Intersection of immune and oncometabolic pathways drives cancer hyperprogression during immunotherapy

Abstract

Immune checkpoint blockade (ICB) can produce durable responses against cancer. We and others have found that a subset of patients experiences paradoxical rapid cancer progression during immunotherapy. It is poorly understood how tumors can accelerate their progression during ICB. In some preclinical models, ICB causes hyperprogressive disease (HPD). While immune exclusion drives resistance to ICB, counterintuitively, patients with HPD and complete response (CR) following ICB manifest comparable levels of tumor-infiltrating CD8⁺ T cells and interferon γ (IFNγ) gene signature. Interestingly, patients with HPD but not CR exhibit elevated tumoral fibroblast growth factor 2 (FGF2) and β-catenin signaling. In animal models, T cell-derived IFNγ promotes tumor FGF2 signaling, thereby suppressing PKM2 activity and decreasing NAD⁺, resulting in reduction of SIRT1-mediated β-catenin deacetylation and enhanced β-catenin acetylation, consequently reprograming tumor stemness. Targeting the IFNγ-PKM2-β-catenin axis prevents HPD in preclinical models. Thus, the crosstalk of core immunogenic, metabolic, and oncogenic pathways via the IFNγ-PKM2-β-catenin cascade underlies ICB-associated HPD.

Keywords: FGF2; IFNγ; PD-L1/PD-1 pathway; T cell immunity; complete response; glycolytic metabolism; hyperprogressive disease; immune checkpoint blockade; oncogenesis; β-catenin.

Figure 1:. Rapid cancer progression occurs in a subset of patients during immunotherapy.

A, Overall survival (OS) of patients with metastatic melanoma (Cohort 1) stratified by therapy type; inset, 3-month OS; immunotherapy n = 251, targeted therapy n = 138, Restricted mean survival time (RMST) at 3 months, Hazard ratio (HR) = 0.95, P < 0.0001 by log-rank test. B, Progression-free survival (PFS) of patients with metastatic melanoma stratified by therapy type. Inset, 3-month PFS. Progression-free RMST at 3 months HR = 0.89, immunotherapy n = 251, targeted therapy n = 138, P < 0.0001 by log-rank test. C, OS of patients with metastatic NSCLC stratified by therapy type; Inset, 3-month OS; immunotherapy n = 279, chemotherapy n = 96, RMST at 3 months, HR = 0.94, P < 0.0001 by log-rank test. D, PFS of patients with metastatic NSCLC stratified by therapy type. Inset, 3-month PFS; immunotherapy n = 279, chemotherapy n = 96, Progression-free RMST at 3 months, HR = 0.94, P < 0.0001 by log-rank test. E, PFS of metastatic NSCLC patients treated with immunotherapy or chemotherapy, pooled analysis of Keynote-042, Poplar, and Checkmate 227 randomized control trials. Progression-free Log-rank HR at 3 months = 0.616, P = 0.0336. F, Hazard ratios for 3-month PFS of metastatic NSCLC patients treated with immunotherapy or chemotherapy, pooled analysis of Keynote-042, Poplar, and Checkmate 227 randomized control trials. Two-sided t test, P = 0.0152. G, OS of patients with metastatic melanoma treated with ICB (Cohort 1) stratified by timing of progression, other n = 146, rapid progression (PFS < 3 months) n = 53, Landmark analysis (3 months) hazard ratio (HR) = 0.291, P < 0.0001 by log-rank test. H, OS of patients with metastatic NSCLC treated with ICB (Cohort 2) stratified by timing of progression, other n = 113, rapid progression (PFS < 3 months) n = 67, Landmark analysis (3 months) hazard ratio (HR) = 0.3251, P < 0.0001 by log-rank test. I, Waterfall plot showing change of tumoral burden from initiation of therapy to first surveillance imaging in melanoma patients treated with indicated therapy, dotted line > 50% increase in tumor burden, immunotherapy n = 200, targeted therapy n = 96, Chi-square = 19.53, P < 0.0001. Data are shown as percentage change. J, Waterfall plot showing change of tumoral burden from initiation of therapy to first surveillance imaging in NSCLC patients treated with indicated therapy, dotted line > 50% increase in tumor burden, immunotherapy n = 212, chemotherapy n = 68, Chi-square = 5.133, P = 0.0235. Data are shown as percentage change. K-L, Representative cross-sectional (lower) and 3D reconstructed (upper) computed tomography (CT) images of a patient with metastatic melanoma (K) and a patient with NSCLC (L) with HPD preceding receipt of immunotherapy (left), at baseline preceding immunotherapy (middle), and at first reassessment following immunotherapy (right). M-N, Longitudinal tumor burden assessment in melanoma (M) or NSCLC (N) patients who progressed while receiving ICB stratified by pattern of response. Baseline- cross sectional imaging immediately prior to ICB initiation. Pre-therapy- Imaging assessment prior to baseline evaluation. On therapy- next surveillance scan after baseline assessment. Melanoma patients with PD (progressive disease, per RECIST 1.1, n = 48) and HPD (hyperprogressive disease, per Champiat et al., n = 21); NSCLC patients with PD (n = 77) and HPD (n = 26), interrupted time series regression, Data are shown as mean ± s.d., P value indicated. O. OS of metastatic melanoma patients (Cohort 1) stratified by best response, complete response (CR) n = 31, partial/stable disease (PR/SD) n = 58, progressive disease (PD) n = 48, and hyperprogressive disease (HPD) n = 21. Log-rank test, HPD vs PD HR = 0.3058, ***P < 0.001. P. OS of metastatic NSCLC patients (Cohort 2) stratified by best response, CR n = 7, PR/SD n = 77, PD n = 77, and HPD n = 26. Log-rank test, HPD vs PD HR = 0.25, ***P < 0.001. See also Figure S1 and Tables S1–S3.

