IKKα-deficient lung adenocarcinomas generate an immunosuppressive microenvironment by overproducing Treg-inducing cytokines

Significance

This study reveals that impaired IKKα expression or activity in lung cancer enhances differentiation of protumorigenic Treg cells through a TNF/TNFR2/NF-κB signaling pathway in both human and mouse lung ADC. Depletion of one of the molecules that are required for Treg cell induction represses lung ADC development. Thus, the components that interfere with this particular Treg differentiation provide targets for the generation of TME-modifying therapies.

Abstract

The tumor microenvironment (TME) provides potential targets for cancer therapy. However, how signals originating in cancer cells affect tumor-directed immunity is largely unknown. Deletions in the CHUK locus, coding for IκB kinase α (IKKα), correlate with reduced lung adenocarcinoma (ADC) patient survival and promote Kras^G12D-initiated ADC development in mice, but it is unknown how reduced IKKα expression affects the TME. Here, we report that low IKKα expression in human and mouse lung ADC cells correlates with increased monocyte-derived macrophage and regulatory T cell (Treg) scores and elevated transcription of genes coding for macrophage-recruiting and Treg-inducing cytokines (CSF1, CCL22, TNF, and IL-23A). By stimulating recruitment of monocyte-derived macrophages from the bone marrow and enforcing a TNF/TNFR2/c-Rel signaling cascade that stimulates Treg generation, these cytokines promote lung ADC progression. Depletion of TNFR2, c-Rel, or TNF in CD4⁺ T cells or monocyte-derived macrophages dampens Treg generation and lung tumorigenesis. Treg depletion also attenuates carcinogenesis. In conclusion, reduced cancer cell IKKα activity enhances formation of a protumorigenic TME through a pathway whose constituents may serve as therapeutic targets for KRAS-initiated lung ADC.

Cancer-related genomic aberrations generate neoantigens, which drive tumor-directed immune responses that affect tumor progression. Identifying cancer-cell-intrinsic alterations that shape tumor-directed immunity (TDI) is important for choosing and improving treatments for lung cancer, which remains the leading cause of cancer-related mortality (1). Immune checkpoint blockade has improved 5-y survival rates for patients with nonsmall cell lung carcinoma (NSCLC), which includes lung adenocarcinoma (ADC) and squamous cell carcinoma (SCC) (2, 3). Previously, we found that the CHUK gene, coding for IKKα, is a suppressor of lung ADC (4). Oncogenic KRAS mutations and CHUK nonsense mutations or homozygous and hemizygous deletions are present in ∼35% and 25% of human lung ADC patients, respectively (4–6). Patients with lung ADC whose tumors contain both CHUK hemizygous deletions and oncogenic KRAS mutations die earlier than patients with ADC with KRAS mutations alone (4, 6). IKKα inactivation in mouse lung epithelium results in spontaneous ADC formation and enhanced Kras^G12D-initiated lung ADC development (4). Urethane-induced lung ADC incidence and ADC weights were also higher in mice lacking IKKα in type II lung epithelial cells compared to wild-type (WT) counterparts (7). Ikkα deletion in pancreatic epithelial cells enhances Kras^G12D-induced pancreatic ductal ADC pathogenesis by interfering with the completion of autophagy, which results in chronic pancreatitis (8, 9).

Reduced IKKα activity increases expression of inflammatory cytokines and growth factors and enhances macrophage infiltration into sites of Ras-initiated pancreatic and skin carcinomas (10, 11). However, it remains to be determined how ADC cell-derived inflammatory responses influence cancer immunoediting to impact tumor elimination or escape (12). An adoptive CD4⁺ T cell TDI has the potential to eliminate Ras-initiated tumors at an early stage. Indeed, CD4⁺ T cells antagonize Kras-initiated lung ADC development, while tolerance-related Treg cells dampen TDI (13). We hypothesized that impaired IKKα expression or activity unleashes Kras^G12D ADC-intrinsic signaling pathways that blunt TDI and allow ADC to escape immunosurveillance and undergo further progression. Because reduced IKKα expression or activity worsens survival of patients with lung ADC (4), we investigated how IKKα deficiency affects TDI in lung ADC with the hope of identifying new targets for therapeutic intervention.

Here we describe shared patterns of increased monocyte-derived macrophage and Treg-cell scores and regulators of macrophage and Treg differentiation in human and mouse lung ADC lacking IKKα. ADC-intrinsic Ikkα ablation up-regulated expression of regulators of macrophage recruitment and Treg differentiation, thereby converting the tumor microenvironment (TME) from cancer restrictive to cancer supportive. A major effector of ADC-intrinsic IKKα deficiency is a Treg-promoting TNF/TNFR2/NF-κB signaling cascade.

Results

IKKα Deficiency in Human Lung ADC Correlates with Increased FOXP3⁺ Treg and Monocyte-Derived Macrophage Scores.

