WHAT IS ALREADY KNOWN ON THIS TOPICHOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICYIntroduction

Ewing sarcoma (ES) is a malignant pediatric bone and soft tissue tumor. Despite multiple therapeutic approaches including surgery, chemotherapy and radiation over the last 40 years, patients with recurrent metastatic ES have a dismal average 5-year overall survival of less than 30%,1 largely secondary to therapy resistance within the tumor microenvironment (TME). New therapeutic paradigms are urgently needed and represent an unmet need.2 3

Natural killer (NK) cells are innate immune cells highly cytotoxic to tumor cells, including ES.4 Unlike T cells, NK cells do not require prior sensitization to target tumor cells. NK function is regulated by a complex balance of inhibitory and activating signals mediated by its inhibitory and activating receptors, respectively.5 ES cells express high levels of MICA/B, ligands to activating NK receptor NKG2D, and negligible levels of human leukocyte antigen (HLA) class I molecules, ligands to inhibitory NK receptors killer Ig-like receptors (KIRs) of NKG2A.4 6 These expression patterns shift the balance toward NK cell activation and render ES cells highly sensitive to NK-mediated killing.4 6

However, NK cell number and function are low in ES patient tumors, in large part due to the immunosuppressive TME. The mechanisms of TME resistance to NK cell tumor immunity include small numbers of active NK cells, poor NK cell function, activation and persistence, lack of specific targeting, among others.7 To address low NK cell number and function, our group has developed a genetically engineered feeder cell K562-mbIL-21-4-1BBL to expand peripheral derived NK cells.8 This approach achieves 35,000 fold expansion of NK cells with high functional activation and preserved telomere length.8 To further enhance the cytotoxicity of NK cells and facilitate specific targeting of tumor cells, we and others have engineered NK cells to express CARs against cancer targets.9–12 However, due to the scarcity of tumor-associated antigens in ES,13 few NK CARs have been developed to target ES. Recently, expression of CAR against the ganglioside antigen GD2 in activated and expanded NK cells has been shown to increase NK cytotoxicity against GD2+ES cells in vitro.14 However, in the preclinical setting, adoptive transfer of GD2-CAR-NK cells failed to eliminate GD2+ES xenografts.14 Moreover, the efficacy of CAR-NK cells in limiting ES tumor metastasis is largely uninvestigated. Additional strategies and the development of CAR-NK cells against new cancer targets that are critical in tumor metastatic spread may potentially enhance the antitumor activity of NK cells and circumvent solid tumor resistance.

Melanoma cell adhesion molecule (MCAM) is a cell surface protein overexpressed in common pediatric cancers including ES.13 MCAM is highly expressed in the embryo but minimally expressed in mature normal tissues.15 Depletion of MCAM has been shown to inhibit ES cell migration; high MCAM expression is correlated with poor prognosis in ES clinical cohorts.16 All these features make MCAM a promising target for immunotherapy for patients with metastatic ES. Indeed, a fully humanized antibody targeting MCAM, ABX-MA1 was found to inhibit spontaneous pulmonary metastasis in an orthotopic mouse model of osteosarcoma.17 However, the anti-tumor effects of MCAM targeted cellular immunotherapy have not been previously investigated in ES.

NK tumor immunity is limited by their poor persistence in vivo. NKTR-255 is an engineered polymer-conjugated recombinant human IL-15 receptor agonist with a high affinity to IL-15Rα and a significantly longer half-life compared with natural IL-15 (27 vs <1 hour). NKTR-255 induces the NK cell IL-15 signal transduction pathway (pSTAT5), stimulating NK cell proliferation, survival and cytotoxicity.18 19 NKTR-255 alone and/or in combination with various therapeutic monoclonal antibodies are currently under clinical investigations in patients with hematological malignancies (NCT04136756 and NCT03233854) and solid tumors (NCT04616196) as reported by our group and others.20 21

TAMs are the most abundant infiltrating immune cells in ES TME22 and confer a poor prognosis for ES patients.23 Cross-talks between TAMs and NK cells regulate their functional activities in the TME, resulting in immune activation or suppression.24 Here, we found that TAMs play a critical role in mediating NK cell resistance in ES. Magrolimab (MAG) is an investigational monoclonal antibody against CD47 that blocks the macrophage “don’t eat me” signal25 and reactivates macrophage phagocytic and antitumor activities.26 In the current study, we engineered an ex vivo expanded CAR-NK cell targeting MCAM and combined it with NKTR-255 and MAG to overcome the ES TME resistance to the NK/CAR-NK cell therapy and thereby facilitate increased antitumor efficacy against metastatic ES.