Figure 2:. Immunogenic and oncogenic pathways correlate in patients with HPD.

A. Immune gene signature analysis of patients receiving immunotherapy who developed a complete response (CR) or hyperprogressive disease (HPD) per Champiat et al. in Cohort 3, individual patients are shown; P-values were generated from multivariate mixed effect linear models controlling for biopsy site (fixed effect) and disease type (random effect). B-C. Representative immunofluorescence staining (B) and quantitation (C) for baseline tumor infiltrating CD8+ T cells in melanoma patients with indicated response to therapy. Frequency of positive cells is shown; CR, n = 20, HPD, n =12. Two-sided t-test. D-E. Representative immunofluorescence staining (D) and quantitation (E) for baseline tumor infiltrating CD8+ T cells in NSCLC patients with indicated response to therapy. Frequency of positive cells is shown; CR, n = 20, HPD, n =12. Two-sided t-test. F. Oncogenic gene signature analysis of patients receiving immunotherapy who developed a complete response (CR) or hyperprogressive disease (HPD) in Cohort 3, individual patients are shown; P-values were generated from multivariate mixed effect linear models controlling for biopsy site (fixed effect) and disease type (random effect). G-J. Multiplex immunofluorescence staining was conducted in tumor tissues from patients with melanoma (G, H) and NSCLC (I, J). Representative images showed FGF2, MYC, and CD133 expressing tumor cells in patients with HPD and CR (G, I). Percentages of single or double positive tumor cells are shown in patients with HPD and CR (H, J). Mean and interquartile range shown. Melanoma patients with CR (n = 20) and HPD (n =12); NSCLC patients with HPD (n = 5) and CR (n = 6). Two-sided t-test. See also Figure S2 and Table S4.

Figure 3:. CD8⁺ T cells drive cancer hyperprogression via IFNγ. See also Figure S3.

