Resistance of melanoma to immune checkpoint inhibitors is overcome by targeting the sphingosine kinase-1

Abstract

Immune checkpoint inhibitors (ICIs) have dramatically modified the prognosis of several advanced cancers, however many patients still do not respond to treatment. Optimal results might be obtained by targeting cancer cell metabolism to modulate the immunosuppressive tumor microenvironment. Here, we identify sphingosine kinase-1 (SK1) as a key regulator of anti-tumor immunity. Increased expression of SK1 in tumor cells is significantly associated with shorter survival in metastatic melanoma patients treated with anti-PD-1. Targeting SK1 markedly enhances the responses to ICI in murine models of melanoma, breast and colon cancer. Mechanistically, SK1 silencing decreases the expression of various immunosuppressive factors in the tumor microenvironment to limit regulatory T cell (Treg) infiltration. Accordingly, a SK1-dependent immunosuppressive signature is also observed in human melanoma biopsies. Altogether, this study identifies SK1 as a checkpoint lipid kinase that could be targeted to enhance immunotherapy.

Conflict of interest statement

N.M. has worked as an investigator and/or consultant and/or speaker for BMS, MSD, Amgen, Roche, GSK, Novartis, Pierre Fabre. B.S. has worked as an investigator, consultant and speaker for BMS. The authors declare that they have no other conflict of interest.

Figures

Fig. 1. SPHK1 expression inversely correlates with survival after ICI therapy.

aSPHK1 expression in human nevi (n = 17) compared to primary (P) melanoma (n = 45), and in primary melanoma (n = 25) compared to metastatic (M) melanoma (n = 44) based on the Oncomine database. Data were compared using Mann–Whitney test. The bottom-most and topmost horizontal lines, the lower and upper hinges, and the middle line of the boxplots indicate the minimum and maximum values, the 25th and 75th percentiles, and the median, respectively. b Percentages of cancer cells positive for SPHK1 mRNA staining in metastatic melanoma tissues of 32 patients prior anti-PD-1 treatment (Low: ≤ 50% of tumor cells are positive (black points); High: > 50% of tumor cells are negative (red points)). c Representative mRNA staining of low and high SPHK1 expression. Skin (P1,P3) or lymph node (P2,P4) biopsies from patients (P). Percentages (%) indicate the proportion of cancer cells positive for SPHK1 mRNA staining. Large and small blue lines represent 200 and 20 μm, respectively. d Progression-free survival and e overall survival curves of patients with >50% of melanoma cells positive for SPHK1 (red line; n = 11) or <50% (black line; n = 21). Survival times were calculated from the first day of the cycle of anti-PD-1 post biopsy. Statistical significance was determined by log-rank test.

Fig. 2. SK1 downregulation increases the CD8 + /Treg ratio and reduces tumor growth.

a SK1 enzymatic activity was measured in Yumm cells transfected with a control shRNA (shCtrl, black column) or two SK1-targeted shRNA (white and gray columns) (means ± SEM of 5 independent experiments). b Cell proliferation (n = 4). Results represent means of 3 independent experiments. c shCtrl (black points), shSK1(1) or shSK1(2) (white points) Yumm cells were injected in wild-type (WT) C57BL/6 mice (n = 12 mice/group). Arrows indicate tumor regression of shSK1 tumors. d shCtrl or shSK1(1) Yumm cells were injected in NOD scid gamma (NSG) immunodeficient mice (n = 5 mice/group). Tumor volume was determined at the indicated days. Tumor volumes, presented as means ± SEM, are representative of at least two independent experiments. e, f shCtrl or shSK1(1) Yumm cells were injected in C57BL/6 mice, and TIL content was analyzed by flow cytometry on day 11. Percentages of CD8 + (upper panel) and CD4 + Foxp3 + (Treg; lower panel) T cells and CD8 + /Treg ratio were calculated. e Representative stainings; values indicate the percentage of cells in the quadrant. f Each symbol represents an independent tumor (n = 9 mice/group). Results are representative of at least two independent experiments. Data were compared using Kruskal–Wallis test with Dunn’s correction (a, b), two-way ANOVA test (c, d) or Mann–Whitney test (e and f).

