Using a genomic strategy, researchers led by Ohio State University's David Carbone have identified a rare genetic mutation that they believe might characterize a small sub-group of lung cancer patients who could be super-responders to the cancer drug Nexavar (sorafenib).
In a paper published this week in the Journal of Clinical Investigation, the researchers described how, in a 300-patient Nexavar trial, they used whole-genome sequencing and RNA sequencing to identify the ARAF S214C mutation in a single lung cancer patient and zeroed in on that mutation as the likely reason for her extraordinary response to Nexavar.
The work demonstrates why outlier patients in clinical studies deserve to be scrutinized more than they currently are and how these individuals can advance understanding of complex diseases.
The finding regarding the ARAF S214C mutation needs to be prospectively investigated in other patients who harbor mutations in ARAF or the related RAF1 gene. Nonetheless, in addition to potentially uncovering a new oncogenic marker in lung cancer, Carbone and his colleagues have demonstrated a viable strategy that drug developers can employ to systematically discover subpopulations of patients they might have otherwise overlooked and are who are highly responsive to marketed or investigational therapies.
"In many clinical trials, even those with very low response rates, there is a subset of patients that do very well," Carbone, director of the James Thoracic Center at OSU, told PGx Reporter. In the Nexavar trial, called ECOG 2501, he noted that 3 percent had an objective response when the data was not enriched by biomarkers, but one patient had an especially good and long response. "That kind of a result is often ignored … [But] we are proposing [in this paper] that the response would be much higher in patients with ARAF mutations," Carbone said.
As trial after clinical trial has disproven the one-size-fits-all approach to cancer drug development, pharmaceutical companies are increasingly searching for biomarkers that can identify smaller and smaller subsets of cancer patients who respond to their drugs. A decade earlier, few drug developers would have been eager to develop a therapy that targets ALK-positive non-small cell lung cancer tumors, which occur in between 3 percent and 5 percent of NSCLC patients. Today, after the launch of Pfizer's Xalkori (crizotinib) in 2011, a number of pharma companies are also developing ALK inhibitors, pursing the same patient subset.
In fact, the development program for Xalkori was sparked by research that identified the unique characteristics of an outlier group of lung cancer patients. Pfizer pursued Xalkori in ALK-positive NSCLC after Japanese researchers reported in 2007 that five out of 75 NSCLC patients they examined had EML4-ALK fusion transcripts and these "individuals were distinct from those harboring mutations in the EGFR gene."
Moreover, after launching Xalkori for patients with ALK-positive tumors, Pfizer has continued to investigate even smaller subsets of patients who can benefit from the drug, such as individuals with ROS1 and RET gene fusions, which occur in between 1 percent to 2 percent of NSCLC patients. Pfizer is also exploring novel testing approaches that can identify responders to its drug. Researchers from the drug giant and Seoul's Samsung Medical Center recently published a paper demonstrating that a multiplex molecular barcoding test gauging ALK, ROS1, and RET fusions is just as accurate as assays using fluorescent in situ hybridization or immunohistochemistry, but more sensitive and efficient that these standard methods.
Although drug development paradigms are undergoing significant transformations within pharmaceutical R&D divisions, in Carbone's view there still isn't systematic effort put behind probing why outlier patients respond particularly well to a specific agent. In development studies, if only a few patients respond to a drug, even if they have a very good response, "the drug is called a failure and it is pretty much dropped," Carbone said. "In fact, sorafenib is pretty much dropped in lung cancer right now."
Nexavar, an inhibitor of VEGFR, Raf, and a number of other kinases, is co-developed and co-marketed by Amgen subsidiary Onyx Pharmaceuticals and Bayer HealthCare. The drug is currently approved in the US as a treatment for inoperable liver cancer, advanced kidney cancer, and radioactive, iodine-resistant, advanced thyroid cancer. The sponsors had been trying to advance Nexavar as a lung cancer treatment, but it failed to meet primary endpoints in two Phase III non-small cell lung cancer trials.
In 2010, Bayer/Onyx reported that Nexavar did not improve overall survival compared to placebo for first-line NSCLC patients in the NExUS trial, and then in the 2012 MISSION study, the drug in combination with chemotherapy compared to placebo failed to extend the lives of advanced NSCLC patients who had received two or three prior treatments. In both trials, the drug seemed to extend progression-free survival. For both trials Bayer/Onyx did not report prospectively looking at response in molecularly-defined NSCLC populations, although the sponsors collected tumor and serum samples for half of those enrolled in the MISSION study.
Subsequent retrospective analysis has not revealed clear leads in terms of predictive biomarkers of response for Nexavar in lung cancer. In 2012, researchers presented analysis from the MISSION study suggesting that patients with EGFR mutations lived longer than those with the normal form of the gene while on Nexavar, but contradicted earlier findings that KRAS mutation may also predict response. At the time, investigators reasoned that the response seen in EGFR mutated lung cancer patients was likely due to the drug's inhibition on BRAF.
