Keywords

Introduction

Immunotherapy exploits the body’s immune system to fight cancer. The explosion in the number of ongoing cancer immunotherapy (CIT) trials reflects the great enthusiasm and potential for a cancer cure the field is believed to hold. Strikingly, more than 800 different immunotherapy clinical trials are currently underway to evaluate anti-PD-L1/PD-1 and other immunomodulatory agents alone or in combination with existing or new molecular entities in cancer patients [1,2,3]. To date, immunotherapy has shown efficacy in numerous tumor types including melanoma, non-small cell lung cancer (NSCLC), renal cancer, bladder cancer, colon cancer, and head and neck cancers [4]. These successes are only just the beginning. In the years ahead, new drug combinations will allow us to unlock patient populations and new indications that are currently unresponsive to the existing CIT treatments. As additional therapeutic options come to the market, it will be critical to identify and co-develop novel predictive biomarkers that will facilitate matching the right patient with the right drug combination.

The first wave of immunotherapy trials compared CIT monotherapy versus standard-of-care chemotherapy and established efficacy in cancer patients with advanced and metastatic disease that failed previous lines of therapy. The current focus (second wave) of clinical trials is to augment the success seen with monotherapy CIT through the combination of immunotherapy with other modalities, such as chemotherapy, radiation, or with other immunomodulatory agents. Promising results and positive trials have now been reported in NSCLC and renal cell carcinoma, where anti-PD-L1 or anti-PD-1 inhibitors were combined with chemo- and/or anti-angiogenic therapy or immune doublet therapies such as the combination of PD-1 targeted agents with CTLA4 inhibitors [5, 6]. A third wave of trials will undoubtedly combine novel multiple immunomodulatory agents with anti-PD-L1/PD-1 with the intention to replace cytotoxic chemotherapy.

While clinical success has been observed across multiple tumor types and with multiple agents, only a minority of patients have benefited from durable and long-lasting responses. It will be critical in the coming years to develop a deeper understanding of this phenomenon in order to facilitate smarter and science-driven drug combinations. Biomarkers have been and will continue to define patient populations that respond best to therapy (predictive biomarker), that are prone to poor clinical outcome (prognostic biomarker), or that are at risk of higher drug-associated toxicity (safety biomarker). Predictive biomarkers have garnered the most attention, as they are commonly used to guide treatment decisions for patients. Importantly, predictive biomarkers have increased clinical trial success rate, accelerated market access, decreased clinical development costs, and ultimately saved patients from receiving drug from which they would not benefit. To ensure biomarker results are robust and reproducible, biomarker assay development requires extensive analytical validation. As discussed in other chapters of this book, the emergence of novel high throughout technologies, such as mass spectrometry and gene expression profiling, provides an exciting opportunity for comprehensive biomarker profiling and co-development of companion diagnostics with new therapies [7, 8]. The remainder of this chapter focuses on predictive biomarkers, both approved as well as those still under development, for currently approved CIT drugs.

Immune checkpoint blockade therapies have transformed our approach to cancer treatment. The checkpoint inhibitors currently approved for clinical use are listed in Table 29.1. Based on the consistent clinical activity seen with anti-PD-L1/PD-1 inhibitors, the first-generation diagnostic tests focused on detecting PD-L1 expression in the tumor tissue. Tissue-based, immunohistochemical (IHC) tests were developed and employed in clinical development programs. PD-1, the receptor for PD-L1, is a member of the B7-CD28 super family and is expressed on numerous cell types, including activated T cells, B cells, and NK cells [28]. The interaction of PD-1 with its two ligands, PD-L1 and PD-L2, which can be expressed on various cell types within the tumor microenvironment (including tumor and various types of immune cells), may lead to downregulation of a potential antitumor immune response [29]. PD-L1 IHC assays have been granted regulatory approval as companion or complementary diagnostic tests (see details below) for PD-1/PD-L1 targeted therapies. Additional biomarkers such as those evaluating gene expression signatures in tissues, tumor mutational burden, and others have been evaluated in an exploratory fashion and may achieve regulatory approval at some point in the future [8, 30].

