NEW YORK – Testing for somatic copy number alterations at a specific locus on chromosome 9 in squamous head and neck cancer cells may identify the patients most likely to benefit from immune checkpoint inhibitors, according to a new study published Thursday in the Proceedings of the National Academy of Sciences (PNAS).
The study, led by Teresa Davoli, an assistant professor in the department of biochemistry and molecular pharmacology at NYU Langone Health, and Scott Lippman, director of the Moores Cancer Center at the University of California, San Diego, found that amplifications and deletions of copy numbers in genes at 9p24.1 in head and neck cancer cells are associated with significantly improved survival in patients receiving immunotherapy versus chemotherapy.
The researchers focused specifically on patients with human papillomavirus (HPV)-negative head and neck squamous cancer (HNSC), though both Davoli and Lippman pointed out that whole-exome analyses suggested that the biomarker may also have predictive value in several other squamous cancers. The authors focused on HNSC in this research in part because patients have access to checkpoint inhibitors, but few respond.
Merck's checkpoint inhibitor Keytruda (pembrolizumab) is approved by the US Food and Drug Administration as both first- and later-line treatment for HNSC, and Bristol Myers Squibb's Opdivo (nivolumab) is approved for later-line HNSC. Neither drug is approved with the aid of a biomarker to try to identify best responders, however. "With head and neck cancer, we end up giving virtually 100 percent of patients with recurrent disease an anti-PD-1 drug, and 15 percent benefit," Lippman said. "We focused very deeply on this in one severe type of head and neck cancer."
Headed into the current study, Davoli and Lippman had previously conducted research suggesting that 9p somatic copy number deletions contributed to an "immune-cold tumor microenvironment phenotype" in HPV-negative HNSC. However, questions remailed as to which genes or genetic regions in 9p were the main drivers of this phenotype.
To answer this question, they performed a series of multiomic analyses. First, they used whole-exome and whole-transcriptome data from The Cancer Genome Atlas (TCGA) to determine the frequency of somatic gain and loss at 9p24.1, 9p21.3, and the 9p region — that is, the short arm of chromosome 9. Among 343 HPV-negative HNSC patients for which they had TCGA data, they determined that copy number loss frequencies were 34 percent, 48 percent, and 42 percent for 9p arm, 9p21.3, and 9p24.1, respectively. Meanwhile, the respective copy number gain frequencies were 14 percent, 17 percent, and 22 percent, respectively for 9p arm, 9p21.3, and 9p24.1.
The researchers also studied the link between somatic copy number alterations of each chromosomal region on the genome and the immune score or CD8 T-cell level. For this analysis, they used RNA expression of cytotoxic markers GZMH, PRF1, CD3E, CD247, CD2, GZMK, and NKG7 to determine the immune score. Here, they found a strong negative association between 9p24.1 loss and the immune score, and a positive association between 9p arm gain — particularly at 9p24.1 — and the immune score as well as CD8 T-cell levels.
After performing these analyses and several others using whole-exome sequencing and whole-transcriptome sequencing datasets, Davoli, Lippman, and colleagues applied their findings to a retrospective patient cohort to test whether the biomarker could predict responses to anti-PD-1 checkpoint inhibitor therapy.
The researchers used a real-world dataset from Caris Life Sciences, which included whole-transcriptome profiles and survival data on 894 HPV-negative advanced HNSC patients, who received Keytruda or Opdivo as first-line treatment or received chemotherapy but no immunotherapy. Evaluating every gene in the 9p21.3 and 9p24.1 bands, they first determined that only one gene at the 9p21.3 threshold was associated with survival after immunotherapy, whereas the same was true for nine genes at 9p24.1.
"These data … suggest that 9p24.1 is a relative hotbed of immune regulatory genes," wrote the authors in the PNAS paper. Indeed, the HNSC patients with an increase in the number of copies of the genes at 9p24.1 lived for an average of 30 months after receiving checkpoint inhibitors, versus 11 months for those with lower copy numbers of the genes at 9p24.1.
Importantly, the 9p24.1 biomarker did not also predict for improved survival after chemotherapy alone, which supported its predictive, rather than purely prognostic, value for specifically guiding checkpoint inhibitor treatment.
The genes with the strongest correlation with immunotherapy survival on this "hotbed" were JAK2 and CD274. Combined, transcript down-regulation or up-regulation of these genes predicted checkpoint inhibitor resistance or sensitivity, respectively. Lippman acknowledged that the CD274 gene encodes for the PD-L1 protein. PD-L1 expression is commonly used to assess whether patients with different tumor types might benefit from checkpoint inhibitors, though not HNSC patients, and physicians have often bemoaned its flaws, underscoring the need for more reliable and consistent biomarkers.
