PGE2 inhibits TIL expansion by disrupting IL-2 signalling and mitochondrial function

Abstract

Expansion of antigen-experienced CD8⁺ T cells is critical for the success of tumour-infiltrating lymphocyte (TIL)-adoptive cell therapy (ACT) in patients with cancer¹. Interleukin-2 (IL-2) acts as a key regulator of CD8⁺ cytotoxic T lymphocyte functions by promoting expansion and cytotoxic capability^2,3. Therefore, it is essential to comprehend mechanistic barriers to IL-2 sensing in the tumour microenvironment to implement strategies to reinvigorate IL-2 responsiveness and T cell antitumour responses. Here we report that prostaglandin E2 (PGE₂), a known negative regulator of immune response in the tumour microenvironment^4,5, is present at high concentrations in tumour tissue from patients and leads to impaired IL-2 sensing in human CD8⁺ TILs via the PGE₂ receptors EP2 and EP4. Mechanistically, PGE₂ inhibits IL-2 sensing in TILs by downregulating the IL-2Rγ_c chain, resulting in defective assembly of IL-2Rβ-IL2Rγ_c membrane dimers. This results in impaired IL-2-mTOR adaptation and PGC1α transcriptional repression, causing oxidative stress and ferroptotic cell death in tumour-reactive TILs. Inhibition of PGE₂ signalling to EP2 and EP4 during TIL expansion for ACT resulted in increased IL-2 sensing, leading to enhanced proliferation of tumour-reactive TILs and enhanced tumour control once the cells were transferred in vivo. Our study reveals fundamental features that underlie impairment of human TILs mediated by PGE₂ in the tumour microenvironment. These findings have therapeutic implications for cancer immunotherapy and cell therapy, and enable the development of targeted strategies to enhance IL-2 sensing and amplify the IL-2 response in TILs, thereby promoting the expansion of effector T cells with enhanced therapeutic potential.

Fig. 1. The PGE₂–EP2/EP4 axis is associated with decreased IL-2-mediated TIL expansion.

a, Representation of the translational research pipeline of a phase I melanoma TIL-ACT trial. Figure created with BioRender.com. b, Correlation between IL-2 signalling score from pseudobulked patient CD8⁺ TILs and total numbers of TILs at pre-REP (on day 11) from patients with melanoma enrolled in the TIL-ACT trial (n = 13). c, Representation of Reactome pathways that are anti-correlated with IL-2 signalling score from pseudobulked patient CD8⁺ TILs (no correction for multiple testing applied) (n = 13). PCC, Pearson’s correlation score. d, Correlation between IL-2 signalling score and eicosanoid ligand-binding receptor score from pseudobulked patient CD8⁺ TILs (n = 13). e, Violin plot of PGE₂ signature score in tumour-reactive versus non-tumour-reactive CD8⁺ TILs subsets that expanded or did not expand in the cell therapy product. The number of cells is shown at the top of the graph. Box plots display smallest and largest values in the dataset, box hinges represent first and third quartiles with the centre as median and whiskers extend to 1.5× the interquartile range from the first and third quartiles. EM-like, effector memory-like; Pex, precursor exhausted; Tex, terminal exhausted; ISG, IFN-stimulated gene. f, Correlation per patient between baseline PGE₂ levels in the supernatant of expanding TILs from tumour fragments and total numbers of pre-REP TILs in a phase I solid tumour TIL-ACT trial (n = 18). g, Correlation per patient between PGE₂ levels in the supernatant of expanding TILs from tumour fragments of breast (n = 20) and melanoma (n = 12) and total numbers of pre-REP TILs. One-way ANOVA with Tukey’s multiple comparisons test (e); two-Sided Spearman’s correlation (b,d,f,g); or Pearson’s correlation (c). Biological replicates represent individual patients, with exact numbers listed in each panel. NS, not significant (P ≥ 0.05).

