What causes cancer?

Tobacco smoke, most people would say. Probably too much alcohol, sunshine or grilled meat; infection with cervical papillomaviruses; asbestos. All have strong links to cancer, certainly. But they cannot be root causes. Much of the population is exposed to these carcinogens, yet only a tiny minority suffers dangerous tumors as a consequence.

A cause, by definition, leads invariably to its effect. The immediate cause of cancer must be some combination of insults and accidents that induces normal cells in a healthy human body to turn malignant, growing like weeds and sprouting in unnatural places.

At this level, the cause of cancer is not entirely a mystery. In fact, a decade ago many geneticists were confident that science was homing in on a final answer: cancer is the result of cumulative mutations that alter specific locations in a cell's DNA and thus change the particular proteins encoded by cancer-related genes at those spots. The mutations affect two kinds of cancer genes. The first are called tumor suppressors. They normally restrain cells' ability to divide, and mutations permanently disable the genes. The second variety, known as oncogenes, stimulate growth—in other words, cell division. Mutations lock oncogenes into an active state. Some researchers still take it as axiomatic that such growth-promoting changes to a small number of cancer genes are the initial event and root cause of every human cancer.

For the past few years, however, prominent oncologists have increasingly challenged that theory. No one questions that cancer is ultimately a disease of the DNA. But as biologists trace tumors to their roots, they have discovered many other abnormalities at work inside the nuclei of cells that, though not yet cancerous, are headed that way. Whole chromosomes, each containing 1,000 or more genes, are often lost or duplicated. Pieces of chromosomes are frequently scrambled, truncated or fused together. Chemical additions to the DNA, or to the histone proteins around which it coils, somehow silence important genes, but in a reversible process quite different from mutation. And scans of the genomes of malignant cells within tumors have found that they typically harbor myriad rare mutations rather than a handful of common genetic alterations.

The accumulating evidence has spawned new hypotheses that compete with the standard dogma to explain what changes come first and which aberrations matter most in the decadelong transformation of a cell and its descendants from well-behaved tissue to invasive tumor. The challengers dispute the dominant view of the disease as the product of a defined genetic state. They argue that it is more useful to think of cancer as the consequence of a chaotic process, a combination of Murphy's Law and Darwin's Law: anything that can go wrong will, and in a competitive environment, the most prolific variants will dominate.

Despite that shared underlying principle, the new theories make different predictions about what kind of treatments will work best. Some suggest that many cancers could be prevented altogether by better screening, changes in diet, and new drugs—or even by old drugs, such as aspirin. Other theories cast doubt on that hope.

Marks of Malignancy

A WORKABLE THEORY of cancer has to explain both why it is predominantly a disease of old age and why we do not all die from it. A 70-year-old is roughly 100 times as likely to be diagnosed with a malignancy as a 19-year-old is. Yet most people make it to old age without getting cancer.

Biologists estimate that more than 10 million billion cells must cooperate to keep a human being healthy over the course of an 80-year life span. If any one of those myriad cells could give rise to a tumor, why is it that fewer than half the population will ever contract a cancer serious enough to catch a doctor's attention?

One explanation is that a cell must acquire several extraordinary skills to be malignant. “Five or six different regulatory systems must be perturbed in order for a normal cell to grow as a cancer,” asserts Robert A. Weinberg of the Whitehead Institute at the Massachusetts Institute of Technology. In a 2002 review paper, he and William C. Hahn of the Dana-Farber Cancer Institute in Boston argued that all life-threatening cancers manifest at least six such deranged abilities. (Although Weinberg is one of the founding proponents of the standard paradigm, even those who challenge that theory tend to agree with this view.)

For example, cancer cells continue dividing in situations in which normal cells would quietly wait for a special chemical signal—say, from an injured neighbor. Somehow they counterfeit these progrowth messages. Conversely, tumor cells must ignore “stop dividing” commands that are sent out by the adjacent tissues they squeeze and by their own internal aging mechanisms.

