Skip to main content
Premium Trial:

Request an Annual Quote

Immunotherapy Response Prediction Improves With TMB Correction, Other Markers

Premium

NEW YORK – A team led by researchers at Johns Hopkins University has identified several tumor and immune system features that appear to boost checkpoint immunotherapy response predictions in lung cancer. Notably, those features include a corrected tumor mutational burden (TMB) estimate with a computational approach that takes tumor purity into account.

Based on the improved response predictions in more than 100 non-small-cell lung cancers (NSCLC), the investigators believe their corrected TMB algorithm may set the stage for more accurate checkpoint immunotherapy response predictions in other tumor types.

"This study does have significant implications beyond lung cancer," said Victor Velculescu, an oncologist and co-director of cancer biology at Johns Hopkins University School of Medicine who is also affiliated with the Sidney Kimmel Comprehensive Cancer Center and the Bloomberg-Kimmel Institute for Cancer Immunotherapy. "[Tumor purity is] something that's a fundamental feature of all cancers."

While the TMB thresholds needed to guide treatment in various tumor types are yet to be determined, Velculescu is confident that there is "predictive power to this [corrected TMB] biomarker."

"Once you have an accurate readout of mutations, and you correct for that count using tumor purity, and add in additional features, you can start to say, 'Let's do a trial where we specifically evaluate this level of TMB,'" he explained.

There has long been a rationale for treating high-TMB tumors with checkpoint immunotherapy drugs. After all, tumors peppered with mutations should be more prone to producing distinct antigens, or neoantigens, that can be flagged as foreign by a patient's immune system.

Even so, studies and clinical trials that have attempted to predict immune checkpoint blockade response based on TMB have been far from conclusive. At the European Society for Medical Oncology's annual meeting last September, for example, investigators shared mixed results from studies that attempted to use high TMB to predict immunotherapy responses in advanced NSCLC patients.

"Tumor mutational burden is a very inaccurate and imperfect biomarker of response to immune checkpoint blockade," Valsamo Anagnostou, a lead author of the new study and an oncology researcher affiliated with the Sidney Kimmel Comprehensive Cancer Center and the Bloomberg-Kimmel Institute for Cancer Immunotherapy, explained. "We know from multiple studies … that there are patients with high TMB who don't respond to immunotherapy and there are also patients with low TMB that do respond to immunotherapy."

"There's a lack of standardization, and a lack of understanding of what the threshold would be for TMB," Velculescu added, explaining that mutation calling strategies and several other factors are expected to impact how TMB is defined.

In particular, Anagnostou, Velculescu, and their colleagues suspected that the proportion of tumor cells in a given biopsy sample, known as tumor purity, could impact their ability to accurately pick up mutations and make TMB estimates, particularly when those mutations are measured with sequencing approaches that have less-than-ideal sensitivity. While samples from some cancer types often have good tumor purity, it often drops off in tumors that are more difficult to biopsy, for example  in the lung, pancreas, and other parts of the body.

"We can't accurately count the mutations if we don't realize that there's a mixture of tumor cells and normal cells," Velculescu said.

With that in mind, the team first set out to understand whether TMB estimates based on exome or targeted sequencing were accurately reflecting alteration levels in tumors from breast cancer, melanoma, bladder cancer, and other cancer types, using data for almost 5,500 NSCLC tumors assessed for the Cancer Genome Atlas (TCGA) project and smaller studies — a set that included samples with varying tumor purity levels.

"We did a pretty comprehensive analysis, where we first looked at more than 3,500 tumor samples from TCGA and more than 1,500 samples from a published cohort of immunotherapy-treated patients," Anagnostou said. "There, we asked the question, 'Do the TMB values correlate with tumor purity when TMB is derived from whole-exome sequencing or from targeted NGS?'"

Their analysis, published in Nature Cancer this week, supported that tumor purity can impact TMB estimates, which fall in accuracy as tumor purity declines. That was particularly true in clonally heterogeneous tumors.

