GAS6-based CAR-T cells exhibit potent antitumor activity against pancreatic cancer

The human pancreatic cancer cell lines ASPC1, BxPC3 and PANC1, human embryonic kidney 293 T cells (HEK-293 T), mouse embryonic fibroblast cell line NIH 3T3, and mouse breast cancer cell line 4 T-1 used in this study were maintained in our laboratory. We obtained the human pancreatic cancer cell line MIA PaCa2 and mouse hepatoma carcinoma cell line Hepa1-6 from the cell bank of the Chinese Academy of Sciences (Shanghai, China). The ASPC1-gemcitabine-resistant cell line was purchased from FENGHUISHENGWU Co. Ltd. (Hunan, China). All cell lines were authenticated by STR, and mycoplasma contamination was routinely tested by qPCR.

Luciferase-labeled cells were established by infection with pTomo-CMV-luciferase-IRES-Puro lentivirus followed by selection with puromycin (1.0 mg/mL, Gibco, USA) for 2 weeks. All cells were cultured in DMEM containing 10% fetal bovine serum (Gibco, USA), 100U/mL penicillin, and 100 mg/mL streptomycin (Gibco, USA). The culture medium of Hepa1-6 cells also contained 1.0 mM sodium pyruvate (Gibco, USA) and Gluta-MAX™ (100× , Gibco, USA), and gemcitabine-resistant ASPC1 cells were cultured with 1.0 ug/mL gemcitabine. Suspended cell spheres derived from PANC1 and MIA PaCa2, named PANC1-CSC and MIA PaCa2-CSC, respectively, were established by culturing in serum-free stem cell medium composed of DMEM/F12 (Gibco, USA), EGF (20 ng/mL, PeproTech., USA), bFGF (20 ng/mL, PeproTech., USA), and B27 (1× , Gibco, USA).

CAR comprising the CD8 signal peptide, extracellular domain (amino acids 261–678) of human GAS6, the CD8 hinge spacer and transmembrane domain, CD137 (4-1BB), and the CD3ζ endo domains was cloned into the pTomo-Puro plasmid (Addgene, USA) between the AgeI and NheI restriction sites. An mKATE2 sequence was fused to the CAR via a T2A peptide to monitor the transduction efficiency. The same vector sequence without extracellular domain of GAS6 was used as a control (Mock).

To construct TAM-shRNA plasmids, the target sequences were cloned into a pLKO.1-Puro vector obtained from Addgene between the AgeI and EcoRI restriction sites. The target sequences were as follows: shAXL #1, CGAAATCCTCTATGTCAACAT, #2, CGAAAGAAGGAGACCCGTTAT; shTYRO3 #1, GGAGAGGAACTACGAAGAT CG, #2, GCATCAGCGATGAACTAAAGG; shMERTK #1, GCTCAATCAGTGTAC CTAATA, #2, GCATTGGTGTTTCCTGCATGA. Expression plasmids containing AXL (EX-Z7835-Lv105), TYRO3 (EX-A0969-Lv105), and MERTK (EX-Z8208-Lv105) were purchased from iGene Biotechnology Co., Ltd. (Guangzhou, China).

For lentiviral packaging, plasmids were transfected into HEK-293 T cells with the packaging plasmids pCMV-dR8.91 and pMD 2.G (Addgene) at a ratio of 5:2.5:1. The supernatants were collected and filtered through a 0.45-μm filter (Millipore, Bedford, MA) to remove cellular debris and centrifuged at 25,000 rpm for 2.5 h to obtain the virus precipitation.

Human T cells were isolated from healthy donor blood using the RosetteSep™ Human T-Cell Enrichment Cocktail (STEMCELL, Canada) and cultured in advanced 1640 medium (Gibco, USA) containing 10% FBS (Gibco, USA) with 200 U/mL IL-2 (Invitrogen, USA) and Gluta-MAX™ (100× , Gibco, USA). To generate CAR-T cells, T cells were activated by CD3/CD28 dynabeads (Life Technologies, USA) for 72 h followed by incubation with lentiviral particles at an approximate MOI of 100 with lentiBoost (1.0 μg/mL, Sirion Biotech, Germany) for 24 h. The CAR-T cells were applied for experiments on day 3 after transduction.

Monkey T cells isolated from rhesus monkeys by density gradient centrifugation (Ficoll-Paque) were activated by nonhuman primate T-cell activation/expansion kit (Miltenyi Biotec) and cultured in RPMI-1640 medium containing 10% FBS with 200 U/mL IL-2 and Gluta-MAX™. The activated T cells were transduced with lentiviral particles of GAS6-CAR to prepare CAR-T cells.