A-D. YUMM1.7 tumor bearing C57BL/6 mice were treated with control (IgG) or PD-L1 antibody. Tumor growth curves were plotted (A). FACS analysis showed tumor T cell infiltration (B, C) and tumor Myc and Cd44 expression (D). MFI, mean fluorescence intensity. E. YUMM1.7 tumor bearing C57BL/6 mice were treated with control (IgG) or PD-L1 antibody. Day 14^th after tumor inoculation, the indicated genes expression in tumors was determined by qPCR. F-G. YUMM1.7 tumor bearing C57BL/6 mice were treated with control (IgG) or CD8 antibody. Tumor growth curves were plotted (F), and tumor Myc and Cd44 expression (MFI) were determined by FACS (G). H-I. YUMM1.7 tumor bearing mice were treated with control (IgG), PD-L1 antibody, CD8 antibody, or the combination of PD-L1 and CD8 antibodies. Tumor growth curves were plotted (H) and end point tumor weight (I) were scaled. J. PLC2.4 cells were inoculated in C57BL/6 wild type mice. Mice were treated with control (IgG) or PD-L1 antibody. Tumor growth curves were plotted. K-L. Wild type and Ifngr1 KO (K) or Stat1 KO (L) YUMM1.7 cells were inoculated in C57BL/6 mice. Tumor growth curves were plotted. M. Wild type or 2 single clones of Stat1 KO YUMM1.7 cells were inoculated in C57BL/6 mice. Tumor growth curves were plotted. N-O. PLC2.4 wild type and Ifngr1 KO (N) or Stat1 KO (O) cells were inoculated in C57BL/6 wild type mice. Tumor growth curves were plotted, n = 5 (N), n = 6 (O). P. Wild type or 2 single clones of Stat1 KO PLC2.4 cells were inoculated in C57BL/6 mice. Tumor growth curves were plotted. Q. Wild type or Stat1 KO YUMM5.2 cells were inoculated in C57BL/6 mice. Tumor growth curves were plotted. In all panels n=5 unless otherwise indicated. Data are shown as mean ± s.d., two-tailed t-test. See also Figure S3.

Figure 4:. IFNγ reduces NAD⁺ to activate β-catenin acetylation.

