At the annual American Society of Clinical Oncology meeting last June, Bristol-Myers Squibb (BMS) researchers presented data on a cohort of patients not responding to the company’s approved checkpoint inhibitor nivolumab (Opdivo). Layering on a novel immunotherapy antibody was effective in half of patients tested, the team reported, with no major increase in side effects compared to nivolumab alone.1 Each of the therapies aims to unleash an immune cell–fueled tumor attack by targeting a molecule that normally suppresses T-cell activation—programmed death 1 (PD-1) in the case of nivolumab, and lymphocyte-activation gene 3 (LAG-3) in the case of the new, investigational antibody. The combination worked particularly well in patients whose T cells displayed LAG-3 on their surface. “We now have a population that we can sensitize to immunotherapy that was resistant to anti-PD-1 treatment,” says Nils Lonberg, who heads the immune oncology and targeted drug discovery efforts at BMS in Redwood City, CA.
BMS has several newer checkpoint inhibitors, targeting other immune pathways, that trigger T cells to home in on tumors, and company researchers are accumulating data on combining each with nivolumab. “We focus on both innate and acquired immunity pathways to treat more patients with tumors we know can respond to immunotherapy, and also to open up other cancer types to immunotherapy,” says Lonberg. “From basic-science principles, what we look for first in a combination are drugs with nonredundant mechanisms.”
Other companies, including those with their own FDA-approved checkpoint inhibitors, such as AstraZeneca, Merck, and Roche, are taking similar approaches. The US Food and Drug Administration (FDA) approved the first checkpoint inhibitor antibody—ipilimumab (Yervoy), which targets cytotoxic T-lymphocyte antigen 4 (CTLA-4)—for advanced melanoma in 2011. Five other checkpoint inhibitor antibodies followed—six in total—targeting the PD-1 pathway for numerous cancer types. In 2015, the first and thus far only FDA-approved combination of two immunotherapies hit the US market: nivolumab plus ipilimumab for metastatic melanoma patients.
Combining multiple treatments for patients with recalcitrant cancers is not a new concept. Among the first pairings of anticancer drugs were two or more different chemotherapies. As drug companies developed additional types of cancer drugs, combinations of different modalities—including chemotherapy, radiation, targeted small molecules, and eventually immunotherapies—followed. (See illustration below.) “There has long been a feeling that drug combinations will be needed to have the type of impact in cancer patient care that we would like to see,” says David Hyman, a medical oncologist who specializes in early drug development at the Memorial Sloan Kettering Cancer Center in New York City.
Only a handful of cancer drug combos have so far been approved by the FDA, in part because many of the tested combinations were conceived largely at random.
But only a handful of cancer drug combos have so far been approved by the FDA, in part because many of the tested combinations were conceived largely at random—an inefficient approach given the dizzying number of approved and investigational therapies that could be combined. In fact, of the hundreds or even thousands of novel combos currently in clinical trials, many, if not most, were designed based on little more than convenience, depending on what drugs a company owns, says Charles Swanton, a cancer geneticist at The Francis Crick Institute in London. “My view is that there are too many trials, especially immunotherapy ones, being conducted in a serendipitous manner,” he says. “It’s more about, ‘We’ve got these two cancer drugs, so let’s put them together and see what happens.’”
Only recently have researchers adopted more-systematic approaches. One method that’s growing in popularity is the use of high-throughput screens that allow researchers to quickly evaluate interactions between different cancer therapies to predict which might form a successful combo. Alternatively or in addition, some researchers are relying on knowledge of the underlying biology to determine which therapies are likely to make the strongest pairings, as is the case for BMS’s checkpoint inhibitor combos. “We have to start from fundamental principles of tumor biology,” says Swanton. “Once we know this information, then we can start to come up with rational combo approaches.”
Ross Camidge, a thoracic oncologist at the University of Colorado Denver, agrees. “Our chances of successful combination therapies are only as good as the science going into the selection of the combinations.”
