The present study was approved by the Kawasaki Medical School Ethics Committee (no. 1673-3), and written informed consent was obtained from all patients for the use of specimens. This study included patients with NSCLC who received platinum-based chemotherapy followed by surgery at the Kawasaki Medical School Hospital between January 2006 and October 2013. Histologic diagnosis was confirmed using hematoxylin and eosin (H&E) staining according to the WHO 2004 criteria (22) and the IASLC/ATS/ERS classification of lung adenocarcinoma (23). Pathological stages were defined according to the 7th edition of the TNM classification (24).

NSCLC tissue sections were stained according to a previously described protocol (14). The expression levels of MICA/B, ULBP-2/5/6, PD-L1, and HLA class I were determined by immunohistochemical staining (IHC) using a mouse anti-MICA/B antibody (1:50 dilution; clone no. D-8; Santa Cruz Biotechnology, Dallas, TX, USA), goat anti-ULBP-2/5/6 polyclonal antibody (polyclonal, 1:20 dilution, cat. no. AF1298; R&D Systems, Minneapolis, MN, USA), mouse monoclonal anti-PD-L1 antibody (1:100 dilution; clone no. SP142; Spring Bioscience Corp., Pleasanton, CA, USA), or mouse monoclonal anti-HLA class I (HLA-A, B and C) antibody (1:100 dilution; clone no. EMR8-5; Medical Biological Laboratories Co., Ltd., Nagoya, Japan) followed by poly-HRP-conjugated goat anti-mouse/rabbit secondary antibody (cat. no. K5007; Dako, Santa Clara, CA, USA) or anti-goat HRP-DAB Cell & Tissue Staining kit (cat. no. CTS008; R&D Systems). To assess the expression level of each marker, the slides were evaluated by two investigators (RO and AM) who were blinded to the corresponding clinicopathological data. The intensity scoring for staining was defined as follows: 0, no staining; 1+, weak staining that was visible only with high magnification; 2+, moderate staining (between 1+ and 3+), and 3+, strong staining that was visible with low magnification.

The human NSCLC cell lines A549 and PC-9 were obtained from Riken BRC through the National Bio-Resource Project of MEXT (Tsukuba, Japan) and the IBL Cell Bank (Gunma, Japan), respectively. The genotypes of the cell lines were identified with the PowerPlex 16 STR System (Promega, Fitchburg, WI, USA). Both cell lines were maintained as previously described (25). For cell culture experiments, cisplatin (Wako Pure Chemicals Industries Ltd., Osaka, Japan) and tofacitinib (Selleck Chemicals, Houston, TX, USA) stock solutions were prepared in dimethyl sulfoxide (DMSO) (Sigma-Aldrich; Merck KGaA, Darmstadt, Germany), whereas IFNγ (PBL Assay Science, Piscataway, NJ, USA) stock solution was prepared in phosphate-buffered saline (PBS) (−).

Previously, the pharmacokinetics study showed that the platinum concentration in plasma ranged from 1.7 to 9.6 µg/ml in patients treated with 50–100 mg/m² of cisplatin (26), which was equivalent to cisplatin concentration of 8.7–49.2 µM. Based on this report, both A549 and PC-9 cells were treated with 10 µM cisplatin thrice every 1–2 weeks to establish the NSCLC cell models by repeated exposure to cisplatin. Second or third treatment with cisplatin was undergone after the cells reached 50% confluency in a cell culture dish.

The WST-1 assay was performed using the Cell Proliferation Reagent WST-1 (Roche Diagnostics, Basel, Switzerland) as previously described (25). In brief, NSCLC cells were cultured in triplicate wells of 96-well flat-bottomed plates with 1–100 µM of cisplatin in culture medium for 48 h, then WST reagent were added to the wells according to the manufacturer's protocol. The colorimetric reaction was measured with the spectral scanning multimode reader Varioskan Flash (Thermo Fisher Scientific, Inc., Waltham, MA, USA). Cisplatin-mediated inhibition of cell proliferation was calculated as described previously with the following formula: 100× (absorbance of the wells for the cells treated with cisplatin/absorbance of the wells for the cells treated with DMSO) (25).