Fig. 2. PGE₂–EP2/EP4 signalling restricts IL-2 signalling in TILs by deregulating the IL-2R complex.

a, Relative TIL count following treatment with PGE₂ at various doses for 5 days (n = 5). FC, fold change. b, Relative TIL count following treatment for 72 h with PGE₂ at different doses of IL-2 (n = 6). Ctrl, control. c, Relative CD8⁺ TIL count following treatment for 72 h with PGE₂, EP2/EP4 antagonists (EP2/4), or combined treatment (n = 4). d, Surface expression of IL-2Rα, IL-2Rβ and IL-2Rγ_c in CD8⁺ TILs treated with PGE₂ and EP2/EP4 antagonists for 72 h (n = 4). MFI, mean fluorescence intensity. e, Relative IL-2Rγ_c expression in CD8⁺ TILs treated with PGE₂ for 2 h, or treated with PGE₂ for 2 h and then re-exposed to medium without PGE₂ for 70 h (n = 3). f, IL-2Rγc expression in unstimulated T cells treated with PGE₂, the calcium chelator BAPTA, the cAMP antagonist Rp-8-CPT, ionomycin or combined treatment for 2 h (n = 4). g,h, Flow cytometry image of IL-2Rα, IL-2Rβ and IL-2Rγ_c expression (g; representative of four biological replicates) and colocalization of IL-2Rβ and IL-2Rγ_c in CD8⁺ TILs (h) upon 24 h treatment with PGE₂, assessed by ImageStream (n = 4). A 7 μm scale bar is shown at bottom left of each row. i, FRET analysis of IL-2Rβγ_c in TILs treated with PGE₂ for 24 h (n = 6). j, Relative mRNA expression of indicated genes in unstimulated T cells, RA T cells and TILs (n = 4). k, IL-2 signalling in RA T cells treated with PGE₂ for 48 h and subsequently stimulated with IL-2 or IL-2v for 15 min (representative of 3 biological replicates). l, pS6 levels in CD8⁺ TILs treated for 2 h with PGE₂ and subsequently stimulated for 30 min with IL-2, anti-CD3 or anti-CD3 plus anti-CD28 (anti-CD3/CD28) (n = 3). Data are mean ± s.d. Paired two-tailed t-test (h,i); one-way ANOVA with Dunnett’s post hoc test for multiple comparisons (a–f,j,l). Independent biological samples were used; exact numbers of biological replicates are listed in each panel. pJAK3, pS6, pAKT, STAT1, STAT3, pSTAT3, JAK1, pJAK1 and STAT5 were run on separate gels for blotting.

Fig. 3. PGE₂ rewires metabolism of TILs.

a, Heat map of normalized expression, z-scored by row, of the top differentially expressed Hallmark signatures (P < 0.05) between unstimulated and RA CD8⁺ T cells treated with PGE₂ for 24 h or untreated. P values (left column) indicate significance of differences between control and PGE₂-treated RA CD8⁺ T cells in three patients (n = 3). P values (false discovery rate, Bonferroni-corrected) were calculated by applying GSEA on the average expression per group. b,c, Violin plot representation of fold changes in reaction rates of the inferred metabolic states for RA CD8⁺ T cells (b; n = 3) and CD8⁺ TILs (c; n = 1) upon 24 h exposure to PGE₂. ETC, electron transport chain; PIP, phosphoinositide. d, Heat map representation of polar metabolites in CD8⁺ TILs upon PGE₂ treatment (n = 4). P values (left column) were calculated using two-tailed paired t-test for the peak areas of the corresponding metabolites. e, ATP quantification by ELISA in CD8⁺ TILs treated for 24 h with PGE₂, EP2/EP4 antagonists or combined treatment (n = 5). Data are mean ± s.d. One-way ANOVA with Dunnett’s post hoc test for multiple comparisons (e). Independent biological samples were used; exact numbers of biological replicates are listed in each panel.

Fig. 4. PGE₂ increases oxidative stress in TILs by impairing the IL-2–mTOR–PGC1α axis, leading to ferroptosis.

