MAGE-A1 in lung adenocarcinoma as a promising target of chimeric antigen receptor T cells

From the CTA database in our previous study [22], we retrieved LUAD-related data and screened candidate CTAs by score ranking (normalized expression > 3%). Then, we searched the GTEx Portal database (https://​www.​gtexportal.​org) to further identify appropriate CTAs that are only expressed in testes and not in normal tissues. Next, we inspected the GeneCard database (http://​www.​genecards.​org) to filter suitable CTAs that are expressed in the cytomembranes of cancer cells (expression confidence > 3). Moreover, we employed The Cancer Genome Atlas (TCGA) data (https://​cancergenome.​nih.​gov) to validate the RNA expression levels of eligible CTAs in LUAD tissues and corresponding noncancerous tissues (expression fold change > 10). Finally, we checked the Human Protein Atlas database (http://​www.​proteinatlas.​org) to ensure CTA protein expression in LC.

A tissue microarray (TMA) containing 90 cases of normal human tissue samples was purchased from Outdo Biotech Co., Ltd. (Shanghai, China). Simultaneously, five LUAD tissue samples and corresponding non-cancerous tissue samples were collected from the Department of Thoracic Surgery, Nanjing Medical University Affiliated Cancer Hospital. A TMA containing 93 cases of LUAD was also purchased from Outdo Biotech Co., Ltd. (Shanghai, China) [23]. Important clinical parameters were collected along with the LUAD TMA. Written informed consent was obtained from the patients for the publication of this study and the use of any accompanying images. The study protocol was approved by the Ethics Committee of Nanjing Medical University Affiliated Cancer Hospital, and all experiments were performed following the approved guidelines of Nanjing Medical University.

Four LUAD cell lines (PC9, H1299, GLC82, A549) and the human embryonic kidney 293T cell line (HEK-293T) were preserved in our lab and enrolled in the present study. The human melanoma A375 cell line was purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). The human normal bronchial epithelial (HBE) cell line was kindly provided by Professor. Erbao Zhang from the Department of Epidemiology and Biostatistics, Nanjing Medical University, to serve as the non-cancerous cell line. Peripheral blood mononuclear cells (PBMCs) derived from a healthy donor were collected by Ficoll-Hypaque density-gradient centrifugation conducted by the Jiangsu Blood Center. Medium with recombinant human interleukin-2 (IL-2) 300 U/ml was used for the expansion of T cells.

MAGE-A1 expression was thoroughly examined in LUAD cell lines and tissue samples. For the qPCR, the sequences of the primers are listed in Additional file 7: Table S2. For the western blotting analysis, two types of primary monoclonal antibodies were obtained from Abcam (ab193330, ab243935, Abcam, Cambridge, MA, USA). The protocols of the qPCR test and western blotting analysis were described previously [24, 25]. The immunofluorescence test was conducted following the protocols described in our previous study [26]. Cells were incubated with FITC-labeled human anti-MAGE-A1 antibody (Abcam, ab212590) in the dark. 4′-6-diamidino-2-phenylindole (DAPI, Biotium, Hayward, CA) was used for nuclear staining. The ubiquitous Desmoglein 2 (DSG-2) was employed as a positive control (Abcam, ab150372). Immunohistochemistry (IHC) analysis was performed as previously described [27, 28]. TMA sections were incubated with mouse monoclonal anti-MAGE-A1 antibody (Abcam, ab193330). The secondary antibody used was horseradish peroxidase-conjugated anti-mouse antibody. Phosphate-buffered saline (PBS) was used as a negative control.

The overexpression and short-hairpin RNA (shRNA)-mediated knockdown lentivirus plasmids and packaging vectors were prepared as previously described [29]. Full-length MAGE-A1 was inserted into the lentivirus pLenti-EF1a-EGFP-P2A-Puro-CMV-MCS vector (Obio Technology, Co., Ltd., Shanghai, China). The detailed sequences of the three shRNAs and related siRNAs used in this study are listed in Additional file 7: Table S2. shRNA targeting MAGE-A1 (shMAGE) or scrambled shRNA (shCT) were cloned into pLKD-CMV-G&PR-U6-shRNA (Obio Technology). PC9 cells were then infected with MAGE-A1 overexpression (OEMAGE) or shMAGE viruses. After viral transfection, MAGE-A1 expression was evaluated by qPCR and western blotting analyses. Then, stable OEMAGE and shMAGE PC9 cell lines were confirmed by puromycin selection and prepared for further experiments.

CCK-8, wound healing, and Transwell assays were performed in OEMAGE and shMAGE PC9 cell lines, respectively, to detect the malignant behaviors of MAGE-A1 in LUAD, including its effects on cell proliferation, cell migration, and cell invasion, as described before [30].

