When the Target Isn’t Really the Target: One Way Cancer Drugs Fall Out of Clinical Trials
Ninety-seven percent of potential new cancer drugs never make it to market, dropping out of clinical trials when they don’t meet measures of safety or efficacy.
“Why that is, we don’t really know. But I think that this extremely high failure rate suggests that there are some fundamental issues in how new drug targets are studied and how new drugs are characterized,” said molecular biologist Jason Sheltzer, PhD, an Independent Fellow at the Cold Spring Harbor Laboratory on Long Island, NY.
He decided to investigate, and uncovered the potential power of publishing negative evidence. The work fits in with Open Access week here at Public Library of Science.
The team reports in Science Translational Medicine on using the gene editing tool CRISPR-Cas9 to test whether 10 experimental cancer drugs work exactly how their developers predicted they would. And they found a tunnel vision in the way that drugs are targeted that might explain why certain patients do not respond as hoped.
Like an arrow that hits a tree rather than the bullseye, some cancer drugs may not actually reach their targets – but many studies weren’t designed to reveal this. And so when results look promising, the drug candidate progresses through the FDA labyrinth.
Sheltzer’s strategy was straightforward: use CRISPR to remove the purported target, and if the drug still works, then the target isn’t really the target. Perhaps preclinical research that identified a molecule as the drug target was halted, the scientists concluding success, when something fit the bill.
The researchers tested drugs that are either in clinical trials or once were, or are in preclinical studies (animals or human cells) – not cancer drugs currently on the market. The experiments were done on standard cell lines from cancer patients.
“The idea for many of these drugs is that they block the function of a certain protein in cancer cells. We showed that most of these drugs don’t work by blocking the function of the protein that they were reported to block,” Sheltzer explained.
Using CRISPR provided greater precision in interrogating potential drug targets than the older method, RNA interference. RNAi “knocks down” gene expression rather than snipping out a gene like CRISPR can.
Might a small molecule bind more than one type of target, like shooting arrows that hit trees and bushes as well as the bullseye? And sometimes what seems to be a valid drug target in vitro isn’t exactly what happens in a body.
But a drug can make it to market without anyone knowing exactly how it works. That’s the case for selective serotonin reuptake inhibitor (SSRI) anti-depressants. The cartoons in ads depict neuromuscular junctions with the drug keeping serotonin in synapses longer by binding the reuptake proteins, presumably offsetting a deficit behind the symptoms. But googling SSRIs returns “the exact mechanism of action of SSRIs is unknown.”
From Slash-and-Burn to Hitting Targets
The new cancer drugs work in a few ways. Some of them zero in on molecules specific to cancer cells. These include:
- protein receptors for growth factors or hormones
- enzymes critical for cell division, like cyclins and kinases
- immune checkpoint inhibitors, which lift the dampening of the immune response that cancer cells exert.
These targeted drugs offer an alternative or adjunct to traditional drugs that broadly kill many types of rapidly-dividing cells, not just the cancer cells.
The targeted drugs began with Herceptin in 1998, its inventors recently honored with a Lasker award. The FDA approved another hugely successful targeted cancer drug, Gleevec, in just a few months in 2001. Today melodramatic ads pitch the new arsenal of cancer treatments: Zelboraf, Tafinlar, Keytruda, Opdivo.
But targeted drugs can fail if a new mutation alters the target or cancer cells find an alternate pathway that hikes cell division rate.
The researchers took a dual experimental approach based on logic:
- Remove the target (such as a cell surface protein) from cancer cells. If the cells divide, the target isn’t essential.
- Add the drug to cancer cells that have had the target removed. If the cells die, then the drug is hitting something else.
Part of the confusion, I think, is semantic. Sometimes we deem a chemical interaction “off-target” if it doesn’t do what we designed it to. Maybe our expectations were wrong. To be more unbiased, some researchers alter the language, calling the reliance of a cancer cell on a particular protein an “addiction” and investigating to seek “druggable cancer dependencies.”