Fig. 3. SK1 silencing reduces Treg accumulation in tumors.

Control shRNA (shCtrl; black points) or SK1-targeted shRNA (shSK1(1); white points) Yumm melanoma cells were injected intradermally in C57BL/6 mice, and TIL content was analyzed by flow cytometry on day 11 (n = 9 mice/group). Percentages (%) of CD8 + T cells (a) and Treg cells (b) expressing Ki67. c Proportion of IFNγ + CD8 cells after PMA and Ionomycin stimulation. d Percentage of PD-1 + CTLA-4 + (upper panel) and PD-1 + TIM-3 + (lower panel) cells. e MFI of CTLA-4, PD-1 and TIM-3 among CD8 + TILs. Percentage (f) and Mean Fluorescence Intensity (MFI; g) of CTLA-4 in Treg. Each symbol represents an independent tumor (n = 9). h Levels of Foxp3, Tgfb1, Il10, Ccl17, Ccl22 and Ido1 transcripts in shCtrl (black boxes) or shSK1(1) (white boxes) tumors were quantified by RT-qPCR. The bottom-most and topmost horizontal lines, the lower and upper hinges, and the middle line of the boxplots indicate the minimum and maximum values, the 25th and 75th percentiles, and the median, respectively. Results are representative of two independent experiments (n = 8 mice/group). Samples were compared using Mann–Whitney test.

Fig. 4. SK1 inhibition enhances the efficacy of anti-CTLA-4 or anti-PD-1 therapy.

Control shRNA (shCtrl) or SK1-targeted shRNA (shSK1(1)) Yumm cells were injected on day 0, and then mice were treated with isotype (iso) control antibody (black lines; shCtrl n = 10 mice, shSK1(1) n = 11 mice), anti-CTLA-4 (red lines; shCtrl n = 11 mice, shSK1(1) n = 11 mice), or anti-PD-1 (blue lines; shCtrl n = 11 mice, shSK1(1) n = 12 mice). a Individual growth curves are depicted for each tumor. Inserts: numbers indicate percentage (%) of tumor-free mice at day 26. b Kaplan–Meier survival curves with log-rank test (iso, circles and black lines; anti-CTLA-4, triangles and red lines; anti-PD-1, squares and blue lines. Full and empty symbols correspond to shCtrl and shSK1(1), respectively). The arrow indicates a second orthotopic injection of shSK1 cells. c Percentages of CD4 + Foxp3 + among CD4 T cells (left panel) or total cells (middle panel) and CD8/CD4 + Foxp3 + T-cell ratio (right panel) at day 11 (iso, black/white boxes; anti-CTLA-4, red boxes; anti-PD-1, blue boxes. Full and empty symbols correspond to shCtrl and shSK1(1), respectively) (n = 9 mice/group). dCCL17 and CCL22 mRNA expression in tumors. Results are representative of two independent experiments (n = 5 mice/group). c, d Samples were compared using Kruskal–Wallis test with Dunn’s correction.

Fig. 5. SK1 targeting enhances ICI efficacy in various cancer models.