The cytotoxicity of CAR-T cells was tested using a Luciferase Assay System (Promega, E1501) at variable effector-to-target (E/T) ratios of 0.5:1, 1:1, 2:1, and 4:1. Briefly, 2 × 10³ target cells per well were seeded in 96-well plates with 100 μL medium, and an equal volume of effector cells was added. After 24 h of coculture, the supernatant was collected and used to determine the concentrations of IFN-γ (Invitrogen, KHC4021) and TNF-α (Proteintech, KE00154). The cells were then lysed for luciferase assay according to the instructions of the manufacturer, and the cytotoxicity of CAR-T cells was calculated as ratio to tumor cells incubated with non-transduced T (NT) cells. The results were expressed as means and standard deviations for triplicate assays.

Western blotting was performed as described previously [20]. Harvested cells were lysed in RIPA buffer, and protein concentrations were quantified using BCA protein assay kits (Beyotime, Shanghai, China). The total protein lysates were separated by 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis and transferred to PVDF membranes (Millipore, Billerica, MA, USA). The membranes were blocked in Tris-buffered saline with 5% non-fat milk and 0.5% BSA for 1 h, prior to incubation with primary antibodies overnight at 4 °C and incubation with horseradish peroxidase (HRP)-conjugated secondary antibodies for 1.5 h at room temperature. Blots were visualized with chemiluminescent HRP substrate (Millipore). Detailed information of antibodies used in this experiment is listed in Additional file 1: Table S1.

The cells (1 × 10⁶) were fixed in 4% formaldehyde for 15 min at room temperature. After washing, the cells were incubated with primary antibodies for 1 h and then the fluorescent secondary antibodies for 30 min at room temperature (Additional file 1: Table S1). Finally, the cells were analyzed by BD LSRFortessa Flow cytometry (BD Biosciences), and data were analyzed using FlowJo software version 10 (TreeStar, Inc.).

Total RNA and genomic DNA were extracted as described previously [20, 21]. qPCR assays were performed with SYBR Selected Master Mix (Thermo Fisher, USA). The comparative cycle time (Ct) method was used to determine differences between samples, and the expression of target genes was normalized to 18S rRNA or GAPDH (2^−△△Ct). The primer sequences are listed in Additional file 1: Table S2.

5 × 10⁵ PANC1-luciferase cells were suspended in PBS containing 20% Matrigel (BD Bioscience) and subcutaneously injected into the right flank of six-week-old female NOD/ShiLtJGpt-Prkdc^em26Cd52Il2rg^em26Cd22 /Gpt (NCG) mice (GemPharmatech Co. Ltd., China). Mice were intraperitoneally injected with 150 mg/kg D-luciferin (BioVison, 7903-1G) after anesthesia with 1.5% isoflurane, and tumor progression was determined using an in vivo imaging software (IVIS) system (Guangzhou Biolight Biotechnology Co., Ltd., aniview100). The mice were randomly divided into two groups according to bioluminescent signals at 3rd day and treated with 1 × 10⁷ Mock T cells or GAS6-CAR-T cells by tail vein injection. Bioluminescent signals were subsequently measured every 7 days.

To establish the PDX model of pancreatic cancer, 3 × 3 mm blocks of patient-derived pancreatic tumor tissues were implanted in the right flank of six-week-old female NCG mice. After 14 days, the mice were randomly divided into two groups and treated with an injection of 1 × 10⁷ Mock T cells or GAS6-CAR-T cells. Tumor size was measured twice a week using a digital caliper, and tumor volume was calculated using the following formula: (major axis of tumor) × (minor axis of tumor)²/2. The mice were euthanized when the tumor volume reached 1000 mm³. This experiment was completed with the assistance of Sichuan Kang Cheng Biotechnology Co. (Chengdu, China).

To detect multicolor immunofluorescence, we performed this experiment using Opal™ Multiplex IHC Assay (Akoya Biosciences, USA) that allowed to use any standard unlabeled primary antibody, including multiple antibodies raised in the same species.

Tissues were fixed with 4% paraformaldehyde, dehydrated with gradient ethanol, and embedded in paraffin. Tissues slides were dewaxed and dehydrated, boiled in citrate buffer (pH 6.0) for antigen retrieval, and blocked using 5% normal goat serum at room temperature for 1 h. Then, the slides were incubated at 4 °C overnight with the following primary antibodies and incubated for 1 h with the corresponding HRP-conjugated secondary antibodies (Additional file 1: Table S1) and TSA Plus Fluorescein Reagent (1:50) for 10 min. Finally, nuclei were stained with DAPI. Fluorescent images were taken using a confocal microscope (Nikon, Japan), and representative microscopy images were shown.

