Transcriptional signature of rapidly responding NK cells reveals S1P5 and CXCR4 as anti-tumor response disruptors

Abstract

Natural killer (NK) cells are prototypic cytotoxic innate lymphocytes that can kill target cells, such as tumor cells, in the absence of antigen-restriction. Peripheral NK cells exhibit a high degree of heterogeneity. Here, we set out to broadly assess intrinsic modulators of NK cell degranulation in an unbiased manner. We stimulated human primary blood-borne NK cells pre-treated with different cytokine regimens with the HCT116 human colon cancer cell line and used detection of lysosome-associated membrane glycoprotein 1 (LAMP1) as an identifier of rapid NK cell degranulation. RNA sequencing of FACS-sorted LAMP1^hi NK cells showed CXCR4 and S1PR5 were top down-regulated genes. Using compounds that modulate activity of CXCR4 and S1P receptor family members S1P1 and S1P5, we confirmed they play an important immunosuppressive role in NK cell cytotoxicity. Mechanistically, engagement of CXCR4 and S1P1/5 receptors triggered phosphorylation of p42 and Ca²⁺ influx. CXCR4 activation promoted S1P5 upregulation and vice versa, and joint activation of both receptors amplified the defect NK cell degranulation. Intriguingly, in tumor samples the expression of both receptors and the synthesis of their ligands themselves appear to be coordinately regulated. Together, these data suggest that specifically and simultaneously targeting CXCR4 and S1P5 activity in the tumor microenvironment (TME) could be a beneficial strategy to unleash full cytotoxic potential of cytotoxic NK effector cells in the tumor.

Fig. 1

LAMP1 is detected at the cell surface of stimulated NK cells and can be modulated by IL-2 and TGF-β. In all instances, NK cells were treated for 48 h with the indicated cytokine regimens (IL-2 20 ng/mL, TGF-β 10 ng/mL or vehicle). For degranulation assays, pre-stimulated NK cells were co-cultured with HCT-116 at a 1:1 E: T ratio for 2 h. A Flow cytometry of LAMP1 levels in unstimulated (0 h) and HCT-116-challenged (2 h) NK cells. B Percentage of LAMP1⁺ cells in resting and HCT-116-challenged NK cells pre-treated with cytokine regimens. C) Survival of HCT-116 cells after 72 h co-culture with NK cells. D Graphic abstract of the strategy to identify new intrinsic modulators of NK cell activation created with BioRender.com. E Flow cytometry histograms of LAMP1 signal in cytokine-treated NK cells after stimulation with HCT-116 for 2 h (RNA-seq samples). F Secreted IFN-γ after 4 day co-culture of HCT-116 and NK cells. G Survival of HCT-116 cells after 96 h co-culture with NK cells. Signal was normalized to confluence in HCT-116 mono-culture wells. For all charts (B, E–G) every data point corresponds to the average value in each independent donor and bars the mean ± SD and Two-way ANOVA statistical tests were conducted. Dunnett’s test was applied for multiple comparisons except for the viability assessment (G), for which Tukey’s test was used. Symbols represent ns = not significant, *p-value < 0.05, **p-value < 0.01, ***p-value < 0.001, ****p-value < 0.0001.

Fig. 2

Bulk RNA-seq produced a comprehensive list of differentially regulated genes in LAMP1⁻ and LAMP1^hi NK cells. A Principal Component (PC) analysis plot based on gene expression values. Every point represents a sample belonging to an independent donor. Volcano plots of genes differentially expressed in LAMP1^hi against LAMP1⁻ cells in untreated NK cells (PBS) (B), LAMP1^hi versus LAMP1⁻ cells in cells treated with TGF-β (C) and LAMP1^hi cells treated with IL-2 versus vehicle (D). Genes highlighted in orange pertain to the top 50 differentially regulated genes according to log2Fold Change and -log10 adjusted p-value (-log10(padj)) in each comparison. E Dot plot of significantly differentially regulated genes in each comparison. Top 100 pathways across all comparisons were selected based on perturbation (effect size) and adjusted p-value. Degree of activation of each pathway is indicated by color of the dots and their size illustrates the fraction of intersection with genes included in each pathway.