All cancerous cells have serious problems of some sort with their DNA, and as they double again and again, many cells in the resulting colony end up far from the blood vessels that supply oxygen and nutrients. Such stresses trigger autodestruct mechanisms in healthy cells. Tumor cells find some way to avoid this kind of suicide. Then they have to persuade nearby blood vessels to build the infrastructure they need to thrive.

A fifth derangement that almost all cancers acquire is immortality. A culture of normal human cells stops dividing after 50 to 70 generations. That is more than enough doublings to sustain a person through even a century of healthy life. But the great majority of cells in tumors quickly die of their genetic defects, so those rare few that survive must reproduce indefinitely if the tumor is to grow. The survivors do so in part by manipulating their telomeres, gene-free complexes of DNA and protein that protect the ends of each chromosome.

Tumors that develop these five faculties are trouble, but they are probably not deadly. It is a sixth property, the ability to invade nearby tissue and then metastasize to distant parts of the body, that gives cancer its lethal character. Local invasions can usually be removed surgically. But nine of every 10 deaths from the disease are the result of metastases.

Only an elite few cells in a tumor seem to acquire this ability to detach from the initial mass, float through the circulation and start a new colony in a different organ from the one that gave birth to them. Unfortunately, by the time they are discovered, many cancers have already metastasized—including, in the U.S., 76 percent of lung cancers, 55 percent of colorectal cancers, and 37 percent of breast cancers. By then the prognosis is frequently grim.

The Order of Disorder

DOCTORS COULD CATCH incipient tumors sooner if scientists could trace the steps that cells take down the road to cancer after the initial assault to their DNA by a carcinogen or some random biochemical mishap. Researchers broadly agree on the traits of the diseased cells that emerge from the journey. It is the propelling force and the order of each milestone that are under active debate.

The dominant paradigm for 30 years has been that tumors grow in spurts of mutation and expansion. Genetic damage to a cell deletes or disrupts a tumor suppressor gene—RB, p53, BRCA2 and APC are among the best known—thereby suppressing proteins that normally ensure the integrity of the genome and the process of cell division. Alternatively, a mutation may increase the activity of an oncogene—such as HER2/NEU, c-fos or c-erbB3—whose proteins then stimulate the cell to reproduce.

Changes to cancer genes free the cell of one or more normal restraints, allowing it to outbreed its neighbors. The cell passes abnormalities in its DNA sequence on to its descendants, which become a kind of clone army that grows to the limits of its capacity. Eventually another random mutation to a cancer gene looses another shackle, initiating another burst of growth.

Cells normally have two copies of every chromosome—one from the mother, the other from the father—and thus two copies, or alleles, of every gene. (In males, the single X and Y chromosomes are notable exceptions.) A mutation to just one allele is enough to activate an oncogene permanently. But it takes two hits to knock out both alleles of a tumor suppressor gene. Four to 10 mutations in the right genes can transform any cell. Or so the theory goes.

The mutant-gene paradigm gained almost universal acceptance because it explained very well what scientists saw in their experiments on genetically engineered mice and human cell cultures. But new technologies now allow researchers to survey large fractions of the genomes of cancerous and precancerous cells taken directly from people. Many recent observations seem to contradict the idea that mutations to a few specific genes lie at the root of all cancers.

Unexplained Phenomena

IN 2003, for example, a team led by Michael F. Clarke, then at the University of Michigan at Ann Arbor and now at Stanford University, reported that it had identified distinguishing marks for a rare subset of cells within human breast cancers that can form new tumors. As few as 100 cells of this type quickly spawned disease when injected into mice lacking an immune system. Tens of thousands of other cells, harvested from the same nine breast malignancies but lacking the telltale marks, failed to do so. These were the first tumor-initiating cells ever isolated for solid tumors, and later studies confirmed that for many common human cancers, just a small fraction of the cells in a tumor are responsible for its growth and metastasis—and so for the illness and death of the patient. The discovery of such cells, which some began calling “cancer stem cells,” posed a problem for the mutant-gene theory of cancer. If mutations, which are copied from a cell to its progeny, give tumor cells their deranged powers, then shouldn't all clones in the army be equally powerful?