Materials and methods

Additional methods are detailed in online supplemental materials.27–31

DNA constructs

The MCAM Ab single chain variable fragment (scFv) sequence was generously provided by Bin Liu (University of California at San Francisco)32 and the scFv DNA (M1) was codon optimized and synthesized (Integrated DNA Technologies), followed by subcloning in frame with a linker (eskygppcppcpm), CD28TM (NP_001230007.1, aa 34–59), 4-1BB (NP_001552, aa 213-255), and CD3ζ (NP_000725, aa 52-163) into the pcDNA3 vector to generate a second generation anti-MCAM-CAR. The CAR construct was further optimized by addition of a 2bgUTR.150A sequence generously provided by Carl June and Yangbing Zhao (University of Pennsylvania)33 at 3’ end of the construct. The clustered regulatory interspaced short palindromic repeats (CRISPR)/Cas9 knockout constructs for MCAM were created by cloning the CRISPR guide RNA against MCAM (5’-GTTGCATGACCTGAAACGGG-3’ (KO1) and 5’-AGGAGGCGGCTATCGCTGCG-3’ (KO2)) into the lentiCRISPRv2 vector (Addgene plasmid #52961).34 Guide sequences were designed using the Broad Institute sgRNA designer tool (https://portals.broadinstitute.org/gpp/public/analysis-tools/sgrna-design).

RNA-seq analysis

Total RNA was extracted from tumors using RNeasy mini kit (Qiagen) after homogenization of tumors by GentleMACS tissue dissociator (Miltenyi Biotec) according to manufacturer’s instructions. RNA was treated with DNase and RiboZero capture beads (Illumina) to deplete DNA and ribosomal contaminants, respectively. RNA was then used as input for Illumina TruSeq Stranded Total RNA library preparation kit and sequenced on NovaSeq 6000 to obtain ~60M reads per sample. To prevent any bias that would occur from aligning to one genome at a time, RNA-seq reads were aligned using STAR (STAR_2.6.1c) to a concatenated reference comprised of the Homo sapiens (GRCh38.p12 assembly) and Mus musculus (GRCm38.p6 assembly) genomes. The resulting alignments were separated into species-specific bam files, converted back to fastq files and processed through individual human and mouse pipelines, thus creating individual alignments and transcript counts per genome. DESeq2 (V.1.30.1) was used to normalize expression values and to identify differentially expressed genes (DEGs) between groups.35 A threshold for DEGs between the two groups was set to an absolute value of fold change ≥1.5 and a false discovery rate of ≤0.10.

Bioluminescence-based in vitro cytotoxicity assay

Bioluminescence (BLI)-based in vitro cytotoxicity assays were performed as we have previously described29 with minor modifications. Briefly, luciferase-expressing tumor cells (5×104) were incubated with effector cells (NK/CAR-NK) at different effector-to-target (E:T) ratios (0.2:1, 0.5:1, and 1:1) in DMEM media supplemented with 10% FBS in 96-well tissue culture plates at 37°C for 4 hours before D-firefly luciferin potassium salt (LUCK-1G, Goldbio, St Louis, Missouri, USA) was added to the cells and BLI was measured with a luminometer (Molecular Devices Multifilter F5 plate reader). In CAR-NK and NKTR-255 combination cytotoxicity assay, MCAM-CAR-NK cells were cultured in RPMI1640 media supplemented with or without NKTR-255 (40 ng/mL, Nektar Therapeutics, San Francisco, California, USA) for 72 hours in the absence of IL-2 before incubating with luciferase- expressing tumor cells.

