Human specimens containing 94 cases of NSCLC tissues with adjacent normal lung tissues (> 5 cm from the lesion) for immunohistochemical analysis were collected from February 2013 to December 2014. Fresh surgical specimens of NSCLC tissues and paired adjacent normal lung tissues for testing CARD9 expression and location by western blot, polymerase chain reaction (PCR), and immunofluorescence were collected from July 2018 to December 2018. All these samples were obtained from the Huadong Hospital, Fudan University. The inclusion criteria were as follows: patients with primary lung cancer without other tumor diseases; all patients who underwent radical resection had no chemotherapy or radiotherapy before surgery; absence of cancer within 5 years before surgery. Patients without accurate medical records were excluded from this study. The specimens were staged according to the American Joint Committee on Cancer and the histological type were classified according to the World Health Organization classification. Paraffin blocks containing formalin-fixed specimens were available for histological examination and immunohistochemistry (IHC) from 94 NSCLC tissues and paired adjacent normal lung tissues. The follow-up began from the date of surgery and the deadline was April 1, 2017. Overall survival (OS) was defined as the time from surgery to death or from surgery to the deadline for the surviving patients. Progression-free survival (PFS) was defined as the time from surgery to the first progression or death.

All sections were heated at 60℃ for 30 minutes and then de-waxed in xylene and rehydrated in a series of graded ethanol. The endogenous peroxidase activity was blocked with 3% hydrogen peroxide for 10 minutes and subsequently microwave antigen retrieval was performed with ethylenediaminetetraacetic acid buffer (1 mmol/L, pH 8) for 30 minutes. After being blocked with the goat normal serum for 10 minutes, the sections were incubated with rabbit monoclonal anti-CARD9 antibody overnight at 4℃ followed by incubation with horseradish peroxidase (HRP)–conjugated goat anti-rabbit secondary antibody for 1 hour at room temperature. The immunostaining was performed by 3,3-diaminobenzidinetetrahydrochloride and Mayer’s hematoxylin counterstain was applied. Negative controls consisted in the absence of primary antibody.

The level of CARD9 expression was performed by semiquantitative analysis as described previously [11]. The intensity of immunostaining was evaluated by the degree of color and then was scored as none (0), weak (1), moderate (2), and strong (3). The distribution of positive tumor cells was defined as follows: 0, 0%; 1, 1%-20%; 2, 21%-40%; 3, 41%-60%; 4, 61%-80%; and 5, 81%-100%. Both of them were evaluated by two independent observers who were unaware of the clinicopathologic features of the patients. The expression of CARD9 in all samples were finally scored by multiplying the intensity and the percentage ranging from 1 to 9. The immunohistochemistry (IHC) scores of CARD9 expression in NSCLC tissues were divided into two levels: low (0-3) and high (4-12).

The fresh tissues were frozen in tissue freezing medium at –80℃ immediately and was sliced into 4 μm section for immunofluorescence (Leica CM1950, Jena, Germany). Briefly, frozen tumor sections were fixed by Immunol Staining Fix Solution (Beyotime, Shanghai, China) at room temperature for 10 minutes and washed by phosphate buffered saline (PBS). Then it was blocked with PBS solution containing 5% goat serum and 0.2% Triton X-100 at room temperature for 30 minutes and incubated with primary antibody mixture overnight at 4℃. Anti-CD68 was used to stain macrophages, and anti–epithelial cell adhesion molecule (EpCAM) was used to stain NSCLC cells (S1 Table). After washing by PBS, freezing slices were incubated with fluorescent secondary antibody mixture at room temperature in dark for 1 hour followed by three times PBS washing. DAPI was used for nuclei staining (Beyotime) for 5 minutes. After washing, slices were covered by Antifade Mounting Medium (Beyotime) and coverslip. Confocal laser scanning microscope (Leica TCS SP8) was used for detection.