A. Correlation between the IFNγ and oncogenic signaling gene scores in lung adenocarcinoma in TCGA datasets. The expression of the indicated oncogenic gene signaling scores was plotted based on the top and bottom 25^th percentiles of the IFNγ gene signaling scores. P value by two-tailed t-test. B. Correlation between the IRF1 and MYC signaling gene scores in several cancer types in TCGA datasets. The expression of the MYC gene signaling scores was plotted based on the top and bottom 25^th percentiles of the IRF1 gene signaling scores. P value by two-tailed t-test. LUAD, lung adenocarcinoma; BLCA, Bladder urothelial carcinoma; KIRC, Kidney renal clear cell carcinoma; OV, ovarian cancer. C-D. Based on the RNA-seq datasets (GSE99299), the indicated gene expression was shown in A375 (C) or A549 (D) cells in the presence or absence of IFNγ. n = 2. E. A375 cells were transfected with β-catenin signaling reporter TOP-FLASH or the mutant vector (FOP-FLASH). Cells were treated with IFNγfor 48 hours. Results are expressed as the relative luciferase activity. Data are shown as mean ± s.d., n = 3, P value by two-tailed t-test. F. A375 cells were treated with IFNγ and recombinant DKK1 or Wnt-C59 (C59) for 48 hours. The indicated gene expression was determined by qPCR. n = 3. G. YUMM1.7 cells were treated with IFNγ and DKK1 or Wnt-C59 (C59) for 48 hours, followed by 3D-sphere culture. Spheres were counted on day 7 after sphere culture. Data are shown as mean ± s.d., n = 6. Two-tailed t-test. H-I. YUMM1.7 cells were treated with IFNγ and DKK1 or Wnt-C59 (C59) for 48 hours. Percentages of Cd44⁺ (H) or Cd133⁺ (I) cells were determined by FACS. Data are shown as mean ± s.d., n = 3. Two-tailed t-test. J. Wild type or 2 clones of CTNNB1 KO A375 cells were treated with IFNγ for 48 hours. The indicated gene expression was determined by qPCR. n = 3. K. Flag-tagged β-catenin expressing A375 cells were treated with IFNγ for 24 hours. Co-IP was performed with Flag antibody. Acetylated-lysine and β-catenin were detected in the IP products. 1 of 2 Western blots shown. L. Flag-tagged β-catenin expressing A375 cells were treated with Salermide for 10 hours. Acetylated-lysine and β-catenin were determined in the Co-IP products with Flag antibody. 1 of 2 blots shown. M. A375 cells were treated with Salermide (Saler) or Sirtinol (Sirti) for 24 hours. The indicated gene expression was determined by qRT-PCR. n = 3. N. A375 cells were treated with IFNγ for 24 hours. SIRT1 and GBP1 (positive control) proteins were detected by Western blot. 1 of 2 Western blots shown. O. A375 cells were treated with IFNγ for 24 hours. NAD⁺ levels were determined by quantitation kit. Data are shown as mean ± s.d., n = 3. Two-tailed t-test. P. Flag-tagged β-catenin expressing A375 cells were treated with IFNγ, in the presence or absence of 0.7 mM nicotinamide riboside (NR), for 24 hours. Acetylated-lysine and β-catenin were detected in the Co-IP products with Flag antibody. 1 of 2 Western blots shown. Q. A375 cells carrying TOP-FLASH were treated with IFNγ, in the presence or absence of 0.7 mM nicotinamide riboside (NR), for 24 hours. Luciferase activity of the β-catenin signaling reporter was determined. Data are shown as mean ± s.d., n = 3. Two-tailed t-test. R. A375 cells were treated with IFNγ and nicotinamide riboside (NR) for 24 hours. MYC and GAPDH proteins were determined by Western blot. 1 of 3 Western blots shown. S. A375 cells were treated with IFNγ and nicotinamide riboside (NR) or β-nicotinamide mononucleotide (NMN) for 48 hours. The indicated gene expression was determined by qPCR. n = 3. T. Schematic diagram showing that IFNγ reduces NAD⁺ to suppress SIRT1-mediated β-catenin deacetylation, thereby activating β-catenin. U-V. Wild type or CTNNB1 K345R mutant A375 cells were treated with IFNγ. 24 hours after treatment, acetylated β-catenin was determined by Western blotting following Flaĝ-catenin Co-IP (U). 48 hours after treatment, the indicated gene expression was determined by qPCR (V). 1 of 2 Western blots shown, n = 3 for qPCR. W-X. Wild type or CTNNB1 K345R mutant A375 cells were treated with Salermide (Saler). 12 hours after treatment, acetylated β-catenin was determined by Western blotting following Flag-β- catenin Co-IP (W). 24 hours after treatment, the indicated gene expression was determined by qPCR (X). 1 of 2 Western blots shown, n = 3 for qPCR. See also Figure S4.

Figure 5:. IFNγ regulates PKM2 phosphorylation to alter NAD⁺/ β-catenin signaling.

A. A375 cells were treated with IFNγ for 24 hours. Seahorse analysis showed the extracellular acidification rate (ECAR) in control cells and IFNγ-treated cells in the presence of glucose, oligomycin or 2-DG. Data are shown as mean ± s.d., n = 3. B. A375 cells were treated with IFNγ for 24 hours. Catalytic activities of glycolysis rate-limiting enzymes were determined by quantitation kits. Data are shown as mean ± s.d., n = 3. Two-tailed t-test. C. Schematic diagram showing the glycolysis pathway and the NAD⁺/ NADH balance. D-E. A375 (D) or YUMM1.7 (E) cells were treated with IFNγ for 24 hours. Phosphorylated or total protein levels of PKM2 were detected by Western blot. 1 of 2 Western blots shown. F-G. A375 cells were treated with IFNγ, in the presence or absence of DASA-58, for 48 hours. Lactate production (F) or intracellular levels of NAD⁺ (G) were determined by quantitation kit. Data are shown as mean ± s.d., n = 3. Two-tailed t-test. H. A375 cells carrying Flag-tagged β-catenin were treated with IFNγ, in the presence or absence of DASA-58. Acetylated-lysine and β-catenin were detected in the Co-IP products with Flag antibody. 1 of 2 Western blots shown. I. A375 cells carrying TOP-FLASH were treated with IFNγ, in the presence or absence of DASA-58, for 24 hours. Relative luciferase activity was determined. Data are shown as mean ± s.d., n = 4. Two-tailed t-test. J. A375 cells were treated with IFNγ, in the presence or absence of DASA-58, for 24 hours, β- catenin signaling genes and IFNγ signaling gene (GBP1) (positive control) were determined by qRT-PCR. Data are shown as mean ± s.d., n = 3. Two-tailed t-test. K. A375 cells were treated with IFNγ, in the presence or absence of DASA-58, for 48 hours. MYC and GAPDFI proteins were determined by Western blot. 1 of 2 Western blots shown. L. YUMM1.7 cells were treated with IFNγ, in the presence or absence of ML-265, for 48 hours. Surface expression of Cd44 was determined by FACS. Data are shown as mean ± s.d., n = 3. Two-tailed t-test. M. YUMM1.7 tumor bearing C57BL/6 mice were treated with anti-PD-L1, ML-265, or the combination of anti-PD-L1 and ML-265. Tumor growth curves were plotted, n = 5 / group. See also Figure S5.

