Construction of self-driving anti-αFR CAR-engineered NK cells based on IFN-γ and TNF-α synergistically induced high expression of CXCL10

Highlights

• •In this study, we found for the first time that the third-generation αFR-CAR-NK92 effector cells targeting αFR successfully constructed in the previous stage of our research group can produce the chemokine CXCL10 after co-culture with ovarian cancer target cells, and no relevant studies have reported this phenomenon in CAR-T/NK cell therapy.
• •In ovarian cancer cells, we investigated the signaling pathway that TNF-α or/and IFN-γ induces the secretion of CXCL10, which was not been reported before.
• •In cancer cells, we studied the mechanism by which TNF-α and IFN-γ synergically induce the increase of CXCL10 secretion in cancer cells, which has not been reported before.
• •In this study, we used the CXCL10/CXCR3 axis for the first time to construct the fourth-generation CXCR3-αFR-CAR-NK92 cells that co-express the chemokine receptor CXCR3
• •We verified the killing function and chemotaxis of CXCR3-αFR-CAR-NK92 cells targeting ovarian cancer cells through in vitro and in vivo experiments.
• •The CXCR3-αFR-CAR-NK92 cells constructed by us improved the immunotherapy efficacy of αFR-CAR-NK92 cells, providing a new strategy for immunotherapy of ovarian cancer. In particular, it is beneficial to personalized treatment for this patient group of high-grade serous ovarian cancer immune response type (that is, tumor tissue high expression of CXCL10), and to a certain extent, alleviate the difficulty of transport obstruction when CAR-NK92 cells treat solid ovarian cancer.

Abstract

Introduction

Ovarian cancer is the most malignant gynecological tumor. Previous studies have demonstrated that chimeric antigen receptor (CAR)-engineered NK-92 cells targeting folate receptor α (αFR) (NK-92-αFR-CAR) can specifically kill αFR-positive ovarian cancer cells. However, the migration barrier restricts antitumor effects of CAR-engineered cells.

Objectives

To elucidate the mechanism by which NK-92-αFR-CAR cells induce the secretion of chemokine CXCL10 during killing ovarian cancer cells. It is speculated that NK-92-αFR-CAR-CXCR3A can target αFR and have chemotaxis of CXCL10, and they may have stronger killing effect of ovarian cancer.

Methods

Study the mechanism of CXCL10 expression strongly induced by TNF-α and IFN-γ combined stimulation in ovarian cancer cells. Construct the fourth generation of NK-92-αFR-CAR-CXCR3A cells, which were co-expressed CXCR3A and αFR-CAR. Evaluate the killing and migration effects of NK-92-αFR-CAR-CXCR3A in vitro and in vivo.

Results

RNA sequencing (RNA-seq) first revealed that the expression level of the chemokine CXCL10 was most significantly increased in ovarian cancer cells co-cultured with NK-92-αFR-CAR. Secondly, cytokine stimulation experiments confirmed that IFN-γ and TNF-α secreted by NK-92-αFR-CAR synergistically induced high CXCL10 expression in ovarian cancer cells. Further signaling pathway experiments showed that IFN-γ and TNF-α enhanced the activation level of the IFN-γ-IFNGR-JAK1/2-STAT1-CXCL10 signaling axis. Cytotoxicity experiments showed that NK-92-αFR-CAR-CXCR3A cells could not only efficiently kill αFR-positive ovarian cancer cells in vitro but also secrete IFN-γ and TNF-α. Higher migration than that of NK-92-αFR-CAR was detected in NK-92-αFR-CAR-CXCR3A using transwell assay. NK-92-αFR-CAR-CXCR3A effectively killed tumor cells in different mouse xenograft models of ovarian cancer and increased infiltration into tumor tissue.

Conclusion

This study confirmed that IFN-γ and TNF-α secreted by αFR-CAR-engineered NK cells can synergistically induce high expression of CXCL10 in ovarian cancer cells and constructed self-driving αFR-CAR-engineered NK cells that can break through migration barriers based on CXCL10, which may provide a new therapeutic weapon for ovarian cancer.

Graphical abstract

Introduction

Ovarian cancer is the most malignant gynecological tumor that threaten women's health worldwide. Approximately 70 % of ovarian cancer patients are in the advanced stage at initial diagnosis, with a 5-year survival rate of 30 % for distant-stage disease. This is due to the insidious onset of ovarian cancer and the lack of specific screening methods and effective therapeutic strategies [[1], [2], [3]].

In recent years, accumulating preclinical studies have suggested that the use of chimeric antigen receptor (CAR)-modified immune cells, such as CAR-T and CAR-NK cells, could be conducive for ovarian cancer therapy [4,5]. Several CAR-T cell clinical trials in ovarian cancer are underway, targeting different antigens such as NY-ESO-1, HER2, αFR, MSLN, MUC16, and p53. NK cells have been demonstrated to be optimal loading cells for CAR. In addition to providing "off-the-shelf" live cell therapy, CAR-NK cells are widely available in the source, including peripheral blood, umbilical cord blood, human embryonic stem cell-derived NK cells, and the NK92 cell line [6,7]. In numerous clinical trials [8,9], CAR-NK cells have proven to be safer than CAR-T cells. In our previous study, CAR-engineered NK-92 cells targeting folate receptor α (αFR) (NK-92-αFR-CAR) cells were found to be highly targeted and specifically kill ovarian cancer cells in vitro and in vivo, suggesting that CAR-NK is a potential treatment for ovarian cancer [10].

However, the heterogeneity of target antigens, inherent negative immune regulatory mechanism of the tumor microenvironment, migration barrier of CAR-engineered immune cells, and other factors may lead to poor efficacy of CAR-engineered immune cell therapy in solid tumors [11]. Since the migration of CAR-modified immune cells to the tumor microenvironment is the first step in exerting antitumor effects, overcoming the migration barrier of CAR-modified immune cells has aroused researchers' interest. It has been demonstrated that the expression level of chemokines is high in the tumor microenvironment [12], which suggests that the expression of chemokine receptors could promote the migration of CAR-T/NK cells to the tumor microenvironment. Sapoznik et al. found that CXCR1 overexpression enhanced the migration ability of NK cells as well as its killing effect on tumors [13]. Importantly, several studies have constructed CAR-T/NK cells co-expressing chemokine receptors, such as CXCR1, CCR2b, and CXCR2. Compared with CAR-T/NK cells without chemokine receptors, CAR-T/NK cells expressing chemokine receptors can migrate and infiltrate into tumor tissues more effectively, followed by a stronger anticancer effect [12,14,15], suggesting that innovative combinations of chemokines that promote CAR-NK therapy should be explored in ovarian cancer.

In this study, we found that IFN-γ and TNF-α secreted by NK-92-αFR-CAR synergistically induced high CXCL10 expression in tumor cells when NK-92-αFR-CAR killed αFR-positive ovarian cancer cells. In addition, the JAK/STAT1 pathway mediated IFN-γ-induced CXCL10 expression, and TNF-α upregulated CXCL10 and IFN-γ receptor 1 (IFNGR1) expression through the NF-κB pathway. Increased IFNGR1 expression can further enhance the activation of IFN-γ the IFNGR/JAK/STAT1/CXCL10 signaling axis. Previous studies have shown that under the chemotactic effect of CXCL10, CD4+ Th1 cells and CD8+ cytotoxic T cells migrate into tumor tissues, kill tumor cells, and further upregulate the expression of CXCL10, which subsequently recruits more T cells and NK cells to participate in the antitumor immune response [16,17]. Thus, we hypothesized that co-expression of CXC-chemokine receptor 3A (CXCR3A), the receptor of CXCL10, and CAR in NK-92 cells could be used to construct self-driving CAR-NK-92 cells. Specifically, NK-92 cells were modified with CAR targeting αFR to kill αFR-positive ovarian cancer cells with high CXCL10 expression, and CXCL10 could then bind to CXCR3A on NK-92 cells and induce more CAR-engineered NK-92 cells to migrate to the tumor site to exert tumor-killing function. To verify this, we explored the factors causing the high expression of CXCL10 in ovarian cancer cells and its underlying mechanism using enzyme-linked immunosorbent assay (ELISA), RT-qPCR, western blotting, and luciferase reporter assays. In NK-92 cells, CXCR3A was co-expressed with CAR. In vitro and in vivo experiments showed that CAR-NK-92 cells could not only kill αFR-positive ovarian cancer cells efficiently but also migrate into tumor tissues in large numbers under CXCL10 chemotaxis. In conclusion, our study demonstrated for the first time that IFN-γ and TNF-α secreted by anti-αFR CAR (αFR-CAR) engineered immune cells could synergistically induce CXCL10 expression in tumor cells, and that constructing self-chemotactic NK-92-αFR-CAR cells co-expressing CXCR3A could be a novel and potential immunotherapy strategy for ovarian cancer treatment.

Materials and methods

Cell lines and culture

The human ovarian cancer cell lines SK-OV-3 and OVCAR3, human epidermoid carcinoma cell line A-431, and human embryonic kidney cell line 293T were generously provided by the Stem Cell Bank, Chinese Academy of Sciences (Shanghai, China). All cells were cultured in the recommended medium supplemented with 10 % fetal bovine serum (FBS). The human NK cell line, NK-92, was purchased from the American Type Culture Collection (ATCC, Rockville, USA). NK-92 cells were maintained in NK-92 complete medium (CELLCOOK, Guangzhou, China) with 200 IU/mL recombinant human interleukin-2 (IL-2) (SinoBiological, Beijing, China). All cell lines were cultured at 37 °C with 5 % CO₂ in a humidified incubator. Except for the NK-92 complete medium, all other media and FBS were purchased from VivaCell (Shanghai, China).

