CyTOF and scRNAseq experiments generate large amounts of information that are only interpretable using bioinformatic tools. We used a dimensionality reduction algorithm to create single-cell maps of tumor-infiltrating CD8+ T cells that revealed a very interesting cluster of cells expressing the integrin E or CD103. CD103 was a molecule that we had already been studying as part of Genentech’s Etrolizumab program (an antibody targeting the Beta 7 integrin subunit that pairs with E); however, we did not know that it would turn out to be a molecule of interest in the context of tumor immunotherapy. CD103 broadly defines intraepithelial lymphocytes, including tissue-resident memory T cells or TRM. We found CD8+ tumor TRM cells to be enriched for activation and proliferation markers while also expressing high levels of immune checkpoints such as PD-1 and CTLA4, which are typically associated with a dysfunctional or “exhausted” state. This was confusing to us as we could not resolve how cells that are exhausted express high levels of Ki-67, a classic biomarker of proliferation. It seemed like we were capturing cells locked in a death match with the tumor, but they obviously were not winning, or going to win, without intervention.
Based on these observations, we turned to bulk RNAseq data sets from atezolizumab (anti-PD-L1) clinical trials to ask whether levels of CD103 gene expression correlated with clinical response. In lung and bladder cancer patients treated with atezolizumab, CD103 performed as well as, and sometimes better, than CD8 in predicting clinical response to anti-PD-L1 therapy. Furthermore, CD103 was highly enriched in inflamed tumors characterized by deep intratumoral CD8+ T-cell infiltration relative to excluded tumors, in which CD8+ T cells largely sit in the periphery of the tumor.
Nearly three years ago, many publications started to come out that focused on tissue-resident memory T-cell biology in the tumor microenvironment, suggesting that these cells or at least a more functional precursor of them may be the answer to the question of which T cells are the “executioners” of cancer immunotherapy. Furthermore, a subset of tumor TRM co-expressing CD103 and CD39 has been found to have the most activated T cell phenotype in tumors. Thus, CD39 became “the new CD103”. Importantly, a number of recent studies demonstrate that these CD39+ TRM cells are reactive to tumor antigens in coculture experiments. Tumor reactivity is the essential feature of any T cell capable of killing tumors, thus phenotyping in the absence of confidence in tumor reactivity is fairly meaningless. In light of these studies, we investigated tumors by single-cell and T-cell receptor sequencing data (scRNAseq, scTCRseq) to probe the relationship between cell identity and clonal expansion within the tumor. We observed that in tumors, the CD8+ T cells with the largest clone sizes (i.e., the most proliferated in response to antigen) were indeed characterized by a TRM phenotype that had hallmarks of proliferation, activation, exhaustion, and cytotoxic functionality. Thus, in one simple model, the T cells that are capable of reacting against and rejecting tumor cells following immunotherapy are in the tumor, to begin with, and look like TRM cells.
Although this model is straightforward and compelling, it may not be correct. Many prominent investigators are skeptical of this simple model and believe that exhausted T cells within solid tumors are incapable of being reinvigorated. In this alternative model, the T cells that kill the tumor come from somewhere else (probably the tumor-draining lymph node) following immunotherapy. A recent study of checkpoint inhibition in melanoma patients suggests that a high tumor content of cells with an exhausted phenotype actually predicts a lack of response to immunotherapy and that enumeration of T cells expressing the transcription factor TCF7, marking stem cell memory T (TSCM) cells, is the best T-cell biomarker of clinical response. These studies, however, do not demonstrate that TCF7+ cells are tumor-reactive. Thus, it remains unclear precisely which T cells reject tumors, but our study points to intratumoral TRM cells being at least a key component of the process, perhaps by propagating release of tumor antigens that subsequently draws in new T cell clones.
Written by: William O’Gorman, Romain Banchereau, Avantika S. Chitre, & Sanjeev Mariathasan, Genentech Inc.
Read the Abstract