Fig. 5: NRF2-mediated PTGIR expression drives CD8⁺ T cell exhaustion.

a, Correlation between NRF2 activity and Ptgir expression in CD8⁺ T cell subsets. Relative Ptgir mRNA expression (log₂(fold-change)) versus NRF2 pathway enrichment scores (NFE2L2.V2) in CD8 T cells from L. monocytogenes infection (T_n, T_eff and T_mem cells) or CD8⁺ TILs isolated from early (<21 d) versus late (>21 d) stage liver tumors (data from Philip et al.¹⁹). Expression values were normalized to T_eff cells. Linear regression indicates goodness of fit (R² = 0.92). b, Relative PTGIR expression in WT CD8⁺ T cells transduced with either an EV control or PTGIR-expressing plasmid (+PTGIR) and Keap1^−/− CD8⁺ T cells (transduced with EV). PTGIR peptide (GFTQAIAPDSR) levels were determined by mass spectrometry and normalized to ACTN4 levels (n = 3 technical replicates). c, B16-OVA melanoma tumor growth in mice that received either no T cells (HBSS) or WT control (WT + EV), PTGIR-expressing (WT + PTGIR) or Keap1^−/− (Keap1^−/− + EV) OT-I T cells at 7 d.p.t.i. (n = 12 mice per group). d, Kaplan–Meier (KM) survival curves for time to tumor endpoint (>1,500 mm³) for B16-OVA-bearing mice from c. e, Relative expression of Hmox1 and Ptgir mRNA in CRISPR–Cas9-edited WT or Keap1^−/− OT-I T cells. Control WT OT-I cells were edited with an sgScr and transduced with an EV construct (WT sgScr + EV). Keap1^−/− OT-I cells were engineered to express control constructs (Keap1^−/− sgScr + EV) or to delete NRF2 without (Keap1^−/− sgNfe2l2 + EV) or with ectopic PTGIR expression (Keap1^−/− sgNfe2l2 + PTGIR). The mRNA levels were determined by qPCR and normalized to Actnb levels (n = 3 technical replicates). f, B16-OVA melanoma tumor growth in mice that received (on 6 d.p.t.i.) no T cells (HBSS) or adoptive transfer of OT-I cells of the following genotypes: WT control (sgScr + EV), Keap1^−/− (sgScr + EV), NRF2-deficient Keap1^−/− (sgNfe2l2 + EV) or PTGIR-expressing and NRF2-deficient Keap1^−/− (sgNfe2l2 + PTGIR) (n = 12–20 mice per group). g, KM survival curves for time to tumor endpoint (>1,500 mm³) for B16-OVA-bearing mice from f. The data are indicated as mean ± s.d. in b and e for technical replicates from a representative experiment. The data in c and f are indicated as mean ± s.e.m. Statistical significance was calculated via one-way ANOVA with Tukey’s test for b and e, two-way ANOVA for c and f and log(rank test) for d and g. NS, P ≥ 0.05; ^*P = 0.05; ^**P = 0.01; ^***P = 0.001; ^****P = 0.0001.