Table 29.1 Currently approved checkpoint inhibitor therapies and associated diagnostic tests
Full size table

PD-L1 Immunohistochemistry (IHC)

IHC assays that are currently approved to detect PD-L1 in patients for anti-PD-L1/PD-1 therapies differ from each other in numerous ways, ranging from cutoffs defining the PD-L1 positivity to the type of cells (i.e., tumor vs immune) used to determine PD-L1 expression. Four tests currently being used in the clinic (22C3, 28-8, SP142, and SP263) are discussed below. These tests rely on formalin-fixed, paraffin-embedded (FFPE) tissue sections as source material, which is typically available in pathology laboratories for diagnostic and biomarker testing. The different assays are summarized in Table 29.1.

22C3

The PD-L1 IHC 22C3 pharmDx test from Agilent-Dako is a qualitative immunohistochemical companion diagnostic test to identify patients eligible for treatment with pembrolizumab (KEYTRUDA®; Merck). It is approved for patient identification in first-line (1L) and second-line and beyond (2L+) NSCLC and third-line gastroesophageal cancer. In 1L and 2L+ NSCLC, this test was approved based on the results of phase III randomized studies KEYNOTE-024 and KEYNOTE-010, respectively [21]; patients with PD-L1 expression on at least 50% (1L) and 1% (2L+) of tumor cells showed clinical benefit from pembrolizumab. While PD-L1 assessment is limited to tumor cells in NSCLC, it includes both tumor and stromal cells in gastroesophageal cancer. In the KEYNOTE-059 study, PD-L1 expression as determined by 22C3 was categorized as the PD-L1 combined positive score (CPS) defined as the percentage of PD-L1-expressing tumor and infiltrating immune cells relative to the total number of viable tumor cells. Patients with a CPS of ≥1% derived clinical benefit from treatment with pembrolizumab.

28-8

The PD-L1 IHC 28-8 pharmDx test from Agilent-Dako is a qualitative immunohistochemical test approved as complementary diagnostic assay to identify patients, who may derive clinical benefit from treatment with nivolumab (OPDIVO®; Bristol-Myers Squibb) in second-line NSCLC and first-line metastatic melanoma based on the results of the phase III randomized studies, CheckMate-057 and CheckMate-067, respectively. Both trials showed improved survival of patients receiving nivolumab compared to patients on the control arm; the survival benefit was more pronounced in patients with PD-L1-positive tumors (defined as tumors with ≥1% of tumor cells showing complete or incomplete membranous staining of any intensity). PD-L1 protein expression in this assay is defined as the percentage of tumor cells exhibiting positive membrane staining at any intensity; staining on stromal/immune cells is not considered. Unlike the companion diagnostic test 22C3, which is required for patient eligibility for pembrolizumab, 28-8 is a complementary diagnostic test for nivolumab and is intended to aide in the clinical decision-making process [31].

SP142

The SP142 PD-L1 IHC assay developed by Ventana is approved as a complementary diagnostic test for the use of atezolizumab (TECENTRIQ®; Genentech/Roche) in patients with NSCLC and urothelial carcinoma. Both tumor cells (TC) and tumor-infiltrating immune cells (IC) are evaluated for PD-L1 expression in this assay. PD-L1 TC results are expressed as the percentage of tumor cells staining positive at any intensity, and IC results are expressed as the percentage of the area of viable tumor occupied by PD-L1-positive immune cells. The complementary diagnostic approval for NSCLC was based on a randomized phase III trial (OAK) comparing atezolizumab with standard-of-care chemotherapy in patients who had failed first-line therapy; a survival benefit to atezolizumab was observed across all levels of PD-L1 expression but was greatest for patients with the highest PD-L1 expression (defined as expression on ≥50% TC or ≥10% IC) [26]. For urothelial carcinoma, approval as a complementary diagnostic was based on a single-arm, two-cohort phase II trial for patients who had failed platinum-based chemotherapy. Treatment benefit from atezolizumab was greatest in patients with tumors expressing PD-L1 on ≥5% of IC [27]. Most recently, atezolizumab in combination with bevacizumab showed positive results in a phase II trial for metastatic renal cell carcinoma. Patients with PD-L1-positive tumors on ≥1% of IC derived greater benefit than the intention-to-treat population (https://cancerletter.com/articles/20180209_7/).

