Tumor mutational burden as a tissue-agnostic biomarker for cancer immunotherapy

KEYWORDS:

• Cancer immunotherapy
• tissue-agnostic biomarker
• tumor mutational burden

Tumor mutational burden (TMB), the total number of somatic mutations per coding area of a tumor genome[1], has emerged as a biomarker predictive of response to immune checkpoint inhibition. Conventionally, anti-cancer treatment was determined by primary site and histology; in recent years, evidence suggests the role of common biomarkers to guide treatment in a tissue-agnostic manner. In June 2020, the FDA granted accelerated approval to pembrolizumab for the treatment of unresectable or metastatic solid tumors with high TMB (10 or more mutations per megabase) as determined by the FoundationOne CDx assay[2]. This represents a second time the FDA has approved a cancer immunotherapy for a tissue-agnostic indication, subsequent to the approval of pembrolizumab in microsatellite instability-high (MSI-H) or mismatch repair deficient (dMMR) solid tumors in 2017[3]. While the TMB-based approval was an exciting breakthrough, experts have expressed reservations in solely using TMB, at a cutoff of 10 mutations per megabase, as a predictive marker for all solid tumors.

The FDA approval was largely based on results from a prospective exploratory biomarker analysis of the KEYNOTE-158 (KN-158) study. The primary endpoint of the study, objective response rate (ORR) to pembrolizumab, was 29% in TMB-high tumors, 28% in TMB-high excluding MSI-H/dMMR tumors, and 6% in non-TMB-high tumors[4]. The results suggest that TMB is predictive of immunotherapy response, independent of MSI/dMMR status and programmed death-ligand 1 (PD-L1) expression. In this article, we discuss caveats and future directions of tissue-agnostic approach of cancer immunotherapy.

First, in the KN-158 study, only 7 out of 10 tumor types demonstrated a superior ORR in the TMB-high versus non-TMB-high group[4]. At the specified cutoff of 10 mutations per megabase, TMB-high anal cancer and mesothelioma had lower ORRs than TMB-low groups. Prevalence of TMB-high status in each tumor-type-specific cohort contributed to the reliability of ORR. Half of the cohorts (including biliary, mesothelioma, neuroendocrine, salivary, and thyroid cancer) had five or fewer patients with high TMB, leading to spuriously high (100%, 2 out of 2 in thyroid cancer) or low (0%, 0 out of 1 in mesothelioma) ORRs.

Results above suggest that the benefit of immunotherapy for TMB-high versus non-TMB-high tumors vary across tumor types. In an analysis to evaluate the relationship between TMB and ORR across multiple tumor types, Yarchoan et al observed a significant correlation between TMB and ORR, yet different tumor types respond better or worse than predicted by TMB[1]. This variation in tumor immunogenicity may be explained by other immune-related mechanisms not captured by TMB, such as viral antigen expression and antigen presentation efficiency[5]. For example, the presentation of viral antigens may confer a higher immunogenic response in Merkel cell carcinoma[6]. In contrast to MSI-H/dMMR status, TMB is a continuous quantitative biomarker. Thus, it is important to determine the pertinent threshold for defining TMB-H and whether that threshold is the same across tumor types.

Second, there was no demonstrable benefit of using pembrolizumab over standard later-line treatment. About 40% of patients in the analysis had only received one previous line of systemic treatment and could have benefitted from approved second-line treatment[4]. Moreover, the results showed no significant improvement in progression-free survival (PFS) or overall survival (OS) in the TMB-high compared to the non-TMB-high group[4].

Third, the TMB-high study population was predominated by small cell lung (SCLC), cervical, and endometrial cancer patients, with ORRs of 29%, 31%, and 47% respectively[4]. Prior to the TMB era, these three tumor types each had their own tissue-specific FDA indications for immunotherapy[7]. However, it is unclear how to integrate the agnostic TMB indication in the context of these tissue-specific indications. For example, questions that arise include whether SCLC patients will now require TMB testing before using pembrolizumab alone, and whether cervical cancer patients with PD-L1 combined positive score of <1 but TMB high will observe a similar ORR as reported in the study.

Moreover, the FDA approval requires that the TMB-high status for the use of pembrolizumab has to be determined by the FoundationOne CDx assay. TMB assessment is limited by the availability and reliability of the assay. While the gold standard for TMB assessment is whole exome sequencing (WES) by next-generation sequencing (NGS), targeted gene panels such as the FoundationOne CDx have been more commonly adopted in clinical settings due to its lower cost and complexity. However, the inter- and within-assay reliability of this technology still requires optimization. Other commercial NGS panels that estimate TMB vary greatly in terms of gene coverage, types of captured mutations, bioinformatic algorithms, and methods to filter germline variants. The Friends of Cancer Research TMB Harmonization Project observed considerable variability in TMB estimates across 11 NGS panels, both in silico and in formalin-fixed paraffin-embedded tumor samples [8,9]. The inconsistencies were more prominent at higher WES TMB values, particularly in endometrial carcinoma, bladder urothelial carcinoma, and colon adenocarcinoma[8]. The group suggested that calibration using The Cancer Genome Atlas data may represent a viable and robust approach to align across panel TMB scores[9].

Finally, although pembrolizumab is generally well-tolerated, there is a paucity of data showing convincing clinical benefit or improved quality of life. The TMB-based approval of pembrolizumab prompts more widespread use of genomic assays and pembrolizumab in the later lines of treatment across tumor types. Specific to each healthcare system, further research is needed to estimate the cost-effectiveness of this novel approach.

In summary, the KN-158 study has groundbreakingly demonstrated a tissue-agnostic TMB association to ORR to pembrolizumab. Nevertheless, clinical application of these results is limited by the lack of meaningful benefit (in terms of PFS or OS) over alternative treatment, a small sample size, and apparent differential prevalence of TMB-high status and variable response rates of TMB-high groups across tumor types. Importantly, whether pembrolizumab is effective among tumor types not included in the KN-158 study is yet unknown. Although the evidence was relatively stronger in SCLC, cervical and endometrial cancers, we need to be cognizant of the need and the choice of TMB, PD-L1, and MSI testing, as well as their combined interpretation in relation to the expected tumor-specific response rates before recommending to our patients.

As we gradually unravel the complex interplay between cancer cells and the immune system, it is becoming apparent that most biomarkers are imperfect if used separately on their own. In concert with TMB, inflammatory biomarkers such as PD-L1 expression and T-cell inflamed gene expression profile [10] have been shown to provide complementary information that predicts response to immunotherapy. These immune-relevant biomarkers play an important role in predicting not only response to immunotherapy, but also the risk of immune-related adverse events[11]. Our vision for tissue-agnostic cancer immunotherapy lies in the development and validation of a composite predictive score which incorporates testing that is reliable, accessible, and cost-effective.

Expert Opinion Despite the tissue-agnostic approval of pembrolizumab for the treatment of unresectable or metastatic solid tumors with high TMB, experts have expressed reservations in using TMB as a sole predictive marker for all solid tumors. Support for this approval is provided by the KN-158 study, which demonstrated an ORR of 29% in TMB-high tumors and 6% in non-TMB-high tumors. There are several important caveats to this approach, including: varied tumor immunogenicity across tumor types not accounted for by TMB, lack of comparison to a standard treatment arm, unclear integration with existing tissue-specific indications, and limited availability and reliability of TMB assays. Investigation of additional biomarkers is warranted, such as tumor infiltrating lymphocytes, immunoprofiling, T-cell receptor repertoire, and epigenetic signatures. Further research should refine the role and threshold of TMB in concert with other immune-relevant markers to guide the use of immunotherapy.

Declaration if interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This paper was not funded.