Exploring the relationship between reduced IKKα expression and the TME in human lung ADC, we found that low IKKα correlated with increased intratumoral macrophage (CD68⁺, a monocyte lineage marker) and elevated Treg (FOXP3⁺) numbers in a tissue array of 90 human lung ADC (Fig. 1 A and B and SI Appendix, Fig. S1 A and B). Immunohistochemical (IHC) staining revealed an identical pattern of CD68⁺ and CD163⁺ macrophages in human lung ADC and adjacent tissues (Fig. 1C and SI Appendix, Fig. S1C), suggesting that either CD68 or CD163 can be used to detect intratumoral macrophages. We did not observe a significantly correlated distribution between CD8⁺ T cell numbers and IKKα expression in the lung ADC array, although some high IKKα ADC showed increased CD8⁺ T cell infiltration (SI Appendix, Fig. S1 D and E). IKKα expression, which was grouped by median and quartiles, was inversely correlated with Treg and macrophage (including monocyte) scores in human lung ADC cohorts (Fig. 1D, ref. 14, and SI Appendix, Fig. S1 F and G). Furthermore, IKBKB mRNA, which encodes IKKβ that interacts with IKKα and IKKγ to form the IKK complex (15), positively correlated with macrophage scores, though its expression did not show a significant correlation with Treg scores in the same ADC cohort (Fig. 1E) (14). Curiously, IKKα and IKKβ expressions were inversely correlated (SI Appendix, Fig. S1H).

Fig. 1.

Human lung ADCs lacking CHUK/IKKα are highly correlated with elevated TNF/TNFRSF1B-associated FOXP3 Treg-cell and macrophage scores. (A, Left) The array containing 90 human lung ADCs (BCS04017a; US Biomax, Inc.) was divided into two groups of low CD68 (L-CD68)- and high CD68 (H-CD68)-positive cells, analyzed by immunofluorescence (IF) staining. (Right) The correlation between macrophages (CD68) and Treg-cell (FOXP3) numbers was further analyzed. n, case numbers; ****P < 0.0001; Student’s t test. (B) Correlation of IKKα expression and macrophage numbers in the human lung ADC array (BCS04017a) was analyzed by IHC staining. The distribution of IKKα-stained intensities from strong (+++) to negative (−) staining was divided into two groups of L(low)-CD68 and H(high)-CD68. Data represent mean ± SEM. Data were analyzed by χ² test. n, ADC numbers; ***P < 0.001. (C) Comparison of CD68 and CD163 expression patterns in human lung ADC and its adjacent tissue in human tissue arrays (HLung030PG02; US Biomax, Inc.), analyzed by IHC staining. Dark brown, positive cells; numbers on the Top, positions in the tissue array. (Scale bar, 50 μm.) (D) Correlation of CHUK expression levels with macrophage (Left) and Treg-cell (Right) scores in human lung ADCs determined by analyzing the data (14). Using quartile analysis, CHUK expression was divided into high (H-CHUK) and low (L-CHUK) expression levels. CHUK, IKKα; n, ADC numbers per group; P value at the Top of panels, Student’s t test. (E) Correlation of IKBKB expression levels with macrophage (Top) or Treg-cell (Bottom) scores in human lung ADCs determined by analyzing the data (14). Using quartiles and median analyses, IKBKB expression was divided into high IKBKB (H-IKBKB) and low IKBKB (L-IKBKB) expression levels. n, ADC numbers; P value at the Top of panels, Student’s t test. (F) Coexpression of the genes encoding molecules for macrophage and Treg-cell development in human lung ADCs determined by analyzing the data (6). P value at the Top of panels, two-tailed t test; n, ADC numbers; red lines, repression lines; r, Pearson correlation coefficient; −4 (−2 or −1), outliers that were removed from 230. (G) Correlation of H-CHUK or L-CHUK expression levels (using median and quartile analysis to divide CHUK expression groups) with the expression of CSF1, CSF1R, FOXP3, TNFRSF1B, TNF, or CCL22 genes in human lung ADCs (TCGA, PanCancer Atlas, cBioPortal). n, ADC numbers. P value at the Top of panels, Student’s t test.

By analyzing a human lung ADC cohort (6), we found that expression levels of gene pairs encoding CSF1R/CSF1, TNFRSF1B/FOXP3, FOXP3/CSF1R, FOXP3/CSF1, TNF/FOXP3, and CSF1R/CCL22 were significantly correlated (Fig. 1F), indicating an association between TNF/TNFRSF1B-linked Treg and macrophage numbers in human lung ADC. Of note, IKKα amounts were inversely correlated with expression of CSF1, CSF1R, FOXP3, TNFRSF1B, TNF, CCL22, and IKKβ in a dose-dependent manner (Fig. 1G and SI Appendix, Fig. S1I). These results suggest that reduced IKKα expression is associated with increased macrophage and Treg numbers and their regulators in human lung ADC.

IKKα Loss Enhances Kras^G12D-Induced Lung ADC Development and Macrophage and Treg Infiltration.