Statistical analyses were performed using GraphPad Prism version 8.0 (GraphPad Software Inc.). All data are presented as mean ± SD. Statistical differences between two groups were analyzed using Student’s t tests with Welch correction. Statistical differences among three or more groups were analyzed by one-way or two-way ANOVA with Sidak correction. In all statistical analyses, the P values (*P < 0.05, **P < 0.01) were considered significant, ns = not significant.

Laminin g-like domain (LG domain) of GAS6 (amino acids 261–678) binds to the immunoglobulin-like domain of TAM receptors [3, 22, 23]. To target the TAM family, we used the LG domain of GAS6 as the recognition domain in CAR design. The CAR and control vector are graphically represented in Additional file 1: Fig. S1A. Flow cytometry analysis of mKATE2 fluorescence revealed a transduction efficiency of approximately 35% (Additional file 1: Fig. S1B, C). The lentiviral transduction and exogenous CAR expression had no significant effects on the proliferation of T cells (Additional file 1: Fig. S1D).

We detected the expression of TAM proteins in pancreatic cancer cells. As shown in Fig. 1A, AXL, TYRO3, and MERTK were found to be highly expressed in PANC1 and MIA PaCa2 and low levels in BxPC3 and ASPC1, while human embryonic kidney cell line HEK-293 T expressed a relatively higher level of MERTK. To determine whether GAS6-CAR can target the TAM family members, AXL, TYRO3, and MERTK were individually transduced in TAM-low HEK-293 T and BxPC3 cells, and the overexpression was confirmed by western blot analysis (Fig. 1B). Luciferase-labeled AXL-, TYRO3-, or MERTK-overexpressing HEK-293 T and BxPC3 cells were co-incubated with CAR-T cells, and luciferase activity was measured after 24 h incubation. As shown in Fig. 1C, D, GAS6-CAR-T cells efficiently killed cells overexpressing any TAM protein, but had no effect on the control cells. The antigen-stimulated release of the IFN-γ (Fig. 1E, F) and TNF-α (Fig. 1G, H) cytokines was induced in these CAR-T cells. These data demonstrate that GAS6-CAR-T cells can target all three TAM proteins.

We next assessed the capacity of GAS6-CAR-T cells to kill TAM-positive pancreatic cancer cells. Compared to Mock T cells, incubation with GAS6-CAR-T cells led to significantly higher rates of death in the TAM-positive MIA PaCa2 and PANC1 cells in a dose-dependent manner (Fig. 2C, D), but no killing effects on TAM-low ASPC1 and BxPC3 cell lines (Fig. 2A, B). The mRNA transcript levels for IFN-γ, TNF-α, IL-2, and IL-10 were significantly increased (Additional file 1: Fig. S2C, D), and the release of IFN-γ and TNF-α into the supernatant was also significantly increased in the incubation with TAM-positive MIA PaCa2 and PANC1 cells (Fig. 2G, H), while the expression of cytokines did not increase in incubation with TAM-low ASPC1 and BxPC3 cell lines (Fig. 2E, F and Additional file 1: S2A, B). Thus, the GAS6-CAR-T cells appear to specifically target TAM-positive pancreatic cancer cells.

To determine which TAM member was responsible to CAR-T effects, we tested the effects of knocking down individual TAM proteins. We identified and confirmed the silencing efficiency of two shRNA sequences for each TAM protein (Fig. 3A). Both AXL-shRNAs significantly abolished the cytotoxic effects of GAS6-CAR-T cells toward PANC1 and MIA PaCa2 target cells (Fig. 3B). Furthermore, the antigen-specific expression of IFN-γ, TNF-α, IL-2, and IL-10 was significantly reduced (Additional file 1: Fig. S3), and the secretion of IFN-γ (Fig. 3D, E) and TNF-α (Fig. 3F, G) maintained at low level. However, shRNA sequences against TYRO3 or MERTK had no significant effects on the cytotoxicity of GAS6-CAR-T cells (Fig. 3B, C) and the secretion of cytokines (Fig. 3D–G). It suggests that AXL is the main target in tested cell lines, which is probably involved in the different expression of TAM members and highest affinity of AXL to GAS6.