a SK1 enzymatic activity in 4T1-Luc cells stably transfected with a control shRNA (shCtrl;) or a SK1-targeted shRNA (shSK1(1)). Data are means ± SEM of three independent experiments. b Cell proliferation. Results are means ± SEM of four independent experiments. c shCtrl and shSK1(1) 4T1-Luc cells were injected in mammary fat pad of BALB/C mice. Mice were then treated with isotype control antibody (iso) or α-CTLA-4. Metastasis index at day 31 in lung of mice injected with shCtrl (upper panel; full columns) or shSK1(1) (lower panel; empty columns) 4T1-Luc tumors and treated (right panel; red columns) or not (left panel; black and white columns) with α-CTLA-4. Inserts: numbers indicate percentage of metastasis-free mice. Data are representative of two independent experiments. d–f Mice were challenged with untransfected Yumm cells and then treated or not with vehicle (Ctrl; black circles with black full line, n = 11 mice), SKI-I (white circles with black dotted line, n = 12 mice), anti-CTLA-4 (red and white triangles; full symbols with full line (without SKI-I n = 12 mice) or empty symbols with dotted line (with SKI-I n = 12 mice). Data are representative of two independent experiments. d Tumor volumes are presented as means ± SEM. e Kaplan–Meier survival curves with log-rank test. f CD8/CD4 + Foxp3 + T-cell ratio in tumors at day 11 (n = 4). g Mice were challenged with untransfected Yumm cells and then treated or not with vehicle (Ctrl; black squares with black line), SKI-I (empty green circles or triangles with green lines) and anti-PD-1 (full blue (without SKI-I) or empty green (with SKI-I) triangles) (n = 12 mice/group). h Mice were inoculated with MC38 and then treated or not with vehicle (Ctrl; black squares with black line), SKI-I (empty green circles or triangles with green lines) and anti-PD-1 (full blue (without SKI-I) or empty green (with SKI-I) triangles) (n = 5 mice /group). Data are representative of two independent experiments. Tumor volumes are presented as means ± SEM. Samples were compared using Mann–Whitney test (a) Kruskal–Wallis test with Dunn’s correction (b, c, and f) or using two ANOVA test (d, g, and h).

Fig. 6. Reduced PGE2 production induced by shSK1 partially enhances ICI efficacy.

a Volcano plot showing proteins differentially regulated between control shRNA (shCtrl) and SK1-targeted shRNA (shSK1(1 and 2)) Yumm cells. Blue and red points represent under- and overexpressed proteins, respectively. bSphk1 (left panel) aexpression in shCtrl (black column, n = 17 mice), shSK1(1) (white column, n = 17 mice) or shSK1(2) (gray column n = 12 mice) Yumm cells and Ptges (right panel) expression in shCtrl (black column, n = 8 mice), shSK1(1) (white column, n = 8 mice) or shSK1(2) (gray column, n = 5 mice) Yumm cells; n = 5–10 independent experiments. c Intracellular (left panel, n = 5) and extracellular (right panel, n = 4) PGE2 levels were normalized to those in shCtrl; n = 2–6 independent experiments. dPtges (left panel, n = 5) and Sphk1 (right panel, n = 7) expression was analyzed in shCtrl(P) (dark gray columns) or shPGES (clear gray columns) Yumm cells; n = 5–6 independent experiments. e, f Mice were challenged with shCtrl(P) or shPGES cells on day 0, and then treated with isotype control antibody (iso; black lines), anti-CTLA-4 (red lines) or anti-PD-1 (blue lines). e Individual growth curves are depicted for each tumor (n = 10 mice). Insets: numbers indicate percentage of tumor-free mice at day 25. f Kaplan–Meier survival curves with log-rank test. shCtrl(P) and shPGES tumors were represented by black and white squares, respectively. g Levels of Foxp3, Tgfb1, Il10, Ccl17 and Ccl22 transcripts in shCtrl(P) (dark gray columns) or shPGES (clear gray columns) tumors (n = 5). Data, expressed as relative expression, are means ± SEM. The bottom-most and topmost horizontal lines, the lower and upper hinges, and the middle line of the boxplots indicate the minimum and maximum values, the 25th and 75th percentiles, and the median, respectively. Samples were compared using Kruskal–Wallis test with Dunn’s correction (b and c) or Mann–Whitney test (d).

Fig. 7. Human SPHK1 expression correlates with immune suppressive factors.

a Heat map for a selected list of genes indicative of immune suppression exhibiting low (20th percentile) and high (80th percentile) SPHK1 expression in melanoma samples using the TCGA cohort. Red indicates high expression and blue indicates low expression relative to the mean expression of the gene across all samples (n = 136). b Correlations between SPHK1 expression (log2(x + 1) transformed RSEM normalized count) and genes indicative of immune suppression using non-parametric Spearman’s test (n = 342).