To investigate whether IKKα deficiency increases intratumoral macrophage and Treg numbers in mouse ADC, we expressed Kras^G12D (16) and ablated Ikkα in lung epithelial cells (Ikkα^ΔLU) by intratracheally injecting adenovirus–cyclization recombinase (Ad.Cre) (4) to generate Kras^G12D and Kras^G12D; Ikkα^ΔLU mice on a C57BL/6 background. Ad.Cre is expressed only in the cytosolic compartment, disappearing after cell division. Kras^G12D;Ikkα^ΔLU mice developed heavier lungs with significantly higher ADC burden compared to Kras^G12D mice (Fig. 2A and SI Appendix, Fig. S2A). Flow cytometry showed that ADC-associated F4/80⁺ cell numbers or F4/80⁺CD11b⁺ monocyte-derived macrophage numbers normalized to lung ADC weight were higher in IKKα-deficient tumors (Fig. 2 B and C). F4/80⁺ and F4/80⁺CD11b⁺ cell numbers versus CD45⁺ cells were also higher in IKKα-deficient tumors (SI Appendix, Fig. S2B). Treg cell numbers were elevated but CD8⁺ T cells were lower in Kras^G12D;Ikkα^ΔLU lung ADCs compared to Kras^G12D lung ADCs (Fig. 2 B–E and SI Appendix, Fig. S2C).

Fig. 2.

IKKα loss promotes lung carcinogenesis associated with increased Foxp3 Treg-cell and macrophage numbers. (A) An approach (Top) of Kras^G12D activation and Ikkα ablation in lungs of Kras^G12D and Kras^G12D;Ikkα^f/f mice by adenovirus-Cre (Ad.Cre, or Cre) and the lung appearances (Bottom) of Kras^G12D and Kras^G12D;Ikkα^ΔLU mice at 3 mo after Ad.Cre treatment. (B and C) Flow cytometric analyses of F4/80⁺CD11b⁺ (B) or F4/80⁺ (C) cells in Kras^G12D (n = 5) and Kras^G12D;Ikkα^ΔLU (n = 6) lung ADCs. Representative images for flow cytometry on the Left and statistical analyses on the Right are shown. Data represent mean ± SEM (multiple experiments). This result is representative. **P < 0.01; ****P < 0.0001; Student’s t test. (D) Flow cytometric analyses of Treg-cell numbers in Kras^G12D and Kras^G12D;Ikkα^ΔLU lung ADCs. Data represent mean ± SEM (three experiments). **P < 0.01; Student’s t test. (E) IHC staining for % CD8 T cells in Kras^G12D and Kras^G12D;Ikkα^ΔLU lung ADCs, analyzed by n = 3 mice/group; three sections per mouse. Data represent mean ± SD (three repeats). ***P < 0.001; Student’s t test. (F) Lung appearances with ADC (Left) and ADC burden (Right) derived from Kras-CL and Kras^IKKαL ADC cells in WT mice (n = 3/group). This is representative. Data represent mean ± SEM (three experiments). ***P < 0.001; Student’s t test. (G) IHC analysis of % F4/80 (Left), % Foxp3 (middle, n = 3), and % CD8 (Right, n = 4) cells in Kras-CL and Kras^IKKαL ADCs. Data represent mean ± SD (three repeats). *P < 0.05; ***P < 0.001; Student’s t test. (H) Representative images of flow cytometric analyses for % macrophages (F4/80⁺CD11b⁺) in CD45⁺ cells isolated from Kras^IKKαL ADCs in WT mice. (I) ROS (dichlorofluorescein [DCF]) levels in macrophages isolated from Kras-CL and Kras^IKKαL lung ADCs in WT mice (n = 3/group). Data represent mean ± SEM (three repeats). *P < 0.05; Student’s t test. (J) A heat map analyzing the gene expression profiles in macrophages isolated from Kras-CL or Kras^IKKαL lung ADC in WT mice. Statistical analysis for all gene comparison from two groups, P < 0.05; one-way ANOVA test. (K) A summary: increased IKKα-deficient Kras^G12D ADC development is correlated with increased macrophage and Treg-cell numbers. Mϕ, macrophage.

To verify whether IKKα-deficient ADC cells cause similar TME alterations when growing in WT mice, we intratracheally injected mouse Kras-CL and Kras^IKKαL cell lines, derived from Kras^G12D-initiated ADC without or with IKKα deficiency, respectively (4) into lungs of WT mice. Kras^IKKαL ADC cells express low amounts of IKKα and have increased tumorigenic activity compared to Kras-CL cells (4). Kras^IKKαL tumor burden and Kras^IKKαL ADC-associated F4/80⁺ and Treg numbers were higher, but CD8⁺ cells were lower than in Kras-CL ADC (Fig. 2 F and G and SI Appendix, Fig. S2D). Most intratumoral F4/80⁺ macrophages also expressed CD11b, a monocyte marker (Fig. 2H and SI Appendix, Fig. S2E).

Monocyte-derived macrophages isolated from Kras^IKKαL ADC showed elevated reactive oxygen species (ROS) levels and expressed reduced antioxidant genes but more ROS-, inflammation-, and mitogenesis-related genes compared to macrophages isolated from Kras-CL ADC (Fig. 2 I and J and SI Appendix, Fig. S2 F and G). These results show that ADC-intrinsic IKKα deficiency alters the TME to become more immunosuppressive (Fig. 2K).