Human cell lines (human bronchial epithelioid cells [HBE], H1299, A549, 95C, 95D, PC9, and THP-1) were obtained from the Cell Bank of the Chinese Academy of Science (Shanghai, China), and BEAS-2B cells (human normal lung epithelial cells) was obtained from Beyotime. All these cells were maintained in Roswell Park Memorial Institute-1640 (RPMI-1640, Hyclon, Logan, UT) supplemented with 10% fetal bovine serum (FBS; Gibco, Carlsbad, CA) and 1% penicillin/streptomycin (Invitrogen, Carlsbad, CA) at 37℃ in a humidified atmosphere containing 5% CO₂. To establish stable CARD9 silencing cell lines, cells were transfected with GFP/CARD9 ShRNA lentivirus construct (GeneChem, Shanghai, China) (shCAR-D9 group) while cells transfected with the empty lentivirus vectors (shControl group) were considered as a negative control. The stable cell lines were selected with 5 μg/mL puro-mycin (Beyotime) for 1 week and the downregulated level of CARD9 were verified by quantitative realtime PCR (qRT-PCR) and western blot. CARD9 gene was ligated into the FV115 plasmid (Huzbio Technology, Shanghai, China), which was used to generate CARD9 overexpression cell lines. Firstly, 1×10⁶ cells were seeded into 6-well plate, which were transfected with FV115-CARD9 or FV115 plasmid with instruction of Lipofectamine 6000 (Beyotime) after 24 hours. Stably transfected cell lines were selected by puromycin and verified by real-time quantitative reverse transcription polymerase chain reaction and western blot. In the inhibition assay, SB203580 (Cell Signaling Technology, Danvers, MA) dissolved in dimethyl sulfoxide (DMSO; Sigma, St. Louis, MO) was added to culture medium when cells density rea-ched to 50%-70%. The final concentration of SB203580 was 10 nM. Same volume of DMSO was added as the control. Two hours later, cells were re-cultured in the fresh RPMI-1640 or collected for further experiments.

The total cellular proteins were extracted with RIPA buffer (Beyotime). Equal amounts of protein (30 μg/lane) were separated on sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and then transferred to polyvinylidene fluoride membranes (Millipore, Burlington, MA). After blocking with PBS buffer (Beyotime) containing 5% non-fat milk and 0.1% Tween 20 (Beyotime), membranes were incubated with primary antibody over-night at 4℃. Subsequently HRP-conjugated secondary antibodies were incubated for 1 hour at room temperature and developed with enhanced chemiluminesence (Beyotime). All antibodies used were listed in S1 Table.

Total cellular RNA was isolated with TRIzol reagent (Beyotime) following the manufacturers’ instructions. Equal amounts of RNA (1 μg) were used as template to synthesize first-strand cDNA with the PrimeScript RT reagent Kit with gDNA Eraser (Takara, Dalian, China). Then each PCR reac-tion contained a total volume of 20 μL SYBR Premix Ex Taq (2 μL cDNA, 0.4 μL of each specific primer, 0.4 μL of ROX Reference Dye [50×], 10 μL SYBR Premix Ex Taq, and 6.8 μL RNase free water, Takara) was performed with an Applied Biosystems 7500 Fast Real-Time PCR System (Foster City, CA, USA) at the following conditions: 95℃ for 30 seconds, 40 cycles at 95℃ for 15 seconds, and 60℃ for 30 seconds. β-actin was used as an internal control to normalize the expression of target gene. The PCR reaction for each sample was done in triplicate and the experiments were performed for 3 times. The relative quantity of target gene expression was calculated according to 2^-ΔΔCT method. Primers used in this study were as follows: CARD9: F, ATGTCGGACTACGAGAACGAT; R, TGATGCGTGAGGGGTCGAT; β-actin: F, GTGGACATCCGCAAAGAC; R, AAAGGGTGTAACGCAACTA.