Extracellular staining was performed with fluorochrome-conjugated antibodies as previously described (27). The following antibodies were used for staining: PE-labeled MICA (clone no. 159227; R&D Systems), allophycocyanin-labeled MICB (clone no. 236511; R&D Systems), PE-labeled MICA/B (clone no. 6D4; BioLegend, San Diego, CA, USA), Alexa Fluor 488-labeled ULBP-1 (clone no. 170818; R&D Systems), PE- and allophycocyanin-labeled ULBP-2/5/6 (clone no. 165903; R&D Systems), PE-labeled ULBP-3 (clone no. 1166510; R&D Systems), Alexa Fluor 488-labeled ULBP-4 (clone no. 709116; R&D Systems), PE-labeled PD-L1 (clone no. 29E.2A3; BioLegend), allophycocyanin-labeled HLA-A, B and C (clone no. G46-2.6; BioLegend), as well as PE-, allophycocyanin- and Alexa Fluor 488-labeled anti-mouse IgG1κ (clone no. MOPC-21; BioLegend) and IgG2bκ (clone no. MOPC-173; BioLegend) as isotype controls. The cells were assayed using a FACSCanto II flow cytometer (BD Biosciences, San Diego, CA, USA) and analyzed using FlowJo software 6.4.7 (Tree Star, Inc., Ashland, OR, USA). The increase in mean fluorescence intensity (ΔMFI) was calculated as: (MFI with specific mAb-MFI with isotype control)/MFI with isotype control). The relative MFI (rMFI) values were calculated to compare the differences between ΔMFI of a specific treatment and control as: 100× (ΔMFI of a specific treatment/ΔMFI of the control treatment).

Peripheral blood mononuclear cells (PBMCs) were collected from healthy donors by gradient centrifugation with Vacutainer CPT mononuclear cell preparation tubes (BD Biosciences, San Jose, CA, USA). NK cells were purified using a NK cell isolation kit (Stem Cell Technologies, Vancouver, BC, Canada) and incubated with 100 IU/ml IL-2 (Teceleukin; Shionogi & Co., Ltd., Osaka, Japan) for 24 h as previously described (25). The NSCLC cells, with or without repeated exposure to 10 µM cisplatin, were tested for sensitivity to NK cell-mediated cytotoxicity using an LDH release assay kit (CytoTox 96 Non-Radioactive Cytotoxicity assay; Promega) and a Varioskan Flash spectral scanning multimode reader (Thermo Fisher Scientific, Inc.) as previously described (25). To analyze the involvement of NKG2D in the cytotoxicity of NK cells, NK cells were co-incubated with 20 µg/ml of anti-NKG2D blocking antibody (clone no. 1D11; BioLegend) or an isotype-matched control antibody (clone no. 11711; R&D Systems). For the LDH release assay, blood samples were collected only from researchers who were involved with this study; therefore, written informed consent was not required. The Ethics Research Committee of the Kawasaki Medical School approved both the study and this consent procedure (no. 1217-5).

Differences in means were evaluated with the Student's t-test. All analyses were performed at a significance level of 5% (P<0.05) using GraphPad Prism 5 (GraphPad Software Inc., La Jolla, CA, USA).

We collected tissue samples from 10 patients who received more than two courses of platinum-based chemotherapy followed by surgery, and the expression levels of MICA/B and ULBP-2/5/6 were assessed by IHC. The patient characteristics are shown in Table I. Following chemotherapy, MICA/B expression was downregulated in 2 of 10 patients (Patient #1 and #3) (Fig. S1), whereas ULBP-2/5/6 expression was downregulated in 4 of 10 patients (Patient #1, #8, #9 and #10) (Fig. S2 and Table II), which could lead to tumor immunoescape from NK cells. Surprisingly, these findings contrasted the results of our previous in vitro study, which showed that cisplatin enhances NKG2D ligand expression in NSCLC cell lines (14). One important difference between our present and previous studies (14) is the number of exposures to chemotherapeutic reagents: Repeated vs. single, respectively. For this reason, in the subsequent in vitro experiments, we further investigated the effect of repeated exposure to cisplatin on NKG2D ligand expression.

Through IHC, we also evaluated the expression of both PD-L1 and HLA class I, which are important for tumor recognition by T cells. Notably, the expression of PD-L1 was upregulated in 3 of the 10 patients (Patient #4, #5 and #8) (Fig. S3) and that of HLA class I was downregulated in 5 of the 10 patients (Patient #1, #4, #7, #8 and #10) (Fig. S4 and Table II), supporting tumor immunoescape from T cells.