Fig. 3

Activation of CXCR4 hinders degranulation in stimulated NK cells. A CXCR4 expression quantified as normalized counts in the conditions used for RNA-seq. Flow cytometry quantification of CXCR4 signal in NK cells treated with IL-2, TGF-β or combination cytokine regimens at high (10 ng/mL) (B) or low concentration (0.5 ng/mL) (C) after 48 h of treatment. D Flow cytometry histograms indicating levels of CXCR4 at the cell surface of CD56^dim and CD56^bright freshly isolated human peripheral NK cells. E Percentage of LAMP1^hi cells in CXCR4⁻ and CXCR4⁺ NK populations after co-culture with HCT-116 for 2 h following 48 h of cytokine regimens. F Degranulation triggered by 2 h HCT-116 stimulation of NK cells after overnight culture with 100 ng/mL CXCL12 or vehicle. G) Density plots of cell types present in the lower chamber of a transwell plate in a migration assay. NK cells were placed in the upper chamber and left to migrate for 2 h towards HCT-116 cells in the lower chamber in the presence or absence of a 100 ng/mL CXCL12 gradient. NK cells were pre-treated with 10 µM plerixafor or vehicle for 2 h prior to transfer to transwell plate. H Percentage of viable NK cells out of total single cells in the lower chamber. I Percentage of CXCR4⁺ NK cells in the lower chamber of the transwell plate at the end of the experiment. Degranulation of NK cells (J) and CXCR4 levels (K) after 2 h co-culture with HCT-116 following overnight NK cell treatment with CXCL12 and plerixafor. L Cell survival of HCT-116 cells after 24 h of co-culturing with NK cells previously treated with indicated regimens. For all charts, every data point corresponds to the average value in each independent donor and bars the mean ± SD. Adjusted p-values (padj) were calculated using DESeq for RNA-seq gene expression analysis (A). T student tests were used for comparisons between two conditions (D, F, L). One-way ANOVA statistical tests were used in systems with several conditions for a single variable (B, C, H, I, J, K). For these, Dunnett’s test (B, C, H, I) or Tukey’s test (I, J) were used to perform multiple comparisons. The Two-way ANOVA statistical test was performed for analysis containing two variables (E), for which Šídák’s test for multiple comparisons was used. Symbols represent ns = not significant, *p-value < 0.05, **p-value < 0.01, ***p-value < 0.001.

Fig. 4

NK cells pre-treated with S1P1/5 agonist ceralifimod are less reactive to stimulation by tumor cells. A S1P receptor family gene expression quantified in transcripts per million (tpm) in the RNA-seq. B Heat map showing scaled expression of S1P receptor genes in each treatment and LAMP1 classification of NK cells. C Percentage of freshly isolated NK cells binding to S1PF. D Flow cytometry indicating levels of S1P fluorescein binding to CD56^dim and CD56^bright freshly isolated human peripheral NK cells. E Quantification of S1PF binding to NK cells treated with cytokine regimens (10 ng/mL IL-2, 10 ng/mL TGF-β or vehicle) 2 days after treatment. F Percentage of LAMP1^hi cells in S1PF⁻ and S1PF⁺ NK populations after co-culture with HCT-116 for 2 h following 48 h of cytokine treatment. Percentage of LAMP1⁺ NK cells upon 48 h treatment with IL-2 in combination with 1 µM ceralifimod or DMSO and stimulation with SF-539 (G) or HCT-116 (H) cells for 2 h. I Percentage of LAMP1⁺ NK cells after co-culture with HCT-116 for 2 h following 48 h of cytokine, 1 µM ceralifimod and 1 µM inhibitor regimens. J Survival of HCT-116 cells after 24 h of co-culturing with NK cells previously treated with indicated regimens at different E: T ratios. E:T stands for Effector:Tumor cell. For all charts, every data point corresponds to the average value in each independent donor and bars the mean ± SD. T student tests were used for comparisons between two conditions (D). One-way ANOVA statistical tests were used in systems with several conditions for a single variable (E, I). Dunnett’s test was used to perform multiple comparisons. The Two-way ANOVA statistical test was performed for analysis containing two variables (F, G, H, J), for which Šídák’s test (F, G, H) or Dunnett’s test (J) for multiple comparisons were used. Symbols represent ns = not significant, *p-value < 0.05, **p-value < 0.01, ***p-value < 0.001, ****p-value < 0.0001.

