The presence of intratumoral CD8+ T cells correlates with better outcomes in many cancers.1 Nonetheless, tumor-infiltrating CD8+ T cells are highly diverse with respect to their differentiation state and functionality, as well as in terms of their ability to recognize tumor-derived antigens. In fact, a considerable portion of tumor-infiltrating CD8+ T cells do not recognize tumor-derived antigens, including memory T cells (TM) that have been raised in response to infections.2, 3 While tumor reactivity is usually enriched within T-cell populations that express checkpoint inhibitory receptors, such as PD-1, TIM3 and LAG3, a detailed account of the relationship between antigen specificity, the nature of tumor-derived antigens and the acquisition of various states of T-cell differentiation in human cancer has been missing.4 In what represents a technical tour de force, a recently published study by Oliveira et al. has begun to fill this important gap in knowledge by functionally determining tumor reactivity of a large number of T-cell receptors (TCRs) from melanoma-infiltrating CD8+ T cells and linking these to phenotypic and transcriptional hallmarks of exhaustion determined by high-dimensional single-cell profiling.5

In a first step, the authors subjected about 30 000 CD8+ T cells isolated from tumors of four advanced stage melanoma patients to single-cell RNA and TCR sequencing in conjunction with an analysis of surface proteins commonly used to partition major T-cell subsets. Cross-referencing to published data sets revealed 13 distinct clusters across five major populations, which were designated memory (TM) and effector memory cells (TEM), acutely activated cells (TAct), as well as terminally exhausted (TE) and precursor exhausted cells (TPE). Of note, the most abundant TCR clonotypes were primarily associated with exhaustion-related clusters populated by TE, TPE and TAct cells, whereas less expanded TCRs were more often found in memory clusters encompassing TE and TEM cells. This apparent bias provided strong evidence that antigen specificity not only linked to T-cell expansion, but also imparted a bias in transcriptional profiles and presumably, functional or dysfunctional states.

To more precisely understand the link between antigen specificity and various differentiation states of melanoma-infiltrating CD8+ T cells, the authors took on the Herculean task of cloning 172 of the most abundant TCRs from clusters associated with exhaustion or memory, and expressed these in T cells from healthy donors via lentiviral transduction. Using upregulation of the activation marker CD137 as a surrogate for TCR activation, the authors then tested TCR reactivity against autologous patient-derived melanoma cells and Epstein–Barr virus (EBV)-immortalized cell lines. Remarkably, 83% of analyzed TCRs that predominantly mapped to exhaustion-associated clusters were confirmed to be tumor reactive. By contrast, only 10% of TCRs from primary clusters associated with memory phenotypes were tumor reactive. Finally, 22% of the TCR clonotypes that failed to recognize tumor cells were specific for EBV, and these almost exclusively mapped to memory rather than exhaustion-associated clusters (Figure 1). These results were confirmed for tumor-specific T-cell clones from blood and validated separately for TCRs reconstructed from previously published data sets. Thus, tumor-reactive CD8+ T cells displayed a strong bias toward exhaustion-associated transcriptional profiles, although this link was not absolute, as most tumor-specific TCRs were also expressed by a minor portion of cells with memory-associated transcriptomes. Interestingly, when the authors determined the antigenic specificity of 180 TCRs using large peptide pools, this bias was equally evident for TCRs that recognized personal neoantigens or public melanoma-associated antigens.

Tumor-reactive CD8+ T cells display a strong bias toward phenotypes and transcriptional profiles associated with dysfunction. CD8+ T cells recognizing different classes of tumor-derived antigens display a strong yet not absolute bias toward dysfunction-related phenotypes. By contrast, nontumor-reactive bystander cells, some of which are specific for widely prevalent viruses, almost exclusively display hallmarks of functional memory cells. Genes shown next to dysfunction- and memory-related clusters represent examples of the larger pool of differentially regulated genes between tumor-reactive and virus-specific T cells. TE, terminally exhausted T cell; TPE, precursor exhausted T cell; TAct, acutely activated T cell; TCR, T-cell receptor; TEM, effector memory T cell; TM, memory T cell.

Direct comparison of transcriptomes from CD8+ T cells with functionally established reactivity to either tumor-derived or viral antigens revealed differential expression of 98 transcripts, including many genes known to be associated with activation/exhaustion (such as TOX, PDCD1, CTLA4, HAVRC2, ENTPD1 and CXCL13) or memory states (such as TCF7, IL7R and CCR7), respectively.4 Furthermore, reclustering of tumor-reactive TCRs provided a more granular view on different subsets of tumor-reactive T cells. Two-thirds were primarily identified as TE cells and marked by expression of transcripts encoding perforin (PRF1) and granzyme B (GZMB), as well as expression of PD1, TIM3, LAG3 and CD39 proteins, although most TCRs also contributed to one or more distinct clusters. Moreover, about one-tenth of the tumor-reactive clones predominantly displayed characteristics of TAct cells with elevated expression of NR4A1, IFNG and heat shock proteins, whereas only few clones were categorized as proliferating (elevated MKI69 expression), memory phenotype (elevated expression of TCF7, IL7R and genes encoding effector cytokines) or progenitor exhausted T cells (elevated TCF7, CCR7 and HLA-DR expression).

Overall, the study by Oliveira et al. provides the most comprehensive analysis of TCR specificities and their related differentiation states in melanoma to date. Together with data from preclinical models,6 their results strongly support the notion that continuous recognition of tumor-derived antigens is one of the major drivers of T-cell dysfunction in cancer. Importantly, the authors also define a small population of TPE CD8+ T cells. Although developmental trajectories were not analyzed in their study, such cells have recently emerged as an important cellular source replenishing intratumoral CD8+ T cells with varying degrees of functionality and are now considered key targets of checkpoint blockade immunotherapy in cancer.7 Separately, tumor-associated CD8+ T cells with a CD69+CD103+ resident memory phenotype can display strong proliferative potential and antitumor activity and have been associated with better outcomes and response to checkpoint blockade in several cancers, including melanoma.8-10 Interestingly, despite their apparent functionality, these cells often express dysfunction-associated markers, such as PD1 and TIM3, while they lack expression of one of the TPE-defining transcriptional regulators, TCF1 (encoded by TCF7).9, 10 Because expression of CD103 was not analyzed at the protein level, future studies will have to elucidate how resident memory-like T cells relate to the clusters identified in the current study. Regardless, high-dimensional single-cell analyses of the type presented by Oliveira et al., potentially in conjunction with serial analyses of tumor biopsies and spatial information on microanatomical distribution of T-cell subsets, hold great promise for identifying predictive characteristics of patients’ responses to cancer and immunotherapy.

ACKNOWLEDGMENTS

KH is a Rhian and Paul Brazis Fellow in Translational Melanoma Immunology administered by the Peter MacCallum Cancer Foundation. TG is a National Health and Medical Research Council Leadership Fellow.

CONFLICT OF INTEREST

The authors declare no conflicts of interest.

AUTHOR CONTRIBUTION

Katharina Hochheiser: Conceptualization; Writing-original draft; Writing-review & editing. David E Gyorki: Writing-review & editing. Thomas Gebhardt: Conceptualization; Writing-original draft; Writing-review & editing.