Bone Marrow–Derived Macrophages and Treg Cells Support Growth of IKKα-Deficient ADC.

To determine the effect of macrophage-derived ROS on IKKα-deficient ADC development, we transferred WT or Nox2^−/− (Nox2knockout [KO]) bone marrow (BM) into irradiated Kras^G12D and Kras^G12D;Ikkα^ΔLU mice. Lung weights and ADC burden were decreased in chimeric Kras^G12D;Ikkα^ΔLU mice with Nox2KO BM compared to Kras^G12D;Ikkα^ΔLU mice with WT BM (Fig. 3 A and B). Interestingly, lung ADC with Nox2KO BM showed decreased macrophage and Treg numbers and elevated CD8⁺ T cell numbers compared to lung ADC with WT BM, in addition to the expected decrease in ROS (Fig. 3 C and D and SI Appendix, Fig. S3 A and B). By contrast, lung ADC burden and infiltrating BM-derived macrophages, Treg cells, and CD8⁺ T cell numbers in chimeric Kras^G12D mice were not affected by the Nox2 status of the transplanted BM (Fig. 3 E and F and SI Appendix, Fig. S3C). Lung tumor burden and tumoral F4/80⁺ and Treg cell numbers were significantly reduced in Nox2^−/− mice receiving Kras^IKKαL cell injections compared to WT hosts transplanted with the same cells (SI Appendix, Fig. S3D). Thus, ROS produced by BM-derived macrophages may stimulate Treg differentiation to support the development and progression of IKKα-deficient lung ADC.

Fig. 3.

Macrophages paired with Foxp3 Treg cells specifically promote IKKα-deficient lung ADC development. (A and B) Lung ADC burden (% area, A) and lung weight (B) in Kras^G12D;Ikkα^ΔLU mice receiving WT or Nox2KO BM (n = 6 mice/group). g, grams. Data represent mean ± SD. *P < 0.05; ***P < 0.001; Student’s t test. (C and D) IHC analyses of % Foxp3 Treg cells (C, Left, n = 3), % F4/80 macrophages (C, Right, n = 4), and % CD8 cells (D, n = 4) in lung ADCs of Kras^G12D;Ikkα^ΔLU mice receiving WT BM or Nox2KO BM. Data represent mean ± SEM (three repeats). **P < 0.01; ***P < 0.001; Student’s t test. (E) Lung ADC burden (% area) in Kras^G12D mice receiving WT BM (n = 5 mice) or Nox2KO BM (n = 6 mice). ns, not significant; Student’s t test. (F) IHC analyses of % F4/80 macrophages (Left) and % Foxp3 Treg cells (Right) in lung ADCs of chimeric Kras^G12D mice receiving WT BM or Nox2KO BM (n = 3 mice/group; three sections per mouse). Data represent mean ± SD. ns, not significant; Student’s t test. (G) Kras^IKKαL lung ADC burden (% area, Left) in WT mice treated with clodronate-loaded liposomes (Lipos) or a vehicle control (n = 4 for each group). Flow cytometric analyses of % F4/80⁺CD11b⁺ cells (Middle) and % CD8⁺ cells (Right) in CD45⁺ cells associated with Kras^IKKαL lung ADCs derived from WT mice treated with clodronate-loaded Lipos (n = 5/group). Data represent mean ± SEM *P < 0.05; **P < 0.01; ****P < 0.0001; Student’s t test. (H) Flow cytometric analyses of % Treg cells (Left) in CD4⁺ T cells and c-Rel levels (MFI, median fluorescence intensity, Right) in CD4⁺Foxp3⁺ Treg cells associated with Kras^IKKαL lung ADCs of WT mice treated with clodronate-loaded Lipos or a vehicle control (n = 5/group). Data represent mean ± SD. *P < 0.05; ****P < 0.0001; Student’s t test. (I) Kras^IKKαL lung ADC burden in WT and Cd4^−/− (Cd4KO) mice (n = 5/group). *P < 0.05; Student’s t test. (J) Kras^IKKαL lung ADC burden in WT mice treated with an anti-Treg antibody (anti-CD25Ab) or a vehicle control (n = 5/group). Ab, antibody. Data represent mean ± SD. *P < 0.05; Student’s t test. (K) IHC analyses of % F4/80 macrophages (Left) and % Foxp3 Treg cells (Right) in lung ADCs derived from WT mice treated with an anti-CD25 Ab or a vehicle control (Cont) (n = 3/group; three sections per mouse). Data represent mean ± SEM (three repeats). ***P < 0.001; **P < 0.01; Student’s t test. (L) Flow cytometric analyses of % Treg (CD4⁺CD25⁺Foxp3⁺) cells (Left) in CD45⁺ cells isolated from lung ADCs of Foxp3-DTR-gfp mice (DEREG) treated with or without DT. Lung ADC burden of Foxp3-DTR-gfp mice treated with or without DT (Right). Data represent mean ± SEM (three repeats). ***P < 0.001; ****P < 0.0001; Student’s t test. (M) Lungs (Top) and hematoxylin and eosin slides (Bottom) representing lung ADCs with dark spots in Foxp3-DTR-gfp mice with DT or a vehicle control. (N) Western blotting analyses of nuclear c-Rel protein levels of CD4 T cells isolated from Kras-CL ADCs or Kras^IKKαL lung ADCs (SI Appendix, Fig. S3L). The relative activity was measured by comparing the nuclear c-Rel protein intensity with LaminB1 intensity, a nuclear protein loading control. Each sample was isolated from the ADCs obtained from multiple mice. Data represent mean ± SEM (three repeats). *P < 0.05; Student’s t test. (O) Heat maps analyzing gene expression profiles related to c-Rel targets (Right) and to inflammation and Foxp3 regulators (Left) in CD4 T cells isolated from Kras-CL or Kras^IKKαL ADCs in WT mice. Statistical analysis for all listed genes from two groups, P < 0.05; one-way ANOVA test.