Animal studies

All animal studies were performed in accordance with protocols approved by the New York Medical College Institutional Animal Care and Use Committee. Luciferase-expressing A673 cells or ES patient-derived xenograft (PDX) tumors (NCH-EW-1) were implanted into the tibia (2×105 cells/site) or flanks (a 5×5×5 mm piece/site) of 4–6 weeks old female/male NSG mice (Jackson Laboratory). NSG was used to facilitate stable engraftment of human tumor cells and the evaluation of human NK cell antitumor efficacy. After validation of tumor engraftment by Xenogen In Vivo Imaging System (IVIS) or when PDX tumors reach 5–7 mm in diameter, PBS or NK or CAR-NK were injected (1×107 cells/animal, intravenous, once a week for 3–6 weeks) together with or without NKTR-255 (0.3 mg/kg, intravenous, once every 2 weeks) (generously provided by Nektar Therapeutics) and/or MAG (100 µg/animal, i.p. once per day for 12 days) (generously provided by Gilead Sciences, Foster City, California, USA). Before conducting mouse experiments, sample sizes achieving 80% power to detect an effect size >2 were determined at significant level as 0.05. No randomization or blinding was used. Tumor growth was monitored by daily caliper measurement and/or weekly IVIS imaging as we have previously described.36 Mice were followed until death or sacrificed on reaching a tumor size of 1.5 cm in any dimension when the tumors and/or lungs were harvested. For mechanistic study, when xenograft tumors reached 1 cm, Phosphate buffered saline (PBS) or 1×107 of ex vivo expanded NK cells were injected and tumors were harvested 24 hours after injection and dissected into peripheral (distance from center of the tumor: 3–5 mm) and central (distance from center of the tumor: 0–3 mm) sections for subsequent analyses.

In vitro phagocytosis assay

Generation of monocyte-derived macrophages was performed as previously reported37 with minor modifications. Briefly, donor peripheral blood mononuclear cells were isolated by density gradient using Ficoll-Paque PLUS (GE17-1440-03, Millipore sigma) and seeded in RPMI with 10% FBS at a density of 1×107 cells/mL in 10 cm tissue culture treated plates for 6 days in the presence of 100 ng/mL recombinant human M-CSF (130-096-491, Miltenyi Biotec) with replenishment of media on day 4. ES cells were labeled with CellTracker Green CMFDA dye (C2925, ThermoFisher Scientific) and co-cultured with macrophages at a 2:1 ratio together with or without MAG at a concentration of 1 µg/mL in RPMI for 2–4 hours. Cells were harvested and washed with cold PBS and stained with α-CD11b (Miltenyi Biotec, 130-110-554). Flow cytometry was carried out and phagocytosis index was measured as the percentage of CD11b+FITC+ macrophages in the total CD11b+macrophages and normalized to the control condition.

ResultsCharacterization of the immune TME of the NK resistant ES tumors

To identify the mechanisms of resistance to NK cells in the ES TME as demonstrated in online supplemental figure 1A, we treated the ES xenograft tumors with ex vivo expanded NK cells or PBS and harvested the tumors 24 hours after treatment. We performed immunostaining (IHC) and found fewer NK cell infiltration into the ES xenograft tumor compared with neuroblastoma xenograft tumor (online supplemental figure 1B). We hypothesized that the immunosuppressive ES TME inhibited NK cell activity and prevented NK cell infiltration into the tumor. To test this hypothesis, we dissected the NK treated and untreated tumors into peripheral and central tumor sections. Total RNA was extracted separately from the NK treated (T) or untreated (U) peripheral (P) or central (C) tumor sections and subjected to transcriptomic sequencing. We identified 1043 murine and 130 human DEGs when comparing treated peripheral (TP) and treated central (TC) tumor samples, 6 murine and 92 human DEGs in comparison of TP and untreated peripheral (UP) samples, and 216 murine and 3 human DEGs comparing TC and UC (untreated central) samples (figure 1A,B, online supplemental figure 2A,B, online supplemental file 1). By DAVID functional annotation analysis using the DEGs between treated (TP and TC) and untreated (UP and UC) samples, we found that immunity, plasma membrane, adaptive immune response, chemotaxis and granzyme-mediated apoptosis signaling pathway are among the most enriched terms in the DEGs of human genes (online supplemental figure 2C), and intermediate filament and keratin are the two most enriched terms in the DEGs of mouse genes (online supplemental figure 2D). Interestingly, using the DEGs between TP and TC samples, we found that the same enriched terms such as keratin/keratin filament and monocyte chemotaxis in the DEG of mouse genes (figure 1C), and regulation of immune response and NK cell-mediated cytotoxicity in the list of human DEGs (figure 1D). This points to a unique immune signature identified only in the peripheral sections of the treated tumors in which NK cells and mouse monocytes play key roles. An ingenuity pathway analysis (IPA) of the murine and human DEGs further revealed activation of mouse monocytes and crosstalk/interactions between human NK cells and mouse macrophages/monocytes (figure 1E,F).

