2 Clarke Drive
Suite 100
Cranbury, NJ 08512
© 2024 MJH Life Sciences™ and OncLive - Clinical Oncology News, Cancer Expert Insights. All rights reserved.
The wide variability in clinical outcomes among patients undergoing anti–PD-1/PD-L1 immunotherapy has increased interest in finding biomarkers that predict response.
Suzanne L.Topalian, MD
The wide variability in clinical outcomes among patients undergoing anti-PD-1/PDL1 immunotherapy has increased interest in finding biomarkers that predict response. Although tumor expression of PD-L1 is classically used as a biomarker, complexities associated with its expression and measurement in the tissue suggest that PD-L1 status should be used in conjunction with other biomarkers to best predict patients who would benefit.
That is the latest thinking of leading experts in the emerging immune checkpoint blockade immunotherapy field, according to several presentations at the 2016 ASCO Annual Meeting.
“We believe these biomarkers are more highly predictive and potent if they can be used in combination,” said Suzanne L.Topalian, MD, of the Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins School of Medicine and a presenter at the conference. “Moving forward, we’re going to examine these areas of intersection.”
Topalian presented data from a 2012 study showing that more than one-third of patients with PD-L1-positive metastatic melanoma, non—small-cell lung cancer (NSCLC), or renal cell cancer had an objective response to PD-1 blockade, whereas none of the patients with PD-L1–negative tumors responded.1
Another study showed that pembrolizumab improved objective response and overall survival (OS) in patients with NSCLC that highly expressed PD-L1 (≥50% of tumor cells),2 and a pooled analysis of seven studies showed that patients with PD-L1-positive NSCLC (PD-L1 staining on ≥1% of cells) had a greater overall response rate to PD-1 inhibitor therapy than those with PD-L1—negative tumors.3 Taken together, data in these studies suggested that tumor expression of PD-L1 is a strong predictor of response to PD-1 blockade, according to Topalian.
Complexities of Measuring PD-L1
However, Topalian identified more recent data that indicated PD-L1 tumor expression does not predict all responders to anti-PD-1/PD-L1 therapy. A 2015 review of all studies published on PD-1/ PD-L1 blockade showed that approximately 15% of patients with PD-L1—negative tumors responded to anti-PD-1 therapy.4 While this was significantly lower than the response rate in patients who had PD-L1—positive tumors (45%), Topalian stated that PD-L1 status alone is insufficient to exclude patients from receiving anti-PD-1/PD-L1 therapy.Several presenters noted multiple complexities inherent to the measurement and expression of PD-L1. Multiple types of cells in the tumor microenvironment, including infiltrating macrophages and activated lymphocytes, also express PD-L1, and immunohistochemistry (IHC) assays for PDL1 differ in the types of cells (immune-infiltrating cells, tumor cells, or both) they include when scoring PD-L1 expression.
Ravindra Uppaluri, MD, PhD, of Washington University School of Medicine in St. Louis, noted that for patients with recurrent/ metastatic head and neck squamous cell carcinoma in the KEYNOTE-012 trial,5 PD-L1 expression significantly predicted response to pembrolizumab when analyzed as a combined proportion score (expression in tumor and inflammatory cells) but not when analyzed as a tumor composite score (expression in tumor cells only).
Furthermore, he showed that 10 of the 29 patients with PD-L1-positive inflammatory cells and PD-L1-negative tumor cells responded to anti-PD-1 therapy—patients who would have been considered to be PD-L1-negative if only tumor cells had been included.
Additionally, Topalian stated that PD-L1 expression may vary between primary and metastatic lesions in patients with multiple metastases.
She presented a case of a patient with metastatic melanoma who achieved a complete response to anti-PD-1 therapy, yet only had PD-1 expression at one of the three tumor sites. “This patient could have easily been excluded from therapy if [PD-L1 expression] was used as a literal companion diagnostic. We don’t believe that it should be used as a requirement for therapy.”
Topalian further stated that PD-L1 expression is not homogeneous throughout the tumor microenvironment but tends to be concentrated at the boundaries between tumor cells and infiltrating stromal cells. “If you only had a needle biopsy for your diagnostic, you could easily miss the area of PD-L1-positive expression,” she said.
Janice M. Mehnert, MD, of the Rutgers Cancer Institute of New Jersey, also emphasized in her presentation that the highly dynamic nature of the immune system and tumor microenvironment may lead to a change in PD-L1 expression over time, and thus timing of the biopsy relative to initiation of treatment is important.
