Biomarkers and the precise therapies that target them will cure cancer by degrees—and change the way drugs do business.

In the 2010 film adaptation of Alice in Wonderland, the Red Queen inspects a line of frog footmen to determine which frog ate her tarts. All of these footmen sweat and shiver, but it’s the bit of jam at the corner of one footman’s mouth that gives him
away. Off with his head!

This is a biomarker—a physical clue that marks something different from its surroundings. In Alice in Wonderland, the jam biomarker makes the unfortunate frog footman a target, and the Red Queen is his destructor. Researchers at the University of Colorado Cancer Center are using similar biomarkers to paint jam bull’s-eyes on cancer.

“We’ve all heard cancer called the ‘Big C,’ ” says Dan Theodorescu, MD, PhD, director of the CU Cancer Center. “But it turns out that it’s actually made up of literally thousands of little C’s, all of which share characteristics like unchecked growth, but beyond which may be as different as cats and dogs. They’re both mammals, but you can’t expect them both to come when you whistle.”

The realization that the “Big C” is many little C’s, each perhaps identified by distinct, individual biomarkers, also means that the other Big C—“Cure”—is likely to be a series of singles that pick off each subset in turn, rather than the home run that ends the disease entirely.

“It’s basically peeling an onion,” Theodorescu says. “Once you peel off the layer of ALK-positive lung cancer [as the CU Cancer Center did in 2011], you find another layer. Peel off that next layer of cancer and you’re onto another until eventually you’ve peeled away the disease entirely, one subset at a time.”

This rethink of cancer as “cancers” is driving a sea change not only in new, targeted treatments but in how these treatments are developed.

But let’s start with jam—what are the biomarkers of cancer?


A cancer biomarker can be as simple and well-known as the prostate-specific antigen (PSA) test for prostate cancer. A man with high or increasing levels of PSA in his blood has jam on his cheek—off with his prostate! (Caution: oversimplification alert, but you get the point.)

Another well-known biomarker is the Philadelphia chromosome—a mutation that occurs when chromosome number 9 breaks off a bit of itself and inserts it in chromosome number 22. This Philadelphia chromosome (now without the speed governor that should limit its growth) is present in 95 percent of people with chronic myelogenous leukemia (CML).

Once researchers discover a biomarker, they can target it, which is what the drug Gleevec does in CML, attaching specifically to these cells’ broken speed governors. Before Gleevec, CML patients died; after Gleevec, a recent New England Journal of Medicine study shows that 95.4 percent of CML patients are alive eight years after diagnosis.

Researchers around the country are trying to duplicate this success story. What will be the next Philadelphia chromosome and what will be the next Gleevec? Sometimes, though, the answer isn’t quite so simple; sometimes the definition of a disease rests not on the back of one chromosome but on many. Sometimes the signature of a disease is a sprinkling of hints throughout a cancer cell’s genome. Dan Theodorescu hunts these hints.


First some background: the National Cancer Institute (NCI) maintains 60 lines of cancer cells. In the late 1980s, these cells were taken from 60 real human breast, lung, and brain and other human tumors, and since then the NCI has carefully kept the cell lines alive. Like pulling off a piece of sourdough starter for a loaf of bread, when researchers want to experiment on cancer cells, they can split off cells from these 60 lines. What kills cancer? Well, researchers have now tested more than 110,000 chemicals on these 60 cell lines and have a pretty darn good idea of what kills them.

But there’s a twist: Theodorescu specializes in bladder cancer, and there are no bladder cancer cells included in the NCI-60. How can you know what drugs work against bladder cancer, or for that matter any cancer not represented in the NCI-60, if the cells haven’t been hit with the 110,000 drugs? One method would be to get bladder cancer cells, blast them with the drugs and see what happens. (Can you spare many millions of dollars?) Another method would be to look at historical data of how bladder cancers with certain genetic signatures responded to the drugs used to treat them. (If you have the data, Theodorescu and other bladder cancer researchers would love to see it.)