Casting a wide net
In vitro screening of large numbers of drug combinations is one of the approaches to sort through a vast ocean of drug-pairing possibilities. In silico screening methods typically rely on compiling data generated by in vitro experiments and animal studies, then using the data as a basis for computer algorithms to predict promising interactions. But these methods are labor intensive and in vitro drug combination screening is also expensive, which is why they have not been widely adopted. “There are not many academic labs with the capability to do [large-scale] combination screens, and not many pharmaceutical companies are doing it either, for that matter,” says Marc Ferrer, a researcher at the Chemical Genomics Center within the National Center for Advancing Translational Sciences.
To bypass the need for having libraries of drug compounds to physically pair, researchers have been taking advantage of genetic methods, including novel gene-editing techniques, to identify potential drug pairings that kill cancer cells. These approaches can be easier and less expensive than traditional cell-based drug combination screens using multiwall plates. Using guide RNAs to knock out pairs of genes using CRISPR, for example, Stanford University’s Michael Bassik identified pairs of genetic targets that might encourage cancer cell death.2 Researchers can then use databases to search for drugs that bind to and inhibit the proteins encoded by those gene pairs. Possible combos identified through such screening methods require validation in cell culture and animal experiments. With the CRISPR screen, “we’re using a genetic proxy for a drug effect: pairs of genes versus pairs of drugs, which require extensive robotics, plates, lots of time and money,” says Bassik. “We are making assumptions that there are specific drugs for those gene targets, which is often, but not always true.”
In addition, some researchers are going directly to cell culture–based screens to identify promising combos. In 2009, Georgetown University pediatric oncologist and researcher Jeffrey Toretsky identified a novel small molecule that targets an oncogenic fusion protein, EWS-FLI1, found exclusively in Ewing sarcoma, a type of bone cancer. His lab pulled out the molecule from a biophysical screen that tested the ability of thousands of compounds to bind a recombinant EWS-FLI1 protein. Then, using cell culture, Toretsky’s lab tested pairwise combinations of the small-molecule inhibitor with 69 generic cancer drugs. This second screen uncovered a synergy with the chemotherapy drug vincristine (Marqibo, Vincasar PFS),3 a finding that Toretsky and his colleagues confirmed with in vivo data last year, showing that the combination thwarted tumor growth in two Ewing sarcoma xenograft mouse models.4 The human version of the EWS-FLI1 inhibitor, called TK216, is now in a Phase 1 clinical trial for Ewing sarcoma, and the combination will also be tested, says Toretsky.
Such cell culture–based screens are able to relatively quickly parse through large numbers of potential combinations, says Toretsky. Traditionally, however, only chemotherapies, targeted small molecules, and certain targeted antibodies—not immunotherapies—could be screened using cell culture–based screens. “There are many cellular interactions that are not captured in a 2-D monolayer of cells,” says Ferrer.
Because cancer drug combinations are showing promise in clinical trials, Ferrer and his colleagues are trying to devise more-dynamic in vivo screens that better mimic the tumor and its microenvironment. But this is no easy feat. In 2012, his team developed a way to systematically screen many cancer drugs using three-dimensional sphere cultures of tumor cells,5 an approach that identified drug-combination effects that were drastically different than those measured in 2-D cultures.6 Ferrer has used the high-throughput 3-D assay to test dose ranges of drug combinations, and he’s now working to increase the complexity of the cultures by mixing tumor cells with cells from the tumor microenvironment, hoping to eventually include immune cells.
To further narrow the search, many researchers urge forethought on the front-end and reasoning on the backend, examining what is known about how certain drugs work and thinking about mechanisms that might pair well together. “The permutations of potential combinations are endless,” says Samir Khleif of Augusta University’s Georgia Cancer Center. “The best thing that we have in our hands is biology and logic.” Khleif, for his part, is testing currently available immunotherapy drugs in various combinations in animal models based on hypotheses of what pathways might work well together to fight tumors.
Back to biology
For a one-two punch aimed at two different mechanisms driving cancer cell survival, one approach is to go after two targets, each within a different signaling pathway. This is the strategy employed by the BMS researchers who paired the anti-LAG3 and anti-PD-1 checkpoint inhibitors that showed promise in combination in the recent clinical study. And the BMS team is not alone.