Drug resistance is a major factor underlying the failure of cancer treatments, and high levels of TAM proteins are strongly correlated with acquired drug resistance [16]. Therefore, we tested the effectiveness of GAS6-CAR-T cells in killing drug-resistant cell lines. First, we analyzed TAM proteins levels in ASPC1 and ASPC1-gemcitabine-resistant (ASPC1-Gem) pancreatic cancer cell lines. As shown in Fig. 1A and Fig. 4A, TAM proteins were prominent in ASPC1-Gem cells compared to the parental TAM-low cell line ASPC1. We next incubated luciferase-labeled ASPC1 and ASPC1-Gem cells with GAS6-CAR-T cells. CAR-T cells incubated with ASPC1-Gem cells showed a significant lysis activity (Fig. 4B), displayed increased mRNA expression levels for IFN-γ, TNF-α, IL-2, and IL-10 (Additional file 1: Fig. S4), and presented significantly elevated release of IFN-γ and TNF-α cytokines (Fig. 4C, D). Thus, GAS6-CAR-T cells can effectively target pancreatic cancer cell with gemcitabine resistance induced by TAM overexpression.

Cancer stem-like cells (CSCs) are important for tumor initiation, recurrence, metastasis, and major drivers of drug resistance [24, 25]. To test whether GAS6-CAR-T cells can also kill CSCs, we produced cancer cell line-derived CSCs as sphere cultures from PANC1 and MIA PaCa2 cells. The stemness of these CSCs was confirmed by the expression of stem cell markers CD133, CXCR4 and OCT4 (Additional file 1: Fig. S5). Interestingly, TYRO3 and MERTK in the CSCs were highly expressed, while AXL expression in the CSCs was unexpectedly decreased compared to the parental cell lines as detected by flow cytometry and WB (Fig. 1A and Fig. 4E). When incubated with GAS6-CAR-T cells, the viability of pancreatic CSCs was significantly reduced (Fig. 4F, G). This reduced viability was accompanied by increased expression of IFN-γ, TNF-α, and IL-2 (Additional file 1: Fig. S6) and release of IFN-γ (Fig. 4H, I) and TNF-α (Fig. 4J, K). To further validate the roles of individual TAM member expression in lysis, we evaluated the effects of knocking down individual TAM proteins. The silencing efficiency was tested by western blot (Additional file 1: Fig. S7A). Both AXL-shRNAs slightly reduced the cytotoxic effects of GAS6-CAR-T cells on CSCs (Additional file 1: Fig. S7B, C), while shRNAs targeting TYRO3 and MERTK significantly reduced the cytotoxic effects of GAS6-CAR-T cells (Additional file 1: Fig. S7B, C) and the secretion of IFN-γ (Additional file 1: Fig. S7D, E) and TNF-α (Additional file 1: Fig. S7F, G). These data suggest that GAS6-CAR-T cells induce significant cytotoxicity in pancreatic cancer stem-like cells that may attribute more to increased TYRO3 and MERTK.

To test the effects of GAS6-CAR-T cells on tumor growth in vivo, we established a xenograft mouse model by injecting PANC1 cells subcutaneously. Compared to Mock T cells, GAS6-CAR-T cells presented with significantly reduced tumor growth. Moreover, these mice remained completely tumor-free from day 21 after GAS6-CAR-T-cell treatment (Fig. 5A, B).

The copy number of CAR-T-cell DNA in peripheral blood was associated with tumor clearance, and copy numbers remained high for at least 42 days in the GAS6-CAR group compared to Mock controls (Fig. 5C). Previous research has shown that the number of infiltrating T cells within a tumor is closely related to antitumor activity [26]. The human origin of cells in tumor was determined by immunohistochemistry analysis with a commercially available antibody specific for the human mitochondrial marker COX IV [27], and the tumor weight was significantly reduced at day 5 after GAS6-CAR-T cells infusion (Fig. 5D), and increase in CD3 + T-cell infiltration in mice receiving GAS6-CAR-T cells compared to Mock-treated mice (Fig. 5E). These results suggest that GAS6-CAR-T cells can effectively home to target sites and inhibit the growth of tumor xenografts.

Double staining of AXL and OCT4 revealed that OCT4-positive stem cells are significantly reduced in GAS6-CAR-T group (Fig. 5F), thereby indicating that GAS6-CAR-T can target CSCs in vivo to enhance tumor clearance in consistent with the in vitro effects on CSCs.