a, Representative electron microscopy images of RA T cells with and without 24 h PGE₂ treatment (n = 3). Scale bar, 500 nm. b,c, Representative cristae number and length per mitochondrion in unstimulated and RA T cells upon 24 h PGE₂ treatment (n = 3). d, Fold change (relative to control) of mitochondrial potential (indicated by tetramethylrhodamine methyl ester (TMRM)) in CD39⁻ and CD39⁺ CD8⁺ TILs after 24 h PGE₂ treatment (n = 4). e, Fold change (relative to control) of oxidized/reduced glutathione quantified by ELISA in CD8⁺ TILs upon 24 h PGE₂ treatment with or without EP2/EP4 antagonists (n = 4). f, Relative PGC1A and PGC1B mRNA expression in TILs after 48 h PGE₂ (n = 6). g, Relative PGC1A mRNA expression in RA T cells treated with 12 h PGE₂ with or without an mTOR activator (MHY1485) and subsequently stimulated with IL-2 for 15 min (n = 6). h, Cell count of RA OT-1 mouse T cells overexpressing PGC1α (PGC1α OE) upon 72 h PGE₂ exposure (n = 4). i,j, Electron microscopy images (i; representative of three biological replicates) and mean number of lipid droplets per cell in unstimulated T cells and RA T cells upon 24 h PGE₂ exposure (n = 3). Scale bar, 3 μm. k, Lipid peroxidation in CD8⁺ TILs after 48 h PGE₂ exposure (n = 3). l, Fold change in GPX4 mRNA expression (exp.) in CD8⁺ TILs after 48 h PGE₂ (n = 5). m, Frequency of viable TILs after 72 h treatment with PGE₂ and indicated concentrations of Fst1 (ferroptosis inhibitor), MCC905 (pyroptosis inhibitor), z-vad-fmk (apoptosis inhibitor) or necrostatin 1S (necroptosis inhibitor) (n = 3). Two-way ANOVA with Dunnett’s post test; *P < 0.05, **P < 0.01, ***P < 0.001. n, MDA quantification in CD8⁺ TILs upon 48 h PGE₂ exposure with or without EP2/EP4 antagonists, using ELISA (n = 5). o, Relative cell count of TILs upon 72 h PGE₂ exposure with or without NAC or vitamin E (VITE) (n = 6). Data are mean ± s.d. Paired two-tailed t-test (k,l); one-way ANOVA with Dunnett’s post hoc test for multiple comparisons (b–h,j,n,o). Independent biological samples were used; exact numbers of biological replicates are listed in each panel.

Fig. 5. Blockade of the PGE₂–EP2/EP4 axis increases TIL expansion, fitness and tumour reactivity.

a, Relative number of pre-REP TILs from different solid tumours treated with IL-2, IL-2 plus PGE₂, IL-2 plus EP2/EP4 antagonists or IL-2 plus ketorolac (COXi) during the first 48 h of culture (n = 13). b, Relative number of REP TILs in the COXi group from different solid tumours (n = 8). c, Uniform manifold approximation and projection (UMAP) projection of 34-parameter CyTOF data showing sub-clustering of control and COXi REP TILs from three patients with melanoma (n = 3). d, Violin plots showing expression of indicated proteins detected by CyTOF in five different clusters in control and COXi REP TILs. e, Fold change frequency of CD8⁺ REP TILs per cluster in control and COXi groups (n = 3). f,g, Frequency of TCF1⁺CD39⁻ (f) and TOX⁺TCF1⁻ (g) TILs as a percentage of CD8⁺ TILs in control and COXi REP TILs (n = 7). h, Fold change of mitochondrial DNA copy number in REP COXi TILs (n = 5). i,j, Relative cell count of tumour-reactive CD8⁺ (n = 3) (i) and CD4⁺ (j) TILs at the end of REP step (n = 3). Tumour reactivity was assessed via 41BB surface expression upon co-culture with autologous tumours. k, Fold change in frequency of TNF⁺, IFNγ⁺ or TNF⁺IFNγ⁺ CD8⁺ REP TILs as a percentage of CD8⁺ TILs upon overnight co-culture with autologous tumour cells (n = 3). l,m, Tumour growth kinetics (l) and survival curve (m) for mice treated with COXi and control REP TILs in a Winn assay transfer (n = 6 mice per group). n,o, Frequency as a percentage of total cells (n) and representative images (o) of intratumoral human CD45⁺ cells upon treatment of COXi or control REP TILs (n = 3 or 4 mice per group). Scale bars, 100 μm. Data are mean ± s.d. One-way ANOVA (a,e,k,n) or two-way ANOVA with Dunnett’s post hoc test for multiple comparisons (l,m); Cohen’s D test (d) (Supplementary Table 5); or paired (f–h) or unpaired (b,i,j) two-tailed t-test. Independent biological samples were used; exact numbers of biological replicates are listed in each panel.