Last year, Karen Cichowski’s lab at Harvard Medical School published results indicating that two targeted therapies, each of which binds to a molecule in a different pathway, can together cause enough oxidative stress in tumors in mice to kill cancers that are driven by the Ras oncogene.7 The two oral drugs—one an inhibitor of the mechanistic target of rapamycin (mTOR) and the other an inhibitor of a histone deacetylase (HDAC)—are each individually approved for some tumor types. Several human trials, including a Phase 2 study in certain blood cancers, are testing the combination.
Other researchers are looking to harness such dual action in a single drug. Scientists at the Massachusetts division of Germany-based Merck KGaA are testing in mouse models a single antibody fusion protein, M7824, that simultaneously binds to the PD-1 ligand PD-L1 and traps transforming growth factor beta (TGF-β), a soluble cytokine protein that increases in abundance in patients with cancer. In results published earlier this year, the researchers reported that mice with breast and colorectal cancers treated with M7824 survived longer than those treated with either an anti-PD-L1 antibody or TGF-β trap binding alone.8 M7824 is currently being tested in Phase 1 trials for advanced solid tumors.
Other drug combinations are born by pairing compounds that hit the same signaling pathway, to stave off resistance that can arise when treating with either drug alone. (See “How Cancers Evolve Drug Resistance,” The Scientist, April 2017.) One such example is the small molecule trametinib, which was initially tested in combination with the already approved drug dabrafenib for patients with advanced melanoma. Both drugs target the Ras signaling pathway, which is a driver of cancerous growth in the 40 percent of melanoma tumors with an activating mutation in the BRAF gene. Dabrafenib targets the B-raf protein itself, while trametinib targets MEK, a downstream kinase. The combination decreased the risk of death from melanoma by 31 percent compared with dabrafenib alone9 and was approved by the FDA in January 2014. Recently, researchers at the Netherlands Cancer Institute uncovered two distinct populations of cells within drug-resistant melanomas. One consisted of cells expressing low levels of AXL, a receptor tyrosine kinase, and sensitive to B-raf and MEK inhibitors. The second population expressed high levels of AXL, was resistant to a B-raf plus MEK inhibitor combination, but was sensitive to a novel drug called an antibody-drug conjugate that binds to AXL on the surface of the tumor cells. The team showed that a triple combination targeting both cell populations was more effective than the standard combination, resulting in durable responses in patient-derived xenografts from resistant melanomas.10
Sometimes the logic behind a potential drug combo is not as simple as targeting the pathways inside tumor cells. When it comes to immune checkpoint therapies, which don’t target the tumor cells directly but rather the immune system, clinical studies have revealed that patients are most responsive if they have already started to mount an antitumor response. Thus, some researchers are now looking to layer additional drugs on top of an immunotherapy to transform a nonresponsive immune system to a tumor-responsive one.
Even when researchers think they have a solid hypothesis for a two-drug combination, biology can throw
them for a loop.
Earlier this year, for example, researchers at the University of Ottawa found they could slow tumor growth in mouse models of triple-negative breast cancer that are typically unresponsive to an immune checkpoint therapy by treating them with a Maraba rhabdovirus that sensitized the animals to an anti-PD-1 antibody.11 And when the combination was coupled with tumor resection, up to 90 percent of the animals had zero evidence of disease. A version of the Maraba virus expressing the neoantigen MAGE-A3 is currently being tested in a Phase 2 clinical trial for patients with advanced lung cancer. Meanwhile, a group of U.K.-based researchers showed this year that a similar combination of an oncolytic human reovirus plus an anti-PD-1 antibody resulted in an antitumor response in mouse models of brain cancer.12
Even when researchers think they have a solid hypothesis for a two-drug combination, biology can throw them for a loop. One cautionary tale is that of an anti-PD-1 antibody plus an OX40 agonist that stimulates the proliferation and expansion of T cells. Two recent studies, from Bernard Fox’s group at Oregon Health & Science University’s Knight Cancer Institute and Khleif’s lab at Augusta University, demonstrated that while, on paper, the combination should have been at least twice as effective in stimulating T cells as either treatment alone, it instead caused T cells to die in several mouse models of various tumor types. Indeed, two early-phase clinical trials combining OX40 and PD-1-targeting antibodies initiated prior to these publications have not panned out.13,14
As it turned out, the researchers were able to produce a synergistic effect, compared to an anti-PD-1 antibody alone, but only by giving the mice the OX40 antibody first, then treating them with the anti-PD-1 a few days later; reversing the order was ineffective. “Immunotherapy is the way of the future in cancer treatment, but the path is not straightforward,” says Khleif. “When you treat with one immunotherapy, you are targeting an entire biological system, and the treatment changes that system in a way that adding a second immunotherapy results in an unexpected result.”
