Identification of an Unfavorable Immune Signature in Advanced Lung Tumors from Nrf2-Deficient Mice

Lung carcinogenesis exacerbated in Nrf2^−/− mice challenged with vinyl carbamate

To determine the effects of vinyl carbamate on Nrf2 KO mice on a BALB/c background, female WT and KO mice were injected i.p. with two to four doses of vinyl carbamate, using protocols previously used with Nrf2 KO mice (65) or with vinyl carbamate (46). Age-matched cohorts of Nrf2 WT and KO mice (four mice per group per time point) were harvested beginning at 20 weeks after initiation and then every 4 (24, 28, and 32) weeks. More lung tumors were consistently visible over time (20–36 weeks) both (i) in Nrf2 KO mice versus WT mice and (ii) in mice injected with vinyl carbamate four times versus two times in both WT and KO mice (average number = 2.7 ± 0.7 tumors with two doses vinyl carbamate vs. 6.6 ± 0.6 tumors with four doses in WT mice, p < 0.05; 7.95 ± 0.7 tumors with two doses vs. 10.9 ± 1.0 tumors with four doses in Nrf2 KO mice, p < 0.05). Although the number and size of the tumors increased over time in both groups, these parameters were consistently higher in the Nrf2 KO mice (Fig. 1A). At 20 weeks, 4–12 tumors were visible on the surface of lungs in Nrf2 KO mice, whereas almost no visible tumors (0–1) were observed at this time in WT mice. By 32 weeks, an average of 6.5 ± 0.95 tumors per lung were found in the WT mice and 12.25 ± 2.6 tumors were found in the Nrf2 KO group. For all time points, the number of grossly visible surface tumors was significantly (p < 0.05) higher in Nrf2 KO mice than in WT mice (Fig. 1B), in the entire lung (unfixed) or in the intact left lung (after formalin fixation but before sectioning).

OPEN IN VIEWER

FIG. 1.

Nrf2 deficiency promotes lung carcinogenesis. Nrf2 KO and WT mice were injected with vinyl carbamate between 6 and 8 weeks of age. Lungs from age-matched Nrf2 KO and WT mice were harvested for a period of 20–40 weeks after initiation with the carcinogen to examine tumor burden. (A) Representative images of lung sections stained with hematoxylin and eosin; tumors are darkly stained purple areas (10 × magnification). (B) The average number of tumors on the surface of the entire lungs or on the left lung after fixation in formalin were counted, n = 31–42 mice per group (C). Average number of tumors (C) and total tumor volume (D) per slide were calculated in both Nrf2 WT and KO groups. n = 20 mice per group. *p < 0.05. (E) Representative images of lung histopathology in WT and KO groups. KO, knockout; Nrf2, nuclear factor (erythroid-derived 2)-like 2; WT, wild-type.

After gross evaluation, each left lung was then sectioned and stained with hematoxylin and eosin. The average number of tumors (Fig. 1C) and total tumor volume (Fig. 1D) per slide were also significantly (p < 0.05) higher in the lungs of Nrf2 KO mice versus WT mice. There were no significant differences in tumor histopathology between groups as almost all of the tumors were high grade (tumefactive architecture, fused trabeculae, and distinct nucleoli and conspicuous mitoses within nuclei) by 20 weeks after initiation. Representative images demonstrating this histopathology are shown in Figure 1E. As expected based on the larger tumor size, the tumors in the Nrf2 KO mice proliferate more rapidly than the WT tumors, as the percentage of proliferating cell nuclear antigen (PCNA) positive cells is significantly (p < 0.05) higher in the Nrf2 KO tumors (Supplementary Fig. S1; Supplementary Data are available online at www.liebertpub.com/ars).