Athymic 4-week-old BALB/c nude mice were purchased from SLAC Laboratory Animal Co., Ltd. (Shanghai, China) and kept under specific pathogen-free (SPF) conditions. In brief, 1.0 × 10⁷ PC9 (OEMAGE and shMAGE) cells were injected into nude mice subcutaneously. After inoculation, the tumor-bearing mice were observed, and tumor size was measured with a Vernier caliper. The subsequent procedures of the tumor growth assay in mice were described previously [26].

The MAGE-A1-CAR (mCAR) was designed to consist of a human CD8α leader, anti-MAGE-A1-scFv, CD8α hinge and transmembrane domain (CD8™), and CD137 and CD3ζ cytoplasmic domains [31, 32]. The anti-MAGE-A1 scFv was determined in our previous study [33], and the detailed amino acid sequence is shown in Additional file 8: Table S3. The fragments encoding the CD8α leader, anti-MAGE-A1 scFv, CD8™, and CD137-CD3ζ were produced by PCR and cloned into the EcoRI and XbaI sites of the lentiviral expression vector pLVX-IRES-ZsGreen (Clontech, USA). All positive clones were confirmed by sequencing analysis.

For lentivirus production, HEK-293T cells were co-transfected with mCAR vector, pMD2.G plasmid (Invitrogen, Carlsbad, CA, USA) and packaging psPAX2 plasmid (Invitrogen). Supernatants containing the lentivirus were collected 48 h and 72 h later. After filtration through a 0.45-μm filter, the lentivirus supernatant was concentrated 30-fold by ultracentrifugation (Amicon Ultra 100 kD, Millipore, USA). 293T cells transfected with CD19-CAR (unrelated-CAR) and untransfected 293T cells (blank) were employed as controls. Then, CD3ζ was selected as the target to test mCAR expression after 293T cell transfection by western blotting analysis.

A sandwich ELISA was performed to evaluate the binding ability of mCAR to MAGE-A1 as described before [34]. Briefly, 96-well plates were seeded with transfected 293T cells (mCAR and unrelated-CAR). Untransfected 293T cells (blank) were used as a negative control. Then, each well was washed and MAGE-A1 antigens were added (Novus Biologicals, Littleton, CO, USA) at different dilutions. Then, the supernatants were collected and added to another 96-well plate, which was preliminarily coated with anti-MAGE-A1 rabbit polyclonal antibody (LS-C327797-200, LifeSpan BioSciences, Seattle, WA, USA), followed by the addition of a primary anti-MAGE-A1 mouse monoclonal antibody (LS-C25368-100, LifeSpan BioSciences) and a secondary anti-mouse antibody. After washing, the optical density at 450 nm (OD450) was measured with an automatic microplate reader (Thermo Fisher Scientific, USA). The supernatant lentivirus titers were detected following the protocol described previously [35, 36].

PBMCs were separated from 10 mL of peripheral blood from a healthy volunteer using lymphocyte separation medium. PBMCs were activated in 24-well plates coated with anti-human CD3 (Life Technologies, Mountain View, CA, USA) and anti-human CD28 antibodies (Life Technologies) at day 0. After 48 h, IL-2 (300 U/mL) was added to stimulate the expansion of the T cells. After 72 h, T cells were transfected with the mCAR lentivirus. Unrelated-CART and control T cells (T) served as controls. At day 7, all T cells were harvested, and the details of the mCART activity and characteristics were examined by flow cytometry. Briefly, the transfection efficiency of T cells expressing CAR was tested by direct GFP (ZsGreen) fluorescence and MAGE-A1-PE staining. Phenotypic characterization and activation of the T cells were determined by staining with CD3, CD4, and CD8. Flow cytometry was performed on a BD FACSCelesta flow cytometer. Data were graphed using FlowJo 7.6 software (Ashland, OR, USA).

Antitumor activity was quantified by LDH release assay, as described previously [37]. mCART, unrelated-CART, and T were co-cultured with LUAD cell lines (H1299, PC9, PC9(sh)) at different ratios (20:1, 10:1, 5:1, 2:1). Then, mCART was co-cultured with different LUAD cell lines (PC9, H1299, GLC82, A549) at a fixed ratio (10:1). The HBE cell line was employed as a control. Unrelated-CART representsCD19-CAR-T cells that are produced and preserved in our lab. The supernatant was analyzed for IFN-γ and IL-2 production using the related ELISA assay kits (eBioscience, San Diego, CA, USA) following the manufacturer’s protocols.

