No effective therapy is available for unresectable soft-tissue sarcomas (STS). This unmet clinical need prompted us to test whether chondroitin sulfate proteoglycan 4 (CSPG4)-specific chimeric antigen receptor (CAR)-redirected cytokine-induced killer lymphocytes (CAR.CIK) are effective in eliminating tumor cells derived from multiple STS histotypes in vitro and in immunodeficient mice.
The experimental platform included patient-derived CAR.CIK and cell lines established from multiple STS histotypes. CAR.CIK were transduced with a retroviral vector encoding second-generation CSPG4-specific CAR (CSPG4-CAR) with 4-1BB costimulation. The functional activity of CSPG4-CAR.CIK was explored in vitro, in two- and three-dimensional STS cultures, and in three in vivo STS xenograft models.
CSPG4-CAR.CIK were efficiently generated from patients with STS. CSPG4 was highly expressed in multiple STS histotypes by in silico analysis and on all 16 STS cell lines tested by flow cytometry. CSPG4-CAR.CIK displayed superior in vitro cytolytic activity against multiple STS histotypes as compared with paired unmodified control CIK. CSPG4-CAR.CIK also showed strong antitumor activity against STS spheroids; this effect was associated with tumor recruitment, infiltration, and matrix penetration. CSPG4-CAR.CIK significantly delayed or reversed tumor growth in vivo in three STS xenograft models (leiomyosarcoma, undifferentiated pleomorphic sarcoma, and fibrosarcoma). Tumor growth inhibition persisted for up to 2 weeks following the last administration of CSPG4-CAR.CIK.
This study has shown that CSPG4-CAR.CIK effectively targets multiple STS histotypes in vitro and in immunodeficient mice. These results provide a strong rationale to translate the novel strategy we have developed into a clinical setting.
Our study has shown that cytokine-induced killer lymphocytes (CIK) engineered with a chondroitin sulfate proteoglycan 4 (CSPG4)-specific chimeric antigen receptor (CAR; CSPG4-CAR) are effective in eliminating many types of soft-tissue sarcoma (STS)-derived cells both in vitro and in immunodeficient mice. These preclinical results provide a strong rationale for the clinical translation of the CSPG4-CAR.CIK–based immunotherapeutic strategy, which we have developed. Patients with unresectable high-grade STS, who respond poorly to checkpoint inhibitor-based immunotherapy may greatly benefit from this novel immunotherapy.
Soft-tissue sarcomas (STS) are rare tumors of mesenchymal origin that affect both children and adults (1). These peculiar tumors encompass multiple histotypes, which are characterized by extremely variable biological and clinical behaviors. Chemotherapy and molecularly targeted approaches may offer transient disease control to patients ineligible for radical surgical resection. However, the overall prognosis remains dismal with a 5-year survival rate of less than 25% (1–3). STS are also poorly responsive to checkpoint blockade–based immunotherapy (3–6). This unmet clinical need prompted us to develop a novel strategy for the treatment of multiple STS histotypes (7).
Adoptive immunotherapy with T lymphocytes redirected by tumor antigen (TA)-specific chimeric antigen receptors (CAR) is one of the most effective therapies in B-cell malignancies (8–10). However, application of CAR.T cells to solid tumors remains challenging (11–15). Here, we have tested the ability of cytokine-induced killer lymphocytes (CIK) engineered with a TA-specific CAR to target tumor cells obtained from multiple STS histotypes in vitro and in vivo. CIK are patient-derived polyclonal T natural killer (NK) lymphocytes endowed with HLA class I–independent antitumor activity, mediated mostly by the interaction of their NKG2D receptor with stress-inducible targets (MIC A/B and ULBPs 1–6) on tumor cells (16–26). To increase their tumor cell specificity, CIK were engineered with a TA-specific CAR. CAR-engineered CIK have been modeled in hematologic malignancies to target either CD19 or CD33/CD23 antigens (27–31), and clinical trials are currently ongoing. We and others studied the use of CAR-expressing CIK in STS (32–35). In our previous study, we demonstrated the preclinical effectiveness of CAR-redirected CIK against CD44v6, which is expressed in about 40% of STS (32). Here, we assessed whether alternative TA could be targeted via CAR-engineered immune cells in STS to further expand their clinical applicability.
The TA, chondroitin sulfate proteoglycan 4 (CSPG4), a cell surface proteoglycan that plays an important role in oncogenic pathways involved in cancer progression and metastatic spread (36–42), was selected as the target for the following reasons. CSPG4 is highly expressed with limited heterogeneity on both differentiated cancer cells and cancer-initiating cells (CIC) in several types of cancer. According to the cancer stem cell hypothesis, CIC play a major role in disease recurrence and metastatic spread, the two major causes of patients' mortality and morbidity. In contrast, CSPG4 is not detectable in normal tissues. Furthermore, CSPG4 is expressed by activated pericytes in the tumor microenvironment (43). As a result, CSPG4 immunotargeting selectively inhibits neoangiogenesis in the tumor microenvironment, thus contributing to the elimination of the vasculature that supports tumor growth (44).
This article describes the in vitro and in vivo ability of CSPG4-specific CAR-redirected CIK (CSPG4-CAR.CIK) to eliminate STS cells following a description of CSPG4 expression on multiple STS cells.