Figure 1

Identification of mechanisms of resistance to NK therapy in ES xenograft tumors. (A, B) Top differentially expressed murine (A) and human (B) genes comparing NK cell treated peripheral and central tumor sections. ES xenograft tumors (N=5 per condition) were harvested 24 hours after the treatment with PBS or ex vivo expanded NK cells and dissected into peripheral and central tumor sections. Total RNA was extracted from the tumor sections and subjected to RNA-seq and subsequent data analyses for differentially expressed genes. Each cell represents the gene expression of each sample. (C, D) DAVID functional annotation analyses of differentially expressed murine (C) and human (D) genes comparing treated peripheral and central tumor samples. Top functional annotation terms were plotted with enrichment scores (−lg (p value)). (E, F) Ingenuity pathway analysis of differentially expressed murine (E) and human (F) genes comparing treated peripheral and central tumor samples. Red and blue colors indicate higher and lower expression in treated peripheral tumor samples, respectively. Octagon represents the regulated functions; other shapes indicate the types of molecules. Solid and dash lines represent direct and indirect interactions. DEGs, differentially expressed gene; ES, Ewing sarcoma; NK, natural killer; PBS, phophate buffered saline.

To further study the ES tumor immune microenvironment, CIBERSORT was used to deconvolute the abundance of immune cell types (online supplemental figure 3A,B). Interestingly, we found that in the peripheral tumor sections, resting instead of activated human NK cells were present, and M1 mouse macrophages were predominant, while in the central tumor sections, M2 mouse macrophages are the most abundant immune cells and human NK cells are barely present (figure 2A,B). All these data suggest that the interaction between mouse macrophages and human NK cells may be part of the mechanism of NK cell resistance in ES.