In fact, most tumors are not masses of identical clones. On the contrary, closer examination has revealed amazing genetic diversity among their cells, some of which are so different from normal human cells (and from one another) that they might fairly be called new species.

In 2006 Bert Vogelstein of Johns Hopkins University and a large team of collaborators used new genetic sequencing technologies to examine more than 13,000 genes in samples taken from 11 breast tumors and 11 colorectal cancers. They found noninherited mutations in 1,149 (nearly 9 percent) of those genes. Even more startling than the sheer number of genes mutated was the lack of consistency from one tumor to the next—or even from one cell to another within the same tumor. Not a single gene was mutated in more than 5 percent of the tumors.

Moreover, some of the most commonly altered cancer genes have oddly inconsistent effects. Vogelstein's group has also reported that the much studied oncogenes c-fos and c-erbB3 are curiously less active in tumors than they are in nearby normal tissues. The tumor suppressor gene RB was shown to be hyperactive—not disabled—in some colon cancers, and, perversely, it appears to protect those tumors from their autodestruct mechanisms.

The “two hit” hypothesis—that both alleles of a tumor suppressor gene must be deactivated—has also been upended by the discovery of a phenomenon called haploinsufficiency. In some cancers, it turns out, tumor suppressors are not mutated at all. Their output is simply reduced, and that seems to be enough to push cells toward malignancy. This effect has now been seen for more than a dozen tumor suppressor genes, and investigators expect to find many more like them. Searching for the mere presence or absence of a gene's protein is too simplistic. Dosage matters.

Beyond Mutation

RESEARCHERS are now looking more closely at other phenomena, besides errors in a gene's DNA sequence, that can dramatically alter the dosage of a protein in a cell. The loss or gain of a chromosome (or part of one) containing the gene can do this. So can tweaks to regulatory factors that control how often a gene is translated into protein, as well as epigenetic phenomena that alter gene activity by reversible means. All these changes are nearly ubiquitous in established cancers.

“If you look at most solid tumors in adults, it looks like someone set off a bomb in the nucleus,” Hahn says. “In most cells, there are big pieces of chromosomes hooked together and duplications or losses of whole chromosomes.” Cancer biologists call such cells aneuploid.

Almost a century ago German biologist Theodor Boveri noticed the strange imbalance in cancer cells between the numbers of maternal versus paternal chromosomes. He even suggested that aneuploid cells might cause the disease. But scientists could find no recurrent pattern to the chromosomal chaos—indeed, the genome of a typical cancer cell is not merely aneuploid but unstable as well, changing every few generations. So Boveri's idea was dropped as the search for oncogenes started to bear fruit. The aneuploidy and massive genomic instability inside tumor cells were dismissed as side effects of cancer, not prerequisites.

But the oncogene/tumor suppressor gene hypothesis has also failed, despite three decades of effort, to identify a particular set of gene mutations that occurs in every instance of any of the most common and deadly kinds of human cancer. So now, Hahn says, “the question is which comes first: mutations or aneuploidy?”

There are at least four competing answers. The newest and most controversial theory—that cancers often arise from the very stem cells that give healthy organs their growth and healing abilities—is discussed elsewhere in this issue [see “Stem Cells: The Real Culprits in Cancer?” by Michael F. Clarke and Michael W. Becker, on page 40]. So we will focus on the three more established ideas. Let us call them the modified dogma, the early instability theory and the all-aneuploidy theory. Encouragingly, all four of these theories seem to be converging as they bend to accommodate new experimental results.

The modified form of the standard dogma revives an idea proposed in 1974 by Lawrence A. Loeb, now at the University of Washington. He and other geneticists have estimated that, on average, random mutation will affect just one gene in any given cell over the course of a lifetime. Something—a carcinogen, reactive oxidants, or perhaps a malfunction in the DNA duplication and repair machinery of the cell—must dramatically accelerate the mutation rate, Loeb argues.