Materials and Methods
Data analysis of CSPG4 RNA expression in The Cancer Genome Atlas
RNA-sequencing expression data were selected and downloaded from the cBioPortal, The Cancer Genome Atlas (TCGA) PanCancer collections (45, 46). The dataset included 251 STS samples: leiomyosarcoma, n = 99; dedifferentiated liposarcoma, n = 58; undifferentiated pleomorphic sarcoma (UPS)/malignant fibrous histiocytoma/high-grade spindle cell sarcoma, n = 50; myxofibrosarcoma, n = 25; malignant peripheral nerve sheath tumor (MPNST), n = 9; and synovial sarcoma n = 10. Another 336 melanomas served as a positive expression control and various epithelial tumors (breast cancer, n = 1,082; pancreatic cancer, n = 176; lung adenocarcinoma, n = 510; and lung squamous cell carcinoma, n = 482) were explored for comparison. RSEM expression values were plotted after log2 transformation with 0.5 jittering on the x-axis using Microsoft Excel.
STS cell lines and STS spheroids
STS cell lines were generated in our laboratory from patient-derived surgical biopsies (47). We received approval for collection of patient samples and the associated informed consent document from the institutional review board (IRB) per Declaration of Helsinki guidelines (protocol no., 225/2015); each patient signed an informed consent. Patient-derived STS cell lines were cultured in either KO DMEM F12 (KO Out DMEM, Gibco BRL) or Iscove's Modified Dulbecco Medium (Sigma-Aldrich), with 10% or 15% FBS, 25 mmol/L HEPES, 100 U/mL penicillin, and 100 U/mL streptomycin (Gibco BRL) in a humidified 5% CO2 incubator at 37°C. Patient-derived melanoma cell line, M14 (48), which does not express CSPG4, was used as a specificity control and cultured in RPMI1640 Medium (Sigma-Aldrich), supplemented with 10% heat-inactivated FBS, 100 U/mL penicillin, and 100 U/mL streptomycin (Gibco BRL) at 37°C in a 5% CO2 incubator. The HT1080 cell line used in this study was originally obtained from the ATCC, and was authenticated by genotype analysis with the Cell ID System (Promega) that compared its profile with those published on the DMSZ database. Adult and neonatal keratinocytes were cultured with the Lonza KGM Gold Keratinocyte Growth Medium Bullet Kit.
Three-dimensional (3D) STS spheroids were generated as a single spheroid per well using ultra-low attachment (ULA) 96-well round bottom plates (Corning) with no additional coating. An STS cell suspension of 500–5,000 cells/100 μL was plated into wells and then centrifuged at 1,000 × g for 10 minutes (33). STS spheroids were assembled in 1–4 days, depending on the target histotypes. We generated GFP+ STS spheroids from cells previously transduced with the pRRL.sin.PPT.hOct4.eGFP.Wpre VSV-G pseudotyped third-generation lentiviral vector.
Generation of CSPG4-CAR.CIK
Supernatants containing retroviral particles encoding CAR specific for the CSPG4 antigen (CSPG4-CAR) or the control vector encoding CAR specific for the CD19 antigen (CD19-CAR), both containing 4-1BB costimulatory endodomains, were generated as described previously (41). We generated CSPG4-CAR.CIK and CSPG4-CAR.T cells from peripheral blood mononuclear cells (PBMC) isolated from patients diagnosed with STS by density gradient centrifugation using Lymphosep (Aurogene). Approval was obtained from the IRB per the Declaration of Helsinki guidelines for the collection of biological samples (tumors and blood) and for patient informed consent releases (protocol no., 225/2015). For CAR.CIK, PBMCs from 8 patients with STS (Supplementary Table S1) were seeded on day 0 in cell culture flasks at a concentration of 2 × 106 cells/mL with IFNγ (Miltenyi Biotec; 1,000 U/mL) in RPMI1640 Medium (Gibco BRL), supplemented with 10% FBS (Sigma), 100 U/mL penicillin, and 100 U/mL streptomycin (Gibco BRL). Following a 24-hour incubation at 37°C, PBMCs were activated by anti-biotin MACSiBead particles loaded with anti-CD2, -CD3, and -CD28 mAbs (Miltenyi Biotec) and human IL2 IS (Miltenyi Biotec, 300 U/mL). To generate CAR.T cells, PBMCs were seeded on day 0 at a concentration of 2 × 106 cells/mL and activated using anti-biotin MACSiBead particles. On day +1, human IL2 IS (Miltenyi Biotec, 50 U/mL) was added. On day +2, PBMCs were transduced with 0.5 mL of retroviral supernatants in retroNectin-coated plates by overnight incubation. Unmodified not transduced (NTD) NTD.CIK and NTD.T cells were used as a paired control. Both CAR.CIK and control NTD.CIK were expanded over 4 weeks, refreshed with IL2 medium (CIK, 300 U/mL) every 2–3 days as needed, and cultured at 1.8 × 106 cells/mL. CAR.T cells were cultured at 1.8 × 106 cells/mL for 1 week, and the IL2 (50 U/mL) medium was refreshed every 2–3 days as needed.