Figure 2

Macrophages negatively regulate NK cells in the immune microenvironment of Ewing sarcoma (ES) tumors. (A) Abundance of M1 and M2 mouse macrophages in the peripheral and central sections of ES xenograft tumors. The abundance of mouse immune cells was deconvoluted using RNA-seq gene expression data and ImmuCC signature matrix by CIBERSORT analysis. The relative counts of M1 (orange) and M2 (blue) macrophages in central and peripheral tumor sections were plotted with each tumor sample. In x-axis sample names, T1C means treated central tumor 1; U1C, untreated central tumor 1; T1P, treated peripheral tumor 1; U1P, untreated peripheral tumor 1. (B) Percentage of resting and activated human NK cells in the peripheral and central tumor sections. RNA-seq gene expression data were used as input matrix and LM22 as signature matrix in CIBERSORT analysis to deconvolute the abundance of human immune cells in the xenograft tumors. The relative counts of resting (blue) and activated (orange) NK cells in central and peripheral tumor sections were plotted with each tumor sample. In x-axis sample names, T1C means treated central tumor 1; T1P, treated peripheral tumor 1. Aliquots of ex vivo expanded NK cells injected into the mice (NK1 and NK2) were used as control samples. (C), Cytokine array analysis for identification of cytokines/chemokines secreted by ES cells and PDX tumor. ES A673 and SKNMC cells and PDX tumor (NCH-EW-1) were cultured in DMEM media for 48 hours before the conditioned media were harvested and assayed by Human Cytokine Antibody Array (Membrane, 42 Targets, Abcam). Fresh DMEM media was used as a negative control. The duplicate black dots at specific positions show positive signals of certain cytokines. Asterisk: internal control; 1: MCP-1/CCL2; 2: IL-8; 3: GRO; and 4: GRO-α. Representative images are shown. The same trend was seen in two independent biological repeats. (D) Quantification of the dot intensity in (C) by ImageJ (https://imagej.nih.gov/ij). Columns represent the relative average density of the dots compared with the internal control. Error bars indicate the SD of duplicate samples in a representative experiment. The same trend was seen in two independent biological repeats. *p<0.05, **p<0.01 (two-tailed Student’s t-test). (E) Comparison of NK cell receptor expression levels on ex vivo expanded NK cells incubated with or without tumor-associated macrophages (TAMs) isolated from ES xenograft tumors. Macrophages were enriched from the single cell suspension of ES xenograft tumors using the anti-F4/80 MicroBeads (Miltenyi Biotec). Ex vivo expanded NK cells were incubated with (orange) or without (blue) isolated TAMs at a ratio of 1:1 for 4 hours followed by evaluation of NK cell receptor expression via flow cytometry. *p<0.05, **p<0.01 (two-tailed Student t-test). Representative results are shown. The same trend was seen in two independent biological repeats. (F) Correlations of immune cells in the TME of ES patient tumors. CIBERSORT analysis was used to deconvolute immune cell composition in ES patient tumors using publicly available ES microarray dataset (GSE17679). CIBERSORT counts were used to analyze the correlations of abundance between various immune cells by Pearson’s correlation method. Red and blue dots represent negative and positive correlations, respectively. The darker the color is, the stronger the correlation is. (G) The relative abundance of NK cells and macrophages in individual tumors are represented as open circles in scatterplots. Pearson correlation coefficient r=−0.41 (95% CI –0.55 to –0.24), p=6.49×10−6. N=98 in GSE17679. PDX, patient-derived xenograft; TME, tumor microenvironment.

To investigate which cytokines/chemokines are secreted by the ES cells to attract mouse macrophages, we performed a cytokine array analysis and found that monocyte chemoattractant protein 1 (MCP-1/CCL2), IL-8 and chemokine (C-X-C motif) ligand 1 (CXCL1/GRO/GRO-alpha) are cytokines commonly highly secreted by ES cells (A673 and SKNMC) and PDX tumor (NCH-EW-1) (figure 2C,D). To further investigate whether TAMs directly affect the activity of NK cells, we harvested xenograft tumors from the NSG mice and enriched TAMs from the single cell isolates and incubated them with ex vivo expanded NK cells. Phenotype analyses of NK cells revealed that the expression levels of NK cell activating receptors (NKG2D, NKG2C, NKp30, NKp44, and NKp46) on ex vivo expanded NK cells were significantly decreased after incubation with TAMs (figure 2E). These data suggest that TAMs from the ES xenograft tumors negatively regulate the activity of ex vivo expanded NK cells.

To investigate whether the negative relationship between macrophages and NK cells we observed in the preclinical studies is also present in human ES patient tumors, we performed CIBERSORT analyses to deconvolute immune cell composition in patient TME using a publicly available ES patient tumor microarray dataset (GSE17679, n=98).30 The correlations between various immune cells were analyzed by Pearson’s method. Consistent with our preclinical data, we observed a strong negative correlation among NK cell populations (activated or resting) and macrophage populations (M0, M1, or M2) in the ES patient tumors (figure 2F, red dots represent negative correlations, and figure 2G, Pearson correlation coefficient r=−0.41, p=6.49×10−6). Interestingly, we performed the same CIBERSORT and correlation analyses using neuroblastoma patient tumor RNA-seq dataset (TARGET-NBL, n=155)31 but did not observe a clear correlation between NK cells and macrophages (online supplemental figure 4).

