Continuous expression of TOX safeguards exhausted CD8 T cell epigenetic fate

CD8 T cell exhaustion is a major barrier limiting anti-tumor therapy. Though checkpoint blockade temporarily improves exhausted CD8 T cell (Tex) function, the underlying epigenetic landscape of Tex remains largely unchanged, preventing their durable “reinvigoration.” Whereas the transcription factor (TF) TOX has been identified as a critical initiator of Tex epigenetic programming, it remains unclear whether TOX plays an ongoing role in preserving Tex biology after cells commit to exhaustion. Here, we decoupled the role of TOX in the initiation versus maintenance of CD8 T cell exhaustion by temporally deleting TOX in established Tex. Induced TOX ablation in committed Tex resulted in apoptotic-driven loss of Tex, reduced expression of inhibitory receptors including PD-1, and a pronounced decrease in terminally differentiated subsets of Tex cells. Simultaneous gene expression and epigenetic profiling revealed a critical role for TOX in ensuring ongoing chromatin accessibility and transcriptional patterns for key Tex gene modules in committed Tex cells. Moreover, when exposed to effector-driving conditions, inducibly TOX-deleted established Tex acquired an altered chromatin landscape with increased accessibility at cytotoxic genes typically accessible in Teff cells, thus undergoing partial reprogramming into a more functional state. Together, these findings suggest that continuous TOX expression in established Tex acts as a durable epigenetic barrier to reinforce the Tex developmental fate by simultaneously maintaining Tex epigenetic commitment while restraining differentiation into Teff. Manipulation of TOX even after Tex establishment could therefore provide a therapeutic opportunity to rewire Tex biology in settings of chronic infection or cancer. The secondary goal of this dissertation was to develop a novel Tex fate-mapping mouse model driven to track the fate of developing Tex and manipulate Tex in a lineage-restricted fashion. Given the selectively high expression of TOX in Tex, compared to other peripheral non-Tex CD8 lineages, we used the Tox locus to drive this model (termed ToxTREx), which was engineered with a T2A-hmKO2-P2A-CreERT2 cassette knocked into the Tox locus after the last exon. We confirmed intact TOX function, hmKO2 reporter detection and TOX-driven Cre recombinase activity in this model and identified further optimization that will be necessary to improve Cre efficiency and specificity of this model for the Tex lineage. Nevertheless, the ToxTREx model could enable insightful studies that address existing and emerging questions in Tex ontogeny, differentiation and function.

Methods
Cells from inducible-Cre (Rosa26CreERT2/+Toxfl/fl P14) mice where TOX was temporally deleted from mature populations of LCMV-specific T exhausted cells after establishment of chronic LCMV infection 5 days post infection were subjected to scRNA and scATACseq coassay,naive cells and WT cells were used as controls.

Analysis pipeline developed by Josephine Giles and vignettes published by Satija and Stuart labs.Transcript count and peak accessibility matrices deposited in GSE255042,GSE255043.

Seurat/Signac was used to process the scRNA and scATACseq coassay data

The processed Seurat/Signac object above was subsequently used for downstream RNA and ATAC analyses as described below:

DEGs between TOX WT and iKO cells within each subset were identified using FindMarkers (Seurat, Signac), with a log2-fold-change threshold of 0, using the SCT assay. DACRs were identified using FindMarkers using the "LR" test, with a log2-fold-change threshold of 0.1, a min.pct of 0.05, and included the number of counts as a latent variable. DEGs and DACRs were filtered with a false discovery rate (FDR) of less than 0.05, using the Benjamini–Hochberg method to adjust P-values. Gene ontology (GO) analysis of DEGs used Metascape (https://metascape.org/) with all expressed genes as the background gene list. AddModuleScore (Seurat, Signac) was used to calculate per-cell gene set enrichment scores and peak set enrichment scores. Gene set enrichment analysis (GSEA) was performed and visualized using clusterProfiler (4.8.1) and enrichplot (1.20.0). Tex-prog and Tex-term transcriptional signatures were generated from the Giles et al dataset (NI 2022), using FindMarkers to identify DEGs between the "Exh-Prog" cluster and the "Exh-Term" and "Exh-TermGzma" clusters, filtered with a FDR of less than 0.05 and log2 fold-change threshold of 0.25. Peak sets of subset transition ACRs were identified using FindMarkers to perform pairwise comparisons between TOX WT Tex-prog and Tex-int, Tex-klr, and Tex-term respectively (using the "LR" test, with a min.pct of 0.05, included the number of counts as a latent variable and cut-off at top 5000 ACRs by log2 fold-change threshold). Peak sets of naïve, Teff, Tmem and Tex-specific ACRs were generated as previously described (Khan et al, Nature, 2019). Bedtools intersect was used to identify overlapping peaks between different datasets. Genome coverage tracks were generated in Signac using CoveragePlot, PeakPlot, and AnnotationPlot. Genomic locations of peaks were determined using ClosestFeature (Signac) and nearestTSS (edgeR), with promoter-TSS defined as -1kB to +100bp. The average local chromatin accessibility at each gene was determined using GeneActivity (Signac). TF motif enrichment was calculated using the JASPAR2020 function getMatrixSet (species 9606) and Signac functions CreateMotifMatrix, CreateMotifObject, and FindMotifs. Open peaks in the clusters of interest were used as background, using AccessiblePeaks and MatchRegionStats (Signac).