Currently, Onyx's website states that Nexavar is being researched in kidney, liver, thyroid, and breast cancer, but doesn't mention lung cancer.
"This ARAF mutation has not been described in lung cancer before," Carbone said. "At least now we have a candidate for another type of genetic abnormality to look for in these patients, and if it holds up in future patients it can really make a difference in their disease."
Identifying a super-responder
In the Phase II ECOG 2501 study, in which Carbone was also an investigator, approximately 300 pretreated NSCLC patients first received Nexavar daily for two months. Among these patients, 81 stable patients were randomized to receive either Nexavar or placebo, and after two months, disease control rates were 54 percent for those on Nexavar and 23 percent for those receiving placebo. Disease control rate measures the percentage of patients who had a complete response, partial response, or stable disease due to a treatment.
However, nine patients experienced a sustained response to Nexavar, and among these patients, Carbone's team decided to focus on one super-responder, a 66-year old female with NSCLC who since her diagnosis in 2002 with Stage IV disease had failed to respond to several treatment regimens, including a combination of gemcitabine and vinorelbine chemotherapies, AstraZeneca's Iressa (gefitinib), and Takeda/Millennium's Velcade (bortezomib). When none of these agents helped stave off her disease, in 2006, facing worsening hypoxia, she had a lobe of her lung surgically removed as a palliative measure and began treatment with Nexavar in the ECOG 2501 trial.
Within two months, she had a near-complete response on Nexavar, and remained progression-free and asymptomatic for five years. In mid-2011, however, doctors discovered that a mass in her right lung had gotten big enough so that they considered her disease to be once again progressing. She stopped taking Nexavar, and began a regimen of Avastin (bevacizumab), carboplatin, and paclitaxel. But she was tired and couldn't breathe well and had to stop taking the drugs after two cycles.
She died in hospice toward the end of 2011. However, noting that this woman had had the most pronounced response to Nexavar in ECOG 2501 – being the last participant in the study when she relapsed – Carbone and colleagues decided to home in on what it was about her disease that made her a super-responder. They used WGS and RNA sequencing on an Illumina platform to try to elucidate the molecular features that might be unique to her lung tumors.
WGS revealed more than 25,000 mutations and 101 non-synonymous mutations impacting the coding regions of 99 genes. But the patient did not have any known cancer-driving mutations in oncogenes such as KRAS, EGFR, BRAF, ERBB2, or PIK3CA. Her tumors also harbored none of the usual cancer-linked gene fusion suspects involving ALK, ROS1, or RET. Germline variant analysis of peripheral blood WGS data didn't reveal any deleterious mutations among the 29 known Nexavar targets.
Carbone and his colleagues landed on ARAF S214C after analyzing the patient's RNA transcripts, which reliably detected only 15 out of the 101 somatic coding mutations and two in-frame fusions picked up by WGS. ARAF was the second most highly expressed sorafenib target gene after KIT, but it was the only gene that harbored somatic mutations. Furthermore, ARAF encodes a protein kinase in the Raf family, into which the more commonly known BRAF and RAF1 genes fall.
Carbone's group ultimately zeroed in ARAF S214C as the driver of this patient's lung cancer and the reason for her response to Nexavar "because it was mutated and not just highly expressed," he explained. "The mutation was in a region that was structurally predicted to be activating."
The researchers reported in the paper that they detected the ARAF S214C mutation in 11 out of 23 tumor reads, but none of the normal DNA reads, and in 48 out of 56 RNA reads. "The fact that the mutation was present in the majority of the RNA sequencing reads, as well as acquired in the tumor versus normal cells, really caused us to go down this path," Carbone said. The RNA sequencing for this study was performed in Carbone's lab at OSU. The whole-genome sequencing was performed by researchers at the Broad Institute and Dana Farber Cancer Institute.
Next, Carbone's team compared this one patient's data to a treatment-naive TCGA cohort, where three out of 564 adenocarcinoma cases had mutations in the 214 codon of ARAF. Additionally, the researchers found three other cases with mutations in the related RAF1 gene. Among these six cases, investigators found no somatic mutations in known lung adenocarcinoma oncogenes in four patients, but two patients with RAF1 mutations had BRAF and KRAS alterations.
In another effort to probe the role of ARAF or RAF1 mutations in lung cancer, Carbone and his team expressed these mutant alleles in tracheobronchial tissue cells and found that the cells expressing ARAF p.S214 variants exhibited changes suggestive of carcinogenesis, such as increased Mek phosphorylation. Subsequent treatment with Nexavar inhibited these changes in the cells.
Given the observed Mek phosphorylation, the researchers tested the cell lines with a MEK Inhibitor, GSK's Mekinist (trametinib), approved last year by the US Food and Drug Administration as a treatment for BRAF-mutated melanoma. In cells expressing ARAF mutations, Mekinist treatment decreased Erk phsphorylation and suggested anti-cancer activity. When cell lines expressing mutations in RAF1 were treated with Nexavar and Mekinist, the drugs exhibited similar effects.