Fig. 5

CXCL12 and ceralifimod signal through Gα_i. A Heat map showing scaled expression of proteins in the S1P receptor signaling cascade. B Gene expression of selected genes involved in S1P signaling quantified in transcripts per million (tpm). C) Immunoblot of selected proteins in the S1P receptor signaling pathway for NK cells treated with vehicle, IL-2 and TGF-β for 48 h. Phosphorylation of p42 and p44 proteins upon stimulation of NK cells with 100 ng/mL CXCL12 for 15 min (D) or 1 µM ceralifimod for 5 and 10 min (E). F Representative histogram and quantification of Flag-tag signal in HCT-116 transfected with Flag-S1P1 and Flag-S1P5 encoding plasmids. G Protein detection of p42 phosphorylation in HCT-116 cells that overexpress Flag-S1P1/5 proteins. Flow cytometry and quantification of NK cells treated overnight with 1 µM ceralifimod and 1 µM S1P5i (H) or 100 ng/mL CXCL12 and 10 µM plerixafor (I) regimens. (J) Detection by flow cytometry of LAMP1⁺ cells within the S1PF⁻ and S1PF⁺ resting NK cell subsets. T-student test was performed to assess statistical significance. Immunoblots show data for a representative from a total of three. For immunoblots, images were cropped to show the approximate expected molecular size of the protein. For all charts, every data point corresponds to the average value in each independent donor and bars the mean ± SD. One-way ANOVA statistical tests were used in systems with several conditions for a single variable (F, H, I). For these, Dunnett’s test was used to perform multiple comparisons. Symbols represent ns = not significant, *p-value < 0.05, ***p-value < 0.001.

Fig. 6

Activation of CXCR4 and S1P1/5 drive upregulation of one another and maximize defects in NK cell degranulation. A Correlation CXCR4 and S1PR5 expression in NK cells used for RNA-seq. Percentage of CXCR4⁺ cells in S1PF⁻ and S1PF⁺ NK cell subsets at the day of isolation (B) or after 48 h treatment with IL-2, TGF-β or vehicle (C). D Western blot of NK cells subjected to 100 ng/mL CXCL12 for 15 min or overnight culture. E Detection of S1PF binding in rested and CXCL12-treated NK cells. F Immunoblot of NK cells treated with 10 ng/mL IL-2 for 48 h different periods of 1 µM ceralifimod exposure. G Detection of CXCR4 in rested and ceralifimod-treated NK cells. H NK cell degranulation measured by LAMP1^hi detection after treatments with CXCL12, ceralifimod or ceralifimod + CXCL12 followed by stimulation with HCT-116 for 2 h. Correlation between gene expression of the pairs: CXCL12 + SPHK1 (I), CXCL12 + S1PR5 (J) and CXCR4 + SPHK1 (K) in normal tissue and tumor biopsies from patients suffering from colon adenocarcinoma. Values were obtained from The Cancer Genome Atlas (TCGA) and plotted using GEPIA 2. Linear correlation was measured using the Pearson coefficient. L Schematic model showing the mechanism of CXCR4 and S1P1/5 desensitization of NK cells (created with BioRender.com). Immunoblots show data for a representative donor from a total of three. For immunoblots, images were cropped to cover the expected molecular size of the protein. For all charts, every data point corresponds to the average value in each independent donor and bars the mean ± SD. T student tests were used for comparisons between two conditions (C, E, G). One-way ANOVA statistical test was used in systems with several conditions for a single variable (H) and Two-way ANOVA was used when two variables were assessed (C). For both, Šídák’s test was used to examine significance in multiple comparisons. Symbols represent ns = not significant, * p-value < 0.05, ****p-value < 0.0001.