Conjugated CD3, CD4, CD8, CD56, PD-1, CXCR3, CXCR4, and CCR7 mAbs (BD Pharmingen) and CD45RO, CD45RA, and CD62L mAbs (Miltenyi Biotec) were used to characterize lymphocytes. A mAb specific for the IgG1/CH2CH3 Spacer (Jackson ImmunoResearch) was used to detect CAR expression. STS cells were stained with conjugated mAbs for the expression of CIK NKG2D ligands, MIC A/B (BD Pharmingen) and ULBPs (R&D Systems), and for the expression of HLA-ABC, PD-L1, and PD-L2 (BD Pharmingen). STS were stained for CSPG4 with mAbs 225.28, 763.74, and D2.8.5-C4B8 (49), which recognize distinct and spatially distant epitopes of CSPG4. Cells were first incubated with CSPG4-specific mAbs (1 μg/mL for all mAbs), then washed, and incubated with rabbit anti-mouse IgG-PE Secondary Antibody (Miltenyi Biotec). Alternatively, CSPG4 was detected with the conjugated anti-human CSPG4-APC mAb (Miltenyi Biotec). CSPG4 molecules expressed on the surface of STS and other cell lines were measured using a quantitative Immunofluorescence Assay (Bangs Laboratories, Inc.). Briefly, cells of interest and calibration beads with increasing amounts of antibody capture capability were labeled simultaneously with the anti-human CSPG4-APC mAb. Labeled cells and calibration beads were analyzed, and a standard regression line was calculated between fluorescence intensity and antigen density, expressed as antibody-binding capacity in molecules per cell. We defined high CSPG4 expression as 2-fold increase as compared with normal keratinocytes. Labeled cells were acquired on FACS Cyan (Cyan ADP, Beckman Coulter SRL) and analyzed using Summit Software.
Tumor cell killing assays
We assessed the tumor-killing ability of patient-derived CSPG4-CAR.CIK and unmodified NTD.CIK in vitro against STS cell monolayers and STS 3D spheroids. In two cases, CIK and STS cell cultures were generated from samples collected from the same patient (S1 and S172), while in all other cases, cytotoxicity assays were performed with HLA-mismatched effector cells. Cytotoxicity assays against STS cell monolayers were performed using flow cytometry or a bioluminescence cell viability assay. In the first case, target cells were stained with either vital dye PKH26 (Sigma-Aldrich) or 5,6-carboxyfluorescein diacetate succinimidyl ester (CFSE; Molecular Probes), according to the manufacturer's protocols. Immune-mediated killing was analyzed by Flow Cytometry (Cyan ADP, Dako) and measured by the DAPI permeability of target cells (PKH26+ or CFSE+ gate). For the bioluminescence method, cytotoxicity was measured with the CellTiter-Glo Luminescent Cell Viability Assay (Promega), in which the number of viable and metabolically active target cells was evaluated by quantifying the ATP in culture. CIK cells were cocultured at different effector to target cell (E:T) ratios (10:1, 5:1, 2.5:1, 1:1, 1:2, and 1:4) in cytotoxicity assays (300 U/mL IL2 medium at 37°C and 5% CO2) for 5 hours (short-term assay) and 48 hours (long-term assay). In selected experiments, we tested the cytotoxic activity at very low E:T ratio (1:8, 1:16, 1:32, and 1:64). Target cells were also tested separately from CIK cells as control to assess their spontaneous mortality. The percentage of STS-specific lysis for each E:T ratio was calculated using the following formula: [(experimental − spontaneous mortality/100 − spontaneous mortality) × 100]. In selected experiments, growth of residual sarcoma cells was investigated 48 hours after the treatment with either CSPG4-CAR.CIK or NTD.CIK.
In 3D assays, STS spheroids stably expressing GFP were seeded one per well in ULA 96-well round bottom plates. CSPG4-CAR.CIK and unmodified NTD.CIK were stained with PKH26 dye and plated at E:T ratio 2:1 in culture medium with 300 U/mL IL2 at 37°C 5% CO2. Fluorescence images were acquired at 12-hour intervals over 96 hours under the same magnification (10×). Killing activity was determined as loss of GFP fluorescence spheroid area (pixel) using fluorescence microscopy (Leica DMI 3000B equipped with Photometrics CoolSnap HQ CCD Camera). Untreated STS spheroids were used to evaluate spontaneous mortality. All images were analyzed with ImageJ software, and percentage of tumor cell lysis was determined by the formula: [100 − treated GFP+ STS spheroid (pixel) × 100/untreated GFP+ STS spheroid (pixel)].
Immunofluorescence analysis of CSPG4-CAR.CIK cell recruitment and infiltration in STS spheroids
GFP+ STS spheroids were cocultured with CSPG4-CAR.CIK or unmodified NTD.CIK cells stained with red dye PKH26 at E:T ratio 2:1 in culture medium (300 U/mL IL2 at 37°C, 5% CO2). Following a 16-hour coculture at 37°C, CIK cells were removed and immunofluorescence acquisition was conducted on the remaining spheroids. Briefly, STS spheroids were washed twice, centrifuged at 300 × g for 3 minutes in PBS, fixed in 4% paraformaldehyde for 1 hour, resuspended with mounting medium, and applied on either glass slides or glass bottom chamber slide wells. STS spheroids were observed using a Leica SP8 AOBS confocal microscope. Next, 80 MHz pulsed white light laser (470–670 nm) was used to excite the fluorochromes in the spheroids. Fluorescence channels were scanned sequentially, and hybrid Spectral Detectors (HyD SP Leica Microsystems) revealed the emissions. Image acquisition of the STS spheroids was performed maintaining the same laser power, gain, offset, and magnification (20×). We generated maximum intensity projections for each analyzed spheroid with LAS X Software (Leica) to quantify CIK cell recruitment and infiltration. Images of the total PKH26 red fluorescence area (μm2) present either at the boundary or inside the spheroid were analyzed using ImageJ software.