RNA sequencing (RNA-seq)

Three replicates were used for each group. For the Control group, 5 × 10⁶ SK-OV-3 cells were inoculated into 100 mm dishes. For the NK-92-αFR-CAR group, 5 × 10⁶ SK-OV-3 cells and 25 × 10⁶ NK-92-αFR-CAR cells were co-cultured in 100 mm dishes for 24 h. The cells in each dish were collected separately, and total RNA was isolated using TRIzol reagent (Life Technologies, Carlsbad, CA, USA) according to the manufacturer's instructions. Subsequent RNA-seq and RNA-seq analyses were conducted using LC-Bio Technology (Hangzhou, China). For cDNA library construction, RNA samples with high purity (OD 260/280 ≥ 2.0) and integrity (RIN > 7) were used. Mapping reads, cDNA library construction, and FPKM calculations were performed as previously described [18]. An absolute fold change >1.5 and P < 0.05 were used to determine differentially expressed genes for analysis.

CAR construction and lentivirus production

A previously constructed third-generation αFR-CAR cell line was used in this study [10]. This αFR-CAR contains a CD8α signal peptide, an anti-αFR scFv derived from human antibody C4, a human IgG1 Fc fragment, a CD8α hinge region, a CD28 transmembrane region, and a composite CD28-CD137-CD3ζ intracellular signaling domain. Additionally, the αFR-CAR and CXCR3A sequences (NCBI Database, NM_001504.2) were concatenated with the 2A sequence to obtain the αFR-CAR-CXCR3A sequence. After codon optimization, the CAR sequences were synthesized and ligated into the lentiviral expression plasmid pLenti (OBiO, Shanghai, China). High-titer mock lentiviral particles, as well as lentiviral particles expressing αFR-CAR or αFR-CAR-CXCR3A, were produced by OBiO. Lentivirus transfection and stable transfectant selection procedures were modified from the previously published protocols [10,19] .

Flow cytometry

Flow cytometry was performed as described previously [20]. Briefly, lentivirus-transfected NK-92 cells and untransfected NK-92 cells were collected and washed with PBS containing 4 % bovine serum albumin (BSA) (Beyotime, Beijing, China). The cells were then centrifuged at 1500 rpm for 5 min, and the pellets were resuspended at a final concentration of 0.02 µg/µL antibody in the dark and incubated for 30 min at room temperature. Allophycocyanin (APC)-conjugated goat anti-human IgG Fc antibody (Jackson ImmunoResearch, West Grove, PA, USA) and APC-conjugated goat IgG isotype control antibody (Abcam, Cambridge, UK) were used to detect αFR-CAR expression on the surface of NK-92 cells. The expression of CXCR3A on the surface of NK-92 cells was detected using phycoerythrin (PE)-conjugated mouse anti-human CD183 antibody and PE-conjugated mouse IgG1 κ isotype control antibody (BD Biosciences, San Jose, USA). Finally, the samples were analyzed using an ACEA NovoCyte flow cytometer and NovoExpress software (Agilent Technologies, Hangzhou, China).

Transwell migration assay

A 24-well Transwell chamber (Corning, NY, USA) was placed in a 24-well culture plate (Jet Biofil, Guangzhou, China), which was referred to as the upper chamber, and the culture plate was referred to as the lower chamber. In lower chambers, 5 × 10⁴ SK-OV-3 cells and SK-OV-3 cells stably overexpressing CXCL10 (SK-OV-3-CXCL10) were seeded separately. In the lower chambers, 5 × 10⁵ NK-92-αFR-CAR cells were mixed with SK-OV-3 cells. After 12 h of culture at 37 °C and 5 % CO₂, NK-92, NK-92-EV, NK-92-αFR-CAR, and NK-92-αFR-CAR-CXCR3A cells were inoculated separately into the upper chambers. After culturing for 6 h, suspended cells in the lower chambers were collected and counted.

Lactate dehydrogenase (LDH) cytotoxicity assay

The cytotoxicity of lentivirus-transfected NK-92 cells and untransfected NK-92 cells was measured using an LDH cytotoxicity assay kit (DOJINDO, Tokyo, Japan). Lentivirus-transfected NK-92 cells and untransfected NK-92 cells were referred to as effector cells, and SK-OV-3, OVCAR3, and A-431 cells were referred to as target cells. The effector and target cells were co-cultured in 96-well plates at different E/T ratios for 18 h. Supernatants were collected. The manufacturer's instructions were followed to detect the LDH content in the supernatants and to calculate the specific cell lysis of the effector cells at different E/T ratios.

Quantitative real-time polymerase chain reaction (qPCR)

The qPCR procedure was performed as previously described [21]. Briefly, total RNA was extracted using TRIzol (Life Technologies) and reverse transcribed to cDNA using the PrimeScript^TM RT reagent Kit with gDNA Eraser according to the manufacturer's instruction (Takara Bio, Dalian, China). TB Green® Premix Ex Taq^TM II (Tli RNaseH Plus) (Takara Bio) and CFX96 Real-Time System (Bio-Rad, Hercules, USA) were used to perform qPCR. Relative mRNA expression was calculated using the relative standard curve method (2^−ΔΔCq) using the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene as an internal control [22]. The primer sequences are listed in Supplementary Table 1.

Enzyme-linked Immunosorbent assay (ELISA)

The ELISA was performed according to the manufacturer's instructions. For CXCL10 detection, the lentivirus-transfected NK-92 cells and untransfected NK-92 cells were co-cultured with SK-OV-3, OVCAR3, or A-431 cells, for 24 h; NK-92-αFR-CAR cells were co-cultured with SK-OV-3 or OVCAR3 cells for 24 h and different neutralizing antibodies (0.02 µg/µL) (Abcam) were added during co-culture; SK-OV-3 and OVCAR3 cells were treated with different concentrations of IFN-γ and TNF-α (SinoBiological) for 24 h and different inhibitors or siRNAs were added during the treatment. Thereafter, the supernatants were collected to measure the levels of CXCL10 using a human CXCL10/IP-10 ELISA Kit (BOSTER, Wuhan, China). For the detection of IFN-γ and TNF-α, the lentivirus-transfected NK-92 cells and untransfected NK-92 cells were co-cultured with SK-OV-3, OVCAR3, or A-431 cells, for 24 h, and then IFN-γ and TNF-α in the supernatants were determined using the corresponding ELISA kits (BOSTER).

Western blot

Standard western blotting methods were used for western blotting. Total protein was extracted using RIPA lysis buffer (Beyotime), and protein concentration was determined using a BCA protein assay kit (Beyotime) according to the manufacturer's instructions. The protein samples were denatured in sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) sample loading buffer (Beyotime) at 100 °C for 5 min, separated by SDS-PAGE, and transferred to polyvinylidene difluoride membranes (Millipore, Billerica, USA). The transferred membranes were blocked with 5 % BSA in Tris-buffered saline with 0.1 % Tween 20 (TBST) and probed with the appropriate primary antibodies at 4 °C overnight. The primary antibodies used in this study were rabbit anti-JAK1 monoclonal antibody (mAb), rabbit anti-phospho-JAK1 mAb, rabbit anti-JAK2 mAb, rabbit anti-phospho-JAK2 mAb, rabbit anti-STAT1 mAb, rabbit anti-phospho-STAT1 mAb, rabbit anti-IFNGR1 antibody, rabbit anti-GAPDH mAb, and the NF-κB pathway antibody sampler kit (Cell Signaling Technology, Beverly, USA). Membranes were then incubated with horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG antibody or HRP-conjugated goat anti-mouse IgG antibody (Cell Signaling Technology) at a room temperature for 1 h. The blots on the membranes were visualized using the BeyoECL Plus kit (Beyotime) and detected using the ChemiDoc XRS+ imaging system (Bio-Rad). Intensity values were analyzed using ImageJ software (NIH, Bethesda, USA) and normalized relative to the values of GAPDH.

Inhibitor reagents

The JAK1/2 selective inhibitor Ruxolitinib and the NF-κB inhibitor BAY 11-7082 were purchased from Selleck (Houston, TX, USA). All reagents were used in accordance with the manufacturer's instructions.

siRNAs, expression vectors, and transfection

STAT1, p65, IFNGR1, and control siRNAs were purchased from RioBio (Guangzhou, China). For siRNA transfection, siRNA (150 nM) was used per array and was performed using Lipofectamine RNAiMAX reagent as per the manufacturer's instructions (Life Technologies). The efficiency of siRNA interference was evaluated using qPCR and the results are shown in Supplementary Figure. 3C. In addition, the IFNGR1 and control expression vectors were purchased from OBiO. These expression vectors were transfected with PolyJet according to the manufacturer's instructions (SignaGen, Rockville, USA).

