Volcano plots a show log2 fold change on the x-axis against negative log10 p-value on the y-axis for IRF8, IRF4, and IRF2 knockout versus control comparisons. IRF8 knockout shows upregulated genes including Isg15, Gzmf, Ifng, and Prf1 and downregulated genes including Tox, Themis, and Bcl2. IRF4 knockout shows upregulated Crtam, Xcl1, and Ccl1 and downregulated Mki67 and Stat3. IRF2 knockout shows upregulated Prf1, Lag3, Ifng, and Gzmk and downregulated Bcl6 and Arid1a. Venn diagrams b show overlaps of upregulated and downregulated genes across IRF2, IRF4, and IRF8 knockouts, where 18 genes are commonly upregulated and 75 genes are commonly downregulated across all three, with IRF8 knockout contributing the largest unique sets of 110 upregulated and 237 downregulated genes. Principal component analysis plots c and d show separation of IRF8 versus IRF4 and IRF8 versus IRF2 control and knockout samples, with PC1 explaining 53.74 and 33.31 percent and PC2 explaining 15.66 and 26.77 percent of variation, respectively. Horizontal bar graphs e and f show top 5 Hallmark pathways per regulon on negative log10 adjusted p-value x-axes, where graph e shows TNF alpha signaling via NF kappa B and estrogen response late as top pathways for Common All Up regulon reaching beyond 4, and graph f shows TNF alpha signaling via NF kappa B and inflammatory response as top pathways across IRF2 and IRF4 and IRF8 specific regulons with values up to 6. Dot plot g shows log2 fold change in IRF knockout for shared direct targets identified from both single cell RNA sequencing differentially expressed genes and Cut and Run data, where Ccl3 shows the largest positive fold change above 1 and Tox and Runx3 show negative fold changes. Genome browser tracks h show scATACseq TIL accessibility and Cut and Run signals for IRF2, IRF4, IRF8, and IgG at the Tox locus, with zoomed panels highlighting three regions of differential accessibility and transcription factor binding.
Multiplexed scRNA-seq reveals convergence of IRF2, IRF4, and IRF8 on TOX-driven exhaustion in CD8 + TILs. (a) Volcano plots from scRNA-seq showing significant DEGs (LogFC>0.3, adjusted P value <0.05) in IRF8 KO (top), IRF4 KO (middle), and IRF2 KO (bottom) P14+ CD8 T cells compared with their co-transferred control-transduced counterpart (paired analysis using multiplexed hashtagging antibodies) from experiment in Fig. 7 f. (b) Venn diagrams showing the numbers of DEGs that are upregulated (left) and downregulated (right) in each IRF KO compared with control (threshold LogFC>0.3, adjusted P value <0.05). (c and d) PCA of scRNA-seq from IRF8 KO versus IRF4 KO (c) and IRF8 KO versus IRF2 KO (d) with their respective co-transferred control-transduced CD8 TILs issued from co-transfer experiments in Fig. 7 f. (e) Overrepresentation of pathways from MSigDB Hallmark gene sets across up- and downregulated genes common to IRF2, IRF4, and IRF8 KO. (f) Overrepresentation of pathways from MSigDB Hallmark gene sets across up- and downregulated genes common to IRF2 and IRF8 KO and to IRF4 and IRF8 KO. (g) Lollipop plot showing shared direct targets of IRF2, IRF4, and IRF8 based on integration of scRNA-seq data from above and CUT&RUN-seq data generated on in vitro chronically TCR-stimulated CD8+ T cells as described in the Materials and methods. Each point represents a gene, with the x-axis indicating log2 fold change (IRF8 KO versus control) and point size representing statistical significance (–log10 FDR). Genes are ordered by effect size. (h) Genome browser tracks showing CUT&RUN signal for IRF2, IRF4, IRF8, and IgG control, together with scATAC-seq accessibility in TILs at the Tox locus. Zoom-in views of selected regions illustrating overlapping IRF-binding and accessible chromatin, supporting direct transcriptional regulation.