Extended Data Fig. 1. PGE₂-EP2/EP4 axis is associated with decreased IL-2 mediated TIL expansion.

a, Violin plot of IL-2 signalling score in tumour reactive / non-tumour reactive CD8⁺ TILs clones that expanded or did not expand in the cell-therapy product. Number of clonotypes and cells are shown. b, Top 35 Reactome pathways negatively associated to IL-2 signalling in CD8⁺ TILs from melanoma patients enrolled in the TIL-ACT trial (n = 13). In red are the pathways associated with PGE₂ signalling. c, Correlation per patients between eicosanoid ligand binding receptors score in CD8⁺ TILs and total number of pre-REP TILs (n = 13). d, Violin plot of PGE₂ signature score in tumour reactive/non-tumour reactive CD8⁺ TILs clones that expanded or did not expand in the cell-therapy product. Number of clonotypes and cells are shown. (a,d):Boxplots display smallest and largest values in the dataset. Box hinges: first and third quartile with centre as median. Whiskers: 1.5x interquartile range (IQR) from the first and third quartiles. Statistical comparisons were performed with one-way ANOVA with Tukey’s multiple-comparisons test (a, d) and two-sided Spearman’s correlation (c) or Pearson’s correlation (b). Independent patients were used as biological replicates with exact numbers listed in each panel.

Extended Data Fig. 2. Characterization of repeatedly activated PBLs and PGE₂-signature.

a, Schematic representation of the protocol for repeated activation of human peripheral blood lymphocytes using serial CD3/CD28 stimulation every 3 days in low dose IL-2 (Created with BioRender.com). b, Representative flow cytometry gates of exhaustion/costimulation markers used to characterize repeatedly activated PBLs. c, Heatmap representation of the frequency of inhibitory/co-stimulatory markers and TOX⁺ cells in unstimulated and repeatedly activated CD8⁺ T cells on day 10. d, Time-course evaluation of inhibitory markers (CD39, Lag3, PD1, TIM3), exhaustion marker TOX and co-stimulatory molecules (CD27, CD28) expression on CD8⁺ T cells during the repeated activation protocol at day 0-3-7-10. Data are represented as median of 3 biological replicates. Statistical comparisons were performed using one-way ANOVA with Dunnett post-hoc test for multiple comparisons. e, Volcano plot of differentially expressed (DE) genes in repeatedly activated CD8⁺ T cells treated with PGE₂ for 24 h. The Y-axis represents log₁₀ p-values, whereas the X-axis represents log₂ fold change in gene expression. Two-sided paired t test with Benjamini Hochberg correction. f, String analysis of the 63 top upregulated genes (p < 0.05, FC > 1) in repeatedly activated CD8⁺ T cells after PGE₂ treatment (PGE₂ signature). N = 3 independent biological samples.

Extended Data Fig. 3. PGE₂-EP2/EP4 signalling restricts IL-2 mediated TIL proliferation by deregulating IL-2R complexes.