Nrf2 deficiency decreases T cell populations during lung carcinogenesis

Next, we investigated immune cell populations in the lungs and spleens of Nrf2 WT and KO mice challenged with vinyl carbamate. The same two lobes of the right lung were always collected, and the fresh tissues were processed for flow cytometry using optimized antibody panels (42). To obtain an overview of the immune cell populations in the lung, the percentages of CD45⁺ immune cells as well as total T cells (CD45⁺ and CD3⁺), CD4 helper T cells (CD45⁺, CD3⁺, and CD4⁺), and CD8 cytotoxic T cells (CD45⁺, CD3⁺, and CD8⁺), B cells (CD45⁺, CD19⁺, and B220⁺), macrophages (CD45⁺, CD11b⁺, and Gr-1⁻), and myeloid-derived suppressor cells (MDSCs) (CD45⁺, CD11b⁺, and Gr-1⁺) were analyzed. Although cohorts of paired, age-matched lungs (four mice per group) were collected 20–36 weeks after initiation with vinyl carbamate, unless otherwise noted, the changes in the Nrf2 KO versus WT lungs were consistent over time, and Nrf2 KO and WT groups were pooled when appropriate.

As shown in Figure 2A, the percentage of CD45⁺ immune cells was significantly (p < 0.05) lower in the lungs of Nrf2 KO mice than that in WT mice, primarily because of changes in T cell populations. Nrf2 KO mice had a significantly (p < 0.05) lower number of total T cells (CD45⁺, CD3⁺, 20–32 weeks; Fig. 2B), CD8 cytotoxic T cells (28–32 weeks; Fig. 2C), and CD4 helper T cells (24–32 weeks; Fig. 2D). CD4 T cells play a central role in immunity, but their contribution to antitumor immunity is complex as different types of CD4 T cells have diverse functions (86). CD8 cytotoxic T cells recognize tumor cells and are essential for the response to immunotherapies such as PD-1/PD-L1 (36).

OPEN IN VIEWER

FIG. 2.

Nrf2 deficiency alters the immune cell populations in the lung and spleen of mice challenged with vinyl carbamate. The immune cell populations in the same two lobes of right lung (A–E) or spleen (F) from Nrf2 KO and Nrf2 WT were analyzed by flow cytometry. Percentages of total immune cells (CD45⁺), total T cells (CD45⁺, CD3⁺), cytotoxic T cells (CD45⁺, CD3⁺, CD8⁺), T helper cells (CD45⁺, CD3⁺, CD4⁺), and macrophages (CD45⁺, CD11b⁺, Gr-1⁻) in the lung (A–E) and myeloid-derived suppressor cells (CD45⁺, CD11b⁺, Gr-1⁺) in the spleen (F) are shown. *p < 0.05. The changes in immune cell populations were confirmed by IHC (G). F4/80 is a macrophage marker.

OPEN IN VIEWER

The percentage of macrophages was significantly (p < 0.05) higher in the lungs of Nrf2 KO mice before 24 weeks (Fig. 2E), but this difference disappeared at later time points. Macrophages play a critical role in promotion of urethane-induced lung carcinogenesis (87). Fewer and smaller lung tumors will develop if macrophages are depleted during tumor initiation and early promotion in this model. There was no significant difference in B cells between the two groups in the lung (data not shown). In the spleen, the only change was a small but significant (p < 0.05; 20–40 weeks) increase in MDSCs in Nrf2 KO mice (Fig. 2F). MDSCs play important roles in immunosuppression. High infiltration of these cells is associated with poor prognosis in cancer patients (32, 56, 75), and the survival and function of MDSCs are regulated by Nrf2 (7). All of the changes in immune populations were confirmed by immunohistochemistry and representative pictures are shown in Figure 2G. As indicated in the pictures, the macrophages and CD8 cytotoxic T cells are located around the periphery of the tumors.

In addition, the decrease of the CD45⁺ population, CD3⁺ T cells, and CD4⁺ T cells was similar at early time points (4 and 8 weeks after initiation; data not shown), whereas the percentage of CD8⁺ T cells was significantly (p < 0.05) higher in the lungs of Nrf2 KO mice than in those of WT mice at early time points (Supplementary Fig. S2). There was no significant difference in immune populations between saline-injected mice and vinyl carbamate-injected mice at these time points (data not shown).