Cell growth assays were performed following the manufacturer’s instruction to test the cell viability. Cells were seeded at a density of 2×10³/well into the 96-well plates and then checked at 0, 24, 48, 72, and 96 hours by the Cell Counting Kit-8 (Beyotime) and the Synergy H1 microplate reader (BioTek, Winooski, VT) at 450 nm. As for clonal formation assay, following transfection and selection, 200 cells were seeded in 6-well plate in triplicates and incubated for 7 days. Then cells were fixed with 10% ice-cold methanol and stained with 0.5% crystal violet solution. Colonies consisting more than 50 cells per colony were counted. Independent experiments were conducted triplicates.

For cell cycle analysis, cells were starved for 12 hours before trypsinization (Beyotime). Trypsinizated cells were washed by cold PBS and centrifuged at 2,000 rpm for 5 minutes. Fifty microliters of PBS was used to resuspend the cells gently, and then 950 μL 75% ethanol was added to fix the cells at 4°C overnight. The fixed cells were washed with cold PBS, centrifuged at 2,000 rpm for 5 minutes and incubated with 500 μL PI/RNase Staining Buffer (BD Biosciences, Franklin Lakes, NJ) at room temperature in dark for 15 minutes. Before analysis, cells were sieved by 200 mesh sieves. For cell apoptosis analysis, cold PBS-washed cells were resuspended by 1× binding buffer and then every 1×10⁵ cells were incubated with 5 μL Annexin-V/7-ADD (BD Biosciences) double staining at room temperature in dark for 30 minutes. Established cells above were tested by flow cytometry (BD FACSAria II). Flow jo 10 and Modfit LT were used to analyze cell apoptosis and cell cycle, respectively.

For wound healing assay, cells were seeded in 6-well plates with 5×10⁶/well. When they grew to 90% confluence, the cell monolayer was scratched with a sterile 200 μL pipette tip. The detached cells were removed by washing with PBS and then the cells were maintained in medium containing 1% FBS and 2 nM Thymidine (CSNpharm, Arlington Heights, IL). Cells migrated to the wounded area were photographed at 0 hour and 24 hours using microscopy (Nikon ECLIPSE Ti, Tokyo, Japan). The experiments were performed in triplicate. For the invasion assay, 50 μL Matrigel (BD Biosciences) was precoated on the upper chambers (Corning Costar, New York, NY) and then 2×10⁴ cells with 200 μL serum-free medium were added. Five hundred microliters of medium containing 10% FBS was added to the lower chambers. After 24 hours of incubation at 37℃, cells on the top of the membrane were removed and cells on the bottom of the membrane were fixed with 10% ice-cold methanol and stained with 0.5% crystal violet. Chambers placed on the glass slide were photographed under microscopy (Nikon ECLIPSE Ti). All of the cells under the chambers were counted without overlapped horizons. The experiments were performed in triplicate.

Cell lysates from successfully transfected H1299 cells or PC9 cells were collected and immunoprecipitated with either 5 μg anti-CARD9 antibody (Santa Cruz Biotechnology, Santa Cruz, CA) or 1 μg anti-mouse IgG antibody (Beyotime) at 4℃ overnight. The cell lysates were then added with 30 μL protein A/G magnetic beads (MedchemExpress, Monmouth Junction, NJ) and rotated for at 4℃ overnight. After washing by PBS containing 0.5% Triton X-100 for 5 times, beads were collected in 15 μL 1× SDS loading buffer and boiled for 5minutes. Then protein samples in SDS were collected for immunoblotting as we described above. HRP-mouse anti-rabbit IgG heavy chain specific was used as the secondary antibody to BCL10 antibody to exclude the influence of light chain.

Statistical analyses and graphs were performed by SPSS ver. 25.0 software (IBM Corp., Armonk, NY) and Prism ver. 7.00 (GraphPad Software, San Diego, CA). The results of Western blot were quantified by Image J (National Institutes of Health, Bethesda, MD). Data under the banks presented as the mean of three results of Western blot. IHC score of CARD9 expression in NSCLC tissues and adjacent normal tissues were analyzed by Wilcoxon rank-sum text. The relationship between CARD9 expression and clinicopathological features was analyzed by the chi-square test and Fisher exact test. The survival curves were made by the Kaplan-Meier method and analyzed by the log-rank test. Binary logistic regression models were used to evaluate the independent prognostic factors. The differences between vector groups and CARD9 groups, shControl groups, and shCARD9 groups were analyzed by Student’s t test. p < 0.05 was considered statistically significant in all the cases.

