Tumor Mutational Burden: A Continually Evolving Biomarker for Immunotherapy
Within recent years, the field of oncology has seen major advancements in regards to therapeutic options for patients. Immunotherapy, which relies on the patient’s own immune system to fight the disease, has demonstrated robust efficacy in the treatment of numerous types of solid tumors and has been shown to produce positive results in some hematological malignancies. However, there is still a void for biomarkers that will fully distinguish responders and non-responders to immunotherapeutic intervention.
In their seminal paper, the concept of mutational signatures was explored by Alexandrov et al. 2013. Their findings suggested that underlying patterns of mutations could be categorized into distinct signatures that could distinguish tumor types and subtypes, and provide insights into potentially effective therapeutic options. In the years following the publication of this paper, tumor mutational burden (TMB) has evolved into a clinically-relevant biomarker for the evaluation of the potential immunogenicity of tumors and for the determination of their likelihood to respond to immune checkpoint inhibition (ICI). This evolution is addressed in Chan et al. 2019, which discusses the critical studies that helped advance the development of TMB. The rationale for TMB as an immunotherapy biomarker is quite simple. Genomic instability is a characteristic of most cancer cells and leads to the accumulation of somatic small and structural alterations in the tumor genome. These somatic mutations can lead to the production of neoantigens, which are tumor-specific, aberrant peptides that are presented on the surface of tumor cells where they can be identified as “foreign” by T-cells and thus lead to the immune system-mediated destruction of said tumor cells. Neoantigens can only be derived from somatic mutations that result in a change in the amino acid sequence of the relevant protein. Such mutations are referred to as being “nonsynonymous” and they can occur anywhere in the protein-coding regions of the genome. Therefore, TMB, which is the total count of nonsynonymous somatic mutations per interrogated megabase of genomic content, serves to act as a surrogate for neoantigen load and, ultimately, tumor immunogenicity (Goodman et al. 2017).
Initial studies proved quite exciting in demonstrating TMB as a tool for predicting response to immune checkpoint inhibition (Snyder et al. 2014, Rizvi et al. 2015). Chalmers et al. 2017 analyzed ~100,000 genomes from various malignancies to evaluate the diverse landscape of TMB. Recent works, such as the study published in JAMA by Singal et al. 2019, investigated over 4,000 NSCLC cases and showed that patients who had a high TMB and were treated with anti-PD-1/PD-L1 had a significantly higher overall survival (16.8 months versus 8.5 months overall survival).
While the utility of TMB has been consistently demonstrated in cancers like NSCLC and melanoma, the benefits of the biomarker remain less clear in other indications such as glioblastoma (GBM) and many hematological cancers. Zhao et al. 2019 found that responders in a GBM cohort, to anti-PD-1 therapy, did not have more nonsynonymous variants. Additionally, recent data from a group at Johns Hopkins University (Anagnostou et al. 2020) showed variations in tumor purity can have profound impacts on the accuracy of TMB estimations. Using a NSCLC cohort treated with anti-PD-1, they determined that a tumor purity correction factor resulted in a more reliable and accurate survival curve separation, thus suggesting that tumor purity-corrected TMB value is likely to be better predictor for response to checkpoint blockade. This points to the need for a finer understanding of the technical and biological nuances of this metric to avoid misclassification.
Obstacles and Challenges of TMB
A number of uncertainties still surround appropriate assessment of TMB, one of which is the identification and filtering of germline mutations. While a matched normal specimen may not always be available for analysis, paired germline sequencing will most definitely reduce the number of false positive somatic mutations that would be included in the TMB count. While bioinformatic pipelines can utilize filters, databases, and quality metrics to compensate for when a germline DNA sample is not acquired, Garofalo et al. 2016 reported an increase in germline false positives in their tumor-only TMB analyses, yet databases such as ExAC aided in alleviating this discrepancy.
Additionally, the issue of what variant types to include in the evaluation and determination of TMB. Many TMB calculation methodologies to date are being defined as either the total number of somatic mutations or as a ratio, somatic mutations per megabase (Mb) sequenced. Additionally, many targeted panel-based approaches include not only nonsynomynous mutations in their TMB count, but also synonymous mutations, the biological appropriateness of which is open to debate. Many of these same methodologies have included only point mutations or single nucleotide variants (SNVs) as part of their calculation. However, studies continue to demonstrate that other variant types may be informative to include in a TMB assessment. For example, Mandal et al. 2019 demonstrated that the insertion and deletion (indel) load was specifically associated with anti-PD-1 response in tumors with mismatch repair deficiency (MMR-d). Likewise, in other mutational landscape assessments like neoantigen load, indels (Turajlic et al. 2017) and gene fusions (Yang et al. 2019) have been shown to be important variant types in understanding patient response to ICI.
An Attempt to Standardize and Harmonize TMB
Until recently, a standard method for reporting of TMB had not been established. In 2020, the Friends of Cancer Research (FOCR) addressed the need to harmonize and establish guidelines for the estimation of TMB (Merino et al. 2020). It was acknowledged that TMB estimation was optimally calculated from WES, but that the advent and utilization of a panel-based approach varied too greatly across the field. There were several factors that were recognized as leading to variability, such as gene coverage, panel size, genes of interest, and bioinformatic workflow. This large in silico assessment performed by the FOCR group, evaluated 11 panels in the context of concordance and deviation from WES estimation of TMB. From this study, a series of practices were recommended for reporting TMB to ensure consistency. The consortium recommendations included the need to report TMB as the number of non-synonymous mutations/Mb to accurately reflect the alterations found in coding content. Analytical validation studies for TMB estimation should be standardized to include assessment of analytical accuracy, precision and sensitivity. Finally, consistency across panels could be ensured through alignment of panel TMB values to WES-derived universal reference standard. Thus, the ongoing FOCR efforts are greatly helping to address and harmonize TMB. In the next phases, they will investigate other factors, such as biological factors including different cancer and specimen types, and will continue to refine their suggested best practices.
ImmunoID NeXT™ TMB Assessment
With the ImmunoID NeXT Platform®, the reported TMB is well-aligned with the recent FOCR-guidance for calculation. Using an enhanced exome-based approach, nonsynonymous somatic mutations (both SNVs and indels) are reported per Mb. In recent data presented at AACR earlier this year, a general concordance was shown from a comparison of FOCR exome-derived TMB from TCGA samples (Merino et al. 2020) and the ImmunoID NeXT exome-derived TMB from the NeXT database for various pan-cancer samples (below; Figure 1). The current biomarkers of today, such as TMB or MSI status, are provided with ImmunoID NeXT, yet it also enables a path towards more novel IO marker identification. To learn more regarding the benefits of the ImmunoID NeXT solution, please visit our website at www.personalis.com
Figure 1: ImmunoID NeXT-derived TMB is exceptionally well-aligned with FOCR exome across tumor types