CIK-cell penetration capability into Matrigel matrix
STS spheroids were collected, washed in PBS, resuspended in 20 μL of liquefied Matrigel (BD Pharmingen) at 4°C, and then plated as droplets in well centers of a 24-well tissue culture plate that had been prewarmed to 37°C. Plates were incubated at 37°C and 5% CO2 for 15 minutes to allow the solidification of the Matrigel domes. Domes were then overlaid with 500 μL of prewarmed medium with 300 U/mL IL2 and cocultured with PKH2-stained CSPG4-CAR.CIK or unmodified NTD.CIK cells (50,000 cells/well) for 5 days. Empty domes were used as controls. At the end of the coculture period, each well was washed twice with prewarmed PBS to eliminate any effector cells outside the domes. Fluorescence microscopy (Leica DMI 3000B with Photometrics CoolSnap HQ equipped with CCD camera) was used to visualize CSPG4-CAR.CIK or unmodified NTD.CIK cell migration at the Matrigel boundary and cells that penetrated into the Matrigel domes. Analysis of the PKH26 red fluorescence dye presence (μm2) was performed with ImageJ software.
In vitro cytokine production
CSPG4-CAR.CIK and CSPG4-CAR.T cells, and unmanipulated NTD.CIK and NTD.T cells were cocultured alone or with tumor cells in RPMI1640 medium with 300 U/mL (CIK) or 50 U/mL IL2 (T cells) at a 2:1 E:T ratio and incubated at 37°C for 48 hours. Concentrations of cytokines in culture supernatant were measured using the Bio-Plex Pro Human Cytokine 9-plex Assay Kit (Bio-Rad Laboratories Inc.) according to the manufacturer's instructions. Each sample was measured in duplicate. Data were acquired and analyzed by Bioclarma (Analysis Service). Granzyme B concentration was measured in supernatants from mixed target/effector cell cultures (ELISA Granzyme B Kit, Diaclone SAS), as recommended by the manufacturer.
In vivo activity of patient-derived CSPG4-CAR.CIK
The antitumor activity of CSPG4-CAR.CIK and unmodified NTD.CIK was evaluated using STS xenograft models in immunodeficient mice. In vivo experiments received approval by the competent committee and internal review board (auth. no., 178/2015-PR). STS xenografts were established in 7- to 8-week-old NOD/SCID/γc−/− (NSG) or NOD/SCID (Charles River Laboratories, SRL) female mice by subcutaneous injection with 1 × 106 cells obtained from three STS [fibrosarcoma (HT1080), leiomyosarcoma (S172), and UPS (S1)]. Autologous CSPG4-CAR.CIK and unmodified NTD.CIK were available for S172 and S1 xenografts. Allogeneic CSPG4-CAR.CIK from unrelated patients with STS were generated following the same protocol used to generate autologous CIK identical and used in the HT1080 xenograft model. When tumors were approximatively 50 mm3 in volume, mice were infused twice a week with 1 × 106 CSPG4-CAR.CIK or unmodified NTD.CIK resuspended in PBS (200 μL), for a total of four infusions. Mice injected with PBS only were used as controls. Treatment and control cohorts included 6 mice each for the leiomyosarcoma (S172) group. For the fibrosarcoma (HT1080) and UPS (S1) xenografts and control groups, each group included 3 mice. In the experiments with fibrosarcoma (HT1080) and leiomyosarcoma (S172), mice were sacrificed at the end of treatment. In the experiment with UPS (S1), mice were sacrificed 14 days after the last CIK infusion. Mice were monitored daily for possible toxicities, while tumor growth was measured weekly with manual caliper. Mice were sacrificed at the end of treatment or if tumor reached 2 cm along the main diameter. In additional experiments, mice engrafted with the S172 cell line (n = 4 mice per treatment) or the HT1080 cell line (n = 5 mice per treatment) were infused with CSPG4-CAR.CIK and unmodified NTD.CIK generated from unrelated donors. In these models, when the tumor volume was approximatively 20 mm3, mice received two infusions (on days 0 and +4) of 3 × 106 CSPG4-CAR.CIK or unmodified NTD.CIK. Mice were sacrificed 11 days after the last dose of cells. Tumor volume was calculated by the following formula: V = 4/3 × π × (a/2)2 × (b/2), where a is the length and b is the width of the tumor.