We previously reported that MICA/B was overexpressed in ~30% of patients with NSCLC (14) and all the tested NSCLC cell lines (25). Although no study has evaluated ULBP-2 expression in tumor tissues from patients with NSCLC, soluble ULBP-2 was detected in the serum of patients with lung cancer (28), and this ligand was expressed in all the tested NSCLC cell lines (25). To investigate whether repeated exposure to cisplatin alters the expression of NKG2D ligands in NSCLC cell lines, NSCLC cells were treated with 10 µM cisplatin thrice, once every 1–2 weeks. First of all, cisplatin resistance was assessed by a WST-1 cell proliferation assay in both the original cells (-original) and cells repeatedly exposed to cisplatin (-C1, -C2, -C3). A549-C cells showed marginal resistance, whereas PC-9-C cells showed marginal sensitivity to cisplatin (Fig. 1).

Next, the expression levels of NKG2D ligands (MICA, MICB and ULBP-1, ULBP-2/5/6, ULBP-3, ULBP-4), PD-L1 and HLA class I in NSCLC cell lines were evaluated by flow cytometry. After exposure to 10 µM cisplatin thrice, the expression of MICA in A549 cells and one of ULBP-2/5/6 in PC-9 cells was significantly upregulated. Additionally, the expression of PD-L1 was significantly upregulated in both the cell lines, and the expression of one of the HLA class I was significantly enhanced in the PC-9 cells. The basal expression of MICA in the PC-9 cell line and the expression levels of MICB, ULBP-1, ULB-3 and ULB-4 in both the cell lines were too weak (ΔMFI <0.5) to analyze (Fig. 2).

To analyze the effect of repeated exposure to cisplatin on NK cell-mediated cytotoxicity, the sensitivity of A549 and PC-9 cells to NK cells was evaluated by an LDH release assay. As expected, NK cell-mediated death increased in the A549 and PC-9 cells after exposure to cisplatin thrice (Fig. 3A). To verify that an NKG2D receptor was involved in the repeated exposure to cisplatin-induced sensitivity to NK cell-mediated death, purified NK cells were pretreated with an anti-NKG2D blocking antibody prior to the assay. In this assay, we selected A549-C2 and PC-9-C3 cells since these cells showed the highest cisplatin-resistance phenotype among the three established phenotypes. The anti-NKG2D blocking antibody inhibited NK cell-mediated lysis of the cisplatin-treated tumor cells, whereas an isotype control antibody had no effect on both cell lines (Fig. 3B). These findings indicated that the NK cell-mediated cytotoxicity of repeated cisplatin exposure was dependent on NKG2D-NKG2D ligand interaction.

To determine the reason for the difference in NKG2D ligand expression in vivo and in vitro, we focused on the immune response of NSCLC cells as another major difference between the two environments is the presence of host immunity. IFNγ is an important cytokine secreted by NK cells to attack tumor cells (29). Previous reports have shown that IFNγ increases PD-L1 expression (15) and decreases NKG2D ligand expression (30) in cancer cells. Thus, we evaluated the effect of IFNγ stimuli on the expression of NKG2D ligands, PD-L1, and HLA class I in NSCLC cell lines. We found that IFNγ significantly downregulated MICA/B in A549 and ULBP-2/5/6 in PC-9 cells, respectively. We previously reported that IFNγ enhanced PD-L1 expression in A549 cells, which was blocked by the JAK inhibitor tofacitinib (31). In line with our previous report (31), the present study demonstrated that IFNγ significantly upregulated PD-L1 in A549 cells and marginally but significantly enhanced PD-L1 expression in PC-9 cells. On the other hand, IFNγ upregulated HLA class I in both cell lines. Moreover, the JAK-STAT inhibitor tofacitinib significantly blocked the IFNγ-induced decrease in MICA/B and blocked the IFNγ-induced increase in PD-L1 in both cell lines (Fig. 4). These findings suggest that both IFNγ-induced decrease in MICA/B and IFNγ-induced increase in PD-L1 are mainly regulated by the JAK/STAT pathway in NSCLC cells. Notably, the IFNγ-treated NSCLC cell lines showed a similar trend for the expression of NKG2D ligands and PD-L1 to that of patients with NSCLC who received platinum-based chemotherapy (Figs. S1–S3), suggesting that the expression levels of NKG2D ligands and PD-L1 after platinum-based chemotherapy are determined not by the chemotherapeutic drug but by IFNγ. This process is one reasonable mechanism that may explain the effect of chemotherapy combined with a PD-1/PD-L1 axis-targeting drug on the improvement of clinical outcome of patients with NSCLC (21).