Ethical approval was obtained from the Clinical Research Ethics Committee of Huadong Hospital (No. 2014K012). All patients signed written informed consent.

To investigate the expression of CARD9 in NSCLC, we first analyzed the difference of CARD9 mRNA between NSCLC tissues and adjacent normal lung tissues using The Cancer Genome Atlas (TCGA) database. Both in lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC), CARD9 mRNA was decreased in tumor tissues compared with paired normal lung tissues (Fig. 1A, p=0.001; Fig. 1B, p < 0.001). To confirm the results, we examined our cohort consisting 94 patients with NSCLC by IHC. CARD9 was mainly located in cell the nucleus and cytoplasm, and IHC score showed CARD9 in carcinoma tissues was lower than in adjacent normal tissues (p < 0.001) (Fig. 1C and D) both in LUAD and LUSC. Consistently, by testing surgical lung resection specimens, CARD9 was found to be decreased in lung tumor tissues in comparison with normal tissues at protein (Fig. 1E) and mRNA levels (p=0.039) (Fig. 1F). CARD9 has previously been identified to be expressed in myeloidderived cells such as macrophages, dendritic cells and MDSCs. To confirm the role of CARD9 in lung tumorigenesis, we used immunofluorescence to testify the cellular location of CARD9. As shown in Fig. 1G, despite weak expression, we can still observe positive CARD9 expression in macrophages (anti-CD68) and lung tumor cells (anti-EpCAM).

Based on TCGA datasets, the level of CARD9 in LUAD was negatively associated with TNM stage, tumor size, lymph node metastasis, and remote metastasis (Table 1). Consistent with this, patients with higher expression of CARD9 in LUAD had better OS than patients with lower expression of CARD9 (hazard ratio [HR], 0.710; p=0.023) (Fig. 2A). However, CARD9 expression in LUSC was only related to lymph node metastasis (Table 2), and OS in LUSC patients was not significantly different regardless of CARD9 expression (HR, 1.233; p=0.152) (Fig. 2B). We surmised that CARD9 had higher clinical value for LUAD. As shown in Table 3, CARD9 was negatively correlated with TNM stage (p < 0.001), histological differentiation (p=0.002), T category (p=0.001), lymph node metastasis (p=0.040), and remote metastasis (p < 0.001). The IHC results also showed decreased CARD9 in Ⅳ stages of NSCLC compared with Ⅰ stages (Fig. 1C). And the group with higher CARD9 expression had better OS (HR, 0.304; p=0.027) (Fig. 2C) and PFS (HR, 0.310; p=0.004) (Fig. 2D) than the group with lower CARD9 expression. We analyzed the relationship between CARD9 and prognosis in patients of LUAD in our cohort. Similarly, patients with lower CARD9 expression had a worse OS (HR, 0.161; p=0.009) (Fig. 2E) and PFS (HR, 0.326; p=0.019) (Fig. 2F). Logistic regression analysis showed that the high expression of CARD9 was an independent protective factor for OS (odds ratio [OR], 0.127; p=0.022) and PFS (OR, 0.126; p=0.004), excluding confounding factors of age, sex, smocking status, and TNM stage. Therefore, we speculated that CARD9 might be a potential predictor of prognosis for NSCLC patients.