STS xenografts were analyzed by IHC. Samples (5-μm thick) were cut from formalin-fixed, paraffin-embedded tissue sections, mounted on slides, and treated as per standard IHC procedures. Tissue sections were deparaffinized with 100% xylene and rehydrated with decreasing concentrations of ethyl alcohol. Antigen retrieval was performed by boiling the sections in 1 mmol/L EDTA (pH 9.0) for 60 minutes. Slides were treated with 3% hydrogen peroxide, 1% BSA (Invitrogen), and 5% normal horse serum in TBS [25 mmol/L Tris (pH 7.4) and 150 mmol/L NaCl] containing 0.1% Tween 20 (Sigma-Aldrich, Inc). To confirm CSPG4 expression, slides were incubated in a closed humid chamber overnight at 4°C with the CSPG4-specific mAb 225.28, 763.74, and D2.8.5-C4B8 (4 μg/mL each) pool. After washing, a secondary anti-mouse IgG xenoantibody was added. Secondary antibodies conjugated to horseradish peroxidase (HRP) were generated and IHC signals were detected with the EnVision1 System-HRP (Dako North America, Inc) and chromogen Diaminobenzidine (DAB) Substrate (DakoCytomation Liquid DAB+ Substrate Chromogen System, Dako). Tissue sections were counterstained with Mayer Hematoxylin (Bio-Optica). We also explored the presence of infiltrated CSPG4-CAR.CIK, unmodified NTD.CIK cells, and apoptotic tumor cells in explanted tumors from treated mice. Tissues were stained according to the manufacturer's protocols with the primary polyclonal antibody anti-CD3 (DAKO) and anti-cleaved caspase 3 (Cell Signaling Technology). Tissue sections were mounted on glass slides and visualized with a DM750 Leica Microscope equipped with Leica ICC50W CCD Camera (LAS EZ3.4.0 software).
All experiments were performed at least twice. Data were analyzed using GraphPad Prism 8.0 (GraphPad Software). Descriptive data are presented as mean values ± SE. To find statistical significance in the comparison of two groups, we used two-tailed Student t tests; for comparison of three or more groups, the data were analyzed by two-way ANOVA with Bonferroni multiple comparison post hoc tests. A P < 0.05 was considered significant. Significance is represented on graphs as *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001.
Generation and characterization of patient-derived CSPG4-CAR.CIK
CIK were efficiently generated from PBMCs collected from 8 patients diagnosed with STS. We genetically engineered CIK to express the CSPG4-CAR that includes the 4-1BB costimulatory endodomain. Three weeks following transduction and ex vivo expansion, the mean expression of CAR on engineered CIK (CSPG4-CAR.CIK) was 48% ± 6% as assessed by flow cytometry (Supplementary Table S1; Supplementary Fig S1A) and was comparable with that obtained using the retroviral vector encoding the CD19-specific CAR (CD19-CAR). Ex vivo expansion of CSPG4-CAR.CIK was 154-fold (27–348) and was comparable with that observed with paired control unmodified CIK (NTD.CIK). The phenotypic characterization of CSPG4-CAR.CIK was comparable with that of paired unmodified NTD.CIK showing CD8+ cells as the main cellular subset (69% ± 4%), of which 39% ± 5% of them coexpressed the CD56 molecule (CD3+CD56+; Supplementary Table S1). The NKG2D receptor was expressed in 66% ± 5%, and 47% ± 0.1% of the cells were CD62L+CD45RA−; the latter represent the effector memory–like population (Supplementary Fig. S1B). The immune checkpoint receptor, PD-1, was expressed on 12% ± 2% in both CSPG4-CAR.CIK and NTD.CIK (Supplementary Table S1).
CSPG4 as a potential CAR target in STS
First, we confirmed that CSPG4 is expressed in multiple STS histotype cells (leiomyosarcoma; dedifferentiated liposarcoma; UPS, malignant fibrous histiocytoma, and high-grade spindle cell sarcoma; myxofibrosarcoma; MPNST; and synovial sarcoma) at levels similar to those found in melanoma cells (Fig. 1A) by in silico analysis with RNA-sequencing expression data from TCGA database. Furthermore, using flow cytometry, we observed CSPG4 expression on the cell surface of 15 of 15 STS cell lines obtained from biopsies of patients affected by different histotypes of advanced STS (UPS, n = 3; gastrointestinal stromal tumor, n = 5; liposarcoma, n = 4; leiomyosarcoma, n = 2; and MPNST, n = 1) and in the HT1080 cell line (fibrosarcoma) from the ATCC (Fig. 1B).
CSPG4 density on STS cell lines was quantified on per cell basis and found variable among STS samples (mean of 321 ± 47 molecules/cell; Fig. 1C). STS cell lines were also confirmed to express variable levels of the main known ligands recognized by the NKG2D CIK receptor (MIC A/B, 30% ± 8%; ULBP1, 1% ± 0.7%; ULBP2/5/6, 71% ± 6%; and ULBP3, 25% ± 9%). HLA class I antigens were highly expressed by all STS cell lines tested (95% ± 2%), along with varying levels of immune checkpoints, PD-L1 (32% ± 9%) and PD-L2 (65% ± 6%; Supplementary Table S2).
CSPG4-CAR.CIK effectively and specifically target STS in vitro
We explored the in vitro antitumor activity of CSPG4-CAR.CIK against 12 STS cell lines. In two cases (S1 UPS and S172 leiomyosarcoma), it was possible to obtain peripheral blood from the patient from whom the STS tumor cell line was generated. At E:T ratios from 10:1 to 1:4, CSPG4-CAR.CIK revealed significantly superior in vitro cytotoxicity against STS (Fig. 2A; Supplementary Fig. S2A), as compared with that reported with unmodified NTD.CIK (n = 29; P < 0.0001). Furthermore, the antitumor activity of CSPG4-CAR.CIK was maintained even at very low E:T ratios (from 1:8 to 1:64; n = 8; P < 0.0001; Fig. 2B).