Because the BM contains multiple haemopoietic cell types, we verified that the effects of NOX2-dependent ROS on lung ADC development were exerted by macrophages. We intratracheally injected Kras^IKKαL cells into WT mice and then depleted host macrophages with clodronate-loaded liposomes, which kill monocyte-derived macrophages (17). In this setting, we could not use Kras-CL cells because they generated very small ADCs with a low number of filtrating macrophages and Treg cells and were poorly responsive to macrophage-generated ROS. Clodronate-loaded liposome treatment significantly reduced lung ADC burden, ADC-associated F4/80⁺CD11b⁺ macrophages, and Treg cell numbers, as well as c-Rel amounts in CD4⁺Foxp3⁺ Treg cells, but increased CD8⁺ T cell numbers compared to vehicle control (Fig. 3 G and H and SI Appendix, Fig. S3E). To determine the role of CD4⁺ T cells, we used Rag1^−/− and Cd4^−/− mice as recipients, which developed heavier lungs with increased ADC burden compared to WT mice after intratracheal Kras^IKKαL cell injection (Fig. 3I and SI Appendix, Fig. S3 F and G). These results suggest that CD4⁺ T cells have antitumor activity. Intratumoral F4/80⁺ cell numbers were elevated in Rag1^−/− and Cd4^−/− hosts compared to WT hosts (SI Appendix, Fig. S3 H and I), suggesting that CD4⁺ T cells may inhibit macrophage infiltration.

To examine the effect of Treg cells on tumorigenesis, we intratracheally injected Kras^IKKαL cells into WT mice and treated the mice with an anti-CD25 antibody to deplete Treg cells (18). Treg reduction attenuated ADC development and reduced intratumoral monocyte-derived macrophages (Fig. 3 J and K and SI Appendix, Fig. S3J). We also depleted Foxp3⁺ cells using diphtheria toxin (DT) in Foxp3-DTR-GFP (DEREG) mice (19) that were intratracheally injected with Kras^IKKαL cells, resulting in dampened lung ADC development (Fig. 3 L and M). These data indicate that Treg cells support lung tumorigenesis in this setting.

TGFβ, a strong Treg-cell inducer, was not highly expressed in Kras^G12D;Ikkα^ΔLU ADC. Because c-Rel promotes Treg cell generation (20) and macrophage depletion reduced its expression in CD4⁺Foxp3⁺ cells (Fig. 3H), we analyzed nuclear c-Rel and expression of its target genes in CD4⁺ T cells isolated from Kras^IKKαL and Kras-CL lung ADC. Kras^IKKαL ADC–derived CD4⁺ T cells showed elevated nuclear c-Rel (Fig. 3N and SI Appendix, Fig. S3 K and L), expressed higher amounts of c-Rel target mRNAs (20, 21), Foxp3 up-regulating molecules, TNF-family–related pathways, IL-1 signaling, and fewer apoptosis-related genes compared to CD4⁺ cells from Kras-CL ADC (Fig. 3O). These results suggest that intratumoral CD4⁺ T cells of IKKα-deficient ADC have higher c-Rel–driven NF-κB activity.

CD4⁺ T Cell c-Rel Signaling and Macrophages Promote Generation of Intratumoral Treg Cells.

To establish a link between intratumoral monocyte-derived macrophages and Treg generation, we cocultured ADC-associated macrophages with CD4⁺CD25⁻ T cells (22) and found that macrophages from Kras^IKKαL ADCs induced higher CD4⁺Foxp3⁺ Treg cell numbers than macrophages from Kras-CL ADCs (Fig. 4A). The macrophages enhanced Treg cell induction in a dose-dependent manner and the induced Foxp3⁺ cells expressed less CD4 (Fig. 4B), which was probably due to elevated CD25 expression (data not shown). Since Nox2KO macrophages decreased intratumoral Treg cell numbers, we examined the role of macrophage-produced ROS in Treg induction. We cocultured CD4⁺CD25⁻ T cells with Nox2KO macrophages, which produced fewer ROS than WT macrophages and were less potent in Treg induction (Fig. 4 C, Left and Middle). To confirm the role of ROS in Treg cell induction, we added N-acetylcysteine (NAC) or apocynin, a ROS inhibitor, to the coculture experiments and found them to attenuate Treg-cell induction by WT macrophages (Fig. 4 C, Right). NAC treatment also reduced the burden of Kras^IKKαL-generated ADC and decreased the number of ADC-associated Treg cells, macrophages, macrophage ROS amounts, and c-Rel expression in intratumoral CD4⁺ T cells (Fig. 4D and SI Appendix, Fig. S4A). Reduced ADC IKKα expression is associated with increased TNF expression and Treg-cell scores (Fig. 1G), and TNF activates NF-κB (23). We therefore tested whether TNF enhances Treg induction in the coculture system by binding to TNFR2 and activating NF-κB in CD4⁺ T cells. Addition of either TNF or H₂O₂ to the macrophage-CD4⁺CD25⁻ T cell coculture system enhanced Treg-cell induction and increased c-Rel expression in both parental CD4⁺CD25⁺ T cells and CD4⁺CD25⁺Foxp3⁺ Treg cells (Fig. 4E).