The third method is the innovative approach that Theodorescu and his co-authors call COXEN, which stands for CO-eXpression ExtrapolatioN. Basically, COXEN looks at cancers not as “breast” and “bone” and “brain” but as signatures of the genes that are mutated in each. In other words, COXEN boils down a cancer cell to the 20-or-so genes that define it—can 20 important genes predict how a tumor will respond to a drug?


You can bet Theodorescu and his coauthors took the question into the lab to kick the tires. First, they mined the NCI-60 to discover which 20-ish genes predict response or resistance to the drugs cisplatin and paclitaxel. Great—and once they had the genetic signature of cancers that were responsive and resistant, they wondered if the same 20 genes might predict response or resistance to these drugs in bladder cancer patients.

In 85 percent of cases with cisplatin and 78 percent of cases with paclitaxel, the answer was yes.

Consider the tires kicked.

Theodorescu did a similar test with breast cancer. Could his group use biomarkers to predict relapse after tamoxifen? Again, Theodorescu mined the NCI-60 to discover which genes predicted response to the drug—which ones were most “up” or “down” in successful and unsuccessful patients. And patients on a clinical trial for the drug had their tumors genetically sequenced before treatment. Would patients whose biomarkers predicted drug resistance relapse? In 71 percent of cases, yes. Based on the signature of these important genes, Theodorescu could predict fairly accurately who would relapse after tamoxifen.

While it’s nifty to have a genetic taxonomy for cancer that predicts how a patient will respond to any given drug, it’s perhaps even niftier to use a disease’s genetic signature to pick the best drug. Here’s how COXEN does it: We know what drugs kill the NCI- 60. We know the genetic signatures of these NCI-60. And we can fairly easily discover the genetic signature of any given cancer. If a tumor’s genetic signature is similar to that of cells in the NCI-60, you can predict what drugs will kill it.

By matching the biomarker profiles of bladder cancer to similar cancer cells in the NCI-60, COXEN predicts a strong bladder cancer drug should be imidazoacridinone—sure enough, when Theodorescu and his team hit bladder cancer cells with imidazoacridinone, more than 60 percent were killed (compared with only 22 percent for the most common bladder cancer drug, cisplatin). Theodorescu hopes imidazocridinone, as identified by the COXEN model, may become a new anchor drug for the treatment of the disease.


So biomarkers allow us to crystalize just the important parts of a disease’s genetic signature and match the signature to similar genes from the NCI-60, which then allows us to predict drug response or pick the best drug—targeted not just at cancer in general, and not just at the cancer’s location, but at the very genes that make up the disease. Dawn Duvall, PhD, and other researchers at the Colorado State University Animal Cancer Center (a member of the CU Cancer Center consortium) asked another interesting question: could COXEN cross species? In other words, do the genetic signatures of specific cancers trump the difference between humans and our best friends?

CSU Animal Cancer Center

Douglas Thamm, Dan Dustafson, Rodney Page and Dawn Duvall at the Colorado State Animal Cancer Center. Photo by Glenn Asakawa.

There’s an interesting intersection of dogs and humans that allowed Duvall and her team to answer that question. That intersection is bone cancer—osteosarcoma. It turns out that on a genetic level, you can’t tell a K9 osteosarcoma from a human one.

And in the realm of osteosarcoma, dogs have the drop on us. “Only about 800 osteosarcomas are diagnosed in humans every year,” Duvall says, “but that number in dogs is 8,000 to 10,000.”

Much of COXEN’s biomarker-driven cancer care is a numbers game. How does a statistical snapshot of one cancer line up with a statistical snapshot of another? Well, you need enough numbers to create crisp statistical pictures. In osteosarcoma, dogs got it and we don’t.

“This might allow us to learn things in dogs we can’t learn in humans,” says Duvall. “If we can develop new treatments in dogs, we can test them in this huge number of dogs and then transfer the knowledge to children who get osteosarcoma.” But first, because of these numbers, osteosarcoma was a great jumping-off point for K9 COXEN—do the genetic signatures of the human NCI-60 predict response and resistance to K9 osteosarcoma?