Noting that Nexavar has activity against multiple protein kinases, Carbone posited that Mekinist, as an inhibitor of MEK1 and MEK2, might prove to be a better option for lung cancer patients with ARAF or RAF1 mutations. "We have data that trametinib worked in vitro. So, I might try trametinib or look around for what's available in terms of a more specific inhibitor that doesn't hit so many kinase targets [as] sorafenib. But sorafenib would certainly be a reasonable fallback."
Carbone said he hadn't reached out to Onyx or Bayer with the hypothesis that lung cancer patients with ARAF mutations might be super-responders to Nexavar.
Although the study authors believe that they may have discovered new lung cancer-associated, druggable markers in ARAF mutations, they wrote in the JCI paper that these mutations shouldn't be considered markers of Nexavar response until there is more data from other super-responders. Carbone estimated that ARAF mutations occur in around 1 percent of lung adenocarcinoma patients. The researchers have looked for ARAF and RAF mutant cell lines in the Cancer Cell Line Encyclopedia, but did not find any.
A systematic approach
Carbone and colleagues have also demonstrated a strategy for homing in on small populations of patients within a larger disease group who have remarkable responses to certain drugs. Systematically studying these super-responders can give new life to existing agents.
"If recurrent but rare mutations underlie the oncogenicity and responsiveness of 'driver-negative' lung adenocarcinomas, they are not likely to be statistically nominated by genome scans of several hundred (or even thousands) of cases," Carbone and his colleagues concluded in the paper. "Our study suggests that a powerful alternative approach to driver mutation discovery may be through the analysis of outlier patient responses and the nomination of driver mutations through the preponderance of genomic, biochemical, and functional evidence."
It's worth noting that when Carbone and his colleagues looked for mutations in the hotspot region containing ARAF and RAF1 in another patient cohort, they were able to identify additional mutations in 4,608 cases with 21 tumor types, including cutaneous melanoma, colorectal cancer, and gastric cancer. This gives researchers reason to look for ARAF and RAF1 mutations in cancer patients across tumor types and study their response to drugs that target this pathway.
"Possibly these kinds of outliers, if you study them, can give you important information about a subset of patients who can benefit from that drug," Carbone said. "We are finding in general in lung cancer now that there are very important subsets that comprise even less than 1 percent and have a dramatic benefit from a specific intervention."
Genomic advances have certainly broken up the once monolithic groupings of lung cancer based on histology. The American Cancer Society estimates that this year, more than 220,000 patients will be diagnosed with lung cancer in the US, and nearly 160,000 will die from the disease. Non-small cell lung cancer accounts for more than 80 percent of lung cancer diagnoses in the US.
In recent years, researchers have identified more and more genomic markers that drive NSCLC. The most prevalent oncogenic drivers in NSCLC patients in the US include KRAS mutations (20 percent) and EGFR mutations (10 percent). Mutations in ALK, ROS1, BRAF, HER2, and PI3K also show up in NSCLC patients with lesser frequency.
Despite these advances, for approximately half of the NSCLC population the molecular markers driving their disease is unknown, likely due to the rarity of the mutation. Focusing on outlier responders is one way to uncover unknown cancer-linked mutations. Carbone suggested that drug developers should routinely collect samples from patients and "put some effort" into evaluating molecular characteristics of people who do unexpectedly well.
"There is a lot of serendipity in this example that we published, but … serendipity favors the prepared mind," he said. "If you ignore these cases, which is done in the vast majority of cases, then you don't learn anything and you don't know what to do with the next patient."
Proponents of personalized medicine often talk of a far-flung future where drugs will be developed for specific molecular pathways and be disease agnostic. A lot needs to change between now and then, not just in terms of business models and drug development paradigms, but also in scientific knowhow.
"Sometimes it's very clear that the same exact lesion in two different tumor types has very similar consequences," Carbone said, offering the example of the similar role that BRAF mutations play in how melanoma and lung cancer patients respond to specific tyrosine kinase inhibiting agents. But colon cancer patients with BRAF mutations don't seem to respond as well to these same agents.
"This is something we have to learn more about, but as a first order hypothesis it's very reasonable to define targets in one disease and assume that they may have a good chance of working in other diseases," Carbone said. "With the development of multiplex testing, it's quite easy now to test for hundreds of abnormalities in a single assay … The incremental cost of adding another gene to a multiplex panel is virtually zero."
Next, Carbone believes researchers should prospectively look for ARAF mutations in cancer patients. Toward this end, at least at OSU, ARAF has been added to the panel of oncogenes that lung cancer patients will be routinely tested for. "If we find new patients with this, then specifically targeting that pathway would be reasonable," he noted. "This is certainly not a demonstrated conclusion … but it's a hypothesis that I think is worth prospectively testing."