Another major bottleneck stems not from biological limitations, but from the fact that biotech and pharmaceutical companies too often would rather test combinations only of molecules they have in-house, says Peter Adamson, professor of pediatrics at the University of Pennsylvania and the Children’s Hospital of Philadelphia. The result is that many new cancer drug combos being trialed are still largely haphazard in nature, driven as much by commercial interests as by underlying biology and compelling preclinical data, he adds. “What’s going on right now in early clinical development is that some companies look at their portfolio of agents, come up with a combination, and then pursue a scientific rationale of varying quality.”
Personalizing combo therapies
Despite the challenges, the cancer research community continues to see drug combinations as the future of therapy. Most researchers agree that successfully reining in a cancer’s growth and spread and extending patient survival will involve a barrage of multiple compounds. And this approach is spreading into the growing field of precision oncology, where researchers are looking in patients’ genomes for clues to which therapies are most likely to be effective.
Several years ago, frustrated by the lack of FDA-approved treatments that offer lasting benefit and by an inability to find an appropriate clinical trial for many of her patients, oncologist Razelle Kurzrock, director of the Center for Personalized Cancer Therapy at the University of California, San Diego (UCSD), decided to start her own customized therapy trial. In her team’s I-PREDICT (Investigation of Profile Related Evidence to Determine Individualized Cancer Therapy) study, patients often receive a custom two- or three-drug combination therapy—either FDA-approved drugs or experimental drugs from clinical trials—to target the specific mutations identified in their tumors.
According to Shumei Kato, a UCSD medical oncologist and a coinvestigator on the I-PREDICT trial, several thousand patients have been through genomic screening, and hundreds are receiving a customized combination of cancer drugs through the I-PREDICT or similar UCSD-led trials. And for some patients, it appears to be working.
On Mother’s Day, 2017, Lisa Darner had muscle spasms and lost consciousness. At her local hospital in San Diego, physicians told her that she had suffered a grand mal seizure—and that she had cancer in several major organs, including the brain, which caused the seizure. She quickly received brain radiation therapy followed by standard chemotherapy for lung cancer, which her oncologists considered to be the most likely primary tumor.
While receiving nonspecific chemotherapy, Lisa opted to also have her tumor biopsy analyzed using a comprehensive genetic panel—not always part of routine cancer care—that homed in on two actionable mutations: an epidermal growth factor receptor (EGFR) gene amplification and an alteration in a cell cycle gene called CDKN2A. Kurzrock and her colleagues at UCSD’s Moores Cancer Center came up with a triple drug combination including palbociclib (Ibrance), a cell cycle kinase inhibitor; a small molecule inhibitor of EGFR, erlotinib (Tarceva); and an antibody that also targeted EGFR, cetuximab (Erbitux). In August 2017, confined to a wheelchair because of her progressing disease, Darner started the custom combo as part of the I-PREDICT trial.
Aside from a rash (a side effect of the drugs), she responded well and is still on the treatment. “My tumors were still growing in August, but by October, scans showed everything was receding or has stabilized,” she says. “There are places where you can’t see a tumor anymore.”
Unfortunately, not all patients are as lucky. “The patient has high expectations from their cancer treatment, but the reality is that it is not always that great,” says Kato. “We’re not saying this is for sure a better approach. It’s a work in progress. But I think that continuing to do the standard-of-care approach, when it’s known not to be beneficial, will not change outcomes for cancer patients. We need to try something different to see a different, better result.”