To find out, Duvall and her CSU collaborators Douglas Thamm, DMV, Rodney Page, DVM, and Dan Gustafson, PhD, had to make their own osteosarcoma predictions, compare them with the COXEN predictions and see if the two were copacetic. “Basically we tried to select groups of tumors that were resistant and compliant, and compare the gene signature of these two groups,” Duvall says. Some dogs survived long after treatment and others didn’t— what genes made these tumors different? “By doing this we found a number of biomarkers,” Duvall says.

And it turns out that, sure enough, once you turn a K9 osteosarcoma into a genetic signature, its drug sensitivity is almost exactly the same as a tumor from the NCI-60 with the same signature.


Cancer care crosses into dogs; could it cross out?

“Actually, this has a strong potential to help Theodorescu’s COXEN advance faster in the human setting,” says Page, director of the CSU Animal Cancer Center. “With dogs, owners and doctors, not insurance companies, decide on what treatments are given, and we don’t have to wait for a drug to fail in order to try a second-line therapy. Also, it’s hard to get people to accept a finding that shows a tumor is sensitive to a drug not on the standard of care list. But with dogs, you can give a promising drug not on a standard of care list and thus give Theodorescu the ammunition he needs to eventually get it put on the list.”

The jump back from CSU’s dogs to the CU Cancer Center’s patients is under way in lymphoma. “It’s prevalent in both dogs and people, making it a great place to start,” says Duvall. Again, it’s all about the numbers game: Once Duvall gets a bead on the genetic signature of K9 lymphoma (in the works), she’ll be able to match that signature to signatures in the NCI-60 to discover which of the 110,000 tested drugs will likely work best with K9 lymphoma. Then, staying always within the bounds of compassionate care, she can test the promising drugs that COXEN picks in CSU’s K9 patients. It’s a short step from a dog-proven drug to humans.


This biomarker-driven cancer care is also forcing academia and industry to step quickly out of the box of drug development mechanics it’s been in since the 1950s. Sure, bigpharm is willing to pony up the $100 million needed to push a big drug through the process of FDA approval—but what about a small, targeted cancer therapy? If a drug is meant to eventually target 100 people a year as opposed to 10,000, will big-pharm still pay for it?

For example, imagine if a biomarker target and a drug to nix it were found for heart cancer. Yes, it killed Eric Carr, drummer of the rock band Kiss, and yes, it’s a leading candidate for Catherine of Aragon’s death in 1536, but only a handful of cases are seen each year in the United States. The same is true as we split big cancers into their smaller subsets: Would big-pharm fund development of a drug to target, for example, a biomarker present in only 0.5 percent of lung cancer patients?

“Personalized medicine, driven in part by matching drugs to biomarkers, is forcing industry and academia to reevaluate how we do business,” says Andrew Thorburn, PhD, deputy director of the CU Cancer Center. What he means is this: don’t expect bigpharm to pay for little drugs. That is, unless academia can give it to industry as a slam dunk.

Thorburn says this slam dunk depends on matching drugs to biomarkers before a drug’s first human trial. “If you give a targeted drug to a general cancer population, the proportion of people who respond might not be high enough to make the drug look effective,” Thorburn says. In this way, drugs that could have helped the 0.5 percent die in development. The fix? Thorburn says it’s in the deep science of picking apart patients’ genomes for predictive biomarkers—the genes that hint at who will and who won’t respond to a drug, and then testing the drug with only this high-responding population. It’s a task, says Thorburn, for which academia and not necessarily industry is best suited.

“That’s why some cancer centers, the CU Cancer Center included, would like to put in place the infrastructure that would allow us to take a drug through phase I clinical trials in-house, in some cases before courting industry involvement,” says Thorburn. “But to do so, we’ll need the generosity and vision of philanthropists.”

The CU Cancer Center has molecular biologists on hand discovering biomarker targets, synthetic chemists making the arrows, and innovative animal researchers taking drugs to the tipping point. “What we don’t have are experts who can make drugs usable for humans or the institutional money to fund clinical trials,” Thorburn says. And this monkey wrench of money in the cogs of drug development mechanics means that a biomarker with potential, matched with a drug that has potential, may stay as that: potential.

Currently academia can produce the frog footmen with jam-smudged lips and the Red Queen to yell “Off with his head!” In most settings, it lacks only the final axe to get the job done: money.