CSPG4-CAR.CIK antitumor activity was superior to that of control CD19-CAR.CIK (n = 5; P < 0.05; Fig. 2C) and was CSPG4 specific because CSPG4-CAR.CIK did not eliminate non-expressing CSPG4 target cells (n = 3; P > 0.05; Fig. 2D). Finally, at E:T ratios that mediated potent antitumor effects, CSPG4-CAR.CIK had no detectable activity against human keratinocytes that show low CSPG4 expression (n = 5; Fig. 2E). CSPG4-CAR.CIK–mediated tumor elimination was strictly dependent on the CSPG4 expression level on tumor cells (n = 7; P < 0.05; Fig. 2F).
We also measured the antitumor activity of CSPG4-CAR.CIK in an STS sample expressing CSPG4, but lacking NKG2D ligands, and found that CAR expression, per se, promotes antitumor activity of redirected CIK independently from NKG2D CIK receptor engagement (P < 0.001; Fig. 2G).
Following treatment with CSPG4-CAR.CIK, we evaluated whether residual tumor cells spared by the cytotoxic effects of CIK may regrow. We observed significant delayed in vitro regrowth of STS cells exposed to CSPG4-CAR.CIK (48 hours after treatment) as compared with tumor cells exposed to unmodified NTD.CIK (n = 2; P < 0.05; Fig. 2H).
A significant and superior (3-fold, E:T 5:1) activity level of STS killing was displayed by CSPG4-CAR.CIK compared with CAR.T lymphocytes. Both effector cells were generated from PBMCs collected from the same patient and expressed comparable levels of CSPG4-CAR molecules (n = 2; P < 0.001; Fig. 2I; Supplementary Fig. S3A and S3B).
In selected experiments, we explored the production of Th1- and Th2-type cytokines and granzyme B by CSPG4-CAR.CIK at baseline and following exposure to STS CSPG4-positive targets. Overall, we observed higher baseline productions of IFNγ, IL6, IL1β, IL4, IL8, GM-CSF, TNFα, IL10, and granzyme B by CSPG4-CAR.CIK that markedly increased following engagement with STS CSPG4-positive targets (n = 4; Supplementary Fig. 4A). The highest cytokine peaks were observed for IFNγ, IL6, IL8, TNFα, and GM-CSF. The most differentially expressed cytokines by CSPG4-CAR.CIK in response to CSPG4-expressing STS as compared with NTD.CIK were: IFNγ (80-fold), IL1β (4.5-fold), IL6 (1.3-fold), TNFα (7.6-fold), GM-CSF (88.5-fold), IL10 (16-fold), IL8 (1.3-fold), IL4 (7-fold), and granzyme B (20-fold). Complete cytokine values, including CAR.T lymphocytes control, are reported in Supplementary Fig. S4B.
CSPG4-CAR.CIK effectively target STS cells in 3D spheroids
We developed the STS spheroid model that mimics tumor 3D and allows exploring CAR.CIK migration in a multidimensional structure. STS spheroids express GFP, which allows tracking their fate by longitudinal imaging. Spheroids were generated for three STS (S1, S5, and S172). Spheroids were coincubated with effector cells at E:T ratio 2:1, and loss of fluorescence over time was considered as a surrogate indication of tumor elimination by CSPG4-CAR.CIK (Fig. 3A and B). CSPG4-CAR.CIK eliminated STS spheroids more effectively than unmodified NTD.CIK (n = 3; P < 0.0001; Fig. 4A–C).
In selected experiments, we also used live imaging to visualize the antitumor kinetics of CSPG4-CAR.CIK against STS spheroids (Supplementary Videos S1 and S2). Furthermore, measurement of the maximum intensity projections of S172 and S5 spheroids after incubation with CSPG4-CAR.CIK or unmodified NTD.CIK indicated that CSPG4-CAR.CIK were present at a higher concentration within the STS spheroids as compared with unmodified NTD.CIK (n = 8; P < 0.05; Fig. 4D–F; Supplementary Videos S3–S6). Finally, in selected experiments, we measured the capability of CSPG4-CAR.CIK to penetrate and migrate toward STS spheroids through Matrigel domes, with the intent of mimicking their dynamics through the extracellular matrix (Fig. 5A). Microscopic inspection indicated that CSPG4-CAR.CIK readily migrated to the membrane boundary and penetrated the Matrigel domes containing STS spheroids more efficiently than unmodified NTD.CIK (n = 5; P < 0.01; Fig. 5B and C).
CSPG4-CAR.CIK controlled tumor growth in vivo
We explored the in vivo antitumor activity of CSPG4-CAR.CIK utilizing three STS xenograft models (HT1080 fibrosarcoma, S172 leiomyosarcoma, and S1 UPS) that were selected because of the different levels of CSPG4 expression. Upon tumor engraftment (∼50 mm3), mice were treated with intravenous infusions of CSPG4-CAR.CIK or unmodified NTD.CIK (Fig. 6A). In two models (S1 and S172) CIK and tumor cells were autologous, while CAR.CIK tested in the HT1080 xenograft model were allogeneic. In the S172 leiomyosarcoma xenograft model characterized by 72% CSPG4 expression and a density of 262 molecules/cell, autologous CSPG4-CAR.CIK, but not unmodified NTD.CIK or vehicle, caused significant delay in tumor growth (P < 0.05; n = 6 mice/group; Fig. 6B). In the HT1080 fibrosarcoma and S1 UPS xenograft models, showing 23% and 95% CSPG4 expression, respectively, but similar densities of target molecules (521 and 499 molecules/cell, respectively), CSPG4-CAR.CIK also delayed tumor growth as compared with controls (P < 0.001, HT1080; P < 0.0001, S1; n = 3 mice/group; Fig. 6C and D). Tumor infiltration by CSPG4-CAR.CIK was confirmed in explanted tumors by IHC (Fig. 6E). Cleaved caspase 3 levels were confirmed to be higher in tumors from mice treated with CSPG4-CAR.CIK as compared with those from NTD.CIK-treated or vehicle-treated mice (Fig. 6F).