Luciferase reporter assay

A dual-luciferase reporter assay system was used for the luciferase reporter assay. Firefly luciferase reporter plasmids encoding the promoter region of CXCL10 gene or IFNGR1 gene (pGL4.10-CXCL10-WT and pGL4.10-IFNGR1-WT), firefly luciferase reporter plasmids encoding transcription factor-binding site mutations in the promoter region (pGL4.10-CXCL10-mtSTAT1, pGL4.10-CXCL10-mtP65, and pGL4.10-IFNGR1-mtP65), control firefly luciferase reporter plasmids (pGL4.10-NC), transcription factor expression plasmids (pcDNA3.1-NC, pcDNA3.1-STAT1, pIRES2-EGFP, and pIRES2-RELA-EGFP), and Renilla luciferase expression plasmids were constructed using OBiO. Appropriate firefly luciferase reporter, transcription factor expression, and Renilla luciferase expression plasmids were simultaneously transfected into 293T cells using PolyJet (SignaGen, Rockville, USA). Additionally, SK-OV-3 and OVCAR3 cells were co-transfected with appropriate firefly luciferase reporter plasmids and Renilla luciferase expression plasmids using PolyJet and subsequently subjected to different treatments. Luciferase activity was detected using a Dual-Lumi kit (Beyotime) and GloMax 20/20 luminometer (Promega, Madison, WI, USA). Transfection efficiency was normalized to the Renilla luciferase activity. The results indicate fold-changes relative to the corresponding control.

Immunohistochemistry

Immunohistochemical staining was performed according to the standard protocols. At the end of the animal study, subcutaneous tumor tissues were isolated and embedded in paraffin. The embedded tissues were sliced into 5 µm sections. Following hydration, a citrate antigen retrieval solution (Beyotime) was used at 95 °C for 20 min for antigen retrieval. The sections were blocked with 10 % goat serum (Beyotime) at 37 °C for 30 min and incubated with mouse anti-human NCAM1 (CD56) mAb (Cell Signaling Technology) at 4 °C overnight. Subsequently, the sections were washed in TBST and incubated for 1 h at 37 °C with an HRP-conjugated goat anti-mouse IgG antibody (Abcam). The sections were stained with a DAB HRP color development kit (Beyotime) at room temperature for 3–30 min, and counterstained with hematoxylin for 5 min. The stained sections were observed under an Olympus CKX31 microscope (Olympus, Tokyo, Japan).

Xenograft model of ovarian cancer

Xenograft models of ovarian cancer were established using 6 to 8-week-old female NOD-Prkdc^scid IL2rg^tm1/Bcgen mice (B-NDG mice) (BIOCYTOGEN, Beijing, China). All animal experiments were reviewed and approved by the Animal Care and Use Committee of Army Medical University (AMUWEC20216441). For the intraperitoneal xenograft tumor model, model establishment, treatment, and bioluminescence imaging were conducted as previously described [10]. For the subcutaneous xenograft tumor model, the mice were inoculated subcutaneously with 3 × 10⁶ SK-OV-3 or SK-OV-3-CXCL10 cells and divided into four groups. One week after subcutaneous inoculation, each group was treated with the appropriate effector cells or PBS at the indicated doses and time intervals. The time for tumor-bearing mice to receive the first treatment was set as day 0, and the tumor size was measured every two days after that. On day 14, the tumor-bearing mice were sacrificed and the tumor tissues were isolated for imaging and immunohistochemistry. The tumor volume (V) was estimated using the formula: V = L × W² × 0.5 (L represents the longest diameter and W represents the shortest diameter). Moreover, euthanasia was performed when the mice appeared to be in distress, such as ruffled fur, impaired ambulation, difficulty breathing, decreased response to stimuli, or evidence of being moribund.

Statistical analysis

The data are presented as mean ± SEM. The SPSS software (version 20.0; SPSS, Chicago, USA) and GraphPad Prism 6.01 software (GraphPad Software, San Diego, CA, USA) were used for statistical analysis. Student's t-test was used for comparisons between two groups, and one-way analysis of variance was used to compare three or more independent groups. Furthermore, the in vivo bioluminescence signal intensity was analyzed using the Kruskal-Wallis test. The log-rank test was used to compare the Kaplan-Meier survival curves. P < 0.05 was considered statistically significant.

Results

Transcriptomic profiling of ovarian cancer cells co-cultured with different effector cells

To understand the effect of NK-92-αFR-CAR cells on the gene expression profile of ovarian cancer cells, transcriptomic analysis was performed on SK-OV-3 cells co-cultured with NK-92-αFR-CAR or NK-92 cells. In Fig. 1A, the volcano plot shows the overall alteration in gene expression in SK-OV-3 cells after co-culture with the two groups. Hierarchical clustering revealed the profile of differentially expressed genes in SK-OV-3 cells after co-culture with the two groups (Fig. 1B). The top 30 enriched signaling pathways of the differentially expressed genes were obtained using Gene Set Enrichment Analysis (GSEA), among which the chemokine-mediated signaling pathway was enriched substantially (Fig. 1C and D). In the chemokine-mediated signaling pathway, 10 genes were upregulated in SK-OV-3 cells upon treatment with NK-92-αFR-CAR cells, and CXCL10 expression was significantly increased (Fig. 1E). Together, these results demonstrate that co-culture with NK-92-αFR-CAR cells could promote CXCL10 expression and activates the chemokine-mediated signaling pathway in ovarian cancer cells.

Fig. 1.

Expression of CXCL10 in ovarian cancer cells after co-culture with NK-92-αFR-CAR cells. (A) Volcano plot of gene expression alternations in SK-OV-3 cells co-cultured with NK-92-αFR-CAR cells. (B) Heatmap of differentially expressed genes in SK-OV-3 cells co-cultured with NK-92-αFR-CAR cells. (C) The top 30 enriched gene sets from GSEA analysis. (D) GSEA enrichment plots of chemokine-mediated signaling pathway gene set. (E) The top 10 core enriched genes involved in chemokine-mediated signaling pathway gene set. (F) CXCL10 levels in the supernatants of ovarian cancer cells and effector cells cultured alone or in co-culture were detected using ELISA assay. (G) The relative expression levels of CXCL10 mRNA normalized to GAPDH mRNA expression in ovarian cancer cells and effector cells cultured alone or in co-culture were detected using qPCR. The data are representative of three independent experiments with similar results. The data in (F) and (G) are expressed as the means ± SEM of triplicate samples. Statistical analysis shows the comparison between the labeled groups. * represents significant difference, **P < 0.01; ns, P ≥ 0.05.

To verify the transcriptome sequencing result showing that CXCL10 expression was induced after the co-culture of tumor cells with NK-92-αFR-CAR cells, we used an ELISA assay to detect CXCL10 abundance in the co-culture supernatant. As shown in Fig. 1E, the co-culture of αFR-positive ovarian cancer cells (SK-OV-3 and OVCAR3) with NK-92-αFR-CAR did not increase the content of CXCL10 in the co-culture supernatant; the level of CXCL10 was most dramatically elevated in the supernatant when co-cultured with NK-92-αFR-CAR cells. However, co-culture with αFR-negative A-431 cells had a negligible effect on CXCL10 expression in the co-cultured supernatants of NK-92, NK-92-EV, and NK-92-αFR-CAR cells. To further clarify that ovarian cancer cells secreted higher CXCL10 when co-cultured, RT-qPCR was used to analyze the mRNA levels of CXCL10 in SK-OV-3 and OVCAR3 cells upon co-culture with NK-92, NK-92-EV, or NK-92-αFR-CAR cells. As shown in Fig. 1F, CXCL10 mRNA increased in SK-OV-3 and OVCAR3 cells after co-culture with these effector cells, and NK-92-αFR-CAR cells induced the highest level of CXCL10. Conversely, the mRNA level of CXCL10 in NK-92, NK-92-EV, and NK-92-αFR-CAR cells did not change significantly when co-cultured with ovarian cancer cells. These results demonstrate that αFR-positive ovarian cancer cells could secrete higher levels of CXCL10 when co-culture with NK-92-αFR-CAR.

IFN-γ and TNF-α secreted by NK-92-αFR-CAR cells synergistically induce CXCL10 expression in ovarian cancer cells

It has been demonstrated that CXCL10 can be induced by IFN-γ expression in U937 cells [23]. In our previous study, TNF-α was shown to induce the expression of CXCL10 in tumor cells [24]. Our previous findings also showed that NK-92-αFR-CAR cells secrete large amounts of IFN-γ and TNF-α when killing αFR-positive ovarian cancer cells [10]. Thus, we hypothesized that NK-92-αFR-CAR cells induced the expression of CXCL10 in co-cultured ovarian cancer cells through the secretion of IFN-γ and/or TNF-α. To confirm this, specific neutralizing antibodies against IFN-γ and/or TNF-α were added to the medium of NK-92-αFR-CAR cells co-cultured with SK-OV-3 or OVCAR3 cells, and then ELISA and RT-qPCR were used to detect the expression of CXCL10 in ovarian cancer cells. The addition of either IFN-γ-specific neutralizing antibody or TNF-α-specific neutralizing antibody decreased CXCL10 expression in SK-OV-3 and OVCAR3 cells co-cultured with NK-92-αFR-CAR cells, as shown in Fig. 2A and 2B. The simultaneous addition of two neutralizing antibodies could reinforce this effect, suggesting that IFN-γ and TNF-α secreted by NK-92-αFR-CAR cells can induce CXCL10 expression in ovarian cancer cells. To explore the induction effect of IFN-γ and TNF-α on CXCL10 expression in ovarian cancer cells, SK-OV-3 and OVCAR3 cells were treated with different concentrations of IFN-γ or TNF-α, and then CXCL10 expression was detected using ELISA. The results showed that compared with untreated cells, treatment with 0.01 ng/ml or 0.1 ng/ml IFN-γ or TNF-α did not affect the CXCL10 content in the culture supernatant of ovarian cancer cells. However, when the concentration of IFN-γ was increased to 1 ng/ml and the concentration of TNF-α was increased to 10 ng/ml, the content of CXCL10 in the culture supernatant of SK-OV-3 and OVCAR3 cells was significantly increased. Furthermore, ovarian cancer cells treated with the combination of threshold concentrations of IFN-γ (1 ng/ml) and TNF-α (10 ng/ml) increased the level of CXCL10 in the culture supernatant compared to treatment alone (Fig. 2C and 2D). These results indicate that IFN-γ and TNF-α secreted by NK-92-αFR-CAR cells synergistically induce high CXCL10 expression in ovarian cancer cells.

