Tumor-derived G-CSF induces an immunosuppressive microenvironment in an osteosarcoma model, reducing response to CAR.GD2 T-cells

Abstract

Sarcomas are rare, mesenchymal tumors, representing about 10-15% of all childhood cancers. GD2 is a suitable target for chimeric antigen receptor (CAR) T-cell therapy due to its overexpression in several solid tumors. In this preclinical study, we investigated the potential use of iCasp9.2A.GD2.CAR-CD28.4-1BBζ (CAR.GD2) T-cells as a treatment option for patients who have GD2-positive sarcomas and we sought to identify factors shaping hostile tumor microenvironment in this setting. GD2 expression was evaluated by flow-cytometry on primary tumor biopsies of pediatric sarcoma patients. GD2 expression in sarcoma cells was also evaluated in response to an enhancer of zeste homolog 2 (EZH2) inhibitor (Tazemetostat). The antitumor activity of CAR.GD2 T-cells was evaluated both in vitro and in vivo preclinical models of orthotopic and/or metastatic soft-tissue and bone sarcomas. GD2 expression was detected in 55% of the primary tumors. Notably, the Osteosarcoma and Alveolar Rhabdomyosarcomas subtypes exhibited the highest GD2 expression levels, while Ewing sarcoma showed the lowest. CAR.GD2 T-cells show a significant tumor control both in vitro and in vivo models of GD2-expressing tumors. Pretreatment with an EZH2 inhibitor (Tazemetostat) upregulating GD2 expression, sensitizes GD2^dim sarcoma cells to CAR.GD2 T-cells cytotoxic activity. Moreover, in mouse models of disseminated Rhabdomyosarcomas and orthotopic Osteosarcoma, CAR.GD2 T-cells showed both a vigorous anti-tumor activity and long-term persistence as compared to un-transduced T-cells. The presence of immunosuppressive murine myeloid-derived suppressor (MDSC) cells significantly reduces long-term anti-tumour activity of infused CAR.GD2 T-cells. Tumor-derived G-CSF was found to be one of the key factors driving expansion of immunosuppressive murine and human MDSC, thus indirectly limiting the efficacy of CAR.GD2 T-cells. Our preclinical data strongly suggest that CAR.GD2 T-cells hold promise as a potential therapeutic option for the treatment of patients with GD2-positive sarcomas. Strategies to tackle hostile immunosuppressive MDSC are desirable to optimize CAR.GD2 T-cell activity.

Keywords: CAR.GD2 T-cells; CXCL8; Chimeric antigen receptor (CAR); EZH2 inhibitor; G-CSF; GD2; ICasp9.2A.GD2.CAR-CD28.4–1BBζ; Myeloid-derived suppressor cells; Sarcoma.

Fig. 1

GD2-synthase (B4GALNT1) mRNA gene expression profile in solid tumours. A Box plot depicting B4GALNT1 expression across cancers and healthy tissues. Data of TGCA and GTEX datasets showed expression level of B4GALNT1 in both normal tissues (green) and tumor patients with solid tumours (red). B B4GALNT1 expression on paediatric sarcoma cancer (Oncogenomics dataset—probe 206435_at) compared to Healthy Tissue (HT) control. C Kaplan–Meier analysis for overall survival in OS (TARGET dataset), RMS (ITCC dataset) and for Ewing’s sarcoma (EWS) (Dirksen dataset) for B4GALNT1 mRNA expression. The percentage of GD2 positive cells D and GD2 MFI E on paediatric sarcoma tissues, evaluated by FACS analysis, compared to PBMC negative control. Two-way Anova was used as multi comparison test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p-value = < 0.0001

Fig. 2

In vitro anti-tumor activity of CAR.GD2 T-cells correlates to GD2 expression on human sarcoma cell lines. (A, top panel) Expression of GD2 in human osteosarcoma (OS) cell lines (143B, MG-63, U-2OS, HOS and SAOS-2); (B, top panel) in embryonal rhabdomyosarcoma (ERMS) cell lines (RD, CT10); (C, top panel) in ARMS cell lines (SCMC, RH4, RH41 and RH30); (D, top panel) in Ewing Sarcoma cell lines (SK-ES-1, A-673) as assessed by flow cytometry. The Isotype control was run for each tumor cell line (white). Five-day co-cultures were performed in 6 to 16 independent experiments, in which GFP + OS (A, bottom panel), ERMS (B, bottom panel) and ARMS (C, bottom panel), Ewing (D, bottom panel) cell lines were co-cultured with NT T or CAR.GD2 T-cells at the effector-target (E: T) ratio of 1:1. Paired t-Test was used as comparison test. *p < 0.05, **p < 0.01, ****p-value = < 0.0001