Fig. 4.

A TNF/TNFRSF1B/c-Rel signaling cascade promotes Foxp3 Treg-cell induction. (A) Flow cytometric analyses of % Treg cells in CD4 T cells from the coculture of WT CD4⁺CD25⁻ T cells with macrophages (macro) that were isolated from Kras-CL or Kras^IKKαL lung ADCs (n = 3/group). Data represent mean ± SEM (three repeats). *P < 0.05; Student’s t test. (B) Flow cytometric analyses of % Treg cells in CD4 T cells from the coculture of WT CD4⁺CD25⁻ T cells and peritoneal macrophages (PM) in a dose-dependent manner (ratio, 2:1 and 1:1) (Right). n = 4. Data represent mean ± SEM (three repeats). ***P < 0.001; ****P < 0.0001; Student’s t test (Left). (C) ROS (dichlorofluorescein [DCF]) levels in WT and Nox2KO PM (Left). Flow cytometric analyses of % Treg cells in CD4 T from the coculture of WT or Nox2KO PM with WT CD4 T cells (n = 3/group, Middle) and from the coculture of WT PM and WT CD4 T cells treated with NAC (n = 4), apocynin (n = 4), and vehicle control (n = 3) (Right). Data represent mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; Student’s t test. (D) Lung ADC burden (% area, Left) in WT mice receiving Kras^IKKαL cells, treated with NAC (n = 5) and vehicle control (n = 4). ROS (DCF) levels in ADC-associated macrophage (Middle), and cytosolic and nuclear c-Rel levels in ADC-associated CD4 T cells detected by Western blotting, from NAC- and vehicle (Cont)-treated WT mice receiving Kras^IKKαL cells (Right). LaminB1, a nuclear protein loading control; α-tubulin, a cytosolic protein loading control. Data represent mean ± SEM (three repeats); **P < 0.01; ***P < 0.001; Student’s t test. (E) Flow cytometric analyses of % Treg cells in CD4 T cells from the coculture of WT PM and WT CD4⁺CD25⁻ T cells supplemented with TNF, H₂O₂, or a vehicle control (n = 3/group, Left), and c-Rel intensity (median fluorescence intensity, MFI) in these CD4 cells and CD25⁺Foxp3⁺ cells (Right). ns, not significant; **P < 0.01; ***P < 0.001. Data represent mean ± SEM (three repeats); Student’s t test. (F) Flow cytometric analyses of % Treg-cell induction in CD4 T cells from the coculture of WT PM and WT CD4 T cells treated with PTXF (50 μg/mL) and control. *P < 0.05. Data represent mean ± SEM (three repeats); Student’s t test. (G) Lung ADC burden (% area, Left) in WT mice receiving Kras^IKKαL cells, treated with PTXF (50 mg/kg) or a vehicle control (n = 4/g/roup). IHC analyses of % Treg cells (Middle) in ADCs and ROS (DCF) levels in ADC-associated macrophages (Right) from WT mice receiving Kras^IKKαL cells, treated with PTXF or a vehicle control (n = 4/group). Data represent mean ± SEM (three repeats); *P < 0.05; **P < 0.01; ***P < 0.001; Student’s t test. (H) Western blotting analyses of cytosolic and nuclear c-Rel levels in CD4 T cells isolated from Kras^IKKαL lung ADCs in WT mice treated with PTXF or a vehicle control (Cont). Each example was obtained from multiple mice. LaminB1, a nuclear protein loading control; α-tubulin, a cytosolic protein loading control. (I) Kras^IKKαL ADC burden (% area, Left) in WT and Tnf^−/− (KO) mice (n = 6/group). Data represent mean ± SD. ****P < 0.0001; Student’s t test. Images of WT and Tnf KO lungs are representative (Right). Dark pink stained spots in hematoxylin and eosin slides represent tumors in the lungs. (J) IHC analyses of % Foxp3 Treg-cell (Left) and % F4/80 macrophage (Right) numbers in Kras^IKKαL ADCs from WT and Tnf KO mice (n = 4/group; three sections per mouse). Data represent mean ± SD. ****P < 0.0001; Student’s t test. (K) TNFR2/TNFRSF1B expression in CD4 T cells isolated from lung ADCs of WT and Tnf KO mice, analyzed by RT-PCR. Data represent mean ± SEM (three repeats). ***P < 0.001; Student’s t test. (L–M) Flow cytometric analyses of % Treg cells in CD4 T cells (L) and c-Rel levels (MFI, M) in Treg cells from the coculture of WT PM with WT or Tnfrsf1b^−/− (KO) CD4⁺CD25⁻ T cells, treated with or without TNF. Data represent mean ± SEM (three repeats). *P < 0.05; **P < 0.01; ***P < 0.001; Student’s test. (N) ROS (DCF) levels in WT and Tnf KO peritoneal macrophages, analyzed by RT-PCR. Data represent mean ± SEM (three repeats). ****P < 0.0001; Student’s t test. (O) TNF expression in WT peritoneal macrophages treated with H₂O₂ or vehicle control. Data represent mean ± SEM (three repeats). *P < 0.05; Student’s t test. (P) Flow cytometric analyses of % Treg cells in CD4 T cells from the coculture of WT PM with WT or Tnfrsf1b^−/− CD4 T cells as well as from the coculture of Tnf KO PM with WT or Tnfrsf1b^−/− CD4 T cells, analyzed by flow cytometry. Data represent mean ± SEM (three repeats). ns, not significant; *P < 0.05; ***P < 0.001; Student’s test. (Q) Lung weight (Left) and Kras^IKKαL ADC burden (Right) in irradiated CD45.1 mice receiving CD45.2 WT or Tnf^−/− (KO) BM (n = 4). Data represent mean ± SD. *P < 0.05; Student’s test. (R) Flow cytometric analyses of ADC-associated CD4⁺Foxp3⁺ cells in CD45⁺ cells (Left) and c-Rel levels (MFI) in Foxp3 cells (Right) in CD45.1 mice receiving CD45.2 WT or Tnf KO BM cells (n = 3). Data represent mean ± SEM (three repeats). *P < 0.05; **P < 0.01; Student’s test. (S) Lung appearances of Tnfrsf1b^−/−;Ikkα^ΔLU;Kras^G12D and Ikkα^ΔLU;Kras^G12D mice. (T) Lung ADC burden (% area, Left) of Tnfrsf1b^−/−;Ikkα^ΔLU;Kras^G12D and Ikkα^ΔLU;Kras^G12D mice (n = 4/group). Flow cytometric analyses of Foxp3⁺ Treg-cell numbers (Middle, n = 3) and CD4 T cell c-Rel levels (Right, n = 3) in lung ADCs of Tnfrsf1b^−/−;Ikkα^ΔLU;Kras^G12D and Ikkα^ΔLU;Kras^G12D mice. Data represent mean ± SEM (three experiments). *P < 0.05; **P < 0.01; Student’s test. (U) Survival rates for lung ADC patients expressing CHUK hemizygous deletions and TNFRSF1B (TNFR2) gain versus patients with ADC expressing CHUK and TNFRSF1B double diploid. The RNA-sequence data were obtained from cBioPortal, TCGA, ref. 6. Logrank test for P value analyses.