Athymic BALB/c nude mice were purchased from SLAC. For LUAD xenograft model establishment and bioluminescent imaging of in vivo tumors, mice were injected with luciferase-expressing H1299 cells with matrix. After inoculation, mice were divided randomly into three groups (mCART group, unrelated-CART group, T group). Treatment was initiated when the xenografts reached volumes of approximately 100 mm³, and mice underwent fully myeloablative radiation. On days 0, 3, and 6, mice received intravenous treatment with mCART (1 × 10⁷), unrelated-CART and T cells. The tumor diameter was measured, and the tumor volume was calculated as described previously [26]. For bioluminescent imaging, mice were injected intraperitoneally with D-luciferin (Gold Biotechnology, St. Louis, MO, USA), and images were recorded on days 2, 5, 8, 13, and 20 by utilizing an IVIS Lumina II (PerkinElmer, Hopkinton, MA, USA). On day 27, all mice were killed, and the xenograft tumors were removed for further analysis. Specifically, CD3 expression was detected by IHC analysis using a primary rabbit monoclonal antibody (Abcam, ab16669). The detailed protocol of IHC analysis was described previously.

First, we retrieved raw LUAD-related data and created a CTA expression heat map with a total of 1019 CTAs (Fig. 1a, Additional file 6: Table S1). Then, we screened 77 candidate CTAs by score ranking (normalized expression > 3%) (Fig. 1b, Additional file 6: Table S1). Subsequently, we searched the GTEx Portal database to further identify 49 CTAs that were only expressed in the testes and not in normal tissues (Fig. 1c, Additional file 1: Figure S1, Additional file 6: Table S1). After that, we inspected the GeneCard database to identify four suitable CTAs that were expressed in the cytomembranes of cancer cells (expression confidence > 3) because the cytomembrane expression of CTAs is important for the construction of CAR-T cells (Fig. 1d, Additional file 2: Figure S2, Additional file 6: Table S1). Moreover, we employed TCGA data to validate 2 CTAs, of which the RNA expression in LUAD tissues was markedly higher than that in the corresponding non-cancerous tissues (expression fold change > 10) (Fig. 1e, Additional file 6: Table S1). In addition, we consulted the Human Protein Atlas database to ensure that the qualified CTAs are positively expressed in LC (Fig. 1f, Additional file 6: Table S1). Finally, MAGE-A1 was selected as the appropriate LUAD-associated CTA from among the original 1019 CTAs (Fig. 1g).

To confirm the expression of MAGE-A1 in LUAD, qPCR and western blotting analyses were performed in LUAD cell lines. In four LUAD cell lines, the results of both qPCR and western blotting analyses showed that MAGE-A1 expression was significantly higher than that in the normal HBE cell line (Fig. 2a, b). Immunofluorescence assay revealed that MAGE-A1 could be stained in MAGE-A1-positive PC9 cell but not in MAGE-A1-negative HBE cell. The human melanoma A375 cell line was employed as a positive control and MAGE-A1-positive staining could also be observed in A375 cell line. Strong staining of MAGE-A1 was mainly localized in the cytomembrane while relatively weak staining of MAGE-A1 was observed in the cytoplasm of cancer cells (Fig. 2c).

We searched GTEx Portal database to preliminarily detect the expression mode of MAGE-A1 in normal human tissue and the data showed that MAGE-A1 was mostly expressed in human testis (Fig. 3a). Further, IHC analysis in normal human TMA confirmed that the MAGE-A1 expression was largely witnessed in human testicle samples while rarely observed in other human tissue samples (Fig. 3b). Then, we collected five LUAD and noncancerous tissue samples, and the data from qPCR and WB tests showed that the expression of MAGE-A1 in LUAD was elevated compared with that in non-cancerous tissues (Fig. 4a, b). After IHC analysis in LUAD TMA, 4 samples of LUAD and 9 samples of non-cancerous tissue in TMA were missing. The results of IHC analysis demonstrated that high MAGE-A1 expression was detected in 49 of 89 (44%) LUAD tissues compared with14 of 78 (18%) non-cancerous tissues, and the difference was highly significant (χ² = 24.36, p = 0.001). The IHC staining for MAGE-A1 expression and its relationships with important clinical characteristics in LUAD patients are presented in Fig. 4c and Table 1. A high level of MAGE-A1 expression was significantly correlated with tumor diameter (p = 0.023) and N status (p = 0.031). A survival analysis was performed, and the results illustrated that MAGE-A1 expression was critically associated with OS in patients with LUAD (p = 0.022) but was not an independent prognostic predictor (p = 0.087) (Fig. 4d and Table 2).

