessary. Some children, even though carrying no such mutation in the germline, can acquire tumor as a result of two somatic events.”

  It was an exquisitely astute hypothesis spun, remarkably, out of statistical reasoning alone. Knudson did not know the molecular identity of his phantasmic anti-oncogenes. He had never looked at a cancer cell to “see” these genes; he had never performed a biological experiment to pin down Rb. Like Mendel, Knudson knew his genes only in a statistical sense. He had inferred them, as he put it, “as one might infer the wind from the movement of the trees.”

  By the late 1970s, Varmus, Bishop, and Knudson could begin to describe the core molecular aberration of the cancer cell, stitching together the coordinated actions of oncogenes and anti-oncogenes. Cancer genes, Knudson proposed, came in two flavors. “Positive” genes, such as src, are mutant activated versions of normal cellular genes. In normal cells, these genes accelerate cell division, but only when the cell receives an appropriate growth signal. In their mutant form, these genes are driven into perpetual hyperactivity, unleashing cell division beyond control. An activated proto-oncogene, to use Bishop’s analogy, is “a jammed accelerator” in a car. A cell with such a jammed accelerator careens down the path of cell division, unable to cease mitosis, dividing and dividing again relentlessly.