Most, but not all, patients with microsatellite-unstable gastric cancer respond to anti–PD-1 therapy. In this issue, Kwon and colleagues show, first, that differences in tumor mutation burden (TMB) may drive this variation in outcomes and, second, that treatment with immune checkpoint inhibitors leads to further immunoediting and a reduction in TMB in responding patients.
See related article by Kwon et al., p. 2168.
The efficacy of immune checkpoint inhibition (ICI) has been most realized in tumors with a high mutation burden (TMB), which may result from external forces (UV light, tobacco-related carcinogens) or internal defects in the DNA repair machinery, such as mismatch repair deficiency. The particular sensitivity of mismatch repair–deficient (MMRD)/microsatellite-unstable (MSI) tumors to ICI led to pembrolizumab, an anti–PD-1 antibody, being the first drug to obtain a tumor-agnostic regulatory approval. Long-term survival is within grasp for some but not all patients with MSI tumors, owing to intrinsic and acquired resistance. Recent studies of resistance mechanisms to ICI in MSI have focused primarily on colorectal cancer. However, gastric cancer, which has a similar proportion of metastatic MSI cancers, is almost comparable to colorectal cancer in terms of global cancer mortality. Until now, the mechanisms of resistance to ICI in MSI gastric cancer were not clearly established.
In this issue of Cancer Discovery, Kwon and colleagues (1), in a prospective clinical trial of advanced MSI gastric cancer treated with anti–PD-1 therapy, comprehensively investigate the landscape of intrinsic resistance to ICI. In doing so, they establish that MSI status in gastric cancer is not dichotomous but a continuous variable, including a spectrum of TMB levels, transcriptomic heterogeneity, and diversity in microenvironment composition accounting for differences in clinical response to treatment.
Using whole-exome sequencing (WES) on 15 MSI gastric cancer pretreatment tumor biopsy specimens, the investigators first established that tissue TMB was the strongest predictor of response to pembrolizumab. Within MSI gastric cancer, the range of TMB detected was wide, from fewer than 20 to more than 50 Mut/Mb. Responding patients had a higher median TMB (38 Mut/Mb) than nonresponders (24 Mut/Mb; Fig. 1). Modeling suggested an optimal cutoff for response of a TMB of 26 Mut/Mb. No patient with a TMB >26 Mut/Mb experienced cancer progression on study, with the caveat that this cutoff would require validation in a larger cohort of patients. The events that drive TMB variability in MSI tumors are not fully elucidated. TMB may be determined by the type of mismatch repair (MMR) gene pathway alteration; MSH6-deficient tumors appear to have higher TMB than tumors that show MLH1/PMS2 loss, although it is not clear whether these processes result in distinct mutational signatures (2). In this study, neither MMR protein nor germline MMR status are examined. Frequent mutations in somatic MMR genes are identified, which may be passengers due to the hypermutated state. A significant proportion of patients also had POLE or POLD1 mutations, but no tumor met criteria for ultra-hypermutation (>100 Mut/Mb), and germline POLE mutations are associated with microsatellite-stable status, indicating that these mutations were somatic and probably did not act as clonal drivers. As most MSI gastric cancers are due to MLH1 promoter methylation and are nonfamilial, it is likely that the heterogeneity of TMB observed is not because of different MMR machinery defects but differences between MLH1-deficient tumors.
TMB has been extensively investigated as a predictor of response to ICI, but many questions remain. Technical factors can significantly affect within-assay reproducibility of TMB assessment, and harmonization between assays remains challenging. Adding further complexity, standard total TMB may not be sufficiently discriminatory for response to ICI; a recent meta-analysis of WES data across >1,000 ICI-treated patients suggested that only clonal TMB (mutation burden present in all cells) rather than total TMB or subclonal TMB is a significant predictor of sensitivity to ICI (3).
High TMB tumors produce more neoantigens, but neoantigen quality is also an important determinant of immunogenicity influencing ICI sensitivity. Although neoantigen prediction was not performed here, increased effector T-cell infiltrate in MSI gastric cancer could reflect neoantigen quality or lack of exclusionary immune mechanisms. Transcriptomic analysis revealed that responsive tumors were more likely to display upregulation of T-cell receptor (TCR) pathway activation and a more diverse TCR repertoire, consistent with findings in other tumors. Interestingly, MLH1-deficient MSI tumors may activate the immune system not just by the presentation of large numbers of neoantigens. Loss of the MutLα subunit in MLH1 deficiency impedes regulatory control of exonuclease 1 during DNA repair, leading to single-strand DNA formation, DNA breaks, chromosomal abnormalities, and the presence of cytosolic DNA (4). This cytosolic DNA itself evokes activation of the cGAS–STING pathway and an immune response (5). In vivo, reversal of MLH1 deficiency results in reduced efficacy of ICI even in high-mutation tumors, again highlighting that TMB alone may be insufficient to drive tumor response.