SP263

The SP263 PD-L1 IHC assay developed by Ventana was first approved as a complementary diagnostic for treatment of patients with advanced urothelial carcinoma with durvalumab (IMFINZI®; AstraZeneca) based on results of single-arm phase II study. The SP263 scoring algorithm captures PD-L1 expression on both TC and IC. A tumor sample is scored as positive if ≥25% of tumor cells exhibit membrane staining of any intensity, if ≥25% of immune cells are positive and occupy >1% of the viable tumor area, or if 100% of IC are PD-L1-positive and occupy 1% of the tumor area (https://www.accessdata.fda.gov/cdrh_docs/pdf16/P160046C.pdf). SP263 has also been CE-marked in EU and can be used to identify NSCLC patients for treatment with nivolumab or pembrolizumab. This approval was based on comparability studies of SP263 with the 28-8 and the 22C3 IHC assays, respectively. In this application only TC are evaluated for PD-L1 expression with SP263 assay.

Limitations of PD-L1 IHC Biomarker Data

The variation in assay platforms, primary antibody clones, and secondary detection reagents among the various PD-L1 assays may yield different results when testing the same tumor for PD-L1 expression. Furthermore, the scoring algorithms used to identify the PD-L1-positive cell types and cutoffs differ among the assays (i.e., tumor cells only vs tumor and immune cells). To evaluate the analytical performance and address market harmonization of the approved PD-L1 IHC assays, a collaborative project called “The Blueprint Programmed Death Ligand 1 (PD-L1) Immunohistochemistry (IHC) Assay Comparison Project” was established [32]. Analytical comparison demonstrated good concordance between the 22C3, 28-8, or SP263 assays with respect to proportion of PD-L1-positive tumor cells; the SP142 assay showed lower proportion of positive tumor cells. Greater variability across the four assays was seen for immune cell staining but without a consistent pattern. Based on the limited sample set, the authors concluded that misclassification of a proportion of tumors with respect to PD-L1 status could occur when using assays and/or scoring algorithms interchangeably [32].

There are several limitations of this first phase of the Blueprint project that could be addressed in subsequent iterations. The cohort of NSCLC cases analyzed was small and not associated with clinical outcome; therefore, the predictive value of each individual assay and associated algorithm could not be evaluated. Pathologist readers (n = 3) did not receive training for interpretation of the assay they were not expert in. Lastly, an orthogonal methodology to verify degree and pattern of PD-L1 expression in the tissues was not attempted. Analysis of larger NSCLC cohorts has now been published by other investigators [33, 34]. It should also be noted that each of the four assays has demonstrated predictive value in pivotal clinical studies and is FDA approved. A recent study compared the performance and predictive value for two PD-L1 assays, 22C3 and SP142, on patients treated with an anti-PDL1 agent atezolizumab in second-line NSCLC; the results demonstrated equivalent survival benefit in patient populations defined as positive for PD-L1 by either assay [35].

Each of the four assays described above performs robustly and reliably when used as prescribed and can enrich for the appropriate patient population. However, the use of an IHC assay in general and PD-L1 as a predictive biomarker for anti-PD-L1/PD-1 targeted therapies specifically comes with challenges [36]. First of all, the tumor microenvironment is a dynamic space with interactions of multiple cell types and assessment of a single marker most likely represents an oversimplification. IHC tends to perform well for binary observations but is much less reliable for the readout of continuous variables; gradients of expression are most likely critical for activation or inhibition of a tumor immune response and might be measured more appropriately through alternate technologies (see below). Common to most cancer immunotherapy trials is the observation that biomarker (PD-L1)-negative patients may respond suggesting that – in the presence of an adequately performing assay – the tissue sample is inadequate (time, location) or the biomarker is imperfect. Below we focus on technologies which characterize the tumor microenvironment more globally and are tested in ongoing clinical trials for their value in identifying patients for checkpoint inhibitor treatment.