a, Frequency of CFSE^high and b, Ki67⁺TILs (% CD8⁺) treated with PGE₂ for 5d in high dose IL-2 (6000IU/ml) (n = 3). c, Relative CD4⁺ count (Fold change to CTRL) of TILs treated for 72 h with PGE₂ +/− EP2/EP4 antagonists (n = 4). d, Representative flow cytometry plots of IL-2Rα/β/γ_c chains surface expression on CD8⁺ TILs treated or not with PGE₂ at 1 or 5uM. e, IL-2Rγ_c, IL-2Rα and IL-2Rβ surface expression in CD4⁺ TILs treated with PGE₂ +/− EP2/EP4 antagonists for 72 hrs (n = 4). f, Relative mRNA expression (Fold change to CTRL) of IL-2RA, IL-2RB and IL-2RG in TILs treated with PGE₂ +/− EP2/EP4 antagonists (n = 4). g, Relative IL-2Rγ_c expression (Fold change to CTRL) in CD4⁺ TILs treated with PGE₂ for 2 h, or for 2h and then re-exposed to media without PGE₂ for 70 h (n = 3). h, Relative surface and total IL-2Rγ_c expression on CD8⁺ TILs (Fold change to CTRL) after 2 h or 72 h PGE₂ treatment (n = 3). i, Time-lapse assessment (left) and representative image (right) of calcium mobilization in TILs upon PGE₂ (representative from 2 biological replicates). j, Colocalization of IL-2Rα/β/γ_c chains by ImageStream in CD8⁺ and k, CD4⁺ TILs treated with PGE₂(n = 4). l, Colocalization of IL-2Rβ and IL-2Rγ_c by confocal microscopy in TILs treated with 24 h PGE₂ (representative from 4 biological replicates). m, Representative dSTORM image of IL-2Rβ and IL-2Rγ_c levels in TILs treated with 24 h PGE₂ (representative of 2 biological replicates). n, Proximity ligation assay (PLA) images and quantification of IL-2Rβ/γ_c in TILs upon 24 h PGE₂ (representative of 3 biological replicates). o, Relative cell count (Fold change to CTRL) of unstimulated or repeatedly activated left) CD4⁺ and right) CD8⁺ T cells upon 72 h PGE₂ at different doses (n = 3). p, Relative frequency (Fold change to CTRL) of Ki67⁺(%CD8⁺) repeatedly activated T cells after PGE₂ for 5d (n = 2). q, Relative cell count (Fold change to CTRL IL-2 low) of repeatedly activated CD8⁺ T cells treated with PGE₂ in the presence of IL-2 or IL-15 for 48 h (n = 3). r, Relative IL-2Rα, IL-2Rβ and IL-2Rγ_c expression in unstimulated and repeatedly activated top) CD4⁺ and bottom) CD8⁺ T cells upon increasing doses of PGE₂ (Fold change to CTRL) (n = 3). s, Densitometry quantification of pJAK3 levels in repeatedly activated T cells from experiment in Fig. 2k (n = 3). Samples derived from the same experiments and gels/blots were processed in parallel. t, pS6 dose response curves in TILs after 30 min stimulation with IL-2 or u, anti-CD3 (n = 3). v, Representative flow cytometry plots of pS6 expression in TILs stimulated with IL-2 or IL-2 and PGE₂. Data are presented as the mean ± S.D. Statistical comparisons were performed using paired (j,k) or unpaired two-tailed t test (l,n,p) or one-way ANOVA (a,b,c,e,f,g,h,o,q,r,s) with Dunnett post-hoc test for multiple comparisons. Independent biological samples were used with exact numbers of biological replicates listed in each panel. p ≥ 0.05, not significant (ns).

Extended Data Fig. 4. PGE₂ rewires metabolism of TILs.

a-b, Gene set enrichment analysis (GSEA) enrichment plot for a, mitochondrial gene expression and b, fatty acid metabolic process in repeatedly activated CD8⁺ T cells after 24 h of PGE₂. The y-axis represents enrichment score and the x-axis shows genes (vertical black lines) represented in the pathways. The coloured band at the bottom represents the degree of correlation of the expression of these genes (red for a high gene expression and blue for a low gene expression). c, Schematic representation of the systems biology approach used to reconstruct metabolic models and infer fluxes and metabolic states consistent with the gene expression profile of repeatedly activated CD8⁺ T cells and TILs treated with PGE₂ for 24 h (created with BioRender.com). d, Heatmap representation of deregulation of metabolic fluxes (row) inferred from the generated metabolic model upon PGE₂ treatment in repeatedly activated CD8⁺ T cells (n = 3). e, Violin plot representation of reaction rate fold changes associated by metabolic subsystems upon PGE₂ treatment in repeatedly activated CD8⁺ T cells (n = 3). f, Representative map of tricarboxylic acid cycle and electron transport chain (ETC) reactions, coloured by flux deregulation (blue for downregulated and red for upregulated inferred state). g-h, Violin plot representation of metabolic task (MT) derived for g, cell growth and h, energy and reactive oxygen species enrichment analysis using gene expression data (blue) or the inferred fluxes (red) in repeatedly activated CD8⁺ T cells upon PGE₂ treatment (n = 3). i, Mitochondrial ROS levels in CD8⁺ TILs in response to PGE₂ (n = 6). Data are presented as the mean ± S.D. Statistical comparisons were performed using paired two-tailed t test. Independent biological samples were used with exact numbers of biological replicates listed in each panel.