Expression of anti-MCAM CAR significantly increased cytotoxic activity of NK cells against MCAMhigh ES cells in vitro

To overcome the resistance in the ES TME and enhance the NK cytotoxic activity and specific targeting of ES cells particularly metastatic ES cells, we engineered ex vivo expanded NK cells to express CAR against MCAM, a novel target that was reported to associate with cancer metastasis.13 16 We analyzed MCAM expression levels in ES A673, TC32, and SKNMC cells and detected high levels of MCAM expression across all the ES cell lines (figure 3A). We engineered ex vivo expanded NK cells to express CAR directed against MCAM (figure 3B) by CAR mRNA electroporation and found that 78%, 70%, 55% and 39% of NK cells expressed CAR at 1, 2, 4 and 6 days postelectroporation, respectively (figure 3C). We performed NK cytotoxicity assays in vitro using A673, SKNMC, TC32 as target cells and found that, compared with the unmodified expanded NK cells (mock), expression of anti-MCAM CAR in expanded NK cells (CAR) significantly enhanced the NK cytotoxicity against these ES cells at various E:T ratios (0.2:1, 0.5:1, 1:1) (figure 3D). Notably, we observed the significant differences between mock and CAR-NK cells at low E:T ratios likely because of the high sensitivity of these tumor cells to NK cells at baseline due to the high expression of NK activating receptor ligands (MIC A/B, CD112, CD155) and low expression of inhibitory receptor ligands (HLA-ABC) on these tumor cells (online supplemental figure 5). Next, we compared mock and CAR-NK cell responses to these tumor cells by quantifying the secretion of cytokines interferon γ (IFN-γ) and perforin and found that CAR-NK cells secreted significantly higher levels of cytokines than mock NK cells when incubated with ES cells (figure 3E,F). To investigate whether the enhanced cytotoxic activity of CAR-NK cells was due to specific targeting of MCAM, we knocked out MCAM in tumor cells by CRISPR/Cas9 (figure 3G) and compared cytotoxic activity of mock and CAR-NK cells against wildtype (WT) and knockout (KO1 and KO2) cells. In the KO tumor cells, we did not observe a significant increase in cytotoxicity with CAR-NK compared with mock NK as we did in the WT tumor cells (figure 3H), demonstrating the increased cytotoxicity of CAR-NK is specific to MCAMhigh tumor cells.

Figure 3

Expression of anti-MCAM CAR significantly increased cytotoxic activity of NK cells against MCAMhigh ES cells in vitro. (A) Expression of MCAM on ES cells. Cells were stained with isotype or MCAM antibody and subject to flow cytometry analyses. Histograms from a representative experiment were shown. Three biological repeat experiments were performed with similar results (n=3). (B) Schematic representation of the MCAM CAR construct. (C) Electroporation of CAR mRNA resulted in CAR expression on ex vivo expanded NK cells. Two micrograms of CAR mRNA per 106 of ex vivo expanded NK cells were used for electroporation using Maxcyte GTx electroporator. CAR expression was detected at 1, 2, 4, 6 days postelectroporation by a biotinylated MCAM protein, followed by FITC-streptavidin staining and flow cytometry analyses. The numbers at the right corner of the dot plots are percent of CAR positive cells. Dot plots from a representative experiment were shown. Three biological repeat experiments (three donors) were performed with similar results. (D) In vitro cytotoxicity of unmodified expanded NK cells (Mock) and anti-MCAM-CAR-NK cells (CAR) against MCAMhigh ES cells A673, TC32, and SKNMC. Mock or CAR-NK cells were incubated with luciferase-expressing ES cells at various effector to target (E:T) ratios (0.2:1, 0.5:1, 1:1) for 4 hours before cytotoxicity was analyzed by addition of D-luciferin to the cells and bioluminescence was measured with a luminometer. Asterisks indicate p<0.05, **p<0.01 (two-tailed Student’s t-test). Error bars indicate the SD of triplicate samples in a representative experiment. The same trend was seen in three independent biological repeats. (E, F), Secretion of cytokines IFN-γ (E) and perforin (F) from mock and anti-MCAM-CAR-NK cells when incubated with ES cells. Mock or CAR-NK cells were incubated with ES A673, TC32 or SKNMC cells at E:T ratio of 0.2:1 for 4 hours and cytokine secretion in the media was analyzed by ELISA. Tumor cells with no NK cell incubation were used as a negative control. *p<0.05, **p<0.01 (two-tailed Student’s t-test). Error bars indicate the SD of triplicate samples in a representative experiment. The same trend was seen in three independent biological repeats. (G) CRISPR/Cas9 knockout of MCAM in ES cells. MCAM expression in wildtype (WT) and knockout (KO1 and KO2) A673, TC32 and SKNMC cells were detected by flow cytometry using MCAM specific antibody. (H) In vitro cytotoxicity of mock and anti-MCAM-CAR-NK cells against MCAM WT and KO ES cells. Mock or CAR-NK cells were incubated with luciferase-expressing MCAM WT or KO (KO1 and KO2) ES cells (A673, TC32, SKNMC) at E:T ratio of 0.2:1 for 4 hours before cytotoxicity was analyzed by addition of D-luciferin to the cells and bioluminescence was measured with a luminometer. *p<0.05 (two-tailed Student’s t-test). Error bars indicate the SD of triplicate samples in a representative experiment. The same trend was seen in three independent biological repeats. ES, Ewing sarcoma; MCAM, melanoma cell adhesion molecule; NK, natural killer.