For many years, he suggested that “early during the genesis of cancer there are enormous numbers of random mutations—10,000 to 100,000 per cell,” but he had little evidence to support the idea. In 2006, however, technology advanced to the point that Loeb was able to test his hypothesis by comparing the rate at which a noncoding portion of the p53 gene suffered small mutations in both normal and malignant human cells. Cancerous cells, his test revealed, harbored anywhere from 65 to 475 mutations per 100 million nucleotides, whereas normal cells had four or fewer. That seems to be a far stretch from the 100,000-fold increase in mutation rate that Loeb had anticipated, but it is nonetheless an important discovery that demands explanation.

Loeb's modified dogma thus may add a prologue to the long-accepted life history of cancer. But the most important plot points in that story would still remain the same: mutations to genes that serve to increase the reproductive success of cells. Mangled and ever changing chromosomes are, in this narrative, mere fortuitous by-products.

Unstable from the Outset

VOGELSTEIN of Johns Hopkins and his former co-worker Christoph Lengauer, now at Novartis, proposed an alternative theory in which chromosomal instability can occur early on. The genetic flux then combines forces with natural selection to produce a benign growth that may later be converted to an invasive malignancy and life-threatening metastases.

In their hypothesis, there are several “master” genes whose function is critical for a cell to reproduce correctly. If just one of these genes is disabled, either epigenetically or by mutation, the cell stumbles each time it attempts the carefully choreographed dance of cell division, muddling some of the chromosomes into an aneuploid state. One result is to increase 100,000-fold the rate at which cells randomly lose one of the two alleles of their genes. For a tumor suppressor gene, a lost allele may effectively put the gene out of commission, either because the remaining copy is already mutated or because of the haploinsufficiency effect. Lengauer and Vogelstein still assume that some cancer genes must be altered before a malignancy can erupt.

Some observations do support the early instability theory. In 2000 Lengauer's laboratory examined colon adenomas—benign polyps that occasionally turn malignant—and observed that more than 90 percent had extra or missing pieces of at least one chromosome. More than half had lost the long arm of chromosome 5, home to the APC tumor suppressor gene, long implicated in the formation of colon cancer. Other researchers have discovered similarly aberrant chromosomes in precancerous growths taken from the stomach, esophagus and breast.

The early instability theory still has some loose ends, however. How can cells with shifty chromosomes outcompete their stable counterparts? Under normal conditions, they probably do not, suggests immunologist Jarle Breivik of the University of Oslo. But in a “war zone,” where a carcinogen or other stressor is continually inflicting damage to cells, normal cells stop dividing until they have completed repairs to their DNA. Genetically unstable cells get that way because their DNA repair systems are already broken. So they simply ignore the damage, keep on proliferating, and thus pull ahead, Breivik hypothesizes.

But what jumbles the chromosomes in the first place? No genes have yet been conclusively identified as master genes, although several strong suspects have surfaced. Beth A. A. Weaver of the University of Wisconsin–Madison and her collaborators may have uncovered a clue in their studies of a gene called CENP-E. The protein produced from this gene is one of several that are crucial for guiding the chromosomes as they replicate and separate during cell division. At a meeting of the Society for Chromosomal Cancer Research in Oakland, Calif., this year, Weaver reported that in mice embryos missing one of their two copies of CENP-E (and thus making only half the normal amount of CENP-E protein), chromosomes quickly went askew. Within a couple of weeks, more than half the embryonic cells were aneuploid. And when these animals grew up, they developed many more spleen and lung cancers than did control mice with normal CENP-E genes.

Aneuploidy All the Way Down

ON THE OTHER HAND, maybe cells can become malignant even before any master genes, oncogenes or tumor suppressor genes are mutated. Peter Duesberg of the University of California, Berkeley, has put forth a third theory: nearly all cancer cells are aneuploid (leukemia being one exception) because they start that way. Lots of things can interfere with a dividing cell so that one of its daughter cells is cheated of its normal complement of 46 chromosomes and the other daughter is endowed with a bonus. Asbestos fibers, Duesberg notes, can physically disrupt the process.