Altered gene signatures in lung tumors from Nrf2 KO versus WT mice

To further investigate the differences between Nrf2 WT and KO mice at a molecular level, lung tumors were dissected out from the lung tissue and pooled (three to four tumors). Total RNA from two independent samples per group was analyzed by RNA sequencing (RNAseq). Significantly (p < 0.05) different expression patterns were detected for 376 genes in tumors from the lungs of Nrf2 WT versus KO mice. Slightly more than half (197 out of 376) of the altered genes in the Nrf2 KO tumors were upregulated compared with WT tumors, and the rest were significantly downregulated (Fig. 3A).

OPEN IN VIEWER

FIG. 3.

Nrf2 deficiency alters global transcription in tumors in the lungs of mice challenged with vinyl carbamate. (A) An unsupervised-hierarchical clustering of Nrf2 WT and KO tumors by differentially regulated genes shows unique transcriptional profile in the two tumor types. The Nrf2 KO tumors have a distinct profile of upregulated (red) and downregulated (blue) genes, which is reversed in the WT tumors. (B) PANTHER analysis was performed on the upregulated genes (left) and downregulated genes (right) with Nrf2 deletion. This analysis identified a number of significantly overrepresented gene ontologies between the WT and KO tumors, especially in the immune system-related genes (GO:0002376) (purple).

Among the differentially regulated genes, the NFE2L2 (Nrf2 gene) was the most significantly downregulated gene in Nrf2 KO tumors, which confirms the deletion of Nrf2. Additional confirmation that Nrf2 was deleted was obtained by genotyping the mice and analyzing lung extracts by Western blotting (data not shown). Differentially regulated genes were then processed to identify over-represented genes using PANTHER analysis (Fig. 3B). As the Nrf2 cytoprotective mechanism was knocked out, the “response to stimuli gene set” (GO:0050896) including extracellular superoxide dismutase (Sod3) and glutathione peroxidase 3 (Gpx3) was downregulated in Nrf2 KO tumors. Notably, the “immune system process” (GO:0002376) was the gene set with the most significant change between groups. As listed in Table 1, a series of cytokines, chemokines, and peptide antigens in this gene set were significantly upregulated in tumors from Nrf2 KO mice. Many of these factors, including Cxcl (1, 3, 12) and Ccl (9, 11, 17) are important for regulation and chemotaxis of immune cells, especially MDSCs, and the activation of M2 macrophages (52, 89). Cytokines can be released in response to inflammation to inhibit tumor development and progression, but alternatively, cancer cells can respond to cytokines that promote growth, inhibit apoptosis, and facilitate metastasis (20). In addition to the striking effects on the immune system, a metastatic signature also emerged. Genes involved in localization (GO:0051179; Fig. 3B) and extracellular matrix (GO:0031012 and GO:0044421; Table 1), which regulate cell migration and metastasis, were differently expressed in these two groups. Overexpression of matrix metalloproteinase (MMP)12, which is upregulated in KO tumors, plays a key role in modulating myelopoiesis, immune suppression, and lung tumorigenesis (61). MMP25, which is elevated in many cancer types and promotes tumor invasion and metastasis through activation of MMP2, was also upregulated in Nrf2 KO tumors (69). Notch signaling (Table 1) is known to crosstalk with the Nrf2 pathway, and aberrant crosstalk between Nrf2 and notch plays an important role in lung carcinogenesis (70). The differential expression of Notch1 in the lung of Nrf2 WT versus KO mice was confirmed by Western blotting (Supplementary Fig. S3). Consistent with the observation of more and larger tumors in Nrf2 KO versus WT mice, cell growth related genes [insulin-like growth factor-binding proteins 5 and 6 (Igfbp5, Igfbp6) and bone morphogenetic protein 6 (Bmp6)] were upregulated and cell death related genes (Srcap and Thsd4) were downregulated in the Nrf2 KO lung tumors (Table 1). P27, an indicator of cell cycle arrest, was significantly downregulated in Nrf2 KO mice versus Nrf2 WT mice (Supplementary Fig. S3). Numerous genes involved in metabolic pathways were also significantly over represented, including heme transport (GO:0015886), negative regulation of cellular amide metabolic process (GO:0034249), and aldonic acid metabolic process (GO:0019520). The whole list of significantly altered genes can be found in Supplementary Table S1. The raw data and processed data are available on Gene Expression Omnibus (GEO) (GSE99338).