To examine the tumor microenvironment, Kwon and colleagues (1) went on to perform single-cell RNA sequencing of a limited number of paired tumor samples before and after two cycles of pembrolizumab; this on-treatment assessment of PD-1 response is a first in MSI gastric cancer. Responding patients had, as expected, higher numbers of infiltrating T cells but also natural killer cells, both of which expanded following anti–PD-1 blockade. Conversely, pembrolizumab-treated patients who did not benefit from ICI exhibited decreased numbers of cells from the innate and adaptive immune system on posttreatment biopsy specimens compared with baseline, in addition to a greater proportion of stromal cells and, specifically, cancer-associated fibroblasts (CAF). In other gastrointestinal tumors, WNT2 secretion by CAFs suppresses the differentiation and immune-stimulating activities of dendritic cells in vitro, providing a potential mechanistic explanation for this negative association and a window of opportunity for intervention using inhibitors of Wnt signaling (6). The importance of Wnt in ICI resistance is underscored by the presence of Wnt pathway gene alterations in nonresponders. Quality as well as quantity of T-cell infiltrate is highlighted, and responders had higher levels of stem-like and non–terminally differentiated T cells, which are more likely to spring into action when stimulated by anti–PD-1 therapy.
In a longitudinal analysis, serial biopsy specimens suggest immunoediting under evolutionary pressure from anti–PD-1 therapy. After pembrolizumab, responding patients exhibited a lower intratumoral TMB, suggesting elimination of high TMB subclones. This could represent acceleration of a process that may have occurred under natural selection in TMB low-MSI tumors. Similar reductions in TMB resulting from ICI treatment have been observed in melanoma, indicating that the overarching mechanism of clonal collapse is not exclusive to MSI tumors (7). MSI gastric cancers that respond to pembrolizumab have decreased posttreatment clonal diversity, which may have implications for salvage therapy following progression on anti–PD-1. In melanoma, a large retrospective cohort suggests that adding anti-CTLA4 antibodies, which expand the T-cell clonal repertoire, to anti–PD-1 therapy following progression on anti–PD-1 monotherapy might be a useful approach (8).
Liquid biopsy is an attractive tool for predictive biomarker development. Here, for the first time in MSI gastric cancer, the authors identified a higher percentage of peripheral blood PD-1–positive effector T cells in patients with durable clinical benefit from pembrolizumab, although the proportion of PD-1–positive T cells did not appear to be reliably associated with response. Blood TMB shows promise as a predictive biomarker in non–small cell lung cancer, and circulating tumor DNA assessment could also be of interest in the future in MSI gastric cancer, with similar caveats with respect to standardization as tissue-based analyses.
The overall efficacy associated with pembrolizumab for patients with MSI gastric cancer in this dataset is as expected (9). Importantly, although PD-L1 staining on tumor and immune cells (combined proportion score) can broadly discriminate between ICI-sensitive and ICI-insensitive non–MSI gastric cancer, PD-L1 assessment alone does not have predictive value in patients with MSI gastric cancer. Similarly, MSI and the presence of Epstein–Barr virus, another driver of immune response in gastric cancer, were mutually exclusive, suggesting that an exhaustive search for other biomarkers associated with ICI sensitivity in MSI gastric cancer may be unwarranted. A limitation of this research is its restriction to Asian patients with MSI gastric cancer; although no evidence has been presented from clinical trials to suggest that Asian patients with MSI gastric cancer have differential sensitivity to ICI compared with non-Asian patients, the prognostic effect of MSI gastric cancer in operable disease is more pronounced in Asian patients, suggesting some degree of interaction between the MSI genotype and host environment (10). Asian patients with gastric cancer appear to derive more survival benefit from ICI in general, although this may result from lower metastatic disease burden rather than any innate enhanced immune responsiveness in Asian patients. As such, the findings of Kwon and colleagues (1) are likely to broadly reflect those in the global MSI gastric cancer population.
ICI has shifted the paradigm for MSI tumors; long-term survival is now a realistic prospect for many patients with MSI gastric cancer. Sadly, for around a third of MSI patients, intrinsic resistance to anti–PD-1 therapy can dash these hopes. In this richly described data set, representing the largest prospective cohort of MSI gastric cancer to date, Kwon and colleagues (1) shine a light on interpatient heterogeneity in TMB and the immune and stromal environment as major factors responsible for ICI sensitivity in MSI gastric cancer. This work also takes the first steps toward understanding the landscape of acquired resistance in MSI gastric cancer treated with anti–PD-1 antibodies. For MSI gastric cancers that are innately immunologically evasive, the challenge for the future will be to develop clinically useful biomarkers to prospectively identify the patients who are likely to require up-front combination therapy to enhance ICI efficacy. These could include dual ICI blockade focusing on CTLA4, LAG3, and/or TIM3 inhibition; Wnt or JAK–STAT inhibition; or combinations targeting coexisting driver genomic alternations (ERBB2, FGFR2, CDH1), all of which are under current clinical investigation.
E. Fontana is an employee of Hospital Corporation of America (HCA) Healthcare UK. E.C. Smyth reports grants and personal fees from BMS; personal fees from AstraZeneca, Merck, Roche, Servier, Astellas, Zymeworks, and personal fees from Beigene outside the submitted work.
The authors thank Gary Doherty, MD, PhD, for his review of the text and helpful comments.