The antitumor activity of CSPG4-CAR.CIK was also verified in additional experiments, in which we reduced the initial tumor burden (∼20 mm3) for both HT1080 and S172 tumors (Supplementary Fig. 5A). In these models, CSPG4-CAR.CIK treatment resulted in significant delay of tumor growth up to 11 days after the end of treatment as compared with controls (P < 0.001, HT1080 and P < 0.0001, S172; Supplementary Fig. S5B–S5E). A complete tumor regression was observed in 2 of 4 mice bearing leiomyosarcoma and 1 of 5 mice bearing fibrosarcoma (HT1080). Of note, antitumor effects were obtained without any macroscopic evidence of toxicity.
STS are tumors for which the clinical impact of targeted therapies remains modest (3). Here, we report that CSPG4 is a clinically relevant target in STS and that CSPG4 expression can be efficiently exploited to eliminate STS tumors by redirecting the specificity of CIK through CAR engineering. We generated compelling evidence supporting the antitumor activity of CAPG4-CAR.CIK using cell lines derived from patients who relapsed after conventional treatments and developing an experimental platform that includes in vitro bidimensional and 3D assays along with three distinct in vivo STS xenograft models.
The survival rate of adults and children with STS remains extremely poor even in the era of checkpoint inhibitors (3, 50, 51). The latter disappointing clinical results are likely to be caused at least, in part, by the “cold” or “immunologically ignorant” tumor microenvironment, low neoantigen load, and defects in HLA class I antigen-presenting machinery in STS (3, 52).
Adoptive transfer of ex vivo–engineered effector T cells, and in particular CAR.T cells, providing HLA-independent tumor recognition may offer the possibility to elicit immune responses in otherwise silent tumors, such as STS. However, the identification of the most appropriate antigen to be recognized by CAR.T cells in STS remains to be defined. Here, we show that CSPG4 holds critical biological features to qualify as a valuable target candidate in STS (38, 41). Using TCGA dataset we found that CSPG4 mRNA is highly expressed in STS across multiple histotypes. Furthermore, we confirmed CSPG4 protein expression on the cell surface of a wide array of STS cell lines spanning multiple STS histotypes and derived from patients who relapsed after conventional treatments. This information is critical in the clinical setting because STS are intrinsically highly heterogeneous, especially in relapsed patients (1). Importantly, CSPG4 expression in all these tumor cell lines was identified by the antibody that we have used to generate the CSPG4-CAR and previously used to show the limited expression of CSPG4 in normal healthy tissues as compared with multiple solid tumors, despite broad CSPG4 mRNA expression in various organs (37, 53). In the perspective of a clinical application, a systematic confirmation of CSPG4 expression by IHC in STS sections that include tumor microenvironment would be warranted.
As for the majority of nonlineage-restricted markers, the density of CSPG4 expression in STS cell lines was variable, but in general much higher than the expression detected in normal keratinocytes. We found a correlation between CSPG4 expression level and antitumor activity by CSPG4-CAR.CIK with minimal activity against normal keratinocytes. The latter finding is consistent with previous data underscoring the importance of antigen density to define a therapeutic window for antigens that are expressed, although at low levels, also by normal tissues (54, 55).
While CAR-redirected effector T cells showed potent antitumor effects in B-cell malignancies, it remains to be defined whether other cell subtypes such as NK cells, NKT cells, or γδT cells may possess intrinsic biologic characteristics to be exploited in solid tumors to enhance the therapeutic index of engineered immune cells (12). Here, rather than exploring different immune cell subsets, we propose that T cells with enhanced cytokine activation during the ex vivo expansion, such as CAR-engineered CIK, may represent a valid cellular platform for the treatment of STS. Dosing and prolonged ex vivo exposure to IL2 induces a mixed T–NK phenotype and function in CIK lymphocytes with variable degree of HLA-independent (NKG2D-mediated) tumor killing ability (56). We reasoned that CAR-engineered CIK would exploit dual tumor cell killing potential, namely antigen independent via NKG2D receptor engagement and antigen dependent via CSPG4-CAR engagement. The utilization of these two mechanisms is expected to ultimately amplify the antitumor effects of CSPG4-CAR.CIK and potentially counteract the escape mechanisms utilized by tumor cells with low CSPG4 expression. We found that CAR.CIK exerted potent antitumor effects against a wide array of STS, which was particularly evident at very low E:T ratios.
