Immunogenomic, single-cell and spatial dissection of CD8⁺T cell exhaustion reveals critical determinants of cancer immunotherapy

Abstract

Tumoural-CD8+T cells exhibit exhausted or dysfunctional states. Contrary to immunotherapy-responsive exhausted-CD8+T cells, the clinical features of dysfunctional-CD8+T cells are disputed. Hence, we conducted large-scale multi-omics and multi-dimensional mapping of CD8+T cell-states across multiple cancer patient-cohorts. This identified tumour-specific continuum of CD8+T cell-states across 6 human cancers, partly imprinted by organ-specific immuno-modulatory niches. Herein, melanoma and glioblastoma enriched prototypical exhausted (CD8+TEXT) and severely-dysfunctional (CD8+TSDF) states, respectively. Contrary to CD8+TEXT, CD8+TSDF displayed transcriptomic and epigenetic effector/cytolytic dysfunctions, and dysregulated effector/memory single-cell trajectories, culminating into maladaptive prodeath stress and cell-cycle defects. Suboptimal antigen-priming underscored CD8+TSDF, which was distinct from immune-checkpoints “rich” CD8+TEXT, reflecting chronic antigen-stimulation. Continuum variation also existed on tumour spatial-level, with convergent (CD8+TEXT-supportive vascular regions) and divergent features (dysfunctional CD4+T::CD8+TSDFcell-to-cell interactions) between melanoma and glioblastoma. Globally, IFNγ-IL2 disparities, paucity of intra-tumoural CD4+/CD8+T cells, and myeloid TGFβ/wound healing responses, distinguished CD8+TSDF-landscape. Within immuno-oncology clinical-trials, anti-PD1 immunotherapy failed to “reinvigorate” CD8+TSDF-landscape, and instead facilitated effector-dysfunction and TGFβ/wound healing. However, cellular immunotherapies (dendritic cell-vaccines, adoptive T-cell therapy) ameliorated assorted CD8+TSDF-landscape disparities, highlighting a roadmap for anti-glioblastoma multimodal-immunotherapy. Collectively, our study comprehensively expands clinical-knowledge on CD8+T cell-exhaustion and suggests that tumour-specific, pre-existing CD8+TEXT/TSDF-states, determine immunotherapy-responses.

Trial watch: Dendritic cell (DC)-based immunotherapy for cancer

“… 273–282 These trends have highlighted the values of biomarker-driven patient pre-selection for the success of anticancer immunotherapy. 283–286 However, unfortunately, there is a severe lack of robust patient pre-selection biomarkers to guide the application of DC-based vaccines, thereby setting them up for failure. This situation needs to rapidly change.…”

The Interface of Tumour-Associated Macrophages with Dying Cancer Cells in Immuno-Oncology

“…Herein, the extreme TAMs heterogeneity, plasticity, human vs. mouse variations and context-dependent activity together serve as major bottlenecks for the therapeutic translation of TAM-targeting immunotherapies. Multi-omics dissections of the human TAM compartment and analyses of conserved mechanisms and ontologies between human and mouse TAMs are together expected to provide crucial insights into the next generations of TAM-targeting immunotherapies [ 173 , 174 ]. Nevertheless, it is necessary for future TAMs-targeting strategies to sufficiently account for the tissue contexture of tumours as well as metastasis.…”

Next-Generation Pathology Using Multiplexed Immunohistochemistry: Mapping Tissue Architecture at Single-Cell Level

“…The distribution of the various cells across a tissue section was determined by computationally breaking up the tissue in smaller "tiles" in which the relative distributions of the identified cell types were determined. The same approach is directly applicable to define tumoral, peritumoral, perivascular and non-tumoral areas (112), and, while these insights are key to define the macrostructure of a tissue, such analysis still remains on the level of one cell type at the time. Once macro-level areas are defined, it becomes key to understand local cellular distributions, an analysis type that is done by performing neighborhood or nearest neighbor analyses.…”