Diagnosis. Chemotherapy. Radiation. Slow painful death. No more. A new era of cancer treatment is dawning. Meet three scientists who are using the revelations of the Human Genome Project to reshape medicine.
When Richard Nixon declared war on cancer in 1971, he had no way of knowing that the fight would turn into medicine's own Vietnam. At the time, cancer seemed like a relatively simple disease. Something in the body was triggering cells to divide abnormally fast; find that mechanism and shut it off, and you'd have a cure. It seemed reasonable enough, but like Agent Smith in The Matrix, cancer turned out to be a malignant rogue with an uncanny ability to survive. It could be burned, poisoned, and eviscerated beyond recognition – only to pop up again somewhere nearby. It could also copy itself at will. In the decades since, scientists have managed to identify a handful of genes that seem to have mutated in cancer patients, but in each case their research led to a paradox. The same genes that triggered the production of cancerous cells also controlled vital life processes. Kill the cancer cell, researchers worried, and you end up halting the thousands of daily cell divisions necessary for survival. How do you delete a piece of the program without crashing the whole thing?
By the 1990s, we were deep in the cancer quagmire. Survival rates hadn't budged in 20 years, and the scientific world appeared chastened. At the nadir, Gina Kolata, a respected science journalist with The New York Times, was vilified for hyping a drug that had so far worked only in mice. The word cure had all but disappeared from the discussion.
| 
Ian White
What we didn't know – couldn't have known before the human genome was sequenced in 2000 – was that cancer just looked like a simple disease. In fact, as we've learned, it's a malady of the genes, created when certain combinations of mutations come together simultaneously. The source of these mutations can be almost anything: inherited defects, toxic chemicals, cigarette smoke, sunlight, viruses, wine. As researchers have only recently discovered, the 33,000 genes that are identically present in each of our cells are active in different combinations, sometimes switching on and off many times a second. They are part of a complex signaling system that tells our cells when to grow or die, according to messages sent by a chain of chemical reactions. If a gene mutates, one or more of the nucleotides in its sequence changes. And because each gene is made of anywhere from several hundred to a few million base pairs, the number of possible mutations and combinations of mutations is almost infinite.
This helps explain why researchers never succeeded in finding a magic bullet to cure cancer. Cancer isn't produced by a single, consistent set of defects. A billion combinatorial roads snake through the genome, and hundreds, possibly thousands, of them arrive at the same place: the first day of the rest of your life in Cancerland.
It's one thing to battle in the dark, believing the fight is roughly equal. Now, the genome age has shined a light on what was once an elusive enemy. Finally, we can see the armies massed against us, a foe of almost impenetrable diversity, and virtually anyone would agree that it doesn't look good. Yet strangely, now that the battle has been joined, cancer researchers have grown almost euphoric. The National Cancer Institute is boldly promising, if not a cure, at least "the elimination of suffering and death due to cancer" by 2015; of more than 20 researchers I spoke with, all believed that the next decade would bring a revolution in cancer medicine.
At the root of this newfound optimism lie the very developments that revealed cancer's true nature in the first place: the sequencing of the human genome and the associated proliferation of new technologies – ranging from DNA chips to high-throughput gene-knockout techniques like RNA interference. (See "5 New Tools for Fighting Cancer," page 104.) Armed with these new weapons, researchers have begun an engagement that will more closely resemble the hunt for elusive al Qaeda operatives than a monolithic Cold War standoff.
The battle against cancer proceeds on three major fronts – drug development, cancer classification, and early detection. I visited researchers working in each area, and all were confident that the odds of victory had finally shifted in their favor. It's a breathtaking reversal after decades of disaffection and doubt. But as the NCI's carefully worded claim of an end to suffering suggests, it comes just as we're finding out how daunting a problem cancer really is.
Brian Druker's lab at Portland's Oregon Health & Science University hardly looks like the place where cancer gets cured. Low-ceilinged and bleak, it's more like a phobic's bathroom. Tinfoil-capped flasks clutter the countertops, next to econo-size boxes of latex gloves. Even the normal signs of lab cheer seem more wan than usual: A scattering of free drug company calendars hang awkwardly next to a lonely poster of chemical reactions detailing "The Hematopoietic Cascade."