Fig. 3

Long-term tumor control of CAR.GD2 T-cells in an in vitro co-culture assay of RMS and OS models under "stressed" conditions. A The panel shows the experimental design of “stressed” co-culture. Residual GFP + RD cells B and CD3 positive cells C were quantified during the “stressed” co-culture with tumor addition every 5 days. D Percentage of CAR + T-cells was evaluated during the “stressed” co-culture by FACS analysis. In particular, data are shown at Day 0 (scattered bar histogram), referring to CAR + T cell percentage before co-culture with tumor cells, and at Day + 5 (white bars), Day + 10 (light grey bars), Day + 15 (dark grey bars) and Day + 20 (black bars) after tumor addiction. E The graph shows the modulation of the distribution of CD4 + and CD8 + subpopulations of CAR.GD2 T-cells before (day 0) and after RD tumor cell addition at day + 5, + 10, + 15 and + 20. Stacked bar graphs show the analysis of T Naïve (striped white bars), TCM (dark grey bars), TEM (light grey bars) and TEMRA (black bars) CD3 + subsets in the long-term “stressed” co-culture with NT T-cells F or CAR.GD2 T-cells G. The panel H shows the experimental design of “stressed” co-culture. Residual GFP + 143B cells I and CD3 positive cells J were quantified during the “stressed” co-culture with tumor addition every 5 days. K Percentage of CAR + T-cells expression was evaluated during the “stressed” co-culture by FACS analysis. In particular, data are shown at Day 0 (scattered bar histogram), referring to CAR + T cell percentage before co-culture with tumor cells, and at Day + 5 (white bars), Day + 10 (light grey bars), Day + 15 (dark grey bars) and Day + 20 (black bars) after tumor addiction. L The graph shows the modulation of the distribution of CD4 + and CD8 + subpopulations of CAR.GD2 T-cells before (day 0) and after RD tumor cell addition at day + 5, + 10, + 15 and + 20. Stacked bar graphs show the analysis of T Naïve (striped white bars), TCM (dark grey bars), TEM (light grey bars) and TEMRA (black bars) CD3 + subsets in the long-term “stressed” co-culture with NT T-cells M or CAR.GD2 T-cells N. Data from four healthy donors (HDs) are expressed as mean ± SED. T-test was applied for the analysis. *p-value ≤ 0.05; **p-value ≤ 0.01; ***p-value ≤ 0.001 and **** ≤ 0.0001.

Fig. 4

In vivo experiments of NSG mice bearing GD2 + sarcoma cells (RD, 143B or U-2OS) treated with NT or CAR.GD2 T-cells. Experimental design (Figure created using Biorender—https://biorender.com) and in vivo bioluminescence imaging time-course of RD-GFP-FF-Luc tumor cells (A), 143B-GFP-FF-Luc (F) and U-2OS-GFP-FF-Luc (K) infused in NSG mice. At the time of tumor engraftment, 10⁶ effector T-cells were administered through i.v. injection. The graph shows bioluminescence analysis of RD (ERMS) (B, C), 143B (OS) (G, H) and U-2OS (OS) (L, M) tumor-bearing mice model treated with NT T-cells (square symbol) or CAR.GD2 T-cells (dot symbol). Overall-Survival (OS) of RD (ERMS) (D) and 143B (OS) (I) tumor-bearing NSG mice treated with NT T-cells (black line) or CAR.GD2 T-cells (blue and red respectively). Two-way ANOVA was applied for the analysis. *p-value ≤ 0.05; **p-value ≤ 0.01; ***p-value ≤ 0.001 and **** ≤ 0.0001. (N) Explanted tumor-volume mass on sacrifice for NT T-cells and GD2.CAR T-cells mice group in U-2OS (OS) tumor model. Two-way ANOVA was applied for the analysis. *p-value ≤ 0.05; **p-value ≤ 0.01; *** p-value ≤ 0.001 and **** ≤ 0.0001. Average of circulating human T-cells evaluated as % of CD45 + CD3 + (NT T-cells, square symbol) and CD3 + CAR + (CAR.GD2 T-cells dot symbol) until NSG-tumor bearing sacrifice for RD (ERMS) (E), 143B (OS) (J) and U2-OS (O) model. T-test and Two-way ANOVA was applied for the analysis. *p-value ≤ 0.05; **p-value ≤ 0.01; ***p-value ≤ 0.001 and **** ≤ 0.0001