Since CARD9 was expressed both in tumor cells and macrophages, we analyzed the baseline expression of CARD9 in NSCLC cell lines (including H1299, A549, 95C, 95D, and PC9), HBE, human normal lung epithelial cells (BEAS-2B), and human leukemic monocyte (THP-1). Because of the high CARD9 production in THP-1, we loaded less concentration of THP-1 protein for easier observation of CARD9 in other cells. Under this circumstance, PC9, H1299, and HBE had detectable CARD9 expression at protein level, among which PC9 had the highest CARD9 expression (Fig. 3A). PCR result showed that, unlike western blot to some extent, 95D cells had comparable CARD9 mRNA to HBE and H1299 cells, which may be due to the regulation of posttranscriptional process (Fig. 3B). Considering CARD9 was also expressed in tumor cells and varied in different NSCLC cell lines, which indicated its potential participation in tumorigenesis, we first focused on the study of CARD9 in tumor cells. We generated H1299 and PC9 cells stably expressed CARD9 or control vector. On the other hand, CARD9-target shRNA successfully knocked down CARD9 in PC9 cells and shControl was served as control (Fig. 3C and D). Since H1299 was relatively scarce in CARD9 production compared with THP-1 and PC9, H1299 shCARD9 was not generated. Wounding healing showed the migratory ability of H1299 and PC9 was inhibited by ectopic CARD9 (p=0.002 and p=0.007, respectively) (Fig. 3E) while decreased CARD9 markedly promoted PC9 migration (p=0.030) (Fig. 3E). Meanwhile, in the matrigel invasion assays, CARD9 was found to block invasiveness capability of H1299 and PC9 (p=0.002 and p=0.010, respectively) (Fig. 3F), and when its expression was suppressed in PC9, invasion was significantly augmented (p=0.040) (Fig. 3F). We also tested epithelial-mesenchymal transition (EMT)–associated protein. In consistent with these results, increased CARD9 was accompanied with decreased N-cadherin and transforming growth factor-β, which were able to facilitate EMT process [12], though E-cadherin and vimentin were unch-anged (Fig. 3C). These suggested that CARD9 acted as an inhibitory factor in cancer progression through weakening tumor migration and invasion ability.

Then we investigated the effect of CARD9 on cell proliferation and apoptosis. Cell growth curve assays showed CARD9-overexpressing H1299 and PC9 had suppressed growth, and CARD9-knockdown PC9 had promoted growth (p < 0.001) (Fig. 4A). Consistently, cell clonal formation assays showed dampened clonal formation in high-CARD9–expressing H1299 and PC9 cells compared with their control counterparts (p=0.036 and p=0.027, respectively) (Fig. 4B), while CARD9-knowdown PC9 cells had elevated clonal numbers (p < 0.001) (Fig. 4B). We further assessed the effect of CARD9 on cell cycle. Ectopic expression of CARD9 in H1299 significantly induced G₀/G₁ arrest, exhibiting increased G0 to G1 phase cells (p=0.008) (Fig. 4C) and decreased S-phase population (p=0.004) (Fig. 4C). In PC9 cells, overexpression of CARD9 also resulted in increased G0 to G1 phase cells (p=0.001) (Fig. 4C) with a corresponding decrease in G2-phase population (p=0.016) (Fig. 4C). On the other hand, knockdown of CARD9 in PC9 cells presented decreased population of G0 to G1-phase (p < 0.001) (Fig. 4C) with elevated population of S- and G2-phase (p=0.001 and p < 0.001, respectively) (Fig. 4C), suggesting that downregulation of CARD9 would drive the transition from G1 to S. Consistent with this, CyclinD1, necessitated for cell cycle progression, was significantly attenuated when CARD9 was overexpressed in H1299 and PC9 cells and was augmented when CARD9 was decreased in PC9 cells (Fig. 4E). In addition, by examining apoptosis by flow cytometry, we found that CARD9 could lead to more late apoptosis in H1299 and PC9 cells (p < 0.001 and p < 0.001, respectively) (Fig. 4D). However, once it was knockdown, population of late apoptosis was consistently reduced (p < 0.001) (Fig. 4D). The early apoptosis is not affected. This supportive role of CARD9 on apoptosis was further evidenced by higher expression of cleaved-caspase-3 accompanied with higher level of CARD9 (Fig. 4E). Taken together, CARD9 could impede NSCLC cell proliferation but enhance apoptosis in vitro.