Expanding Biomarker Search Beyond PD-L1
Immune Signatures
“If the biopsy was performed 5 years ago, is that going to be representative of what is going on now?” she questioned. She also noted that other practical issues, including the instability of PD-L1 epitopes and various affinities of antibodies in the immunohistochemistry assays, increase the difficulty and complexity of obtaining an accurate assessment of PD-L1.Overall, Topalian, Uppaluri, and Mehnert agreed that the inherent complexities with obtaining an accurate PD-L1 measurement and the responses in patients with PD-L1-negative tumors highlight the need to investigate combinations of biomarkers, such as immune signatures, mutational load, soluble and cellular factors, and viral markers.Mehnert said that tumor-infiltrating T cells have been correlated with improved survival in studies of multiple cancer types, including ovarian cancer,6 colorectal cancer,7 melanoma,8 and lung cancer.9
Mehnert showed that in one study, patients with melanoma who responded to pembrolizumab had high levels of CD8-positive T cells, which increased after therapy, at the invasive margins of the tumor.10 By contrast, nonresponders had low levels of CD8-positive cells that did not increase with pembrolizumab.
According to Mehnert, these data suggest that PD-1 blockade induces a response by inhibiting adaptive immune resistance. She proposed that the tumors of responders have a distinct T cell-inflamed phenotype characterized by chemokines, a type I interferon (IFN) signature, and CD8-positive T cell—derived IFN-γ that upregulates PD-L1 and indoleamine-2,3-dioxygenase (IDO). Mehnert said that identifying the combination of factors that characterize this phenotype may help predict patients who would respond to therapy.
Uppaluri also suggested that the predictive value of PD-L1 may increase when used as part of an immune signature. Because IFN-γ regulates expression of both PD-L1 and PD-L2, which were significantly co-expressed in the KEYNOTE-012 trial, Uppaluri suggested that the co-expression indicates an IFN-γ signature that predicts response to anti-PD-1 therapy.
Further analysis showed that a gene expression classifier containing six IFN-γ-regulated genes significantly stratified responders and tended to be a stronger predictor of response than either PD-L1 or PD-L2. Uppaluri demonstrated that in a hypothetical 1000-patient cohort, the gene expression classifier had a positive predictive value of 40% and a negative predictive value of 90%. Although Uppaluri indicated the promise of such immune signatures, he stated that further refinement will be needed to use these and other immune signatures as predictive biomarkers for anti-PD-1 therapy.
Similarly, Jonathan E. Rosenberg, MD, of Memorial Sloan Kettering Cancer Center, presented data from the IMvigor 210 trial11 showing that patients with metastatic urothelial carcinoma who responded to atezolizumab had higher baseline expression of IFN-γ response genes, as well as IFN-γ-inducible major histocompatibility complex (MHC) antigen processing and transport genes.
Soluble and Cellular Factors
The study also showed that immune cell PD-L1 expression was associated with greater antitumor activity, and patients with high expression (≥5%) of PDL1 on immune cells had a greater objective response rate (28%) and OS than patients with low (≥1% and <5%) or no (<1%) immune cell PD-L1 expression. Furthermore, the high immune cell expression of PD-L1 was associated with activation of T cells, as Rosenberg showed that the IFN-γ—inducible Th1 chemokines CXCL9 and CXCL10, as well as CD8-positive T cell infiltration, were positively correlated with immune cell PD-L1 status. “Potentially, simultaneous assessment of these characteristics may define drivers of immune responsiveness and inform potential combination strategies as well as mechanisms of resistance,” he said.Mehnert presented studies that associated immunotherapy response with C-reactive protein, vascular endothelial growth factor, lactate dehydrogenase, immediate lymphocytosis, high absolute lymphocyte count, and increasing eosinophil count during therapy. Although she stated that the small size and retrospective nature of the studies limits their clinical applicability, she suggested that testing multiple serum and peripheral blood factors could be a convenient way to predict immune status within the tumor.
Mutational Load
She cited a study by Weide and colleagues12 that showed patients with high eosinophil count, absence of metastasis outside soft tissue or lung, high absolute lymphocyte count, and low lactate dehydrogenase at baseline had improved OS when treated with pembrolizumab, and having more factors was associated with better prognosis.According to Topalian, tumors with a high mutational burden respond particularly well to PD-1 blockade because mutations produce immunogenic peptides and neoantigens that stimulate the immune system.
Although colorectal cancers generally do not show a strong response to anti-PD-1 therapy, Topalian indicated that the microsatellite instability (MSI)-high genotype, which is present in 15% of colorectal cancer and 2% to 4% of all solid tumors and contributes to high mutational load, is an exception.
A recent study13 showed that 62% and 60% of mismatch repair (MMR)-deficient subsets of colorectal and noncolorectal solid tumors, respectively, had an objective response to anti-PD-1 therapy whereas none of the patients with MMR-proficient colorectal cancers responded. Mehnert also described a case study of a patient with DNA polymerase epsilon (POLE)-mutant endometrial cancer, which is associated with high mutational burden and elevated expression of several immune checkpoint genes, that had an exceptional response to pembrolizumab.14
Although Topalian indicated that MSI assays are relatively standard for most hospitals, they may not detect all cases with high mutational load. Furthermore, Mehnert stated that approximately half of POLE-mutant cancers show MMR proficiency on MSI assays.