Figure 6:. IFNγ induces FGF2 to control PKM2/ NAD⁺/ β-catenin signaling.

A. A375 cells were treated with IFNγ in the presence of Gefitinib (Gef) or Dovitinib (Dov), for 36 hours. MYC protein was determined by FACS. Data are shown as mean ± s.d., n = 3. Two-tailed t- test. B. A375 cells were treated with IFNγ in the presence of Dovitinib (Dov), for 48 hours, β-catenin signaling genes (MYC, CCND1) and IFNγ signaling gene (GBP1) were determined by qRT-PCR. Data are shown as mean ± s.d., n = 3. Two-tailed t-test. C. FGF2 transcripts were quantified by qRT-PCR in IFNγ-treated A375, YUMM1.7 and PLC2.4 cells. Data are shown as mean ± s.d., n = 3. Two-tailed t-test. D. FGF2 protein was determined by Western blot in A375 cells treated with IFNγ. 1 of 2 Western blots shown. E-G. A375 cells were treated with IFNγ, in the presence or absence of FGF2 neutralizing antibody (αFGF2). Phosphorylated (Y105) and total protein levels of PKM2 were determined at 24 hours by Western blot (E). Cellular NAD⁺ levels were quantified at 24 hours by kit (F). MYC expression was determined at 48 hours by FACS (G). 1 of 2 Western blots shown (E). Data are shown as mean ± s.d., n = 3. Two-tailed t-test (F, G). H. Fgf2, Myc or Ccnd1 transcripts were detected by qRT-PCR in PLC2.4 shFlue or shFgf2 cells. Data are shown as mean ± s.d., n = 3. Two-tailed t-test. I. shFluc or shFgf2 PLC2.4 tumor bearing C57BL/6 mice were treated with anti-PD-L1 and isotype IgG. Tumor growth curves were plotted, n = 5 animals. J. MC38 cells were forced expression of Fgf2 (Fgf2^OE). RNA levels of Fgf2, β-catenin signaling genes (Myc, Ccnd1, Cd44), and epithelial marker gene (Cdh1) were determined by qRT-PCR. Data are shown as mean ± s.d., n = 3. Two-tailed t-test. K. Fgf2^OE MC38 tumor bearing C57BL/6 mice were treated with anti-PD-L1 and isotype IgG. Tumor growth curves were plotted, n = 6 animals. See also Figure S6, S7, and Table S5.

Figure 7:. Oncometabolic reprogramming drives cancer hyperprogression during immunotherapy.

Mechanistic scheme of HPD development. IFNγ produced by ICB-activated T cells targets tumor FGF2 signaling, inducing PKM2 phosphorylation at Y105 and decreasing NAD⁺ levels, thereby diminishing SIRT1 activity and lessening β-catenin deacetylation. Consequently, β-catenin signaling pathway is activated resulting in enhanced oncogenic potential and HPD.