Soft-spoken and upbeat, Druker radiates a kind of slow-motion sincerity that seems out of character for someone who for a decade spent 90-hour weeks in the lab. An avid though late-blooming athlete, he has a cycling jersey signed by Lance Armstrong on his office wall. Broad shoulders make Druker's head seem oddly small and narrow, like a loaf of bread balanced on a plank.
It turns out that Druker is the closest thing cancer research has to a hero. A few years ago, he tested half a dozen compounds on leukemia cells grown in a petri dish. One in particular caught his eye. Known only by the license plate-like tag ST1571, it demonstrated the unprecedented ability to kill cancerous cells while leaving healthy cells untouched. The pharmaceutical giant Novartis owned rights to ST1571, and Druker urged the company to develop the drug. But the potential market was small, about 5,000 leukemia patients annually, and the company dawdled for five years before agreeing to conduct a small Phase I trial.
The FDA's Phase I process is intended to test new drugs for their tolerance limit, the point at which toxic side effects occur. The ST1571 study quickly took off in a different direction. "It was like nothing I've ever seen," says Charles Sawyers, one of the doctors who oversaw the research with Druker. "Patients who were literally at death's door got out of bed and started walking around." Within a month of being given the drug, 53 of the original 54 patients had recovered. One ailing TV executive was back at work in two weeks. Side effects were trivial: A few patients got a rash, a few others felt nauseous.
Approved by the FDA in 2001, ST1571, renamed Gleevec, became the cancer war's new beachhead – a position from which to launch more attacks. "Researchers looked at Gleevec and saw the future of cancer treatment," says oncologist Andrew Simpson. This was not about fighting cancer using chemical or nuclear weapons – chemotherapy and radiation. It was about understanding how cancer functions at the cellular level, then building drugs that worked like smart bombs to target only the molecules causing the problem.
What made Gleevec unique was not just how well it worked. It was the first drug to knock out a single, specific aberration that sparked the growth of cancer cells. In this case, the target was a defective protein kinase – a catalyst that acts like a control switch to turn a variety of chemical reactions on or off. (These reactions affect how fast cells divide and when – or whether – they die.) The defective version of the protein, created when a particular pair of genes fuse together, is permanently stuck in the on position, creating a kind of hyperstimulated environment in which cancerous white blood cells grow abnormally and divide frantically.
| 
Richard Ballard (left) Todd Golub, LAB: Whitehead Institute, Mit, GOAL: To map the genetic variations in cancers; Brian Dunker, LAB: Oregon Heath & Science University, GOAL: To develop targeted drugs for treating leukemias; Sudhir Srivastava, LAB: National Cancer Institute, GOAL: To find biomarkers that siginal the earliest stages of cancer.
Druker's solution was to introduce a molecule with just one function: bind to the defective kinase, like a boor monopolizing a busy host at a dinner party. With the kinase blocked, the mutant cells stop multiplying and begin dying off, gradually restoring the body's natural equilibrium.
Because of the way the two genes had fused in this particular leukemia, there was, Druker acknowledges, "a bread crumb trail" leading clearly to the aberrant protein. Cancers in general are much more opaque – black boxes driven by an unknown series of mutations and rogue proteins. But Druker's work suggests that if the tools of genome science could be used to find similar Achilles' heels in other cancers, it would be possible to create precisely targeted, highly effective drugs. It would, in effect, change the way we treat cancer.
Even a short while ago, finding those crucial weaknesses would have taken decades of trial and error, if they were located at all. "In the old days," Druker says, "by which I mean eight years ago, we knew about a handful of genes that might be involved in some cancer processes. Now we know about hundreds."