Cytokines are upregulated in tumors and lungs from Nrf2 KO mice

Both the flow cytometry data and RNAseq data revealed the important role of Nrf2 on the immune system in this lung cancer model. To confirm these results, four cytokines listed in Table 1 were selected to validate ex vivo: Cxcl1, Cxcl12, Ccl9, and Csf1. These cytokines and chemokines are all highly relevant in cancer. Chemokine (C-X-C motif) ligand 1 (Cxcl1) attracts MDSCs (CD11b⁺ and Gr1⁺) into the tumor, which then produce chemokines that enhance tumor survival and inhibit CD8 cytotoxic T cells. It also recruits tumor-associated macrophages (TAMs) and cancer-associated fibroblasts. High Cxcl1 levels are associated with higher metastatic potential and poor prognosis (52, 74, 84). Chemokine (C-X-C motif) ligand 12, Cxcl12, which is the sole ligand of CXCR4 (C-X-C chemokine receptor type 4), provides an attractive niche for tumor cell migration and colonization. It is highly expressed not only in primary lung cancer cells but also in the brain, liver, bone marrow, and adrenal glands, which are all common sites for lung cancer metastasis (24, 77, 78, 82). Chemokine (C-C motif) ligand 9 (Ccl9) is highly induced in myeloid cells by transforming growth factor-β signaling and is also secreted by premetastatic tumor cells to recruit more myeloid cells, which enhances tumor cell survival and metastasis (81). Colony stimulating factor 1 (Csf1) is involved in cancer growth, survival, and metastasis. Blockade of CSF1/CSF1R reprograms TAMs to enhance antigen presentation, thus improving the therapeutic effects of immunotherapies (1, 19, 30, 89).

To validate the bioinformatics predictions (Fig. 4A) of the upregulation of these cytokines and chemokines, aliquots of the same RNA samples analyzed by RNAseq were used for real-time PCR. All four cytokines were significantly upregulated in Nrf2 KO versus WT tumors (Fig. 4B–E, p < 0.05). To confirm these changes, total RNA was isolated from lung tissue, and real-time PCR was used to evaluate relative gene expression of the cytokines. Notably, significant increases in expression of these cytokines in Nrf2 KO lungs were still detected (Fig. 4F). Cxcl1, Csf1, Cxcl12 (24 weeks), and Ccl9 (32 weeks) were significantly upregulated in the lungs of Nrf2 KO mice (Fig. 4F). The expression of Cxcl1 and Cxcl12 protein was detected by enzyme-linked immunosorbent assay (ELISA). Cxcl1 (24–32 weeks; Fig. 4G) and Cxcl12 (32 weeks; Fig. 4H) are significantly upregulated in the Nrf2 KO lungs compared with those in the WT lungs.

OPEN IN VIEWER

FIG. 4.

Differential gene expression of cytokines in the tumors and lungs of Nrf2 WT versus KO mice. (A) Predicted gene expression levels (FPKM) based on RNAseq data of lung tumors from Nrf2 WT and KO mice for Cxcl1, Ccl9, CXcl12, and CSF1. The same tumor RNA analyzed by RNAseq was used to confirm the cytokine gene expression using real-time PCR (B–E). *p < 0.05 WT versus KO. In (F), lung tissues from mice challenged with vinyl carbamate were harvested at three time points, 24–32 weeks after initiation. Total RNA was extracted from the lung and relative gene expression of Csf1, Cxcl1, Cxcl12, and Ccl9 was analyzed by real-time PCR. n = 4 lungs per group at each time point. *p < 0.05 WT versus KO at 24 weeks (Csf1, Cxcl1, Cxcl12) or 32 weeks (Ccl9). The production of Cxcl1 (G) and Cxcl12 (H) in the lung was detected by enzyme-linked immunosorbent assay. *p < 0.05 WT versus KO at 24–32 weeks with Cxcl1 and 32 weeks with Cxcl12. Data are presented as mean ± standard error of the mean. Ccl9, chemokine (C-C motif) ligand 9; Csf1, colony stimulating factor 1; Cxcl1, chemokine (C-X-C motif) ligand 1; Cxcl12, chemokine (C-X-C motif) ligand 12; FPKM, fragments per kilo bases per million reads; RNAseq, RNA sequencing.