Fig. 2.

Synergistic induction of IFN-γ and TNF-α secreted by NK-92-αFR-CAR cells on CXCL10 expression in ovarian cancer cells. (A) CXCL10 levels in the supernatants of ovarian cancer cells co-cultured with NK-92-αFR-CAR cells in the presence or absence of specific neutralizing antibodies were detected using ELISA assay. IgG2a represents isotype control antibody, αIFN-γ represents anti-IFN-γ neutralizing antibody, and αTNF-α represents anti-TNF-α neutralizing antibody. (B) The relative expression levels of CXCL10 mRNA normalized to GAPDH mRNA expression in ovarian cancer cells co-cultured with NK-92-αFR-CAR cells in the presence or absence of specific neutralizing antibodies were detected using qPCR. (C) CXCL10 levels in the supernatants of ovarian cancer cells treated with IFN-γ or TNF-α at different concentrations were determined using ELISA assay. (D) The relative expression levels of CXCL10 mRNA normalized to GAPDH mRNA expression in ovarian cancer cells treated with IFN-γ (1 ng/mL) and TNF-α (10 ng/mL) alone or in combination were detected using qPCR. All data are expressed as the means ± SEM of three independent experiments. Statistical analysis shows the comparison between the labeled groups. * represents significant difference, *P < 0.05; **P < 0.01; ns, P ≥ 0.05.

IFN-γ induces CXCL10 expression in ovarian cancer cells by activating JAK1/2-STAT1 signaling pathway

As IFN-γ stimulation can activate the intracellular JAK1/2-STAT1 signaling pathway to produce various biological effects [25], we attempted to determine whether JAK1/2-STAT1 signaling mediates IFN-γ-induced CXCL10 expression in ovarian cancer cells. Ruxolitinib (a selective JAK1/2 inhibitor) was used to pretreat SK-OV-3 and OVCAR3 cells before stimulating them with IFN-γ, and CXCL10 expression was detected using ELISA and RT-qPCR. The results showed that the induction effect of IFN-γ on CXCL10 was disrupted upon the inhibition of JAK1/2 activity by ruxolitinib (Fig. 3A and 3B). Western blotting showed that JAK1, JAK2, and STAT1 protein phosphorylation levels increased after IFN-γ stimulation in ovarian cancer cells, and pretreatment with ruxolitinib significantly reduced the phosphorylation levels of these proteins (Fig. 3C and Supplementary Figure. 1). Moreover, knockdown of STAT1 expression using small interfering RNA (siRNA) decreased CXCL10 expression induced by IFN-γ stimulation (Fig. 3D). Thus, these results suggest that IFN-γ could induce CXCL10 expression in ovarian cancer cells through activation of the intracellular JAK1/2-STAT1 signaling pathway.

Fig. 3.

Mediated signaling pathways of IFN-γ-induced and TNF-α-induced CXCL10 expression in ovarian cancer cells. (A) CXCL10 levels in the supernatants of Ruxolitinib-pretreated ovarian cancer cells stimulated with IFN-γ were determined using ELISA assay.The concentration of IFN-γ is 1ng/ml. (B) The relative expression levels of CXCL10 mRNA normalized to GAPDH mRNA expression in Ruxolitinib-pretreated ovarian cancer cells stimulated with IFN-γ were detected using qPCR. (C) Phosphorylation levels of JAK1/2-STAT1 signaling pathway in Ruxolitinib-pretreated ovarian cancer cells stimulated with IFN-γ were detected using Western Blot. p- means phosphorylated. (D) The relative expression levels of CXCL10 mRNA normalized to GAPDH mRNA expression in siRNA-transfected ovarian cancer cells stimulated with IFN-γ were detected using qPCR. si-NC represents control siRNA, si-STAT1 represents STAT1 siRNA. (E) and (L) Schematic representation of the constructed firefly luciferase reporter plasmids. (F) IFN-γ-induced CXCL10 promoter activity in the plasmid-transfected ovarian cancer cells. (G) STAT1-induced CXCL10 promoter activity in the plasmid-transfected 293T cells. (H) CXCL10 levels in the supernatants of BAY 11-7082-pretreated ovarian cancer cells stimulated with TNF-α were determined using ELISA assay. (I) The relative expression levels of CXCL10 mRNA normalized to GAPDH mRNA expression in BAY 11-7082-pretreated ovarian cancer cells stimulated with TNF-α were detected using qPCR. (J) Phosphorylation levels of NF-κB signaling pathway in BAY 11-7082-pretreated ovarian cancer cells stimulated with TNF-α were detected using Western Blot. (K) The relative expression levels of CXCL10 mRNA normalized to GAPDH mRNA expression in siRNA-transfected ovarian cancer cells stimulated with TNF-α were detected using qPCR. si-P65 represents P65 siRNA. (M) TNF-α-induced CXCL10 promoter activity in the plasmid-transfected ovarian cancer cells. (N) P65-induced CXCL10 promoter activity in the plasmid-transfected 293T cells. All data are expressed as the means ± SEM of three independent experiments. Statistical analysis shows the comparison between the labeled groups. * represents significant difference, **P < 0.01; ns, P ≥ 0.05.

Next, we investigated whether STAT1, a transcription factor in the JAK1/2-STAT1 signaling pathway, can directly regulate the transcription of CXCL10 in IFN-γ-stimulated ovarian cancer cells. The region (GTTCTAGGAAC) in the CXCL10 promoter could be targeted by STAT1 through JASPAR prediction (http://jaspar.genereg.net/). Luciferase reporter plasmids containing the wild-type CXCL10 promoter (pGL4.10-CXCL10-WT) and a mutated STAT1 binding site (pGL4.10-CXCL10-mtSTAT1) were constructed (Fig. 3E). Two reporter plasmids and the control plasmid (pGL4.10-NC) were transfected into ovarian cancer cells. After IFN-γ stimulation, the luciferase activity was detected in each group. The results showed that ovarian cancer cells transfected with pGL4.10-CXCL10-WT exhibited enhanced luciferase activity compared to cells transfected with pGL4.10-NC. However, luciferase activity in ovarian cancer cells transfected with pGL4.10-CXCL10-mtSTAT1 did not change significantly (Fig. 3F). In addition, STAT1 overexpression increased the luciferase activity of 293T cells transfected with pGL4.10-CXCL10-WT, but not pGL4.10-CXCL10-mtSTAT1(Fig. 3G). Collectively, these results indicate that IFN-γ activates the JAK1/2-STAT1 signaling pathway and promotes CXCL10 gene transcription in ovarian cancer cells.

TNF-α induces CXCL10 expression in ovarian cancer cells through the activation of intracellular NF-κB signaling pathway

In our previous study, we found that the NF-κB signaling pathway is involved in TNF-α-induced CXCL10 expression in colorectal cancer cells [23]. To determine whether TNF-α induces CXCL10 expression in ovarian cancer through this regulatory axis, we pretreated SK-OV-3 and OVCAR3 cells with BAY 11-7082, a selective NF-κB inhibitor, stimulated cells with TNF-α, and then examined its effect on the expression of CXCL10 in ovarian cancer cells. According to ELISA and RT-qPCR results, TNF-α stimulation could upregulate the expression of CXCL10 in ovarian cancer cells, but inhibition of the NF-κB signaling pathway by BAY 11-7082 could decrease the expression level of CXCL10 (Fig. 3H and Fig. 3I). Western blotting showed that the phosphorylation levels of IKK-α/β, IKB-α, and p65 were significantly increased after TNF-α stimulation in ovarian cancer cells, but treatment with BAY 11-7082 could reverse the phosphorylation levels of these proteins (Fig 3J and Supplementary Figure. 2). Furthermore, p65 knockdown resulted in a significant decrease in CXCL10 expression compared to that in the control group (Fig. 3K). These results suggest that TNF-α induces CXCL10 expression in ovarian cancer cells by activating the intracellular NF-κB signaling pathway.

Next, we explored the regulation of CXCL10 expression by transcription factor p65 in the NF-κB signaling pathway. Two regions, CGGAATTTCC and GGGAAGTCCC, in the CXCL10 promoter were predicted to be p65 binding sites. Luciferase reporter plasmids containing the CXCL10 promoter sequence with p65 binding sites (pGL4.10-CXCL10-WT) or mutant (pGL4.10-CXCL10-MTP65) were constructed (Fig. 3L). As shown in Fig. 3M, TNF-α stimulated the intracellular luciferase activity of SK-OV-3 and OVCAR3 cells transfected with pGL4.10-CXCL10-WT, but not those transfected with pGL4.10-CXCL10-MTP65. As expected, p65 overexpression promoted luciferase activity in cells transfected with pGL4.10-CXCL10-WT, but not those transfected with pGL4.10-NC or pGL4.10-CXCL10-MTP65 (Fig. 3N). These results demonstrate that TNF-α induces CXCL10 transcriptional expression by activating of the NF-κB signaling pathway in ovarian cancer cells.