Because MAGE-A1 was upregulated in LUAD, the biological role of MAGE-A1was explored by CCK-8, wound healing and transwell assays in the PC9 cell line. As shown in Fig. 5a, we successfully constructed MAGE-A1 knockdown (shMAGE) and MAGE-A1 overexpression (OEMAGE) models. shMAGE drastically inhibited PC9 cell proliferation, migration, and invasion, while OEMAGE significantly augmented PC9 cell proliferation, migration, and invasion (Fig. 5b–d). Then, shMAGE and OEMAGE PC9 cells were subcutaneously injected into nude mice. As shown in Fig. 5e, the xenograft tumors that developed from OEMAGE PC9 cells grew significantly faster than those that developed from shMAGE PC9 cells. Consistently, the weight (Fig. 5f, Additional file 3: Figure S3) and volume (Fig. 5g) of MAGE-A1 knockdown tumors were much lighter and smaller than those of MAGE-A1 overexpression tumors at 48 days after cell inoculation. These results indicate the promotional function of MAGE-A1 in LUAD tumorigenesis.

The structure of mCAR is shown in Fig. 6a, consisting of a signal peptide leader sequence of CD8α, MAGE-A1-scFv, the hinge spacer, and the transmembrane region of CD8α, the costimulatory molecule CD137 intracellular domain and the CD3ζ signaling moieties. Then, the western blotting analysis was used to detect CD3ζ expression to illustrate the outcome of mCAR generation, and the results confirmed the successful construction and expression of mCAR in 293T cells after transfection. 293T cells transfected with unrelated-CAR were used as a positive control, and 293T cells without transfection were employed as a negative control (Fig. 6b). The results of the sandwich ELISA further implied that compared with 293T cells transfected with unrelated-CAR or untransfected 293T cells, 293T cells transfected with mCAR could specifically bind the uncombined MAGE-A1 antigen, which indicates that the MAGE-A1-scFv contained in mCAR could expectedly recognize MAGE-A1 antigen (Fig. 6c). The lentivirus titer was 1 × 10⁸ TU/mL after detection (Additional file 4: Figure S4). Then, the lentiviral vector encoding mCAR or unrelated-CAR was used to transfect CD3/CD28-activated T cells from a healthy donor. After 7 days of stimulation, fluorescence-activated cell sorting (FACS) analysis demonstrated that the transfection efficiencies of mCART and unrelated-CART by GFP (ZsGreen) were 77.0% and 74.3%, respectively. In comparison, MAGE-A1-PE staining showed that the transfection efficiency of mCART was 65.2%, which was significantly higher than that of unrelated-CART (1.22%) (Fig. 6d). Then, the phenotype of the stimulated T cells after transfection was further determined by FCM analysis. One week after co-culture in the presence of CD3/CD28 antibodies, more than 90% of the sorted cells were CD3-positive, and 80% of the sorted cells were CD8-positive in mCART as well as in unrelated-CART (Fig. 6e). The results strongly suggested that T cells were successfully infected with the lentiviral vector containing the mCAR and that the characteristic mCART was verified.

When co-cultured with LUAD cell lines, mCART mediated significant cell-killing activity in a dose-dependent manner. As shown in Fig. 7a, the tumor-inhibitory rate of mCART in the H1299 and PC9 (MAGE-A1 positive) cell lines was progressively upregulated along with the increase in the E:T ratio of mCART. mCART, with a 20:1 ratio, showed the most effective cell killing activity. In comparison, mCART showed highly ineffective cell-killing ability in MAGE-A1-negative cell lines (HBE and PC (shMAGE)), even though the E:T ratio of mCART was elevated. Then, a fixed E:T ratio of mCART was chosen, and mCART also illustrated significant tumor-inhibitory efficacy for all MAGE-A1-positive LUAD cell lines (Fig. 7b). In all the cell viability assays, unrelated-CART and T showed no cell-killing activities, regardless of the E:T ratio selected or the cell type used. Moreover, mCART co-incubated with LUAD cells caused a large release of cytokines, including IFN-γ and IL-2. In contrast, the release of IFN-γ and IL2 remained unchanged in the unrelated-CART group and T group (Fig. 7c, d). The above data clearly showed the potent tumor-inhibitory role of mCART in MAGE-A1-positive LUAD cells.

The xenograft tumor models produced by inoculation of athymic nude mice with H1299 cells were constructed to investigate the anti-tumor function of mCART, following the protocol shown in Fig. 8a. Bioluminescent imaging of xenograft LUAD derived from luciferase-expressing H1299 cells illustrated a substantial effect on the tumors upon mCART administration (Fig. 8b, c). The tumor growth curve also confirmed that mCART led to a progressive and critical reduction in tumor burden (Additional file 5: Figure S5A). The mean body weight of nude mice in the three groups showed no significant difference (Additional file 5: Figure S5B). On day 27, all mice were sacrificed, and the xenograft tumors were removed and analyzed. The tumor volume (Fig. 8d), weight (Fig. 8e), and morphology (Fig. 8f) further confirmed that mCART can specifically target and significantly inhibit MAGE-A1-positive LUAD xenograft growth in vivo. IHC analysis of CD3 expression in xenograft tumors highly proved that mCART was able to infiltrate into tumors and exert tumor-inhibitory effectiveness (Fig. 8g).