Gene Expression Signatures

Clinical benefit to CIT therapies has been observed in patients that lack PD-L1 expression (i.e., patients negative by PD-L1 IHC tests). This observation suggests a complex biology underlies the immune response and that a single-parameter assay may be insufficient to predict patient outcome to CIT. Furthermore, it highlights the need for continued biomarker discovery to more accurately identify patients that benefit most from CIT therapy [26]. Gene signature profiles are being studied as predictive biomarkers in trials using checkpoint inhibitors for a variety of indications. At the foundation of many of these signatures are T-cell- and immune biology-related genes that are thought to represent T-cell infiltration and pre-existing immunity within the tumor microenvironment. Gene expression assay outputs have the upside of a quantitative continuous variable and are not limited by subjective assessment that accompanies pathological assessment by IHC. In a phase II trial evaluating ipilimumab in advanced melanoma, increased numbers of tumor-infiltrating lymphocytes (TILS) and increased expression of FoxP3 and IDO by IHC were associated with clinical activity. Gene expression analysis in these samples also found increases in immune-related genes such as granzyme B, perforin-1, and T-cell receptor subunits in samples during treatment; however, such changes did not reach levels of statistical significance [37]. Likewise, a recent neoadjuvant ipilimumab melanoma study discovered that high expression of immune-related genes in baseline tumor samples predicted clinical benefit [38].

While there has not been a gene signature assay yet approved by regulatory agencies, favorable associations and outcomes have been similarly reported in a wide range of anti-PD-L1/PD-1 therapies. The phase II POPLAR and phase III OAK studies, both evaluating efficacy of atezolizumab vs docetaxel in 2L+ NSCLC patients, revealed a novel association between a T-effector gene signature profile and clinical benefit (PFS and OS) to atezolizumab [39, 40]. A similar association between the T-effector gene signature and clinical benefit to atezolizumab in combination with bevacizumab and chemotherapy was observed in a recently reported phase III trial, IMpower150 [41]. Similar results were also reported for durvalumab, whereby a baseline IFN-γ gene expression signature was associated with improved clinical outcomes in durvalumab-treated advanced NSCLC cancer patients [42]. Yet another independent study identified and validated gene signatures related to IFN-γ signaling and activated T-cell biology for pembrolizumab across multiple distinct tumor types including melanoma, head and neck squamous cell carcinoma (HNSCC), and gastric cancer [43]. This T-cell-inflamed gene expression profile (GEP) suggested that a tumor microenvironment characterized by antigen presentation, active IFN-γ signaling, and cytotoxic effector activity is potentially responsive to PD-1 checkpoint blockade [43]. Altogether, gene expression profiling affords promise that more accurate and sensitive RNA-based next-generation diagnostic assays will be available in the near future for CIT patients.

Tumor Mutational Burden

Tumor mutational burden (TMB) is yet another biomarker that has demonstrated potential as a predictive biomarker for CIT patients [44]. TMB measures the number of mutations per coding area within a tumor genome. All cancers are caused by somatic mutations, which are attributed to a number of factors including malfunctioning of the DNA replication machinery, exposure to various mutagens (e.g., tobacco, asbestos, UV light), DNA modification, defects in DNA replication, etc. As the adaptive immune system requires foreign antigen presented on MHC to initiate a T-cell-directed immune response, patients with highly mutated tumors are more likely to generate and present a neoantigen that can be subsequently recognized by immune cells to mount an effective antitumor response.

Alexandrov et al. conducted an exhaustive study of the mutation data set across thousands of tumors from 30 cancer types with the goal of evaluating average mutation burden across indications. Interestingly, the actual number of mutations per megabase varied on average across tumor type with the highest mutation burden observed in CIT responsive indications, including melanoma, NSCLC, and bladder cancer [45] (Fig. 29.1). Together with subsequently reported clinical correlation between TMB and efficacy of CIT, this study opened a novel diagnostic opportunity to test mutation burden as a predictor of clinical response to CIT therapy [45].