We examined c-Rel’s involvement in Treg generation and tumorigenesis by using pentoxifylline (PTXF), which inhibits c-Rel activity, although it is not a c-Rel–specific inhibitor (20). PTXF treatment significantly reduced ex vivo Treg induction (Fig. 4F) and reduced Kras^IKKαL-generated lung ADC burden, as well as intratumoral Treg and macrophage numbers, macrophage ROS amounts, and CD4⁺ T cell c-Rel activity, while increasing CD8⁺ T cell numbers (Fig. 4 G and H and SI Appendix, Fig. S4 B–D). These results suggest that NF-κB/c-Rel enhances Treg-cell induction and contributes to tumorigenesis. To rule out an effect of PTXF on macrophages or ADC cells, we treated Rag1^−/− mice inoculated with Kras^IKKαL cells with PTXF. The inhibitor did not alter lung weights and ADC burden in Rag1^−/− mice (SI Appendix, Fig. S4 E and F), suggesting that PTXF acts on T cells.

To test TNF’s involvement in Treg differentiation and tumorigenesis, we intratracheally injected Kras^IKKαL cells into lungs of WT and Tnf^−/− mice. Both lung ADC burden and intratumoral Treg-cell and macrophage numbers were reduced in Tnf^−/− mice along with increased CD8⁺ T cell numbers compared to WT (Fig. 4 I and J and SI Appendix, Fig. S4G). Notably, CD4⁺ T cells isolated from Tnf^−/− ADC expressed reduced amounts of Tnfrsf1b, which encodes TNFR2 (Fig. 4K), suggesting that TNFR2 is the main CD4⁺ T cell TNF receptor mediating Treg-cell induction. Notably, TNFRSF1B mRNA is highly correlated with CD4 mRNA expression in human lung ADC (SI Appendix, Fig. S4H) (24, 25). To determine the role of TNFR2 in Treg-cell induction, we cocultured WT macrophages with WT or Tnfrsf1b^−/− CD4⁺CD25⁻ T cells. Tnfrsf1b^−/− CD4⁺ T cells gave rise to fewer Treg cells and lower c-Rel expression in Foxp3⁺ cells compared to WT CD4⁺ T cells (Fig. 4 L and M). In addition, Tnfrsf1b^−/− CD4⁺ T cells responded poorly to TNF, exhibiting reduced Treg-cell induction and lower c-Rel expression than WT CD4⁺ T cells (Fig. 4 L and M). Nonetheless, residual TNF-stimulated Treg induction in cocultures of WT macrophage and Tnfrsf1b^−/− CD4⁺ T cells suggested that TNF may also work via TNFR1 in the absence of TNFR2. Because expression of CD4 and TNFRSF1A, which encodes TNFR1, was not highly correlated in human lung ADCs (6), TNFR2 may be the more important TNFR in this setting.