The activity of CSPG4-CAR.CIK was confirmed in three independent STS xenograft models and showed clearly gain of function by CIK cells expressing the CAR as compared with unmodified NTD.CIK, further supporting the evidence that NKG2D and CAR engagement is not mutually exclusive in CAR-engineered CIK. Our study, using relatively simple 3D assays to explore the dynamic features of CAR.CIK ex vivo in a more complex structure compared with liquid culture, also uncovered remarkable migratory and infiltrative abilities of CAR-engineered CIK. These properties further underline the potential impact of these cells in solid tumors. It is important to acknowledge that clinical data indicate that CIK are more terminally differentiated as compared with activated and expanded T cells. As a result, their longevity in vivo upon adoptive transfer may be limited (32, 57). However, clinical data also indicate that the manufacturing of CIK both unmanipulated or CAR engineered is very robust. This property allows logarithmic expansion of cells in a relatively short period of time and storing of CIK for multiple infusions. Furthermore, the initial evidence of a lower IL1β and IL6 production by CAR.CIK as compared with paired “conventional” CAR.T lymphocytes, may further support a favorable safety profile.
Overall, our data support CSPG4 as a valuable CAR target for STS and the use of engineered CIK to express the CSPG4-CAR for the development and implementation of a novel and effective immunotherapeutic strategy for the treatment of patients with advanced/relapsed high-grade STS.
Disclosure of Potential Conflicts of Interest
G. Grignani reports personal fees from Lilly and Merck; grants and personal fees from Bayer, PharmaMar, and Novartis; and personal fees from Eisai outside the submitted work. L. D'Ambrosio reports other from GlaxoSmithKline (advisory board), PSI CRO Italy (advisory board), PharmaMar (meeting participation), Celgene (meeting participation), Novartis (editorial activity), and Lilly (meeting participation) outside the submitted work. E. Vigna reports other from METIS Precision Medicine B-Corp (company founder) outside the submitted work. G. Dotti reports grants from Cell Medica (research sponsor agreement), Bluebird Bio (research sponsor agreement), and Bellicum Pharmaceutical (research sponsor agreement); personal fees from Bellicum Pharmaceutical (scientific advisory board), MolMed S.p.A. (scientific advisory board), and Tessa Therapeutics (consultant) outside the submitted work; as well as a patent for CSPG4.CAR pending owned by the University of North Carolina and Massachusetts General Hospital. D. Sangiolo reports grants from AIRC (Associazione Italiana Ricerca sul Cancro), Ministry of Health, FPRC (Fondazione Piemontese per la Ricerca sul Cancro) ONLUS, and University of Torino during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.
V. Leuci: Conceptualization, formal analysis, supervision, investigation, visualization, writing-original draft, writing-review and editing. C. Donini: Conceptualization, formal analysis, supervision, investigation, visualization, writing-original draft, writing-review and editing. G. Grignani: Resources, funding acquisition, writing-review and editing. R. Rotolo: Investigation, writing-original draft, writing-review and editing. G. Mesiano: Resources, investigation, writing-review and editing. E. Fiorino: Resources, investigation, writing-review and editing. L. Gammaitoni: Resources, investigation, writing-review and editing. L. D'Ambrosio: Resources, writing-original draft, writing-review and editing. A. Merlini: Resources, visualization, writing-review and editing. E. Landoni: Resources, investigation, writing-review and editing. E. Medico: Data curation, formal analysis, investigation, writing-original draft, writing-review and editing. S. Capellero: Investigation, writing-review and editing. L. Giraudo: Investigation, visualization, writing-review and editing. G. Cattaneo: Investigation, writing-original draft. I. Iaia: Investigation, writing-review and editing. Y. Pignochino: Resources, investigation, writing-original draft. M. Basiricò: Investigation. E. Vigna: Resources, writing-original draft, writing-review and editing. A. Pisacane: Resources, validation. F. Fagioli: Writing-review and editing. S. Ferrone: Conceptualization, resources, writing-original draft, writing-review and editing. M. Aglietta: Resources, writing-original draft, writing-review and editing. G. Dotti: Conceptualization, resources, writing-original draft, writing-review and editing. D. Sangiolo: Conceptualization, resources, supervision, funding acquisition, writing-original draft, writing-review and editing.
This study was supported by fundings from “Associazione Italiana Ricerca sul Cancro” IG-2017 n. 20259 (to D. Sangiolo), IG-2019 n. 23104 (to G. Grignani); Ricerca corrente Progetto CAR-T RCR-2019-23669115 (to D. Sangiolo and E. Medico); FPRC ONLUS 5 × 1000, Ministero della Salute 2015 Cancer ImGen (to D. Sangiolo, G. Grignani, and E. Medico); FPRC ONLUS 5 × 1000 MIUR 2014 (to G. Grignani and L. D'Ambrosio); Ministero della Salute (GR-2011-02349197 to D. Sangiolo), University of Torino Fondo Ricerca Locale 2017 (to D. Sangiolo); Ricerca Corrente Ministero Salute 2020; and Fondazione per la ricerca sui tumori dell'apparato muscoloscheletrico e rari Onlus CRT RF = 2016–0917 (to G. Grignani). S. Ferrone was supported by NIH grants R01DE028172 and R03CA216114 and by DOD grant W81XWH-16-1-0500. V. Leuci has received a fellowship from the Fondazione Nicola Ferrari ONLUS and M. Basiricò has received a fellowship from ADISCO ONLUS. The authors sincerely thank Joan Leonard (Leonard Editorial Services, LLC) for the linguistic revision and editorial assistance.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.