TNF-α enhanced the activation level of IFN-γ-IFNGR-JAK1/2-STAT1-CXCL10 signal axis by upregulating IFNGR1 expression

As CXCL10 expression was higher in ovarian cancer cells stimulated with the combination of IFN-γ and TNF-α than in cells stimulated alone (Fig. 2D), we investigated whether the JAK1/2-STAT 1 and NF-κB signaling pathways were involved in the synergistic induction of high expression of CXCL10 by IFN-γ and TNF-α. Western blotting was used to detect the activation levels of the JAK1/2-STAT 1 and NF-κB signaling pathways after treatment of SK-OV-3 and OVCAR3 cells with stimulation by IFN-γ, TNF-α, and IFN-γ + TNF-α. Compared to TNF-α stimulation alone, which increased the protein phosphorylation levels of IKK-α/β, IKB-α, and p65, IFN-γ + TNF-α stimulation did not significantly alter the phosphorylation levels of these proteins (Fig. 4A). However, the phosphorylation levels of JAK1, JAK2, and STAT1 were elevated in IFN-γ + TNF-α-stimulated ovarian cancer cells compared to IFN-γ stimulation alone, indicating that the combined stimulation of IFN-γ and TNF-α could enhance the activation level of the JAK1/2-STAT1 signaling pathway in ovarian cancer cells compared to IFN-γ stimulation alone.

Fig. 4.

Mediated signaling pathway of TNF-α-induced IFNGR1 expression in ovarian cancer cells. (A) Phosphorylation levels of JAK1/2-STAT1 and NF-κB signaling pathways in ovarian cancer cells stimulated with IFN-γ or/and TNF-α were detected using Western Blot. p- means phosphorylated. (B) Phosphorylation levels of JAK1/2-STAT1 signaling pathway in siRNA-transfected ovarian cancer cells stimulated with IFN-γ and TNF-α. IFN-γ + TNF-α represents IFN-γ and TNF-α combined to treat. si-NC represents control siRNA, si-IFNGR1 represents IFNGR1 siRNA. (C) The relative expression levels of CXCL10 mRNA normalized to GAPDH mRNA expression in siRNA-transfected ovarian cancer cells stimulated with IFN-γ and TNF-α were detected using qPCR. (D) The relative expression levels of IFNGR1 mRNA normalized to GAPDH mRNA expression in siRNA-transfected ovarian cancer cells stimulated with IFN-γ and TNF-α were detected using qPCR. si-P65 represents P65 siRNA. (E) Schematic representation of the constructed firefly luciferase reporter plasmids. (F) IFN-γ and TNF-α-induced IFNGR1 promoter activity in the plasmid-transfected ovarian cancer cells. (G) P65-induced IFNGR1 promoter activity in the plasmid-transfected 293T cells. (H) The relative expression levels of CXCL10 mRNA normalized to GAPDH mRNA expression in ovarian cancer cells transfected with siRNA and expression vector were detected using qPCR. OV-NC represents control expression vector, OV-IFNGR1 represents IFNGR1 expression vector. All data are expressed as the means ± SEM of three independent experiments. Statistical analysis shows the comparison between the labeled groups. * represents significant difference, *P < 0.05; **P < 0.01; ns, P ≥ 0.05.

IFNGR1 is a membrane receptor for IFN-γ that binds to IFN-γ and transmits activation signals to downstream intracellular signaling pathways. As shown in Fig. 4A and Supplementary Fig. 3A, IFNGR1 expression significantly increased in cancer cells in the presence of TNF-α stimulation (either TNF-α stimulation alone or IFN-γ + TNF-α stimulation). Thus, we hypothesized that during combined IFN-γ and TNF-α stimulation of ovarian cancer cells, TNF-α could upregulate the expression of IFNGR1 and enhance the activation level of the IFN-γ-IFNGR-JAK1/2-STAT1-CXCL10 signaling axis. To verify this hypothesis, the activation level of the JAK1/2-STAT1 signaling pathway was detected upon IFNGR1 knockdown in ovarian cancer cells stimulated by IFN-γ + TNF-α. Results showed that IFNGR1 depletion significantly reduced the phosphorylation levels of JAK1, JAK2, and STAT1 (Fig. 4B and Supplementary Figure. 3B). Subsequently, CXCL10 expression was detected using RT-qPCR after IFNGR1 knockdown. Compared with the IFN-γ + TNF-α group, IFNGR1 knockdown significantly decreased CXCL10 mRNA levels (Fig. 4C). These results indicate that when IFN-γ and TNF-α were combined to stimulate ovarian cancer cells, the upregulation of IFNGR1 expression level could enhance the activation level of the intracellular JAK1/2-STAT1 signaling pathway and then significantly promote CXCL10 expression.

As the activation level of the NF-κB signaling pathway and IFNGR1 expression increased in ovarian cancer cells upon TNF-α stimulation (TNF-α stimulation alone and IFN-γ + TNF-α stimulation) (Fig. 4A), we speculated that TNF-α might induce IFNGR1 expression through the NF-κB signaling pathway. To confirm this, the mRNA level of IFNGR1 was measured in ovarian cancer cells stimulated with IFN-γ and TNF-α upon p65 knockdown. As shown in Fig. 4D, the knockdown of p65 expression induced a significant decrease in IFNGR1 mRNA expression. Furthermore, the sequence of AGGATTTTCC in the IFNGR1 promoter was predicted as the p65 binding site. Luciferase reporter plasmids containing the wild-type IFNGR1 gene promoter sequence (pGL4.10-IFNGR1-WT) and p65 binding site mutation (pGL4.10-IFNGR1-MTp65) were constructed (Fig. 4E). The reporter plasmids and control plasmids (pGL4.10-NC) were transfected into SK-OV-3 and OVCAR3 cells, respectively. After IFN-γ + TNF-α stimulation, the luciferase activity was detected in each group. The results showed that the luciferase activity of ovarian cancer cells transfected with pGL4.10-IFNGR1-WT but not pGL4.10-IFNGR1-MTP65 was significantly enhanced after IFN-γ + TNF-α stimulation (Fig. 4F). In addition, p65 overexpression promoted luciferase activity in cells transfected with pGL4.10-IFNGR1-WT but not pGL4.10-NC or pGL4.10-IFNGR1-MTP65 (Fig. 4G). These results indicate that TNF-α induces the transcriptional expression of IFNGR1 by activating the NF-κB signaling pathway in ovarian cancer cells stimulated by the combination of IFN-γ and TNF-α.

To further confirm that TNF-α upregulates the expression of IFNGR1 and then enhances the activation of the IFN-γ-IFNGR-JAK1/2-STAT1-CXCL10 signaling axis during IFN-γ + TNF-α stimulation of ovarian cancer cells, CXCL10 expression was detected upon p65 knockdown, followed by IFNGR1 overexpression. Compared to IFN-γ + TNF-α stimulated cells, downregulation of p65 significantly decreased CXCL10 mRNA levels in ovarian cancer cells, whereas overexpression of IFNGR1 restored CXCL10 expression (Fig. 4H). Fig. 8 shows a schematic depicting the mechanism of CXCL10 secretion by NK-92-αFR-CAR cells after co-culture with ovarian cancer cells. These results indicate that on the one hand, IFN-γ and TNF-α induce the expression of CXCL10 in ovarian cancer cells through JAK1/2-STAT1 signaling pathway and NF-κB signaling pathway, respectively. On the other hand, TNF-α up-regulates the expression of IFNGR1 through the NF-κB signaling pathway and enhances the activation level of the IFN-γ-IFNGR-JAK1/2-STAT1-CXCL10 signaling axis, which jointly mediates the high expression of CXCL10 in ovarian cancer cells.

Fig. 8.

Schematic of positive feedback for self-driving of NK-92-αFR-CAR-CXCR3A cells based on IFN-γ and TNF-α synergistically induced high CXCL10 expression. In the antitumor process, IFN-γ and TNF-α secreted by NK-92-αFR-CAR-CXCR3A cells synergistically induced high expression of CXCL10 in ovarian cancer cells through JAK-STAT1 and NF-κB signaling pathways, respectively. Under the action of CXCL10, NK-92-αFR-CAR-CXCR3A cells continue to chemotactically migrate into tumor tissues, which can not only make ovarian cancer cells secrete more CXCL10, but also further enhance the antitumor effects.

Design and construction of self-driving anti-αFR-CAR (αFR-CAR) modified NK-92 cells co-expressing CXCR3A

Considering that CXCL10 acts as a chemokine that induces directed migration of the corresponding receptor CXCR3A-positive cells, we constructed self-driving NK-92-αFR-CAR cells co-expressing CXCR3A (NK-92-αFR-CAR-CXCR3A). When NK-92-αFR-CAR-CXCR3A cells kill αFR-positive ovarian cancer cells, IFN-γ and TNF-α secreted by the NK-92 cells cause tumor cells to overexpress CXCL10, which binds to CXCR3A on NK-92 cells and recruits more NK-92-αFR-CAR-CXCR3A cells to the tumor microenvironment. More NK-92-αFR-CAR-CXCR3A cells that kill tumor cells increase IFN-γ and TNF-α secretion, resulting in the formation of a positive feedback from NK-92-αFR-CAR-CXCR3A cells. To achieve this, the sequence co-expressing αFR-CAR and CXCR3A (αFR-CAR-CXCR3A) was designed, as shown in Fig. 5A. The self-cleaved polypeptide 2A sequence was used to link the αFR-CAR sequence with the CXCR3A sequence. Among these, the previously constructed third-generation αFR-CAR was selected for this study [10].

