The problem

Over the past decade, immune checkpoint blockade (ICB) has emerged as one of the most promising approaches for the treatment of an increasing number of tumor types1. Tumor mutation burden (TMB) — the number of protein-altering somatic mutations in the cancer genome — has been used as a proxy for tumor foreignness, a feature that is important for responses to immunotherapies. However, with the exception of mismatch repair-deficient tumors2, TMB is an imperfect biomarker of immunotherapy response. Several studies have shown that mutation quality rather than sheer quantity contributes to the efficacy of ICB in a cancer lineage-dependent manner3. Importantly, the elimination of immunogenic mutations and neoantigens through chromosomal deletions during tumor evolution is linked to the emergence of acquired resistance to immunotherapies4. Thus, we explored whether the propensity of mutations to be retained or lost mediates sustained neoantigen-driven immune responses and differentiates responding from non-responding tumors in the context of cancer immunotherapy.

The discovery

We defined two distinct classes of mutations that are less likely to be lost in cancer cells during tumor evolution: mutations contained in single-copy regions of the genome, and mutations present in multiple copies per cell. First, we quantified the abundance of these ‘persistent’ mutations across 9,242 tumors from The Cancer Genome Atlas. We then evaluated the association between persistent tumor mutation burden (pTMB) and response to ICB treatment in 8 cohorts of over 500 patients with head and neck cancer, melanoma, mesothelioma, and non-small-cell lung cancer (NSCLC). To better understand the dynamics that govern mutation loss under selective pressure of ICB treatment, we comparatively analyzed samples from patients with NSCLC before ICB and at the time of acquired resistance, focusing on retention rates of persistent mutations. Finally, via transcriptomic analyses, we evaluated the impact of persistent mutations in differential phenotypes of the tumor microenvironment.

Our analyses showed that ~10% of mutations within the overall TMB are persistent mutations; this fraction was dependent on the cancer lineage. When comparing pTMB to TMB, we found a reclassification rate of 33% across cancers. Notably, in the ICB cohorts, pTMB distinguished responding from non-responding tumors better than TMB, loss-prone mutations or tumor aneuploidy (Fig. 1a–d). Furthermore, longitudinal tracking of mutations during ICB suggested a significantly lower rate of loss for persistent mutations compared (Fig. 1e). Further analyses revealed the tumor microenvironment of high pTMB tumors was highly enriched for interferon-γ and inflammatory response-related gene sets prior to therapy, a phenomenon that was enhanced during the course of ICB. Collectively, our findings support the notion that persistent mutations are a small yet biologically distinct and clinically relevant subset within the overall TMB with the potential to function as uneditable targets of adaptive immune responses and drive sustained immunological tumor control.

Fig. 1: pTMB and response to immune checkpoint blockade.
figure 1

a–d, pTMB better distinguishes ICB response from melanoma (a), non-small-cell lung cancer (b), mesothelioma (c), and non-responding head and neck cancer (d) compared to TMB. e, During treatment with ICB, a significantly lower fraction of persistent mutations (multi-copy and only-copy) were lost in tumor samples at the time of acquired resistance compared to loss-prone mutations (odds ratio = 61.46, P < 2.2 × 10-16). © 2023, Niknafs, N. et al., CC by 4.0.

Full size image

The implications

Our findings support the idea that mutations carry differential weights within the overall TMB and that a small number of mutation-associated neoantigens can be sufficient to elicit sustained antitumor responses4,5. Persistent mutations may function as a key driver of immunological tumor control that cannot be bypassed via neoantigen elimination under the selective pressure of immunotherapy. We argue that in addition to the overall TMB, the copy number states of the tumor and the sequence alteration load contained in these genomic regions should be considered in clinical decision making.

Prospective validation of these findings is warranted. Notably, we assessed persistent quality by integration of mutation coordinates with somatic copy number profiles in an agnostic manner without consideration of the function or clinical significance of the altered protein. It is therefore plausible that persistent mutations in oncogenes or tumor suppressor genes could carry differential weights, which suggests that the definition of persistent mutations can be further fine-tuned.

Translation of the current findings in clinical cancer care requires further prospective validation, as well as determination of accurate lineage-specific thresholds to enable classification of tumors into high versus low pTMB in each histology. Although in silico simulations supported the feasibility of pTMB assessment from gene panel sequencing, future steps include computing pTMB from targeted next-generation sequencing to further demonstrate its clinical utility.

Noushin Niknafs & Valsamo Anagnostou Johns Hopkins University School of Medicine, Baltimore, MD, USA.

Expert opinion

“While some studies suggest that quality rather than quantity of neoantigens determines immune surveillance and immunotherapy response, this is the first work to my knowledge that has looked at copy loss as a determinant of effective neoantigens. Overall, the methods are straightforward and clear, and the results are compelling”. Hannah Carter, University of California San Diego, La Jolla, CA, USA.

Behind the paper

We have long thought about ‘ranking’ mutations within the overall TMB based on their potential to drive immunological tumor control in the context of immunotherapy. Back in 2016, we discovered that mutation loss via chromosomal deletions drives resistance to immunotherapy; it took us several years to flip the question and ask: are there mutations that are hard to lose due to an intrinsic fitness cost to the tumor? Our ‘eureka’ moment came after discovering that a higher number of mutations in single-copy regions identifies mesothelioma tumors as more likely to regress with immunotherapy. Considering these mutations are evolutionarily ‘hard-to-lose’, we focused on additional alterations that are more likely to be retained during tumor evolution, notably mutations contained in multiple copies. Together, persistent mutations impose an evolutionary bottleneck that cancer cells cannot readily overcome and may thus drive sustained tumor control in immunotherapy. V.A.

From the editor

“This work stood out to us because it proposes the interesting concept of rethinking the clinical utility of the TMB, taking into consideration single-copy regions in the tumor that cannot be eliminated without affecting the fitness of the tumor cell. This study opens up the possibility of using the new metric to predict response to immunotherapies.” Editorial Team, Nature Medicine.