Biomarkers to identify positive responders to checkpoint inhibitors have proven a challenging task for drug developers. While several clinical trials have tried to identify a programmed death-1 (PD-1) or programmed death ligand-1 (PD-L1) expression–dependent response, it’s been an uphill task. During a clinical session at the annual meeting of the American Society of Clinical Oncology (ASCO), researchers were tasked with sharing their data on any breakthroughs or leads with biomarkers for these agents.
Mismatch Repair Deficiency in CRC
During his talk, Programmed Death-1 Blockade in Mismatch Repair Deficient Colorectal Cancer, Luis A Diaz, Jr, MD, medical oncologist, Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins, provided an update on the progress of using mismatch repair (MMR) deficiency as a marker for predicting response to PD-1 receptors. A presentation at ASCO last year by his group generated a lot of interest, because it indicated that a patient’s MMR status can be used to predict their response to the PD-1 inhibitor pembrolizumab, in colorectal cancer (CRC).1
Microsatellite instability resulting from genetic and epigenetic MMR is responsible for the development of CRC, Diaz said, adding that a majority of patients who participated in their trial were young and had Lynch syndrome or hereditary colon cancer.
MMR-deficient colon cancers are densely infiltrated with CD8+T cells and regress when treated with anti–PD-1 antibodies. This antitumor response is thought to be potentiated by somatic mutations, which when expressed as proteins, result in immunogenic neo-antigens that can be recognized by the patient’s immune system.
The current study recruited patients diagnosed with CRC who were either deficient (n = 28) or proficient (n = 25) in MMR. Patients were treated with the anti–PD-1 antibody, pembrolizumab, at a dose of 10 mg/kg, every 2 weeks. The median age of MMR-deficient participants was significantly younger (49 years) compared with those who had MMR-proficient tumors (62 years).2
An immediate biochemical response to treatment was observed in those with MMR-deficient tumors, measured as the levels of carcinoembryonic antigen. At 30 months follow-up, median overall survival (OS) in the MMR-proficient cohort was 5.98 months, while the MMR-deficient cohort is yet to reach a median OS. Additionally, progression-free survival (PFS) in MMR-proficient patients was 2.3 months, but PFS was not reached in the MMR-deficient patients. The objective response rate was 0% and 57% in the MMR-proficient and MMR-deficient patients, respectively, while the disease control rate was 16% and 89%, respectively.
Diaz said that 50% of patients presented with complete and durable response. Five of the 28 MMR-deficient patients had reached the 2-year mark following initiation of treatment and were no longer being treated with pembrolizumab. “They are on active surveillance,” Diaz said.
He had several thoughts on what this data would mean in the long term:
Is it time to think of treating MMR-deficient tumors with anti–PD-1 agents in a histology-independent manner?
How do we evaluate the management of patients who have a stable response following 2 years on a PD-1 inhibitor?
Do we need to figure the molecular etiology of primary and secondary resistance in these tumors?
The discussant for the session, Alexandra Snyder Charen, MD, medical oncologist, Memorial Sloan Kettering Cancer Center, wondered about the assessment of mutation load in the clinical setting. “Would it be possible to use genetic panels such as next-generation sequencing (NGS) panels in the clinic?” While mutation load determined using NGS is a potential biomarker, the limited sample size makes it hard to assess the actual utility and value.
However, there is a trend indicating that there is a critical threshold for mutation burden for specific disease (including melanoma)—higher mutation burden can improve patient response. “Why does mutation burden matter?” Charen asked. Mutations create neo-antigens, which create abnormal proteins which are then presented to the immune system by antigen presenting cells (APCs).
Charen pointed out several questions that remain unanswered:
Does mutation load matter in dual checkpoint blockade–treated patients?
What factors lead to primary and acquired resistance in tumors expected to respond to checkpoint blockade?
Do they upregulate other checkpoints or are the APCs modified or missing pathways?
Charen is hopeful that peripheral testing, using blood-based biomarkers could help make progress in the field. However, a significant challenge remains integration of this multivariable data in a statistically and biologically meaningful manner in the clinic.
NGS Could Be the Answer
Another approach to identifying a biomarker for checkpoint inhibitors is the use of NGS. Douglas Johnson, MD, MSCI, assistant professor of Medicine, Vanderbilt Ingram Cancer Center, presented results of his group’s assessment of somatic mutations in archived samples from patients with melanoma who had been treated with PD-1 inhibitors. The objective was to determine a correlation, if any, between the number and type of somatic mutations and outcomes following PD-1 inhibition.
Johnson said that many elegant studies have been conducted to identify PD-1/PD-L 1 biomarkers to predict response. Some of these strategies include identifying the presence, location, and clonal expansion of infiltrating T cells.
Neo-antigens are produced with increasing mutation load, and in their study Johnson’s team focused on hybrid capture–based NGS, for which they collaborated with Foundation Medicine and Adaptive Biotechnologies for sample analysis. The goal was to study a smaller portion of the genome to get an accurate surrogate for mutation load.
Mutational load captured following hybrid capture–based NGS of coding sequences 236 to 315 were correlated with clinical outcomes and compared with whole genome sequencing, Johnson said. Patients were divided into an initial cohort (median age 55 years) and a validation cohort (median age 62 years). Patients were treated with either nivolumab, pembrolizumab, or atezolizumab, and prior lines of treatment could include BRAF inhibitor, ipilimumab, or chemotherapy.3
Among patients in the initial cohort (n = 32) who responded to anti-PD-1 and anti–PD-L1 agents, higher mutation load was significantly greater in responders compared with nonresponders—responders had a median mutation load of 45.6 mutations/MB, compared with 3.9 mutations/MB among nonresponders. In the validation cohort (n = 33), responders had a median mutation load of 37.1 mutations/MB, compared with 12.8 mutations/MB among nonresponders.
Johnson and his team evaluated specific gene mutations in the patient samples and observed that more number of responders had mutations in NF1, LRP1B, and BRCA2, compared with nonresponders. Significantly, similar to what Diaz presented for CRC, patients with a high mutation load had greater PFS and OS compared with patients with low or medium mutation load.
“Is mutation load a positive prognostic feature?” Johnson asked. Based on their findings, Johnson proposed a potential model for treatment, derived from the mutation load of patients. According to the model, in patients with metastatic melanoma, a high mutation load should be the cue for treatment with anti–PD-1 monotherapy, and low or intermediate mutation load patients should be treated with combinations such as ipilimumab plus nivolumab. EBO