Extended Data Fig. 5. PGE₂ affects mitochondrial fitness, T cell oxidative response and lipid metabolism.

a, Relative mitochondrial DNA copy number in unstimulated, repeatedly activated T cells and TILs after 24 h PGE₂ (Fold change to CTRL) (n = 6). b, Representative Electron Microscopy images and c, representative quantitative plot of mitochondrion number/cell in unstimulated and repeatedly activated T cells upon 24 h PGE₂ (n = 3). d, Representative histogram of mitochondrial potential (TMRM) staining in CTRL versus PGE₂ treated TILs. e, Fold change (relative to CTRL) of mitochondrial potential (TMRM) in repeatedly activated CD8⁺ T cells after 24 h PGE₂ (n = 3). f, Quantification of basal respiration, spare respiratory capacity (SRC) and ATP production in repeatedly activated T cells treated with PGE₂ for 24 h (n = 5). g, Fold change (relative to CTRL) of protein synthesis (OPP) in CD39⁻ and CD39⁺ CD8⁺ TILs treated with PGE₂ for 24 h (n = 5). h, Oxidized (GSSG)/reduced (GSH) glutathione ratio quantified by mass-spectromety in CD8⁺ repeatedly activated T cells treated with PGE₂ for 24 h (n = 5). i, Western Blot (top) and quantification (bottom) of PGC1α in TILs treated with IL-2 and PGE₂ for 48 h (representative of 2 biological replicates). j, PGC1A mRNA relative expression in repeatedly activated T cells treated with PGE₂ or mTOR inhibitor Everolimus for 12 h and then stimulated with IL-2 for 15 min (n = 6). k, PD1, TOX protein expression and EP2 and EP4 gene expression in unstimulated or repeatedly activated murine OT1 T cells (n = 3). l, PGC1A mRNA relative expression in repeatedly activated OTI murine T cells transduced with a PGC1α overexpressing vector (n = 4). m, Oxidized (GSSG)/reduced (GSH) glutathione ratio in response to 72 h PGE₂ in repeatedly activated OTI murine T cells transduced with a PGC1α -overexpressing vector (n = 4). n, Heatmap of free fatty acids relative abundance (%CTRL) measured by mass spectrometry in repeatedly activated CD8⁺ T cells (n = 5) and in o, CD8⁺ TILs treated with PGE₂ for 24 h (n = 4). p, Violin Plot representation (Fold change to CTRL) of lipid droplets metabolic task in repeatedly activated CD8⁺ T cells treated with PGE₂ (n = 3). q, Microscopy image (representative of 4 biological replicates) and r, mean lipid droplets/cells in TILs upon 24 h PGE₂ (n = 4). s, mRNA relative expression of CPT1A, HIF2α, CREB3L3 in TILs treated with PGE₂ for 48 h (n = 5). t, Representative histogram of BODIPY-C11 lipid peroxidation staining in CTRL versus PGE₂ treated CD8⁺ TILs. u, Lipid peroxidation quantification in repeatedly activated CD8⁺ T cells treated with PGE₂ for 48 h (n = 3). v, Western blots of GPX4 protein expression in TILs treated or not with PGE₂ (representative of 2 biological replicates). w, mRNA relative expression of ACSL4, LPCAT3, FSP1 and GLS2 in TILs treated with PGE₂ for 48 h (n = 5). x, Lipid peroxidation quantification (n = 2) and y, Relative cell count (Fold change to CTRL) of TILs treated with PGE₂ +/− NAC or vitamin E (VITE) for 72 h (n = 6). Data are presented as mean ± S.D. Statistical comparisons were performed using paired (f,h,r,u) or unpaired two-tailed t test (e,k) or one-way ANOVA (a,c,g,j,l,m,s,w,y) with Dunnett post-hoc test for multiple comparisons. Results were pooled from 3 experiments with n = 5 mice per group in each experiment (l,m). Independent biological samples were used with exact numbers of biological replicates listed in each panel. p ≥ 0.05, not significant (ns).