Fig. 5.

Construction and expression of αFR-CAR-CXCR3A in NK-92 cells. (A) Schematic representation of αFR-CAR-CXCR3A. (B) The relative expression levels of αFR-CAR mRNA normalized to GAPDH mRNA expression in NK-92 cells were detected using qPCR. (C) The relative expression levels of CXCR3A mRNA normalized to GAPDH mRNA expression in NK-92 cells were detected using qPCR. The data are expressed as the means ± SEM of triplicate samples. (D) Surface expression of αFR-CAR on NK-92 cells was analyzed using flow cytometry. (E) Surface expression of CXCR3A on NK-92 cells was analyzed using flow cytometry. The filled green histograms indicate isotype control, whereas the filled red histograms indicate the αFR-CAR or CXCR3A expression.

The αFR-CAR-CXCR3A gene sequence was inserted into the lentiviral vector for packaging. By infecting NK-92 cells with empty lentiviral particles, lentiviral particles carrying the αFR-CAR gene, or the αFR-CAR-CXCR3A gene, NK-92-EV, NK-92-αFR-CAR, and NK-92-αFR-CAR-CXCR3A cells were obtained, respectively. After puromycin selection, αFR-CAR and CXCR3A expression was detected in modified NK-92 cells. αFR-CAR mRNA was highly expressed in NK-92-αFR-CAR and NK-92-α FR-CAR-CXCR3A cells compared to NK-92-EV and NK-92 cells, but CXCR3A mRNA was only highly expressed in NK-92-αFR-CAR-CXCR3A cells (Fig. 5B and 5C). Similarly, higher levels of αFR-CAR were found on the cell surfaces of NK-92-αFR-CAR and NK-92-αFR-CAR-CXCR3A cells (Fig. 5D), but only NK-92-αFR-CAR-CXCR3A cells expressed higher levels of CXCR3A (Fig. 5E). These results demonstrated that αFR-CAR and CXCR3A were stably co-expressed in the selected NK-92-αFR-CAR-CXCR3A cells.

In vitro antitumor and chemotactic activities of self-driving NK-92-αFR-CAR-CXCR3A cells

Lactate dehydrogenase (LDH) assay was used to evaluate the antitumor activity of NK-92-αFR-CAR-CXCR3A cells in vitro. As shown in Fig. 6A, when co-cultured with SK-OV-3 cells or OVCAR3 cells, NK-92-αFR-CAR-CXCR3A and NK-92-αFR-CAR cells showed a stronger killing effect than NK-92-EV and NK-92 cells, regardless of the effector target ratio (E/T ratio). The higher the E/T ratio, the stronger the killing effect. However, no significant difference was observed in the cytotoxic activity of NK-92-αFR-CAR-CXCR3A and NK-92-αFR-CAR cells against ovarian cancer cells. Furthermore, NK-92-αFR-CAR and NK-92-αFR-CAR-CXCR3A cells showed similar cytotoxic effects on A-431 cells as NK-92 and NK-92-EV cells. These results suggest that, similar to NK-92-αFR-CAR, self-driving NK-92-αFR-CAR-CXCR3A cells can kill αFR-positive tumor cells with high efficiency and specificity in vitro.

Fig. 6.

Specific cytotoxicity, cytokine secretion, and chemotaxis of NK-92-αFR-CAR-CXCR3A cells. (A) Cell killing by NK-92 cells, NK-92-EV cells, NK-92-αFR-CAR cells, and NK-92-αFR-CAR-CXCR3A cells was determined in the LDH cytotoxicity assay after co-culture with target cells at the indicated E/T ratios. (B) and (C) IFN-γ (B) and TNF-α (C) release of NK-92, NK-92-EV, NK-92-αFR-CAR, and NK-92-αFR-CAR-CXCR3A cells in the presence or absence of SK-OV-3, OVCAR3, or A-431 cells using the ELISA assay. (D) CXCL10 release of SK-OV-3 cells and OVCAR3 cells in the presence or absence of NK-92 cells, NK-92-EV cells, NK-92-αFR-CAR cells, or NK-92-αFR-CAR-CXCR3A cells using the ELISA assay. (E) and (F) Chemotaxis of NK-92, NK-92-EV, NK-92-αFR-CAR, and NK-92-αFR-CAR-CXCR3A cells under CXCL10 secreted by SK-OV-3 cells was analyzed by transwell migration assay. All data are expressed as the means ± SEM of three independent experiments. Statistical analysis shows the comparison between the labeled groups. * represents significant difference, *P < 0.05; **P < 0.01; ns, P ≥ 0.05.

The cytokine secretion levels of NK-92-αFR-CAR-CXCR3A cells co-cultured with tumor cells were detected using ELISA. Higher levels of IFN-γ and TNF-α were detected in the co-culture supernatant of NK-92-αFR-CAR-CXCR3A and NK-92-αFR-CAR cells when co-cultured with SK-OV-3 or OVCAR3 cells (Fig. 6B and 6C). However, no differences in IFN-γ and TNF-α abundance in the co-cultured supernatants of the distinct modified NK-92 cells were found when co-cultured with A-431 cells. Moreover, the content of CXCL10 in the co-culture supernatant was detected, and the results were similar to those in Fig. 2A. In the co-culture supernatant of αFR-positive tumor cells and NK-92-αFR-CAR or NK-92-αFR-CAR-CXCR3A cells, high CXCL10 content was detected (Fig. 6D), indicating that NK-92-αFR-CAR-CXCR3A cells could induce CXCL10 expression in tumor cells during the killing process.

To assess the chemotaxis of NK-92-αFR-CAR-CXCR3A cells, two different transwell assays were designed. In the first transwell assay, SK-OV-3 and NK-92-αFR-CAR cells were inoculated into the culture chamber, where SK-OV-3 cells secreted CXCL10 while being killed, as shown in Fig. 6E. NK-92, NK-92-EV, NK-92-αFR-CAR, and NK-92-αFR-CAR-CXCR3A cells were inoculated into a transwell chamber. Changes in cell numbers in the culture chamber, which reflected the chemotaxis of distinct engineered NK-92 cells, were detected. The results showed that more NK-92-αFR-CAR-CXCR3A cells migrated into the culture chamber than the other effector cells (Fig. 6E). In the second transwell assay, only SK-OV-3 cells stably overexpressing CXCL10 (SK-OV-3-CXCL10) were inoculated into the culture chamber, and NK-92, NK-92-EV, NK-92-αFR-CAR, or NK-92-αFR-CAR-CXCR3A cells were inoculated into the transwell chamber, as described above(Fig. 6F). Similar results were found, in which more NK-92-αFR-CAR-CXCR3A cells migrated into the culture chamber than other effector cells. Taken together, these data suggest that CXCL10 can effectively induce the migration of self-chemotactic NK-92-α FR-CAR-CXCR3A cells.

In vivo antitumor and chemotactic activities of self-driving NK-92-αFR-CAR-CXCR3A cells

The antitumor effect of NK-92-αFR-CAR-CXCR3A cells in vivo was evaluated using a subcutaneous tumor model (Fig. 7A). First, each B-NDG mouse was subcutaneously injected with SK-OV-3 cells. One week after inoculation, the mice began receiving an effector cell infusion (denoted as day 0). On days 0, 3, 6, and 9, NK-92-EV, NK-92-αFR-CAR, and NK-92-αFR-CAR-CXCR3A cells were injected into the tail veins of the mice. The volume size of the subcutaneous tumors was measured every day from day 0 to day 14. We found that the tumor volume in the control and NK-92-EV groups increased rapidly, but NK-92-αFR-CAR cells delayed tumor proliferation (Fig. 7B). Notably, tumor volume in mice treated with NK-92-α FR-CAR-CXCR3A cells dramatically declined, implying that NK-92-αFR-CAR-CXCR3A cells killed tumor cells and significantly inhibited tumor growth (Fig. 7C). Furthermore, immunohistochemical analysis of the marker CD56 positive for NK-92 cells showed that compared with the control group, a small number of NK-92-EV cells infiltrated the subcutaneous tumors of mice, while an increasing number of NK-92-αFR-CAR and NK-92-αFR-CAR-CXCR3A cells infiltrated the subcutaneous tumors of mice, with NK-92-αFR-CAR-CXCR3A cells having the highest level of tumor infiltration (Fig. 7D). Similar results were obtained using the subcutaneous tumor model based on the SK-OV-3-CXCL10 cells followed using a subcutaneous tumor model with distinct groups of effector cells (Supplementary Figure. 4A). The tumor volume in each group showed a trend similar to that of the first subcutaneous tumor model (Supplementary Figure. 4B and Supplementary Figure. 4C). The level of tumor infiltration in NK-92-αFR-CAR-CXCR3A cells was also significantly higher than that in other groups (Supplementary Figure. 4D). These results indicate that self-driving NK-92-αFR-CAR-CXCR3A cells could not only kill αFR-positive ovarian cancer cells as effectively as NK-92-αFR-CAR cells in vivo, but they could also migrate into tumor tissues in large numbers during the killing process due to the chemotactic effect of CXCL10.