OPEN IN VIEWER

To further validate the regulation of Nrf2 on cytokine expression and identify which cell type is expressing the cytokines, in vitro studies with both tumor cells and immune cells were performed. VC1 cells, a primary lung tumor cell line extracted from the vinyl carbamate-induced lung cancer model (46), were treated with CDDO-Im (a synthetic triterpenoid and potent Nrf2 activator), and the expression of the cytokines was detected by real-time PCR. Treatment with the Nrf2 activation significantly (p < 0.05) decreased messenger RNA (mRNA) expression of these four cytokines in the lung cancer cells (Fig. 5A). Similarly, in RAW264.7 macrophage-like cells stimulated with lipopolysaccharide (LPS), activation of Nrf2 by CDDO-Im significantly (p < 0.05) decreased Csf1 mRNA expression (Fig. 5B). Cxcl1 and Cxcl12 were expressed at very low levels in RAW264.7 cells, even after treatment with LPS or conditioned media from VC1 cells (data not shown), and there was no significant change in Ccl9 expression after treatment with CDDO-Im (Fig. 5B). These experiments suggest that Nrf2 regulates cytokines in both tumor cells and immune cells and that the expression of different cytokines is cell type dependent.

OPEN IN VIEWER

FIG. 5.

Nrf2 activation decreased cytokine production in both tumor cells and immune cells. VC1 (primary mouse lung cancer cells), (A) and RAW264.7 (macrophage-like cells), (B) were treated with 300 nM CDDO-Im (a potent Nrf2 activator) for 24 h, and RAW264.7 cells were stimulated with 1 ng/mL LPS 20 mins after CDDO-Im treatment. Total RNA was extracted from the cells, and gene expression of Cxcl1, Csf1, Ccl9, and Cxcl12 was detected by real-time PCR. *p < 0.05 CDDO-Im treated group versus control for each cytokine. LPS, lipopolysaccharide.

The regulation of immune response in the mouse model is consistent with data from patients with lung adenocarcinomas

As a striking upregulation of immune responses was found in lung tumors from Nrf2 KO mice, we wanted to verify whether these results were consistent in human patients. We accessed genomic data from lung cancer patients through the TCGA human cancer database, and cBioPortal was used to analyze the genetic alterations in Nfe2l2 and Keap1 within lung cancer patients. As has been reported (71), Keap1 alterations (mutations and deletions) are most frequent in lung adenocarcinoma patients, whereas Nfe2l2 alterations (mutations and amplifications) are more frequent in lung squamous carcinoma. Both loss-of-function alterations of Keap1 and gain-of-function alterations of Nfe2l2 induce high constitutive Nrf2 activity. Oncoprints of Keap1 genetic alterations in lung adenocarcinoma patients (12) (230 samples) and Nfe2l2 genetic alterations in lung squamous carcinoma patients (11) (178 samples) are shown in Figure 6A and C, respectively. Enrichment of gene ontologies was identified through Bayes factor. As shown in Table 2, the expression of immune response-related gene ontologies (GO:0006955, GO:0019882, and GO:0006959) was consistently preserved in human tumors. Then, gene expression of individual cytokines was analyzed, as cytokines were differentially regulated in our mouse model. Cxcl1, 2, 10, 11, 12, 14, Ccl2, 3, 4L1, 13, 22, and 28 and Csf1, 1R, 2RB, and 3R were all significantly downregulated in Keap1-altered adenocarcinoma patients. Cxcl1, 2, 3, 6, 12, 17, Ccl15, 17, 22, 28, and Csf2, 3R were significantly downregulated in Nrf2-altered squamous carcinoma patients. Cxcl1, Cxcl12, and Csf1, the three cytokines upregulated in Nrf2 KO mice with adenocarcinomas, were consistently downregulated in Keap1 loss-of-function patients (Fig. 6B). These results are consistent with the function of Keap1 as a negative regulator of Nrf2, as Nrf2 would be constitutively active in these samples instead of deleted as in the mouse model. Cxcl1 was also downregulated in patients with gain-of-function Nrf2 amplifications and alterations (Fig. 6D). In summary, our findings suggest that Nrf2 regulation of cytokines in lung cancer can be found in both mice and humans.