At the heart of this newfound cornucopia is a technology known as the microarray, a small chip that allows researchers to see how genes behave in a tumor versus in normal tissue. Each of the altered genes – and the proteins they code for – is a potential target for drug development. The trick is figuring out which ones are critical to disabling a cancer. Several hundred aberrant genes is the norm for any given cancer, which makes looking at a microarray a bit like studying the thick cluster of wires leading to a bomb. To defuse the bomb, you need to know which wires to cut, but the microarray doesn't provide a circuit diagram, only clues. A radically overproductive gene is worth focusing on, especially if you happen to know that it's involved in cell division – but this method is hardly comprehensive. "You could imagine a case in which breast cancer cells all share one mutation, the same way all cats have pointed ears," explains UCSF cancer biologist Gerard Evan. "But if you wanted to wipe out the cat population, you wouldn't do it by snipping off the tops of their ears." Similarly, just seeing which genes are active doesn't always tell you which genes are important.
Faced with the new super-abundance of possible drug targets, companies have begun turning to technologies like RNA interference to speed the process of elimination. But perversely, ignorance can sometimes work to drug designers' advantage, as when molecules designed to treat one cancer prove unexpectedly successful on others. In fact, Druker is just back from Chicago, where he was brainstorming with George Demetri, a doctor who has tried Gleevec on patients with gastrointestinal stromal tumors. Stromal tumors are nothing like leukemia, but both diseases are believed to involve defective protein kinases. Both belong to the same family and have similar structures.
To tell whether Gleevec was working, Demetri hooked patients up to a PET scanner, which measures how quickly cells are metabolizing glucose. Onscreen, fast-growing tumors show up as bright spots, while dead tissue appears dark. According to Demetri, the tumors in the Gleevec patients faded from burning orange to black in a single day. "It was crazy," he says. "They just stopped growing."
But even when it's possible to find and block the crucial mutation, as Gleevec does, cancer sometimes manages to route around the barricade. This happened to both Druker's and Demetri's patients, some of whom relapsed just months after bounding out of bed. The problem wasn't that Gleevec had stopped working, but that a few tumor cells had evolved to resist the effects of the drug – the same way bacteria evolve to survive douses of antibacterial soap. Like soap, Gleevec acts as an environmental pressure, wiping out all but a handful of organisms whose genetic makeup makes them naturally more resistant than their kin. The survivors can then thrive.
If nothing else, this explains cancer's eerie, Matrix-like ability to reconstitute itself months after it was supposedly destroyed. "Cancer is a disease of evolution," explains UCSF's Evan. "Mutations don't happen deliberately; but statistically, if you have a lot of cell division and a lot of mutation, you'll end up with some mutants that can survive even a pretty hostile environment." With no competition, those cells will gradually take over.
As a result, researchers have begun looking into drug cocktails that would contain cancer rather than cure it. Druker himself is developing a drug to be used in conjunction with Gleevec, and another for patients with acute myeloid leukemia, which has more than one keystone mutation. For Druker this means tracking down the additional mutations in each case and then finding a drug that can inhibit them without inadvertently blocking other vital functions. (He is concerned that his newest AML drug, which blocks two kinases involved in blood cell development, could be so effective that it would temporarily leave patients with no blood cells at all.)
Oddly enough, Gleevec, which stops chronic myeloid leukemia in weeks, is much less effective on a disease with almost identical symptoms: chronic lymphomatic leukemia. As it turns out, the two are caused by different sets of genetic mutations. This is a vital revelation, given that it affects whether a patient will respond to a particular drug. It's also an increasingly common discovery. Under the microscope of molecular profiling, the old paradigm of cancer as a monolithic disease is steadily crumbling. Cancer, it now appears, is far more multifarious than anyone imagined back in 1971.
Fighting an army of diseases as varied as cancer means first being able to identify the enemy at the molecular level. That's why MIT researcher Todd Golub is putting together what he calls a "global cancer map." The goal: classify all cancers based on their particular pattern of abnormal genetic activity. (Mutations in genes can trigger the production of abnormal proteins, the target of Gleevec, but they can also cause other problems, like radically increasing or decreasing the amount of normal protein production.)