Fig. 5

Only the human OS 143B line, when implanted in NSG mice, induces the expansion of murine MDSCs. A NSG mice were engrafted with six different human sarcoma cell lines. Tumor growth was evaluated by In vivo imaging system (IVIS) for five different orthotopically models (200′000 cell/mice) of five OS cell lines (143B, U-2OS, SAOS-2, MG-63, HOS) and one ERMS cell line (RD). No tumor bearing mice have been used as control group. B Weekly plasma samples were collected from each tumor-bearing mouse and control mice, to investigate circulating mMDSCs using FACS flow-cytometry. The analysis looked for both monocytic MDSCs B and polymorphonuclear (PMN) MDSCs (C) cell subset of CD11b + cells. To ensure accurate comparison and data normalization across samples, counting beads were utilized. *p-value = < 0.05; **p-value = < 0.01; ***p-value = < 0.001; ****p-value = < 0.0001, t test. D-E Quantitative analysis of multiple MSDC ligands was performed using the Luminex assay. Supernatant from tumor cultures (143B, MG-63, U-2OS, HOS, SAOS-2 and RD) D and plasma samples of tumor bearing mice (143B, MG-63, U-2OS,, HOS, SAOS-2 and RD) E were explored using a predesigned panel assay, at 72 h of culture and at time of mice sacrifice respectively. Quantification of the specified cytokines, including CCL2, CCL3, CCL5, CCL7, CX3CL1, CXCL1, CXCL2, CXCL4, CXCL5, CXCL6, CXCL8 (IL8), CXCL13, CXCL14, G-CSF (CSF3), GM-CSF, IFN-gamma, IL10, IL13, IL1b, IL5, MCSF, and TNF-alpha, was normalized on blank control background for the in vitro data (in violet) and no tumor sample for the in vivo data (in blu), and values reported in a Heatmap graph. Data are represented as mean ± standard deviation (SD) from n = 3. F The efficiency of CSF3 and CXCL8 KO in 143B cell lines was assessed by ELLA assay on 72 h cell lines supernatants. (G) Comparative Analysis of Crisped 143B OS Cell Lines for CXCL8 and G-CSF3 (CSF3). The cell lines 143B (wild-type control, orange line), 143B CXCL8^−/− (143B CXCL8 knockout, red line), and 143B G-CSF^−/− (143B G-CSF knockout, blue line), where analyzed for GD2 expression (percentage and GD2 MFI) by flow-cytometry and normalized on the Isotype Control. H The 143B wild-type control (dark grey), 143B CXCL8^−/− (white), 143B G-CSF^−/− (light gray) cells were co-cultured for 5 days with NT or GD2.CAR T-cells at E: T = 1:1 ratio. At the endpoint the percentage of residual tumor was evaluated by FACS as measure of GFP tumor expression. Data are expressed as mean ± SD from 3 donor. Statistical significance was determined using ANOVA test with *p < 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001

Fig. 6

Study of tumor secretome profile and MDSC investigation on sarcoma cell lines. A NSG mice were engrafted with three different human sarcoma cell lines (WT, Crisped 14B OS Cell Lines for CXCL8 and G-CSF3). B Tumor growth was evaluated by In vivo imaging system (IVIS) for all three different orthotopically models (200′000 cell/mice). No tumor bearing mice have been used as control group. Weekly plasma samples were collected from each tumor-bearing mouse and control mice, to investigate circulating mMDSCs using FACS flow-cytometry. The analysis evaluated both monocytic MDSCs C and polymorphonuclear MDSCs D cell subset of CD11b + cells. To ensure accurate comparison and data normalization across samples, counting beads were utilized. *p-value = < 0.05; **p-value = < 0.01; ***p-value = < 0.001; ****p-value = < 0.0001, t test. E Quantification of the specified cytokines, including CCL2, CCL3, CCL5, CCL7, CX3CL1, CXCL1, CXCL2, CXCL4, CXCL5, CXCL6, CXCL8 (IL8), CXCL13, CXCL14, G-CSF (CSF3), GM-CSF, IFN-gamma, IL10, IL13, IL1b, IL5, MCSF, and TNF-alpha, was normalized on blank control background for the in vivo data, and single values reported in a Heatmap graph. Two-way Anova was used as multi comparison test. *p < 0.05, **p < 0.01, ***p < 0.001

Fig. 7

OS 143B cells induce human Mo-MDSC expansion. Schematic representation of PBMC co-cultured with OS tumor cells (A) or with OS conditioned medium (CM) for 72 h (B). The absolute number of PMN-MDSC (left side panel), or Mo-MDSC (right side panel) before (day 0) and after three days of cell–cell contact (C) and cell-free experiments (D) with medium and OS cell lines. Results are shown as box and whiskers graphs showing all points. Friedman matched paired test was applied to compare different cell culture conditions, while Wilcoxon matched-pairs signed-rank test was used for group’s comparison with respect to the absolute number quantified before culture (day0); p values are indicated in the graphs *#: p < 0.05. (E) Co-culture assay with all conditioned medium of OS/PBMC culture was real-time monitored by Incucyte live cell system imaging for five days to assess the tumor clearance by CAR.GD2 T-cells. Total green Confluence associated to the OS tumor cell lines was quantified overtime as a measure of residual tumor. Figure created using Biorender—https://biorender.com. Two-way ANOVA was applied for the analysis. *p-value ≤ 0.05; **p-value ≤ 0.01; ***p-value ≤ 0.001 and **** ≤ 0.0001