Fig. 7.

In vivo antitumor activity and chemotaxis of NK-92-αFR-CAR-CXCR3A cells. (A), (E) Schematic representation of the construction of different mouse xenograft models of ovarian cancer and the corresponding treatment protocols. s.c. represents subcutaneous injection, i.v. represents intravenous injection, i.p. represents intraperitoneal injection. (B) Subcutaneous tumor tissues were isolated for photographing at the endpoint of the animal study. (C) Subcutaneous tumor volumes of each group were estimated every two days to monitor ovarian cancer development. (D) Chemotactic migration of NK-92 cells into subcutaneous tumor tissues were detected using immunohistochemistry. On the top was a representative immunohistochemistry picture of each group. Scale bars: 100 µm. On the bottom was quantification summary of CD56 positive stained cells per field view in each group. Five randomly selected fields of view were quantified for each group. (F) Intraperitoneal tumor development was determined by in vivo bioluminescence imaging at day 0 and day 14. (G) Statistical analysis of bioluminescence signal intensity in each group shown in (F). (H) Kaplan-Meier survival curves of intraperitoneal tumor-bearing mice treated with PBS, NK-92-EV cells, NK-92-αFR-CAR cells, or NK-92-αFR-CAR-CXCR3A cells. The data in (C), (D), and (G) are expressed as the means ± SEM of five mice for all groups. Statistical analysis shows the comparison between the labeled groups. * represents significant difference, *P < 0.05; **P < 0.01; ns, P ≥ 0.05.

To further evaluate the antitumor effect of self-chemotactic NK-92-αFR-CAR-CXCR3A cells in vivo, an abdominal tumor model was constructed using SK-OV-3 cells stably expressing firefly luciferase (SK-OV-3-LUC), and tumor growth was monitored using bioluminescence imaging. As shown in Fig. 7E, each B-NDG mouse was intraperitoneally injected with SK-OV-3-LUC cells. Two weeks after inoculation, the mice began receiving effector cell infusions (denoted as day 0) on days 0, 3, 6, and 9. On day 14, the intraperitoneal bioluminescence signal of the model mice receiving NK-92-αFR-CAR-CXCR3A cells was significantly decreased compared with that of the other groups (Fig. 7F and 7G). Moreover, the survival time of the mice in the NK-92-αFR-CAR-CXCR3A group was significantly longer (Fig. 7H). These results showed that NK-92-αFR-CAR-CXCR3A cells effectively eliminated αFR-positive ovarian cancer cells in vivo.

Discussion

Ovarian, cervical, and endometrial cancers are the three major malignant tumors that threaten women's health, with ovarian cancer having the highest mortality rate [26]. A large number of studies have shown that CAR-engineered immune cells are a promising immunotherapy for ovarian cancer [27]. Previous studies have focused on the construction of CAR-T/NK cells based on novel targets and have demonstrated the effective killing of ovarian cancer cells. Few investigators have focused on other biological changes in tumor cells during the antitumor process of CAR-T/NK cells. Therefore, in this study, transcriptomic analysis was performed on SK-OV-3 cells co-cultured with NK-92-αFR-CAR cells and SK-OV-3 cells cultured alone to explore whether these changes occur. The results showed that the gene expression profile of ovarian cancer cells was significantly altered after co-culture with NK-92-αFR-CAR cells. Among the upregulated genes, CXCL10 expression showed the greatest increase. According to ELISA and RT-qPCR, when NK-92-αFR-CAR cells killed αFR-positive ovarian cancer cells, the tumor cells secreted CXCL10 in large quantities. Luster et al. cloned CXCL10, also known as interferon-gamma-inducible protein 10, from the gene expression products of the human mononuclear cell line U937 after IFN-γ treatment [23]. Previous studies have shown that TNF-α can induce tumor cells to express CXCL10, and that NK-92-αFR-CAR cells secrete a large amount of IFN-γ and TNF-α when killing αFR-positive ovarian cancer cells [24]. Therefore, we hypothesized that IFN-γ or TNF-α might induce high CXCL10 expression in tumor cells when αFR-positive ovarian cancer cells were co-cultured with NK-92-αFR-CAR cells. The results of the IFN-γ- and TNF-α-specific antibody neutralization assay confirmed the hypothesis that IFN-γ and TNF-α secreted by NK-92-αFR-CAR cells can induce high expression of CXCL10 in ovarian cancer cells. Unexpectedly, we found that the level of CXCL10 in the culture supernatant of ovarian cancer cells treated with the combination of threshold concentrations of IFN-γ and TNF-α was significantly higher than that in cells treated with threshold concentrations of IFN-γ or TNF-α alone, suggesting that IFN-γ and TNF-α have a synergistic induction effect on the expression of CXCL10 in ovarian cancer.

Studies have confirmed that IFN-γ induces CXCL10 expression in human salivary gland ductal cells, keratinocytes, glomerular mesangial cells, and breast cancer cells by activating the JAK-STAT1 signaling pathway [28,29]. However, whether this pathway also mediates IFN-γ-induced CXCL10 expression in ovarian cancer cells has not been clearly reported. In this study, we observed that pretreatment of ovarian cancer cells with a JAK1/2 selective inhibitor significantly reduced IFN-γ-induced CXCL10 expression. Western blotting results showed that the protein phosphorylation levels of JAK1, JAK2, and STAT1 in SK-OV-3 and OVCAR3 cells were significantly increased after IFN-γ stimulation. In addition, the siRNA interference assay and luciferase reporter assay showed that STAT1 mediated CXCL10 gene transcription after IFN-γ stimulation in ovarian cancer cells. These results indicate that IFN-γ also induces the expression of CXCL10 through the activation of the JAK1/2-STAT1 signaling pathway in ovarian cancer cells. According to some studies, NF-κB signaling pathway may be involved in the induction of CXCL10 expression by TNF-α since TNF-α usually activates the intracellular NF-κB signaling pathway to produce downstream biological effects [30]. Therefore, this study aimed to investigate the role of TNF-α in ovarian cancer and find out whether TNF-α induces CXCL10 expression through these molecular mechanisms. After BAY 11-7082 was used to inhibit the activation of the NF-κB signaling pathway in ovarian cancer cells, the expression level of CXCL10 induced by TNF-α was significantly decreased. Western blotting results showed that the phosphorylation levels of signal transduction molecules in the NF-κB signaling pathway were significantly increased after TNF-α stimulation in ovarian cancer cells. The siRNA interference and luciferase reporter assays further confirmed that TNF-α initiated CXCL10 gene transcription by activating the NF-κB signaling pathway. These results suggest that TNF-α induces CXCL10 expression in ovarian cancer cells through the activation of the NF-κB signaling pathway. Previous studies have also shown that this pathway mediates TNF-α-induced CXCL10 expression in colorectal cancer cells [24]. In addition, the activation level of the JAK1/2-STAT1 signaling pathway in ovarian cancer cells stimulated by IFN-γ and TNF-α was significantly higher than that stimulated by IFN-γ alone, and the expression level of IFNGR1 was significantly increased, suggesting that when IFN-γ and TNF-α are combined to stimulate ovarian cancer cells, TNF-α may upregulate the expression of IFNGR1, thereby enhancing the activation level of the IFN-γ-IFNGR-JAK1/2-STAT1-CXCL10 signaling axis. The siRNA interference assay confirmed that the upregulation of IFNGR1 expression mediated a further increase in the activation level of the JAK1/2-STAT1 signaling pathway and a large increase in the expression level of CXCL10. Interestingly, the results of the luciferase reporter assay showed that the NF-κB signaling pathway also mediates IFNGR1 gene transcription after TNF-α stimulation in ovarian cancer cells. These results indicated that when IFN-γ and TNF-α were combined to stimulate ovarian cancer cells, IFN-γ and TNF-α induced the expression of CXCL10 through JAK1/2-STAT1 signaling pathway and NF-κB signaling pathway, while TNF-α up-regulated the expression of IFNGR1 through NF-κB signaling pathway. Furthermore, the activation of the IFN-γ-IFNGR-JAK1/2-STAT1-CXCL10 signaling axis was enhanced. These two mechanisms jointly mediate the increased expression of CXCL10 induced by IFN-γ and TNF-α in ovarian cancer cells.