Approved in 2015 for metastatic melanoma, the immunotherapy combination of the anti-CTLA-4 antibody ipilimumab and the anti-PD-1 antibody nivolumab increased the number of patients that responded to nivolumab alone by about 14 percent (NEJM, 373:23-34, 2015). But treatment-related side effects also increased with the combination of two immunotherapies, both of which can also unleash immune cells against healthy tissues. Specifically, 55 percent of patients who received the combination also experienced a greater number of serious treatment-related side effects—such as diarrhea and inflammation of the bowel—compared with 16 percent in the nivolumab-only group.
Layering multiple drugs typically increases the potential for side effects, adding to the challenges of developing promising treatment combos. In addition to the heightened risk of known side effects of either drug, pairing therapies can also uncover dangerous synergies not seen when a treatment is administered as a single drug. When the B-raf inhibitor vemurafenib was combined with ipilimumab in a Phase 1 clinical trial for advanced melanoma patients, for example, patients experienced high liver toxicity not seen with either drug alone, causing researchers to halt the study prematurely.
To make matters worse, such complications are often difficult to predict using animal models; unless a drug combination causes overt toxicity such as organ failure or significant immune–cell depletion in a mouse, the harmful effects of the pairing will likely only emerge in a clinical trial, says Joshua Brody of the Icahn School of Medicine at Mount Sinai in New York City. “Animal models do almost nothing to predict the safety profile of single drugs and drug combinations in humans.”
- P.A. Ascierto et al., “Initial efficacy of anti-lymphocyte activation gene-3 (anti–LAG-3; BMS-986016) in combination with nivolumab (nivo) in pts with melanoma (MEL) previously treated with anti–PD-1/PD-L1 therapy,” ASCO 2017, Abstract 9520.
- K. Han et al., “Synergistic drug combinations for cancer identified in a CRISPR screen for pairwise genetic interactions,” Nat Biotech, 35:463-74, 2017.
- H.V. Erkizan et al., “A small molecule blocking oncogenic protein EWS-FLI1 interaction with RNA helicase A,” Nat Med, 15:750-56, 2009.
- S.K. Zöllner et al., “Inhibition of the oncogenic fusion protein EWS-FLI1 causes G2-M cell cycle arrest and enhanced vincristine sensitivity in Ewing’s sarcoma,” Sci Signal, 10:eaam8429, 2017.
- K L.A. Mathews et al., “A 1536-well quantitative high-throughput screen to identify compounds targeting cancer stem cells,” J Biomol Screen, 17:1231-42, 2012.
- L.A. Mathews Griner et al., “Large-scale pharmacological profiling of 3D tumor models of cancer cells,” Cell Death Dis, 7:e2492, 2016.
- C.F. Malone et al., “mTOR and HDAC inhibitors converge on the TXNIP/thioredoxin pathway to cause catastrophic oxidative stress and regression of RAS-driven tumors,” Cancer Discov, 7:1450-63, 2017.
- Y. Lan et al., “Enhanced preclinical antitumor activity of M7824, a bifunctional fusion protein simultaneously targeting PD-L1 and TGF-β,” Sci Transl Med, 10:eaan5488, 2018.
- C. Robert et al., “Improved overall survival in melanoma with combined dabrafenib and trametinib,” NEJM, 372:30-39, 2015.
- J. Boshuizen et al., “Cooperative targeting of melanoma heterogeneity with an AXL antibody-drug conjugate and BRAF/MEK inhibitors,” Nat Med, 24:203-12, 2018.
- M.-C. Bourgeois-Daigneault et al., “Neoadjuvant oncolytic virotherapy before surgery sensitizes triple-negative breast cancer to immune checkpoint therapy,” Sci Transl Med, 10:eaao1641, 2018.
- A. Samson et al., “Intravenous delivery of oncolytic reovirus to brain tumor patients immunologically primes for subsequent checkpoint blockade,” Sci Transl Med, 10:eaam7577, 2018.
- D.J. Messenheimer et al. “Timing of PD-1 blockade is critical to effective combination immunotherapy with anti-OX40,” Clin Cancer Res, 23:6165-77, 2017.
R.K. Shrimali et al., “Concurrent PD-1 blockade negates the effects of OX40 agonist antibody in combination immunotherapy through inducing T-cell apoptosis,” Cancer Immunol Res, 5:755-66, 2017.