Studying the genomic wiring of response to immune checkpoint blockade

Like Comment
Read the paper

The question that triggered this study was why does tumor mutation burden (TMB) fail to accurately predict responses to immune checkpoint blockade. Tissue TMB estimates are subject to sampling bias and are strongly affected by low tumor purity, however this simple –in principal- concept is completely underestimated in the clinical setting, where physicians often guide their therapeutic decisions based on TMB estimates generated from tumor samples of suboptimal tumor purity. As we dove into the challenging task of assessing TMB in the setting of low tumor purity and clonal heterogeneity it became evident that Oscar Wilde’s saying “the pure and simple truth is rarely pure and never simple” certainly applied to our work. 
Through simulation analyses and in silico dilution experiments of simulated tumor samples we found that the power of detection of genomic alterations especially in the setting of high intra-tumoral heterogeneity diminishes with decreasing tumor purity. Importantly, at the lower end of tumor purity the observed TMB of a TMB-high but clonally heterogeneous tumors becomes indiscernible from the observed TMB of a TMB-low clonally homogeneous tumor. These findings were reproduced through analyses of whole-exome and targeted sequence data from >5,000 tumors, where we found a significant correlation between TMB and tumor purity. This observation was not alleviated by higher sequencing depth. 
As our findings suggested that analysis of low tumor purity tumors is likely to yield inaccurate TMB estimates, we next sought to develop a correction for TMB. We developed a new computational method to estimate a corrected TMB (cTMB) from both whole exome and targeted sequence data. When TMB was adjusted for tumor purity, cTMB more accurately predicted outcome to immune checkpoint blockade. We believe that our approach may prove useful in meaningfully interpreting TMB estimates from next generation sequencing in the clinic. 
But even if TMB is standardized, it remains an imperfect biomarker of response as it captures only one side of the cancer-immune system crosstalk. Next, we focused on studying additional molecular features to determine the genomic wiring of response to immune checkpoint blockade. We performed an unbiased differential enrichment analysis between responding and non-responding tumors to identify genomic alterations selectively linked with clinical outcome. To our surprise, we did not detect recurring mutations in the JAK1/2 genes, co-occurring KRAS/STK11 mutations or mutations in interferon-ɣ related genes in non-responding tumors. Importantly, we found that such differential enrichment analyses may be confounded by TMB and a TMB-adjustment strategy is crucial. Through analyses of sequence and genome-wide structural alterations, we found that tumors, which harbor activating mutations in receptor tyrosine kinase (RTK) genes, are more likely to be resistant to immunotherapy- independent of their TMB. These tumors were more sparsely infiltrated by cytotoxic T cells, suggesting that signaling through receptor tyrosine kinases may be tied into intra-tumoral T cell depletion. 
Ultimately, we developed an integrated model to combine genomic features into a multi-parameter predictor of outcome for patients treated with immune checkpoint blockade. Together, corrected TMB, RTK mutations, smoking-related mutation signature and germline genetic variation in class I human leukocyte antigen (HLA) provided an improved predictor of response to immunotherapy, compared to TMB alone, even the corrected TMB. We envision that such approaches will be incorporated in clinical practice and allow for rational clinical trial design by identifying patients more likely to benefit. 
This work came to fruition through multiple interactions of a truly multidisciplinary team of cancer biologists, bio-informaticians, thoracic oncologists, thoracic pathologists and immunologists and was only possible due to the generosity of our patients.  

Valsamo Anagnostou

Assistant Professor, Johns Hopkins Cancer Center