"Recognizing that this is a problem, we tried to fix it," Anagnostou said, noting that her co-first author Noushin Niknafs was instrumental in developing the algorithm used to solve this problem.

With simulation analyses on tens of thousands of more TCGA tumors with known sequence coverage, tumor purity, clonal composition, and TMB characteristics, the team systematically found correction factors corresponding to tumor purity in tumors assessed with exome  or targeted sequencing.

"For both scenarios [tumors tested by exome or targeted sequencing], we developed correction factors for TMB, and ultimately put those in look-up tables," Anagnostou said, "where one could very simply plug in the observed TMB — the number of observed sequence alterations from whole-exome sequencing, or the TMB estimate from targeted NGS — and then multiply this value by the correction factor to get to the corrected TMB."

From there, the researchers demonstrated that tumor purity-corrected TMB correlated better with outcomes in 104 NSCLC cases than uncorrected TMB estimates.

Most of these patients had received second- or third-line nivolumab (Opdivo from Bristol-Myers Squibb), she noted, though a subset had gotten pembrolizumab (Merck's Keytruda), alone or in combination with chemotherapy, and a few patients had received ipilimumab (marketed as Yervoy by Bristol-Myers Squibb) plus nivolumab.

Whereas TMB poorly predicted immunotherapy response in that cohort, Anagnostou explained, predictions improved considerably using the corrected TMB — an effect that held even when the researchers analyzed patients treated with the same drugs.

She cautioned that "there were still patients with high corrected TMB and patients with low corrected TMB that [both] responded, but the prognostication was improved."

"Even corrected TMB is not a perfect predictor of response," Anagnostou said.

To further improve immunotherapy response prediction, the team searched for other independent contributors to checkpoint immunotherapy response in the lung cancer patients, uncovering response-related immune activity, mutational profiles, and specific pathway alterations.

"Response to immunotherapy is determined by the dynamic interplay or crosstalk between the tumor and the immune system," Anagnostou explained. "It cannot just be the number of mutations, alluding to the number of new antigens presented."

Although tumor PD-L1 levels did not correspond to response in the their NSCLC cohort, the researchers found that the presence or absence of smoking-related mutational signatures, the type of a patient's human leukocyte antigen (HLA) alleles, and the presence of activating mutations in receptor tyrosine kinase (RTK) enzyme-coding genes, such as EGFR, HER2, MET, IGF1R, and FGFR, also seemed to have independent contributions to immune checkpoint response.

"We envision that this approach is applicable to all solid tumors," Anagnostou said. "Our vision is to generate … a tool where the physician would be able to put in the TMB, the tumor purity, and some information about the sequencing that was performed, and then our algorithm is going to run in the background to provide the corrected TMB to the clinician."

At the moment, the researchers are working on follow-up studies to validate the tumor purity correction, along with complementary immunotherapy response predictors, in additional tumor types. Those studies will include further retrospective analyses, though the team hopes the corrected TMB approach will ultimately make its way into prospective clinical trials and the broader clinic.

At least some of the other features are expected to hold as immunotherapy markers in other cancer types, though additional research is needed to confirm that. For example, there is mounting evidence suggesting that activating RTK gene mutations may drive checkpoint immunotherapy resistance in general, Anagnostou said, while germline HLA allele and immune antigen diversity features may also be a common feature across cancers.

"Our goal now is to test, first, the corrected TMB approach and the model as an aggregate in additional immunotherapy-treated cohorts — whether that is additional lung cancer cohorts or cohorts from other solid tumors," Anagnostou said. "We have efforts ongoing to come up with a platform that is going to be end-user friendly, and clinicians could use it to get closer to what the true TMB of a tumor is."

There is a good chance the team will eventually develop a commercially-available version of such a tool, she noted, as the corrected TMB approach has been patented, though that will depend in part on results from future validation work in retrospective and prospective cohorts.

"I think the way this would work in practice is, you would take the improved TMB and you would re-analyze existing data to get a better sense of where the cutoff might be, and then you would design a trial to test that cutoff prospectively — using a corrected TMB with a specific cutoff," Velculescu explained.