At 40, Golub has an impressive résumé. As director of cancer genomics at MIT's Whitehead Institute, he oversees a lab of 25 researchers while also working as an oncologist at Harvard's Dana-Farber Cancer Center. Unlike Druker, he doesn't have the lean look of a man who runs to work every morning, although he exudes a kind of restless tension: mental energy in overflow.
Around 1997, Golub began paying attention to the fact that most cancers were still getting diagnosed the old-fashioned way: by comparing morphological changes in biopsies under a microscope. But cancer cells that look similar under magnification can harbor different genetic mutations altogether. Golub's goal is to use microarrays to sort cancer into genotypes rather than phenotypes, by their genetic signature rather than by their symptoms or appearance. What saves this task from being an academic exercise in taxonomy is the fact that – as Druker discovered – genetically distinct types of the same disease often require different drugs or combinations of drugs.
Golub envisions a world in which a doctor can biopsy a tumor or take a blood sample and then use a microarray to pinpoint the problem. "What's emerging is a scenario where doctors collect molecular information from a patient with cancer and say, 'You have kinase X activated, and a high metastatic profile, so we're going to put you on this new drug. We don't care whether you came into the prostate clinic or the melanoma clinic.'"
Instead of stamping someone with the generic diagnosis "breast cancer," the idea would be to treat a patient's illness according to which genes have gone out of whack. Such a distinction could go a long way to explaining why some cancers respond to certain drugs and others don't. It's also likely to change the way drugs get designed and tested.
For example, the FDA recently approved a drug made by AstraZeneca that works well in just 15 percent of lung cancer patients. These are not the kinds of numbers that would have once impressed regulators, but now it's considered par for the course. There's a growing consensus that the most effective new cancer drugs will end up being narrowly tailored to match the subtypes that a global cancer map would systematically reveal. AstraZeneca's researchers now suspect that they, too, have stumbled on a genotypically distinct variant of lung cancer and are rushing to determine what defines the tumors that respond to the new drug (and which appear mostly in nonsmoking young women). Drug trials, in other words, have turned into research tools, with the potential to reveal subtypes of cancers that researchers never knew existed.
Results like AstraZeneca's have led many researchers to suspect that the picture of cancer will get even more complicated before it arrives at Golub's precision-targeted future. A case in point: Genentech created a breast cancer drug, Herceptin, specifically to treat the 25 percent of patients who overexpress a gene called Her2-neu. But even within this select subset, the drug, which limits the effects of the Her2 gene, works only about a third of the time. So far, no one knows why.
In the US, at least, no one has yet been willing to base treatment decisions on genetic profiling. The story is different in the Netherlands, where doctors have already launched the first large-scale profiling trial. For the study, several thousand women with breast cancer were sorted into two treatment groups based on the results of a 70-gene assay developed by Rosetta Inpharmatics, a subsidiary of Merck. It's a gamble, and many researchers in the US doubt it will pay off.
Golub himself is more optimistic, arguing that a diagnostic tool doesn't have to be 100 percent accurate – just better than what we have now. He also believes that cancers won't dissolve into as many subcategories as people fear. "There will be some rules that cut across all types of cancers," he says. Recently, Golub and colleague Sridhar Ramaswamy sampled tumors from a range of different cancers – breast, lung, prostate, colon – to see if genetic activity could predict whether a tumor was likely to metastasize. They found a correlated cluster of 17 genes. "None of these were genes we suspected of having anything to do with metastasis," Golub says. "It's a result that popped straight out of the microarray statistics."
Developed eight years ago at Stanford, microarrays, also known as DNA chips, are the technological linchpin of Golub's work. A small glass wafer roughly the size of a postage stamp, each chip comes lined with a grid of up to 16,000 probes, single strands of DNA that bind only to their complementary RNA match. Slosh a bit of liquefied tumor over the chip while baking it at 113 degrees Fahrenheit and within a day the tumor genes that match the probes will stick to the chip in amounts proportional to their activity.
Like many of the most exciting new tools in molecular biology, the machines that scan these chips are uninspiring to look at. The model I saw was square and gray, a plastic box enlivened only by three LEDs and a small, nearly invisible door that opened to reveal the chip slot. From the outside, it's about as thrilling as a washing machine.