OPEN IN VIEWER

FIG. 6.

The regulation of immune response by Nrf2 in the mouse model is consistent with data in lung patients. (A) Oncoprint of Keap1 genetic alterations in patients with lung adenocarcinoma (12) (230 samples). Unaltered patients are not shown. Alterations of Keap1 are mainly deletions and mutations. (B) Individual genes (Cxcl12, Cxcl1, Csf1) were downregulated in lung adenocarcinoma patients with Keap1 alterations. p < 0.05. Results are presented as box plot with maximum, third quartile, median, first quartile, and minimum (top to bottom). (C) Oncoprint of Nfe2l2 genetic alterations in lung squamous carcinoma patients (11) (178 samples). Unaltered patients are not shown. Alterations of Nfe2l2 are mainly amplifications and mutations. (D) Cxcl1 was downregulated in lung squamous carcinoma patients with Nfe2l2 alterations. p < 0.05. Results are presented as box plot with maximum, third quartile, median, first quartile, and minimum (top to bottom). Keap1, Kelch-like ECH-associated protein 1; mRNA, messenger RNA; NFE2L2, Nrf2 gene.

In vivo lung carcinogenesis studies

All animal studies were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee at Michigan State University (MSU). Nrf2 ^−/− mice on a mixed C57Bl/6 and AKR background were generated as previously described and received as a generous gift from Dr. Jefferson Chan at the University of California San Francisco (15). These mice were backcrossed onto a BALB/c background for eight generations and were found to be 99% congenic (analysis performed by Jackson Laboratory, Bar Harbor, ME). Age-matched female Nrf2 KO female mice and Nrf2 WT BALB/c mice (JAX) were injected i.p. once a week with two to four doses of vinyl carbamate (0.32 mg per mouse per dose, ∼16 mg per kg per dose), beginning when the mice were 6–8 weeks old. The mice were fed AIN-93G diet (BioServ, Flemington, NJ) throughout the study and were weighed weekly. Age-matched cohorts of Nrf2 WT and KO mice (four mice per group) were harvested from 4 to 8 weeks after initiation with carcinogen in the short-term study or 20–40 weeks after initiation with the carcinogen in the long-term study. Lungs were harvested en bloc, inflated with phosphate buffered saline, and tied off in two perpendicular directions. Left lungs were fixed in neutral buffered formalin for histopathology. Right lungs were not fixed but instead immediately homogenized and processed for flow cytometry (two lobes) or were flash frozen and saved at −80°C (the other two lobes). Tumor parameters were assessed as previously described (44, 46) and include the number, size, and classification of the tumors.

RNA extraction and real-time PCR analysis

Lung tumors were dissected and total RNA was isolated using the RNeasy Mini Kit (Qiagen, Valencia, CA). RNA concentrations are determined by NanoDrop, and complementary DNAs (cDNAs) were synthesized by TaqMan Reverse Transcription reagents (Life Technologies, Carlsbad, CA). Total RNA from lung tissue was isolated with TRIzol (Invitrogen, Carlsbad, CA). Two micrograms RNA was used to synthesize cDNA using SuperScript III reverse transcriptase (Invitrogen). Validated Cxcl1, Ccl9, Csf1, and Cxcl12 primers were purchased from Qiagen. iQ SYBR Green Supermix (Bio-Rad, Berkeley, CA) and the ABI 7500 FAST Real-Time PCR system were used to detect gene expression. The delta–delta Ct method was used to calculate relative gene expression (48). Values were normalized to the reference gene actin and expressed as fold induction compared with WT samples.

ELISA

Lung extracts were homogenized in EBC lysis buffer. Cytokine levels were detected in lung homogenates using ELISA kits and the manufacturer's instructions (R&D Systems).