According to the N-terminus, CXCR3, also known as G protein-coupled receptor 9, can be divided into three different splicing variants: CXCR3A, CXCR3B, and CXCR3 -ALT. Among them, CXCR3A and CXCR3B can bind to CXCL9, CXCL10, and CXCL11, while CXCR3-ALT only binds to CXCL11. Current studies mostly focus on CXCR3A and CXCR3B, and there are still few studies on CXCR3-alt [31]. While CXCR3B is generally limited to vascular endothelial cells, CXCR3A is mainly expressed in CD4+ Th1, CD8+ cytotoxic T, activated B and NK cells, which are at the sites of infection and inflammation [22]. In addition, the biological effects of CXCR3A and CXCR3B binding to the ligands were significantly different [32]. Specifically, CXCR3A binds to ligands and activates downstream signaling pathways, such as PI3K/Akt, Ras/ERK, and PKA, to induce directional cell chemotaxis and proliferation. CXCR3B binding ligand can significantly reduce intracellular DNA synthesis and promote endothelial cell apoptosis, thereby inhibiting angiogenesis [33,34]. In order to confirm the expression of CXCR3A on the surface of human primary NK cells, we obtained human primary NK cells from volunteers using methods that induce PBMC differentiation as previously described [35]. We used flow cytometry to detect CXCR3A expression on human primary NK cells, Indeed, human NK cells selected from PBMC and further amplified with feeder cells (EBV-LCLs) and IL-2 activation did express CXCR3 highly. We used flow cytometry to detect CXCR3 expression on NK cells obtained by this method. The results showed that CXCR3 was high expressed in human primary NK cells after cultured 14 days in vitro (Supplementary Figure. 5A). At the same time, we injected this batch of human primary NK cells into immunodeficient mice, and then extracted mouse PBMC cells 2 days later and 5 days later severally. Flow cytometry showed that CXCR3 expression on the surface of human primary NK cells dropped sharply after being injected into mice 2 days later, and decreased more obviously 5 days later(Supplementary Figure. 5B, 5C and 5D). This suggests that even though CXCR3 expression on NK cells can be high by some induction culture methods in vitro, but the expression of CXCR3 on the surface of amplified NK cells will gradually decrease due to the lack of corresponding stimulant factors after input into the body, thus losing the chemotactic function. Moreover, a large number of self-proliferating NK cells entering the body do not have stimulant factors. CXCR3 may be underexpressed and chemotactic ability may be lost. In this way, continuous expression of CXCR3 in NK cells of the incoming patients will increase the chemotaxis of NK, which is also the value of this study in constructing CAR-NK with tumor chemotaxis by continuous expression of CXCR3 in collaboration with CAR molecules.Therefore, based on the fact that IFN-γ and TNF-α secreted by NK-92-αFR-CAR cells can synergistically induce the high expression of CXCL10 in ovarian cancer cells, we hypothesized that CXCR3A and αFR-CAR could be co-expressed in NK-92 cells (NK-92-αFR-CAR-CXCR3A). When NK-92-αFR-CAR-CXCR3A cells kill ovarian cancer cells, the secreted IFN-γ and TNF-α synergistically induce a high expression of CXCL10 in tumor cells. Under the chemotactic effect of CXCL10, NK-92-αFR-CAR-CXCR3A cells continuously migrated into tumor tissues, enhancing the anticancer effect. Simultaneously, the secretion of IFN-γ and TNF-α was further increased, which in turn synergistically induced tumor cells to express more CXCL10. With the increase in CXCL10 expression, more NK-92-αFR-CAR-CXCR3A cells were recruited to tumor tissues. Thus, positive feedback of NK-92-αFR-CAR-CXCR3A cell self-driving was formed.

To test this hypothesis, the self-cleaved peptide 2A sequence was used to link the αFR-CAR sequence to the CXCR3A sequence in the lentiviral vector. The 2A peptide is a small "self-cleaved" peptide that cleaves the front and back of proteins in a proteolysis-similar manner during translation. NK-92 cells were infected with lentivirus carrying the αFR-CAR-CXCR3A gene. Flow cytometry showed that high expression levels of αFR-CAR and CXCR3A were detected in the selected lentivirus-infected NK-92 cells. This suggests that αFR-CAR and CXCR3A are stably co-expressed in NK-92 cells via the αFR-CAR-CXCR3A gene sequence designed based on self-cleaved peptide 2A. In addition, the LDH cytotoxicity assay and ELISA results showed that NK-92-αFR-CAR CXCR3A cells effectively killed αFR-positive ovarian cancer cells in vitro and secreted IFN-γ and TNF-α, similar to NK-92-αFR-CAR cells. The A431 cells, which do not express αFR, exhibit low cytotoxicity when interacting with effector cells targeting αFR, and almost no secretion of the cytokines IFN-γ and TNF-α. Furthermore, the results of the two designed transwell experiments showed that CXCL10 had a stronger chemotactic effect on NK-92-αFR-CAR-CXCR3A cells than on NK-92 cells with low CXCR3A expression, indicating that CXCL10 can effectively induce NK-92-αFR-CAR-CXCR3A cells for chemotactic activities. Finally, to evaluate the antitumor effect of NK-92-αFR-CAR-CXCR3A cells in vivo, SK-OV-3 or SK-OV-3-CXCL10 cells were used to construct subcutaneous tumor and abdominal tumor models in mice. After receiving the corresponding effector cell infusion, the tumor volume was observed, and the infiltration level of effector cells in the tumor tissues was analyzed using immunohistochemistry. The results showed that compared with NK-92 cells, NK cells, NK-92-EV, and NK-92-αFR-CAR cells, NK-92-αFR-CAR-CXCR3A cells in the three tumor models showed the strongest antitumor activity and the highest level of tumor infiltration. Unlike in vitro killing experiments based on cell co-culture, in animal experiments, various effector cells injected into mice via the tail vein must migrate into tumor tissues to kill tumor cells. Immunohistochemistry showed that only NK-92-αFR-CAR-CXCR3A cells could significantly infiltrate into tumor tissues under the chemotactic effect of CXCL10 during the killing process, which explains why NK-92-αFR-CAR-CXCR3A cells had the best antitumor effect in vivo.

Conclusion

This study not only found that IFN-γ and TNF-α secreted by NK-92-αFR-CAR cells synergically induced the overexpression of CXCL10 in ovarian cancer cells and elucidated the specific molecular mechanism, but also designed and constructed self-driving NK-92-αFR-CAR-CXCR3A cells based on the above results. In vitro and in vivo experiments confirmed that NK-92-αFR-CAR-CXCR3A cells could migrate into tumor tissues in large numbers under the chemotactic effect of CXCL10, thus possessing a strong antigen-specific antitumor activity. These results provide a theoretical basis for the next phase of clinical application of NK-92-αFR-CAR-CXCR3A cells, which may also provide a new immunotherapy for patients with ovarian cancer in the future.

CRediT authorship contribution statement

Min He: Data curation, Investigation, Visualization, Writing – original draft. Xiang Ao: Conceptualization, Methodology. Yu Yang: Investigation, Visualization. Yanmin Xu: Supervision, Validation. Tao Liu: Supervision, Validation. Luoquan Ao: Investigation, Visualization. Wei Guo: Investigation, Visualization. Wei Xing: Investigation, Visualization. Jing Xu: Supervision, Validation. Cheng Qian: Supervision, Validation. Jianhua Yu: Writing – review & editing. Xiang Xu: Writing – review & editing. Ping Yi: Writing – review & editing.

Declaration of competing interest

The authors have declared no conflict of interest

Appendix. Supplementary materials

Supplementary Figure 1 Quantative analysis of protein phosphorylation levels of JAK1/2-STAT1 signaling pathway in Ruxolitinib-pretreated ovarian cancer cells stimulated with IFN-γ using Western Blot. The statistical analysis is a comparison between groups of labeled groups, with * representing significant differences, *P < 0.05, **P < 0.01, ns, P ≥ 0.05.

Supplementary Figure 2 Quantative analysis of the protein phosphorylation levels of NF-κB signaling pathways in BAY 11-7082-pretreated ovarian cancer cells stimulated with TNF-α using Western Blot. The statistical analysis is a comparison between groups of labeled groups, with * representing significant differences, *P < 0.05, **P < 0.01, ns, P ≥ 0.05.

Supplementary Figure 3 (A) Quantative analysis of protein phosphorylation levels of JAK1/2-STAT1 signaling pathway and NF-κB signaling pathway in ovarian cancer cells after IFN-γ or/and TNF-α stimulation. (B) Quantative analysis of JAK1/2-STAT1 signaling pathway in siRNA-transfected ovarian cancer cells stimulated with IFN-γ and TNF-α. (C) Quantative analysis of The efficiency of siRNA interference. The statistical analysis is a comparison between groups of labeled groups, with * representing significant differences, *P < 0.05, ns, P ≥ 0.05

Supplementary Figure 4 In vivo antitumor activity and chemotaxis of NK-92-αFR-CAR-CXCR3A cells. (A) Representation of the construction of xenograft model of SK-OV-3-CXCL10 ovarian cancer cells and the corresponding treatment protocol. s.c. represents subcutaneous injection, i.v. represents intravenous injection. (B) Subcutaneous tumor tissues were isolated for photographing at the endpoint of the animal study. (C) Subcutaneous tumor volumes of each group were estimated every two days to monitor ovarian cancer development. (D) Chemotactic migration of NK-92 cells into subcutaneous tumor tissues were detected using immunohistochemistry. On the top was a representative immunohistochemistry picture of each group. Scale bars: 100 µm. On the bottom was quantification summary of CD56 positive stained cells per field view in each group. Five randomly selected fields of view were quantified for each group. The data are expressed as the means ± SEM of five mice for all groups. Statistical analysis shows the comparison between the labeled groups. * represents significant difference, *P < 0.05; **P < 0.01; ns, P ≥ 0.05.

Supplementary Figure 5 Flow cytometry analysis of CXCR3 expression on human primary NK cells. (A) CXCR3 expression on the surface of human primary NK cells after cultured 14 days in vitro. (B) Expression of CXCR3 on the surface of human primary NK cells 2 days after injection into mice. (C) Expression of CXCR3 on the surface of human primary NK cells 5 days after injection into mice. (D) Statistical analysis of CXCR3 expression for (A), (B) and (C). The filled green histograms indicate isotype control, whereas the filled red histograms indicate the CXCR3 expression. The data in (D) are expressed as the means ± SEM of three mice for all groups. Statistical analysis shows the comparison between the labeled groups. * represents significant difference, *P < 0.05; **P < 0.01; ns, P ≥ 0.05.