Most aneuploid cells are stillborn or growth-retarded. But in the rare survivor, he suggests, the dosage of thousands of genes is altered. That corrupts teams of enzymes that synthesize and maintain DNA. Breaks appear in the double helix, destabilizing the genome further. “The more aneuploid the cell is, the more unstable it is, and the more likely it will produce new combinations of chromosomes that will allow it to grow anywhere,” Duesberg explains.

Unlike the three other theories, the all-aneuploidy hypothesis predicts that the emergence and progress of a tumor are more closely connected to the assortment of chromosomes in its cells than to the mutations in the genes on those chromosomes. Some observations do seem to corroborate the idea.

Thomas Ried, chief of cancer genomics at the National Cancer Institute, has obtained supporting evidence in humans from his investigation of aneuploidy in cervical and colorectal cancers, which he presented at the Oakland conference. “Unequivocally, there are recurrent patterns of genomic imbalances,” Ried avers. “Every single case of [nonhereditary] colorectal cancer, for example, has gains of chromosomes 7, 8, 13 or 20—or a loss of 18. In cervical cancer, aneuploidy of chromosome 3 happens very early, and those cells seem to have a selective advantage.” Ried finds the average number of abnormal chromosomes increasing gradually from 0.2 in a normal cell to 12 in the cells of metastatic colon tumors.

“So I actually think Duesberg is right that aneuploidy can be the first genetic aberration in cancer cells,” Ried says. “But he also argues that no gene mutations are required. This is simply not true.”

Stopping Cancer at Its Roots

NEITHER THE standard dogma nor any of the new theories that challenge it can fully untangle the knotted roots of the 100-odd diseases we call cancer and explain them as variations of a single principle. And all the theories will need to be expanded to incorporate the still mysterious role of epigenetic phenomena, which may be pivotal.

It is important to determine which of the ideas is more right than the others, because they each make different predictions about the kinds of therapy that will succeed best against the most common and lethal cancers. In the standard view, tumors are in effect addicted to the proteins produced by oncogenes and are poisoned by tumor suppressor proteins. Medicines should therefore be designed to break the addiction or supply the poison. Indeed, this strategy is exploited by some newer drugs, such as Gleevec (for rare forms of leukemia and stomach cancer) and Herceptin (for one variety of advanced breast cancer).

But all existing therapies, including Gleevec and Herceptin, fail in some patients because their tumors evolve into a resistant strain. Loeb fears that there may be no easy way around that problem; his latest work, after all, suggests that a tumor of any significant size harbors up to a trillion random mutations. “If I am right, then within any given tumor, there will be cells with random mutations that protect them from any treatment you can conceive,” Loeb says. “So the best you can hope for is to delay the tumor's growth. You are not going to cure it.”

For the elderly—who, after all, are the main victims of cancer—a sufficient delay may be as good as a cure. And even better than slowing the growth of a tumor would be to delay its formation in the first place. If Lengauer, Weaver and others succeed in identifying master genes, then it should also be possible to make drugs that protect or restore their function. The Johns Hopkins group has already licensed some of its cell lines to the pharmaceutical industry to use in drug screening.

Screening of a different kind may be the best approach if the all-aneuploidy theory is correct. There are no known means of selectively killing cells with abnormal chromosomes. But a biopsy that turns up a surfeit of aneuploid cells might warrant careful monitoring or even preventive surgery in certain cases. Duesberg suggests that foods, drugs and chemicals should be tested to identify compounds that cause aneuploidy. And Weaver found that although CENP-E-deficient mouse cells were highly tumorigenic if the cells were moderately aneuploid, cells that were highly aneuploid actually suppressed tumor formation. It may be possible to design drugs that kill cancer cells by exacerbating their aneuploidy.

One day science will produce a definitive answer to the question of what causes cancer. It will probably be a very complicated answer, and it may force us to shift our hope from drugs that cure the disease to medicines that prevent it. Even without a clear understanding of why, doctors have discovered that a daily baby aspirin seems to prevent colon adenomas and to lower the risk of one kind of breast cancer in some adults. The effect is small. But it is a step from chemotherapy toward a better alternative: chemoprevention.