TAM proteins are also highly expressed in M2 macrophages that are the critical cell type in tumor microenvironment [28, 29]. And CD68 is a well-established marker for human macrophages [30]. PDX models have a striking advantage that non-tumor cells in tumor mass are of human origin. Therefore, we choose the pancreatic cancer patient-derived xenograft (PDX) models with the high expression of AXL in both cancer cell and macrophage to test whether GAS6-CAR-T cells can also eliminate cancer cells and macrophages in vivo. The patient-derived pancreatic tumor tissue was chosen according to the expression of markers of tumor cells (CK19) and macrophages (CD68), and AXL detected by immunofluorescence. An AXL-positive sample with a number of macrophages was chosen for xenograft experiment (Additional file 1: Fig. S8A, B). The clinical and pathologic characteristics of the patient are shown in Additional file 1: Table S3. As shown in Fig. 6A, intravenous administration of GAS6-CAR-T cells led to a significant suppression of pancreatic cancer PDX tumor growth, while tumors in the Mock group continued to grow rapidly. Also, the copy number of T-cell DNA in GAS6-CAR group was significantly higher than that in Mock group (Fig. 6B). In addition, GAS6-CAR-T cells could significantly reduce AXL and CK19 double-positive tumor cells (Fig. 6C, D), as well as AXL and CD68 double-positive macrophages (Fig. 6C, E). Overall, these data further strengthen the notion that GAS6-CAR-T cells may be an effective therapy targeting both tumor cells and tumor-associated macrophages.

GAS6-CAR-T cells did not lyse TAM-low ASPC1 and BxPC3 cell lines, suggesting it may spare normal tissues expressing lower AXL relative to cancer tissues. Due to the high conservation of receptor-ligand systems across species, human GAS6 could recognize the mouse TAM proteins at a similar level to human TAM proteins [31]. Therefore, we examined the effects of GAS6-CAR-T cells on murine tumor cell lines. GAS6-CAR-T cells effectively killed the high TAM-expressing mouse hepatoma carcinoma Hepa1-6 and breast cancer 4 T-1 cell lines (Fig. 7A, B) and induced the secretion of IFN-γ (Fig. 7C) and TNF-α (Fig. 7D), while GAS6-CAR-T cells displayed no significant toxicity or induction of cytokine release when exposed to the mouse embryonic fibroblast cell line NIH 3T3, which display low levels of TAM proteins (Fig. 7A–D). However, the mice of PANC1 xenograft and PDX model receiving GAS6-CAR-T cells did not show any obvious side effects, and they remained active and maintained similar body weights to the control group (Fig. 7E, F). Furthermore, while large numbers of CD3 + T cells clustered in the tumor tissues of mice receiving GAS6-CAR-T cells, minimal numbers were observed in the major organs (Additional file 1: Fig. S9A), and no significant tissue damage or structural changes were observed in the major organs (heart, liver, spleen, lung, kidney, and brain) at day 5 or 42 of PANC1 xenograft (Additional file 1: Fig. S9B and Fig. 7G), and PDX model (Fig. 7H) after T-cell treatment in either group. These results suggest GAS6-CAR-T cells have minimal on-target off-tumor effects in vivo.

AXL protein (XP_028695606.1) of rhesus monkeys shares 96.16% homology with human AXL (NP_068713.2), and an AXL inhibitor in phase III clinical trial (AVB-S6-500) containing the extracellular domain of AXL to competitively bind GAS6 had the same affinity for GAS6 of human and cynomolgus monkeys, but not resulted in any obvious side effects in monkey [32]. To further evaluate the safety of CAR-T cells, GAS6-CAR-T cells prepared from rhesus macaques T cells were autologously infused (Fig. 8A). CAR-T cells show modest proliferation in vivo and persist about one month (Fig. 8B). As shown in Fig. 8C, there was no significant fluctuation in diastolic and systolic blood pressure, heart rate, anus temperature, and body weight on days 1, 3, 5, 7, 14, 21, 28, and 35 compared to day 0. GAS6-CAR-T cells did not result in significant changes of blood chemistry parameters, such as alanine transaminase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), γ-glutamine acylase (GGT), total protein (TP), albumin (ALB), globulin (GLO), ALB/GLO (A/G), and total bilirubin (TBIL) for liver function (Fig. 8D); blood glucose (GLU) (Fig. 8E); blood urea nitrogen (BUN) and creatinine (CRE) for kidney function (Fig. 8F); cholesterol (CHOL) and triacylglycerol (TRIG) for serum lipids (Fig. 8G); and creatine kinase (CK) for cardiac function (Fig. 8H). Also, the related blood indicators for red blood cells (RBC) (hemoglobin (HGB), hematocrit (HCT), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC)) (Fig. 8I); platelets (platelet count (PLT), platelet distribution density (PDW), mean platelet volume (MPV), platelet volume ratio (PCT)) (Fig. 8J); white blood cell (WBC) (Fig. 8K); monocyte (Fig. 8L) and lymphocyte (Fig. 8M) did not show any obvious changes. These results suggest GAS6-CAR-T cells show high safety in vivo.