Although whole-exome sequencing was used to determine mutational load in a recent study associating mutational load with clinical benefit and progression-free survival in pembrolizumab-treated NSCLC,15 Douglas B. Johnson, MD, of Vanderbilt-Ingram Cancer Center, stated that the method would be difficult and expensive to implement in the clinic. He presented a method that uses hybrid capture-based next generation sequencing (NGS) of a 236- or 315-gene coding sequence to profile a fraction of the genome and act as a surrogate for mutational load. He and his colleagues analyzed archived samples from patients with melanoma who received anti-PD-1/PD-L1 therapy and showed that the NGS-estimated mutation load was strongly correlated with the whole-exome sequencing results.16
Responders to pembrolizumab had a higher mutational load than did nonresponders, and further stratification showed that patients with a high mutational load had a significantly greater response rate and median OS than patients with intermediate or low mutational load, respectively. Rosenberg described the same NGS assay that was used in the IMvigor 210 trial11 to show that response to atezolizumab was associated with a significantly higher mutational load. Furthermore, the quartile with the highest mutational load had better OS than the quartiles with lower mutational load.
Although these results suggest high mutational burden as a potential biomarker for response to anti-PD-1 therapy, Topalian indicated that the cutoff of mutational load for predicting clinical response is unclear.
Viral Markers
Johnson agreed that mutational load alone should not be used to exclude patients from anti-PD-1 therapy, although he speculated that mutational load could play a role in prioritizing therapy for certain patients. For example, he suggested that monotherapy with PD-1 blockade could be effective for patients with high mutational load and less toxic than combination therapy with ipilimumab and PD-1 blockade.Like the neoantigens produced by hypermutated tumors, viral antigens are strong stimulants of the immune system, according to Topalian. She stated that because virus-positive tumors express PD-1 on infiltrating T cells and PD-L1 on tumor cells and macrophages, they may respond to anti-PD-1/PD-L1 therapy. She cited an analysis of archived tumor samples from Johns Hopkins that showed Merkel cell carcinoma expressing the Merkel cell polyomavirus is more likely to express PD-L1 in an immune-front pattern than virus-negative Merkel cell carcinoma.17
She also described a follow-up study18 showing that patients with advanced Merkel cell carcinoma had a 56% response rate to pembrolizumab. Response rates were similar between viruspositive and virus-negative tumors even though the virus-positive tumors had a lower mutational load.
Enhancing PD-L1 Expression With Nonimmunogenic Therapy
Combining Biomarkers: Strength in Numbers
Topalian suggested that the viral oncoproteins expressed by the virus-positive tumors likely stimulated the immune system much like the mutation-induced neoantigens produced by hypermutated tumors to trigger a response to pembrolizumab. She indicated that further study of anti-PD-1 therapy in this and other types of virus-associated cancers is warranted.Topalian also introduced the use of biomarkers to identify nonimmunologic therapies that enhance PDL1 expression and could be combined with anti-PD-1/ PD-L1 therapy. She cited one study19 showing that selective BRAF inhibition in patients with melanoma who harbor the BRAF mutation was associated with increased CD8-positive T-cell infiltration and tumor PD-L1 expression, which has led to the launching of several clinical trials investigating combination therapy with BRAF inhibition and anti-PD-1 therapy. Taken together, researchers agreed that a combination of biomarkers is likely necessary to predict patients who will respond to anti-PD-1/PD-L1 therapy. Rosenberg showed that although PD-L1 status, The Cancer Genome Atlas (TCGA) subtype, and mutation load were independent predictors of response to pembrolizumab in the IMvigor 210 trial,11 the combination of all three biomarkers improved prediction of response over each individual biomarker or the combination of PD-L1 and TCGA subtype (Figure). “These data highlight the importance of the interaction between the tumor and its microenvironment in understanding response to atezolizumab,” said Rosenberg.
Uppaluri also suggested that developing a cancer immunogram, such as the one proposed by Schumacher and Ribas20 that included mutational load, immune checkpoint status, and MHC status, may be valuable for predicting response to anti-PD-1 therapy. He suggested that further analysis of highly curated databases, such as TCGA, and ongoing trials of immune checkpoint inhibitors will aid in identification of immune signatures that predict response.
Although Topalian agreed that multiple biomarkers should be used to predict response, she indicated that multifactorial markers may be more difficult to regulate and interpret clinically and that currently, no single biomarker can definitively exclude patients from therapy.
With this consideration, Mehnert suggested that without a clear, integral biomarker to predict response, focusing on treating patients first and then analyzing their data will best clarify mechanisms connecting biomarker presence and response to anti-PD-1 therapy at this stage.