The team’s work indicates that what was deemed on-target may really be off-target, and vice versa. Perhaps it’s time to retire those terms.
The First Drug Tested
Earlier, Sheltzer investigated a protein called MELK, to which a company, OncoTherapy Science, is developing an inhibitor, called OTS167. Because MELK (“maternal embryonic leucine zipper kinase”) is abundant in many tumor types, it was presumed to be essential for their growth and therefore a drug target. But when CRISPR removed the gene that encodes MELK protein, nothing happened.
“To our great surprise, when we eliminated these proteins from the cancer cells, they didn’t die. The cancer cells continued to grow just fine, in spite of what had previously been published. They just didn’t care about MELK,” Sheltzer said.
The group published the findings on MELK in 2017, in eLIFE, raising the possibility that OTS167 is perhaps barking up the wrong tree. The drug candidate is in a phase 1 (safety) trial for solid tumors and is recruiting for a phase 1 trial for triple negative and metastatic breast cancer.
The MELK story inspired the group to use their “genetic target-deconvolution strategy” to see whether 10 other drugs were actually hitting their supposed targets. About a thousand cancer patients in total are taking one of these drugs in clinical trials.
Another Misguided Drug Reveals a Novel Target
In the new paper, the investigators question another drug, OTS964, being developed to treat certain lung and breast cancers. In the process, they’ve discovered a new druggable cancer target.
RNAi had indicated that OTS964 targets a protein called PBK. But CRISPR told a different story – cells with PBK gone still succumbed to the drug. “It turns out that the interaction with PBK has nothing to do with how the drug actually kills cancer cells,” Sheltzer said.
To find out how the PBK-targeting drug works, the researchers applied huge amounts of it to cancer cells and then gave the cells time to acquire mutations that would enable them to resist the drug. Cancer genomes are inherently unstable, mutating often. When a mutation renders a cell resistant to a drug, that cell then has an advantage and soon takes over the tumor.
Discovering how a cell circumvents a drug is priceless information.
The resistance experiments revealed that the cancer cell vulnerability that candidate drug OTS964 taps into isn’t PBK after all, but a gene that encodes the protein CDK11. It’s a “cyclin-dependent kinase,” an enzyme that is part of a pathway that leads to cell division.
The FDA has already approved CDK4/6 inhibitors, starting with Ibrance, in February 2015, to treat certain types of breast cancer. CDK11 is a brand new target. And that’s potentially huge.
At a news conference the researchers addressed concerns that their findings will affect people already taking targeted cancer drugs – but they maintained that their work did not discover any approved drugs that were hitting trees instead of bullseyes.
But what about ongoing clinical trials for cancer drugs?
Sheltzer tried to alert folks running the trials. “I filed a FOIA with the FDA to try to get additional information on the safety and efficacy of these drugs. The FDA declined to share that data, and said that it was a trade secret up until the point that these drugs received FDA approval.”
He contacted companies sponsoring clinical trials too, but they wouldn’t disclose any information either.
“I think that the secrecy and the opacity in this drug development process really hurt scientific progress. A lot of drugs tested in cancer patients tragically don’t help cancer patients. If this kind of evidence was routinely collected before drugs entered clinical trials, we might be able to do a better job assigning patients to therapies that are most likely to provide some benefit. With this knowledge, I believe we can better fulfill the promise of precision medicine,” Sheltzer said.
The drug companies would do well to pay more attention to basic scientists who figure out how things work – or don’t work – like Sheltzer. Using CRISPR can enable researchers to “do a better job finding cancer’s central genes and a better job validating a drug’s on-target mechanism of action. We think that that kind of preclinical foundation will help clinicians design better clinical trials to decrease the failure rate of new drugs,” Sheltzer concluded.
It appears that some of the core suggestions made in Lin et al., which is featured in this online article, require reassessment.
Thank you Dai. Your response illuminates the way that science works – that alternate explanations are possible, and that the design of an experiment can influence interpretation of the outcome. Thank you!