But in fact, the box is one of the more crucial inventions to hit cancer research in years. Attached to a computer monitor, it can measure fluorescent markers attached to each of the 16,000 individual probes, producing a display that looks something like a night sky: a black background washed with points of varying brightness, each representing the level of expression for a single gene. (Genes express when they produce the RNA that creates a protein.) Compare the patterns found in tumor cells to ones found in healthy tissue and you end up with a profile of the difference: a mug shot, in effect, of all the genes that are misbehaving in a cancerous cell.
This is exceptionally useful, particularly given that researchers once had to measure activity one gene at a time: laboriously growing cells, extracting their RNA, and running the results out on a specially prepared gel. Microarrays, by contrast, show the precise level of activity of thousands of genes simultaneously. The only problem is that it's almost too much information. "There's genetic variation between two tumors in the same person," Golub says. "Once you begin comparing different people, the amount of noise" – variations that aren't essential to the cancer – "skyrockets." Comparing the profiles of a cancer patient and a healthy patient would reveal thousands of differentially active genes, with no hint as to which ones were related to cancer and which ones were not. To get around this problem, researchers combine the profiles of dozens or even hundreds of patients and note only the genes that are consistently different in each of them.
Understanding the full genetic diversity of cancers is crucial to designing drugs that will properly counter them, but a growing cadre of scientists believes that if we really want to end suffering due to cancer by 2015, we have to find ways to detect nascent tumors earlier. This, at least, is the conviction of Sudhir Srivastava, who heads the Early Detection Research Network at the National Cancer Institute's leafy headquarters in Bethesda, Maryland. Part of a vast government anticancer complex, Srivastava's office is far from the trenches of cancer warfare, locked in a gleaming glass-and-concrete building that contains not a single lab.
Though trained as a scientist, Srivastava these days acts more like a choreographer, harnessing research from dozens of labs nationwide and overseeing the creation of as-yet-hypothetical databases dedicated to the pursuit of biomarkers. Biomarkers are the canaries in the cancer coal mine – faint, physical evidence that the disease is present. The best example: a protein whose level becomes elevated in the blood before a tumor even appears. At present, a mere handful of these have been found. The best known, PSA, for prostate-specific antigen, has been used for some time as a predictor of prostate cancer, but it suffers from a disturbing rate of false positives and, in any case, is currently being evaluated on the question of whether it has helped to save even a single life.
Still, Srivastava and many others argue that early detection is essential. Srivastava himself lost an aunt to cervical cancer, a disease easily detected by a pap smear; in the US at least, it has correspondingly low mortality. This spring, the Nobel Prize-winning cancer researcher Leland Hartwell coauthored an article in Nature Reviews Cancer titled "The Case for Early Detection." In it, he notes that while most new treatments for cancer have failed over the years, there remains a striking correlation between how early a cancer is detected and whether a patient is likely to survive.
This makes sense biologically, since early detection means catching tumors before they spread into neighboring tissues, thereby reducing the chance that malfunctioning cells will migrate to the all-important lymph nodes or into the bloodstream. Cancer also has the effect of spawning even more mutations, at least some of which are likely to prove resistant to a given drug. All these extra mutations aren't necessarily beneficial to the disease – they could even cause a cancer to self-destruct, provided the patient lived long enough – but in the short term, hypermutation makes cancer almost impossible to treat: There are just too many new forms to combat.
The same microarray technology that allows researchers to see genes behaving differently in a tumor has provided early detection researchers with a slew of possible biomarkers. An even newer approach, proteomics, bypasses tumors altogether, searching blood samples for unexpectedly high or low concentrations of key proteins. A pair of early detection researchers, the NCI's Lance Liotta and Emanuel Petricoin of the FDA, recently used one such pattern of proteins to determine whether women in a small test group had early-stage ovarian cancer. The test was 100 percent accurate (in this case, the women had already been diagnosed using existing tests). The next step: confirming that the proteins allow detection at the earliest stages.