Anti-MCAM-CAR-NK cell significantly decreased lung metastasis and extended animal survival in an ES orthotopic xenograft mouse model

To evaluate the efficacy of MCAM CAR engineered ex vivo expanded NK cells against ES in vivo, we employed an orthotopic xenograft mouse model of ES metastasis by transplanting luciferase-expressing A673 cells (A673-luc) into the tibia of NSG mice.36 PBS or mock NK or MCAM-CAR-NK cells were administered via tail vein into the tumor-bearing mice (figure 4A). The frequency of NK/CAR-NK cell injections (once a week) was based on our finding that more than half of the expanded NK cells expressed CAR at day 4 postelectroporation (figure 3C). We found that significantly fewer mice developed lung metastases in the CAR-NK treatment group compared with the PBS control group (43% vs 14%, p<0.01) and the mock NK group (28% vs 14%, p<0.01) (figure 4B,C). These gave the animals in the CAR-NK group a significant advantage in their survival at day 50 compared with the PBS group (0% vs 50%, p<0.05) and the mock NK group (20% vs 50%, p<0.05) (figure 4D). Although treatment with mock NK cell also decreased percent of mice with lung metastases compared with the vehicle (43% vs 28%, p<0.01), it did not contribute to significantly extended animal survival compared with the control group (figure 4B–D).

Figure 4

Anti-MCAM-CAR-NK cells significantly decreased lung metastasis and prolonged animal survival in an ES orthotopic xenograft mouse model. (A) Schematic representation of the animal work schedule. Luciferase-expressing A673 cells (A673-luc) were injected into the tibia of NSG mice on day 0. 24 hours later, PBS or mock or CAR-NK cells were administered through tail vein (107 cells/animal, once a week for 6 weeks). Tumor growth was monitored by IVIS imaging once a week. (B) Effects of NK/CAR-NK treatment on lung metastasis. Lungs were harvested at the endpoint. Graphs show the percentage of mice with pulmonary lesions in each indicated condition. **Fisher’s exact test p<0.01. (C) Representative images of lungs from two of the animals treated with PBS, mock NK, or CAR-NK cells. (D) Kaplan-Meier survival curves for comparison of survival among all groups. Animal survival was followed after therapy initiation using animal death or sacrifice as the terminal event using the Prism program V.8.0 (GraphPad Software). *p<0.05 (log-rank test). N=7 for each group. ES, Ewing sarcoma; IVIS, In Vivo Imaging System; MCAM, Melanoma cell adhesion molecule; NK, natural killer; PBS, phosphate buffered saline.