Extended Data Fig. 6. PGE₂-EP2/EP4 axis blockade increases TIL expansion, fitness and tumour-reactivity.

a, Schematic representation of conventional TIL expansion protocol. TILs are expanded from tumours fragments with IL-2 6000IU/ml for 14-28d (pre-REP phase) and then with IL-2 3000IU/ml, anti-CD3 and feeder cells for 14d (REP phase). b, Correlation between best overall clinical response at 3-months and “PGE₂ signature”, “Eicosanoid ligand binding receptor” and “Prostanoid ligand receptors” signature scores in CD8⁺ TILs from REP-TIL product of melanoma patients enrolled in phase I ACT-TIL therapy trial (n = 13). CR: Complete Response, PR: Partial Response, SD: Stable Disease, PD: Progressive Disease. c, Time-course of PGE₂ concentration (n = 8) and d, PGE₂ concentration at day 7 in the supernatant of expanding pre-REP TILs derived from breast (n = 21), melanoma (n = 12), ovarian (n = 3), and lung (n = 5) tumours. e, Baseline PGE₂ concentration in the supernatant of 48 h pre-REP cultures treated with PGE₂, EP2/4 blockade or Ketorolac/COXi (n = 4). f, Kinetics of expansion of pre-REP TILs treated with IL-2 (6000IU/ml), IL-2 + PGE₂, or IL-2 + Ketorolac/COXi at initiation of the culture (n = 2). g, Frequency of IL-2Rγ_c^high CD8⁺ TILs (%CD8⁺) at day 7 of expansion in presence of different doses of Ketorolac/COXi (n = 2). h, Heatmap representation of TCF7, Myb, PGC1A, PTGER2, PTGER4 mRNA expression in pre-REP TILs expanded with or without Ketorolac/COXi (n = 3). i, Frequency of CD8⁺ TILs per cluster between CTRL and Ketorolac/COXi expanded TILs from mass cytometry (n = 3). j, Heatmap representation of 32 CyTOF markers expression in each cluster. Frequency of CD8⁺ TILs per cluster are depicted at the bottom. k, Representative flow cytometry plots of TOX/TCF1 and TCF1/CD39 of CD8⁺ PBLs or TILs. l, Representative flow cytometry plot of TCF1/CD39 CTRL and COXi-CD8⁺TILs. m, Relative TMRM/mitotracker green ratio (Fold change to CTRL) (n = 3) and n, Lipid peroxidation quantification (n = 3) and o, Relative Oxidized (GSSG)/reduced (GSH) glutathione ratio (Fold change to CTRL) in pre-REP CD8⁺ Ketorolac/COXi TILs (n = 3). p, Representative flow cytometry plots and p, Phenotypic characterization of multimer stained MART-1 tumour-reactive CD8⁺ TILs expanded with IL-2 or IL-2+Ketorolac/COXi (n = 1). q, Schematic representation of autologous tumour cells and TILs co-culture assay. r, Representative flow cytometry plot of 41BB⁺ CD8⁺ TILs in absence (TILs alone) or presence (TILs + Tumour) of tumour cells. s, Relative frequency of tumour-reactive CD4⁺ TILs and t, CD8⁺ TILs at REP in the Ketorolac/COXi (Fold change to CTRL). Tumour-reactive T cells were assessed via 41BB surface staining expression upon co-culture with autologous tumour by flow cytometry staining (n = 3). u, Representative flow cytometry plot of TNFα⁺IFNγ⁺ CD8⁺ TILs in absence (TILs alone) or presence (TILs + Tumour) of tumour cells. v, Relative frequency (Fold change to CTRL) of tumour-reactive Ketorolac/COXi CD8⁺ TILs assessed by IFNγ⁺ expression upon co-culture with autologous tumour line in two out of the three melanoma patients tested (last patient had no IFNγ detected) (n = 2). w, TCRβ repertoire analysis of REP TILs expanded with IL-2 or IL-2+Ketorolac/COXi (n = 3). x, Frequency of CD103⁺ (n = 3) or y, PD1⁺ (%CD8⁺) intratumoural TILs in tumours from mice treated with CTRL or Ketorolac/COXi-expanded REP TILs (n = 3). Data are presented as the mean ± S.D. Statistical comparisons were performed using paired (n) unpaired two-tailed t test (m,o,s,t) or one-way ANOVA (b,e,h,x,y) with Dunnett post-hoc test for multiple comparisons. p ≥ 0.05, not significant (ns). Independent biological samples/patients were used with exact numbers of biological replicates listed in each panel. Panels a and p were created with BioRender.com.