Ideally, these kinds of discoveries would lead to something as simple as an annual blood and urine screen for different types of cancer. One difficulty, however, is proving that a biomarker can actually predict cancer before it's detectable by other means. There's no guarantee that a protein produced in tumors is evident before a tumor even forms, although many researchers believe such early warnings exist. At present, the only way to prove beyond a doubt that a biomarker exists in the blood before a tumor appears is to conduct a long-term trial with thousands of people – drawing blood, checking for the elevated protein, then waiting to see who actually develops cancer and correlating those results with the biomarker data.
It's a daunting prospect, made worse by the fact that chemical levels in the body fluctuate regularly based on diet, sleep patterns, even time of day. Tracking the subtle alterations caused by a nascent cancer is a bit like trying to listen to a walkie-talkie in an electrical storm: It's hard to hear the signal for the noise. Since the cost of making a wrong diagnosis can be high, this is particularly troubling when it comes to detection. No one wants to have his prostate removed or her breast irradiated based on a false alarm. Nevertheless, Srivastava believes that early detection – if it can be made to work – remains the best bet for actually curing cancer rather than simply managing it.
Even now, researchers are split over just how complicated cancer will turn out to be in the end – and some argue that our growing understanding of its molecular biology may also reveal intractable new levels of complexity. (One example: Researchers at UCLA's Jonsson Cancer Center recently investigated how overexpression by the Her2 gene affected the behavior of other genes – and found more than 500 changes.) But for now, optimism still holds sway. And if we succeed, at long last, in charting a tumor's mysterious mechanisms, getting diagnosed with cancer in 2015 could be a distinctly different experience than it is today. Much as AIDS evolved from being a mysterious disease that killed healthy young men in six months to a mostly survivable condition, so cancer might become a manageable illness. "More like an annoying mole in your garden than an alien taking over your body," as one researcher puts it. Either way, Golub believes, students of biology 10 years from now won't be able to imagine a time when cancer was treated without a molecular understanding of how it works. "That's the exciting phase we're in," he says. "It's like a Polaroid. We're beginning to see what's possible." After three decades, the picture is finally coming into focus.
5 New Tools for Fighting Cancer By Joseph Portera Today's trailblazing research and treatment of cancer is fueled by a host of technologies that analyze and manipulate genetic material at the molecular level. Here are the latest weapons.
DNA microarrays For decades, scientists were limited to studying just a few genes in a given experiment. But microarrays, aka gene chips, are changing all that. Using precision robotics, tiny slides are dotted with thousands of DNA samples representing different genes. The stamp-sized chips allow researchers to observe the complex interactions between hundreds (and possibly thousands) of genes that are now linked to cancer. The potential applications for microarrays range from lifting cancer's genetic fingerprints to predicting a patient's response to a drug treatment.
Bioinformatics The Human Genome Project and the technologies that grew out of it have produced an ocean of data. The goal of bioinformatics is to mine that information for meaning. The tools: artificial intelligence, sophisticated search algorithms, and networked databases. By combining genomic and proteomic data from around the globe, researchers can identify cancer markers and even predict survival probabilities.
Proteomics By cataloging the half-million human proteins, researchers in proteomics are seeking to understand their chemical interactions. Long before a tumor forms, cancerous cells produce minute traces of abnormal proteins. A handful of biotech companies are racing to build protein chips – microarrays that will identify telltale cancer proteins, letting doctors detect malignancies and monitor treatment with simple, noninvasive tests. Ultimately, proteomics could uncover new targets for protein-inhibiting drugs.
RNA interference Human cells have a built-in mechanism that fights foreign invaders and regulates gene expression. It's called RNAi, and researchers have figured out how to harness it to short-circuit genetic expression. The DNA itself remains intact, but the cell is unable to produce damaging proteins. The technique has yet to yield any drugs, but it's already being used in the lab as a cheaper, faster way to deactivate particular genes in animals.
High-throughput X-ray crystallography By bombarding crystallized proteins with X rays, researchers are producing highly accurate 3-D models of proteins that play a role in rampant cell division. Armed with this intelligence, drug developers can design precision inhibitors that bind to and deactivate these pernicious proteins.