NKTR-255 significantly improved NK cell survival, sustained NK expansion and enhanced CAR-NK cell cytotoxicity against ES cells in vitro

To further enhance the antitumor efficacy of CAR-NK cells and facilitate improved CAR-NK cell survival and durable persistence, we employed a recombinant IL-15 receptor agonist, NKTR-255. We incubated 14-day expanded NK cells with NKTR-255 in the absence of IL-2 and found that NKTR-255 significantly enhanced the expression levels of NKp30, NKG2D and CD94, the NK activating receptors, but not NKG2A or KIR, the NK inhibitory receptors (figure 5A) compared with the non-treated control. Furthermore, NKTR-255 treated NK cells further expanded while the number of non-treated NK cells gradually decreased over time in the absence of IL-2, demonstrating that NKTR-255 significantly improved NK cell survival and sustained NK expansion in vitro (figure 5B). We next investigated the effect of NKTR-255 on MCAM-CAR-NK cell in vitro cytotoxicity and found that NKTR-255 significantly enhanced the cytotoxic activity of MCAM-CAR-NK cells targeting all three lines of ES cells (figure 5C, p<0.05).

Figure 5

Effects of NKTR-255 on NK/CAR-NK cells and magrolimab (MAG) on macrophage phagocytosis of Ewing sarcoma (ES) cells in vitro. (A) Effects of NKTR-255 on the expression of NK cell receptors. Ex vivo expanded natural killer (NK) cells were incubated with NKTR-255 (40 ng/mL) for various periods of time (0–6 days) in the absence of IL-2 with replenishment of media and NKTR-255 at day 3. Flow cytometry was performed to detect the expression of NK cell receptors (CD94, NKp30, NKG2D, NKG2A, KIR and NKp44). Representative results from 4 days of incubation were shown. The same trend was seen in three independent biological repeats. (B) Effects of NKTR-255 on NK cell survival and expansion in the absence of IL-2 in vitro. NK cells (2.5×106) at 14 days of expansion were incubated with NKTR-255 (40 ng/mL) for various periods of time (0–6 days) in the absence of IL-2 with replenishment of media and NKTR-255 at day 3. Viable cells were counted by the method of trypan blue staining. Error bars indicate the SD of triplicate samples in a representative experiment. The same trend was seen in three independent biological repeats. *p<0.05, **p<0.01 (two-tailed Student’s t-test). (C) In vitro cytotoxic activity of anti-MCAM-CAR-NK cells in combination with NKTR-255 against ES cells. CAR-NK cells were incubated with or without NKTR-255 (40 ng/mL) in the absence of IL-2 for 3 days before incubating with luciferase-expressing ES cells (A673, TC32, SKNMC) at E:T ratio of 0.2:1 for 4 hours. Cytotoxicity was analyzed by addition of D-luciferin and bioluminescent was measured with a luminometer. *p<0.05, **p<0.01 (two-tailed Student’s t-test). Error bars indicate the SD of triplicate samples in a representative experiment. The same trend was seen in three independent biological repeats. (D) CD47 expression on ES cells. CD47 expression levels were detected by flow cytometry on ES A673, TC32, SKNMC cells. (E) CD47 expression on ES PDX tumors. Immunofluorescent staining by CD47 antibody was performed to detect CD47 (green) expression on MSK-3 and NCH-EW-1 ES PDX tumors. The nuclei were counterstained with DAPI (blue). (F) MAG significantly enhanced macrophage phagocytosis of ES cells. Macrophages were derived from healthy donor peripheral blood mononuclear cells by recombinant human M-CSF (100 ng/mL). ES A673, TC32, SKNMC cells were labeled with CMFDA dye (recognized by FITC channel) and co-cultured with macrophages at a 2:1 ratio together with IgG or MAG (1 µg/mL) for 4 hours. Cells were harvested and stained with α-CD11b followed by flow cytometry analysis. The gated population (CD11b+FITC+) in the dot plots (left) show macrophages phagocytosed ES cells. Right, significantly enhanced macrophage phagocytosis of ES cells (A673, TC32, EWS502) by MAG (blue) compared with IgG control (orange). Quantification of the results in three biolo