Toward a routine assay
For oncologists to apply this biomarker research to routinely select HNSC patients for immunotherapy, a validated, widely available test will be necessary, Davoli acknowledged. "It would require some sort of proxy of copy number assessment of the 9p24.1," she said. "Ideally, that would be a genomic test that sequences for the copy number of several genes on 9p24.1." That said, Davoli and colleagues also tested out gene expression as a way to assess this biomarker and found that RNA sequencing was as good, if not better, at predicting response to immune checkpoint inhibitors.
"Patients who had an increased amount of the RNA of genes found on the 9p24.1 had a better response to immunotherapy, while patients who had a lower dosage of the genes expressed on 9p24.1 had a worse response … We showed that an evaluation of the dosage of genes on 9p24.1 can present a potential test to guide clinical decisions," she explained.
Given the latest finding that patients are, as Lippman called them, "exquisitely sensitive" to anti-PD-1 checkpoint inhibitors when they have amplification at 9p24.1, as well as previous studies backing up the findings that deletions at this locus confer resistance, Lippman argued that the biomarker is, in fact, ready for routine use in the clinic now. However, it probably won't be a companion diagnostic any time soon.
Relative to biomarkers that have established use as companion diagnostics for checkpoint inhibitors, like PD-L1 expression, tumor mutational burden, and microsatellite instability status, Lippman said, the overall copy number story, now leading to the 9p findings, has been "a little under the radar" in the field."
Despite that, Lippman said pharmaceutical companies and oncologists have expressed interest in using this biomarker to guide treatment. Commercial labs like Caris already have commercially available tests that include this genomic information, and the biomarker is considered to have level 3 evidence in the eyes of the FDA, meaning that the biomarker has "potential clinical significance," and "may be informational or used to direct patients towards clinical trials for which they may be eligible." Lippman believes that with this latest research, the biomarker could likely approach level 2, meaning the FDA would see it as having definite, rather than potential, evidence of clinical significance.
Davoli and Lippman aren't sure the biomarker will ever move up to level 1, though. This level is reserved for biomarkers that have companion diagnostic status and are "essential for the safe and effective use of a corresponding therapeutic product."
For a biomarker to achieve level 1 companion diagnostic status for a specific drug, it needs to be validated in a prospective study of patients who received that drug. Since the top marketed immune checkpoint inhibitors, Keytruda and Opdivo, are already approved for allcomers in this HNSC patient population and have a strong foothold as standard of care, drugmakers like Merck and BMS aren't likely to repeat their large-scale prospective trials with added biomarker criteria. A large prospective clinical trial denying patients an approved drug on the basis of a level 3 biomarker would not only be unfeasible from a business perspective — since it would mean limiting a market for an allcomer-approved drug — but likely unethical as well, since patients would need to be denied an already-approved drug on the basis of their tumor genomics.
"How can we potentially find a different path for this type of biomarker, which is [validated only] retrospectively, but is still very valuable?" Davoli asked hypothetically. Even though biomarkers without companion diagnostic status can still have tremendous clinical utility, "there doesn't seem to be such a clear path" for integrating them into routine clinical use, she observed.
According to Lippman, the discussion reflects the rapid pace of precision oncology advancements, which can be a challenge as much as a boon. "The biomarker development field is moving much faster than the several years required to design, complete, and analyze the trial results," he said, pointing to the fact that Merck completed its pivotal KEYNOTE-040 trial leading to the HNSC Keytruda approval "before the 9p story began."
Short of companion diagnostics status, Lippman suggested that the 9p24.1 biomarker can achieve complementary biomarker status. In 2015, the FDA approved the first PD-L1 complementary diagnostic as an aid for doctors who wanted more guidance on whether to prescribe Opdivo to their non-small cell lung cancer patients. Unlike companion diagnostics, the FDA doesn't consider complementary tests to be necessary for the safe and effective use of a drug, but such tests can still be useful for personalizing treatment.
While Merck, BMS, and others with PD-1/L1 checkpoint inhibitors aren't likely to repeat pivotal Phase III trials, Lippman pointed out that they could — and indeed are beginning to —provide researchers and test developers with samples from patients treated with these agents in completed clinical trials. This could help further validate new biomarkers, supporting their integration into routine clinical care.
Lippman also noted that some of the big drugmakers are beginning to conduct retrospective biomarker analyses on their pivotal trials themselves. "And that's a big move," he said, "particularly when you're looking at a marker that could have an impact on a $17 billion-a-year drug, and you're talking about 85 percent of patients being resistant. The fact that big pharma is looking at this now is a huge step forward."