Flow cytometry

The same two lobes of the unfixed right lung were harvested from each mouse for flow cytometry. Freshly harvested lung tissue and half of the spleen were homogenized and incubated in digestion media containing collagenase (300 U/mL; Sigma), dispase (1 U/mL; Worthington), and DNAse (2 U/mL; Calbiochem) for 30 min at 37°C. Cells were then passed through a 40 μm cell strainer (BD Falcon). Lysis solution (eBioscience) was used to eliminate red blood cells. Single cell suspensions were stained with 5 μg/mL antimouse Fc block antibody (eBioscience) and two optimized panels of antibodies (42) for 30 min at 4°C. Panel 1: CD45-VioGreen (Miltenyi, clone:30F11, 3 μg/mL), Gr-1-PE (Miltenyi, RB6-8C5, 3 μg/mL), CD11b-FITC (Miltenyi, clone:M1/70, 3 μg/mL), CD19-APC (BioLegend, clone:1D3/CD19, 2 μg/mL), B220-PerCP/Cy5.5 (BioLegend, clone:RA3-6B2, 2 μg/mL). Panel 2: CD4-FITC (Miltenyi, clone:GK1.5, 3 μg/mL), CD3-PE (BioLegend, clone:145-2C11, 2 μg/mL), CD8-PerCP/Cy5.5 (BioLegend, clone:53–6.7, 2 μg/mL). Live cells were determined by propidium iodide staining (BioLegend, 5 μg/mL), and only live cells were included in the analysis. Flow cytometry was performed using an LSR II flow cytometer with DIVA 6.2 software (BD Falcon) and three laser sources (488 nm, 633 nm, and 407 nm); data were analyzed by FlowJo x.10.0.7r2 software (Tree Star).

RNA sequencing

Lung tumors were dissected out from the lung tissue and were pooled (three to four tumors) before isolating total RNA using the RNeasy Mini Kit (Qiagen). The RNA integrity number was measured using the Aligent Bioanalyzer at the MSU Research Technology Support Facility Genomics Core facility. RNAseq and the bioinformatics analysis from two independent pools were performed by Novogene (Sacramento, CA). In brief, reads were processed and mapped. Raw data were stored in FASTQ format. After the removal of adaptors and low-quality reads, reads were mapped to the mm 10 mouse reference genome using TopHat2 (38). Reads that mapped to genes were counted and normalized to gene length, then reported in FPKM (fragments per kilo bases per million reads) as previously described (53). Routine identification of differentially expressed genes was performed using DESeq2 (3). Raw and processed data were deposited on GEO and can be retrieved though record number GSE99338.

Over-representation analysis of differentially expressed genes

Up- and downregulated genes in tumors from NFR2 KO versus WT mice were processed to identify over-represented genes using PANTHER analysis as previously described (51). To identify those over-represented genes upregulated with the alteration of Keap1 in human adenocarcinomas as identified by cBioPortal, significantly enriched genes were queried for over-represented gene ontologies using GATHER (16).

Human data sets

The impact of gene loss in human lung cancer patients was identified using the TCGA pan (10), adenocarcinoma (12), and squamous cell (11) lung cancer data sets. Figures were created using cBioPortal.org (14, 22).

Statistical analyses

The ex vivo experiments were performed in triplicate and were repeated independently at least three times. Unless noted otherwise, data are presented as mean ± standard error of the mean. Results were analyzed using the t-test or one-way analysis of variance (ANOVA) (SigmaStat 3.5). In vivo data were analyzed by one-way ANOVA followed by Tukey test, or one-way ANOVA on ranks and the Dunn test if the data did not fit a normal distribution (SigmaStat 3.5). A p-value <0.05 was considered statistically significant. For RNAseq data, differential expression analysis of two conditions/groups was performed using the DESeq2 R package (3). It provides accepted and routinely used statistical analysis for determining differential expression in digital gene expression data using a model based on the negative binomial distribution. If the readcount of the ith gene in the jth sample is Kij, there is Kij NB(μij, σij2). The resulting p-values were adjusted using the Benjamini and Hochberg's approach for controlling the false discovery rate.