Fig. 29.1
figure 1

Number of somatic mutations (per megabase of genome) across human cancer types. Dots represent individual samples. Red lines correspond to median numbers of mutations in each cancer type. (Reprinted from Alexandrov et al. [45]. With permission from Springer Nature)

Full size image

The commonly used method to measure TMB includes comprehensive genomic profiling using whole-exome sequencing, whereby all genes in the protein-coding region of the tumor genome are sequenced. As a robust alternative, targeted cancer genome panels, such as the FoundationOne® assay (Foundation Medicine), have provided reliable results. This method measures the somatic mutations occurring in selected genes, instead of sequencing the entire exome [27].

Multiple clinical studies have now demonstrated that TMB can predict responses to checkpoint inhibitor immunotherapies across different cancer types, including NSCLC, SCLC, and bladder cancer [46,47,48,49]. TMB was also associated with higher response rate and PFS, but not OS, to nivolumab vs chemotherapy in the exploratory analysis of the recently reported phase III study CheckMate-026 in PD-L1-selected 1L NSCLC patients [50]. The main caveat of TMB analysis in this study was that patients had already been selected based on PD-L1 IHC expression. Nevertheless, these results further solidified the importance of TMB as an important CIT biomarker. Importantly, PD-L1 IHC and TMB appear to be independent predictors of CIT efficacy, further suggesting that multiple biomarkers may be needed to most accurately identify patients benefiting from CIT. TMB can also be reliably measured in blood of cancer patients through NGS approaches on circulating tumor DNA. NSCLC patients with high TMB in their plasma derived an improved PFS benefit to atezolizumab vs docetaxel as demonstrated in an exploratory analysis of studies POPLAR and OAK [51]. Together these data warrant development of blood-based biomarkers for CIT in the near future.

In 2017, FDA approved pembrolizumab for treating solid tumors that are microsatellite instability-high (MSI-H) or mismatch repair deficient (dMMR). This was a seminal approval, as it was the first pan-tumor/pan-tissue agnostic FDA approval of a CIT drug. Approval was based on pembrolizumab efficacy across 15 cancer types, including colon cancer, renal cell cancer, and pancreatic cancer [52]. The efficacy observed in these cancer patients most likely reflects the high level of somatic mutations and mutation-associated neoantigens that trigger the immune system to mount a response against the tumor.

Other Biomarkers

In addition to the three platforms discussed, several other platforms have emerged that aim to profile the tumor immune interface in order to predict response to immunotherapy. These platforms are based on multiplexed transcriptome analysis, protein expression, and genomic variability. Examples include multiparametric flow cytometric immunophenotyping of peripheral blood, T-cell receptor (TCR) sequencing for clonality assessments of tumor-infiltrating lymphocytes (TILs), and assessing presence of CD8+ T cells and TILs within the tumor. These approaches are preliminary and require further investigation but have shown promise in predicting response to CIT checkpoint blockade [8, 53].

Summary

While many patients derive long-term clinical benefit from various CITs alone or in combination with other modalities, a substantial number of patients do not derive such benefit. Therefore, there is a great need to develop and validate predictive biomarkers of response to CIT. Here, we discussed the major biomarker platforms which are FDA approved or being actively pursued for CIT. It must be highlighted that the development of a biomarker test for clinical application is a highly regulated process that involves proper clinical trial design for clinical validation. The regulatory aspects of submission of biomarker assays to the FDA in the USA, as well as regulatory considerations in the European Union and other regions, must be well thought out during the planning and implementation phases of clinical and biomarker development. Ultimately, the approval of well-validated clinical biomarkers can maximize the benefits of CIT while reducing cost and toxicity [54]. Finally, a complex immune biology underlying responses to CIT suggests that using a single biomarker, such as PD-L1 IHC, to identify all patients benefiting from CIT is not possible and that most likely multiple approaches are needed to increase our precision in identifying patients benefiting from CIT [3]. Identifying such more complex biomarkers will be increasingly important to help to select a personalized treatment regimen for patients, as numerous CIT mono- and combination therapies will be becoming available in the future.