Previous studies had found that in anti-infection immunity, CARD9 recruited BCL10 and mucosa-associated lymphoid tissue 1 (MALT1) to form CBM (CARD9-BCL10-MALT1) complex [13,14], activating downstream such as nuclear factor κB (NF-κB) and MAPK signaling [3,6,13,15]. Surprisingly, in our study, we found that MAPK/p38 was inactivated when CARD9 was upregulated and the phosphorylation of p38 was facilitated when CARD9 was inhibited (Fig. 5A). But the other two subunits of MAPK, extracellular signal-related kinase, and JNK (c-Jun N-terminal kinase), exhibited no significant change (Fig. 5A). As for NF-κB signaling, phosphorylation of p50 and p65 were declined in CARD9-overexpressing H1299 and PC9 cells but remained unaltered in CARD9-knockdown PC9 cells (Fig. 5B). To testify whether BCL10 interacted with CARD9, we performed Co-IP in PC9 shControl, PC9 shCARD9, H1299 Vector, and CARD9-overexpressing H1299 cells. As shown in Fig 5C, in PC9 shControl cells, BCL10 was found to bind to CARD9, while in PC9 shCARD9 cells, less CARD9 was pulled down and interacting BCL10 was not detected. In H1299 cells, CARD9 was still undetectable by using CARD9 antibody in H1299 Vector cells, but when CARD9 was overexpressed, BCL10 was found in the elute lysates. These results demonstrated that BCL10 interacted with CARD9. We also searched String database [11] for well-established network proteins of CARD9. It showed that BCL10 was the known protein interacting with CARD9, and in keeping with our results, there were three p38 MAPKs (MAPK11, MAPK12, and MAPK14) associated with CARD9 (Fig. 5D).

Considering the consistent change of p-p38 in our established cell lines, we next explored which cancerous capacities were regulated by CARD9 in a p-p38-dependent manner. We used a p38-kinase–specific inhibitor SB203580 to inhibit phosphorylation of p38 in CARD9-knockdown PC9 cells and DMSO served as control (Fig. 6A). The cell growth curve assays showed a dramatical impeded growth in SB203580-treated PC9 shCARD9 cells (p < 0.001) (Fig. 6B) while the growth of PC9 shControl cells was unaffected by SB203580 (Fig. 6B). The cell clonal formation assays showed no significant difference between PC9 cells and their SB203580-treated counterparts (Fig. 6C). In addition, retarded activation of p38 significantly limited the invasiveness of PC9 shControl and PC9 shCARD9 cells (p=0.009 and p=0.039, respectively) (Fig. 6D). Migration ability was also limited both in shControl and shCARD9 PC9 cells by SB203580 (p=0.021 and p=0.002, respectively) (Fig. 6E). We further explored whether cell apoptosis and cycle progression were controlled by p38 signaling. Of note, late apoptosis in shControl PC9 cannot be induced but even be hindered by SB203580 (p=0.001) (Fig. 6F), though both late and early apoptosis of shCARD9 PC9 cells was indeed increased when intervened with SB203580 (p=0.046 and p=0.013, respectively) (Fig. 6F). This phenomenon also happened in cell cycle progression analysis. We found shCARD9 PC9 cells had accumulated G0 to G1 phase population (p=0.001) (Fig. 6G) and declined S-phase population (p=0.001) (Fig. 6G) when p-p38 was blocked, but shControl PC9 group instead had slightly increased S-phase population with SB203580 treatment (p=0.020) (Fig. 6G). Studies have shown that abrogation of overactivated p38 signaling was able to reverse enhanced G1 to S transition and anti-apoptosis effect in CARD9-knockdown cells but not in normal PC9 cells. This contradictory result may be due to activation of bypass signal when low level of p-p38 was suppressed again in shControl PC9 cells. Taken together, CARD9 impaired abilities of cell proliferation, migration, and invasion by suppressing MAPK/p38 signaling.