Because TNF and H₂O₂ showed similar ability to enhance Treg induction in the coculture system, we hypothesized that TNF and ROS up-regulate each other’s production. Indeed, ROS levels were lower in Tnf^−/− than WT macrophages (Fig. 4N), and H₂O₂ enhanced TNF expression (Fig. 4O). Accordingly, Tnf^−/− macrophages cocultured with WT CD4⁺ T cells induced fewer Treg cells compared to WT macrophages, and the coculture of Tnf^−/− macrophages with Tnfrsf1b^−/− CD4⁺ T cells led to a further reduction in Treg induction (Fig. 4P). To determine the role of macrophage-produced TNF in lung ADC development, we injected CD45.2 WT or Tnf^−/− BM cells into irradiated CD45.1 WT mice that were inoculated with Kras^IKKαL lung ADC cells. The results showed decreased ADC burden, lung weights, and ADC-associated Treg cells in mice reconstituted with Tnf^−/− BM (Fig. 4 Q and R and SI Appendix, Fig. S4I). These experiments also suggested that ADC-associated monocyte-derived F4/80⁺CD11b^High and F4/80⁺CD11b^Low macrophages originate from the BM (SI Appendix, Fig. S4J), further supporting the role of macrophage ROS in TNF production at a level needed for Treg induction.

We compared ADC development in Tnfrsf1b^−/−; Ikkα^ΔLU;Kras^G12D and Ikkα^ΔLU;Kras^G12D mice and found that lung ADC burden was significantly lower in Tnfrsf1b^−/−; Ikkα^ΔLU;Kras^G12D mice than Ikkα^ΔLU;Kras^G12D mice (Fig. 4 S and T, Left). Treg numbers and c-Rel levels in Foxp3⁺ Treg cells were lower in lung ADCs of Tnfrsf1b^−/−;Ikkα^ΔLU;Kras^G12D mice compared to Ikkα^ΔLU;Kras^G12D mice, which were consistent with the result obtained from the coculture system (Fig. 4 T, Middle and Right and SI Appendix, Fig. S4K). Patients with lung ADC with CHUK hemizygous deletions and TNFRSF1B gain die earlier compared to patients doubly diploid for CHUK and TNFRSF1B (Fig. 4U). Patients with lung ADC with CHUK hemizygous deletions and FOXP3 gain showed a lower survival trend compared to patients with ADC who are double diploid for CHUK and FOXP3 (SI Appendix, Fig. S4L). These results suggest that TNFR2 activates c-Rel and stimulates Foxp3⁺ Treg cell differentiation and that immunosuppressive Foxp3⁺ Treg induction accelerates lung carcinogenesis. Accordingly, we postulated that low IKKα lung ADC generates a TNF-rich TME. Indeed, TNF expression was higher in Kras^IKKαL ADC than Kras-CL ADCs (SI Appendix, Fig. S4M).

IKKα Suppresses TNF, CSF1, CCL22, and IL-23A Expression in Lung ADC.

To investigate how IKKα loss up-regulates expression of cytokines and chemokines that shape the immunosuppressive TME (iTME), we used bead-based flow cytometry and detected higher TNF, CCL22, IL-23A, and CSF amounts in sera of ADC-bearing Ikkα^ΔLU;Kras^G12D mice than in Kras^G12D mice with ADC (SI Appendix, Fig. S5A). We confirmed that Kras^IKKαL ADC cells expressed significantly higher levels of Csf1, Tnf, Ccl22, Il23a, and Adam8 mRNAs compared to Kras-CL cells (Fig. 5A). Likewise, low CHUK mRNA was correlated with elevated IL23A and ADAM8 mRNAs in human lung ADC (Fig. 5B). IL-23A inhibits CD8⁺ T cell infiltration, enhances inflammation, and increases tumor incidence (26, 27), and ADAM8 promotes cell proliferation, invasion, and metastasis in human cancers (28, 29). Reduced IKKα also correlated with increased expression of CCL7, CCL8, and ADAM9 in mouse lung ADC cells (SI Appendix, Fig. S5B). Together, these results suggest that IKKα deficiency correlates with elevated expression of cytokines that are important for generation of Treg cells and iTME.

Fig. 5.

IKKα reduction up-regulates transcription of cytokine genes in human and mouse lung ADC cells. (A) Expression of CCL2, CSF1, TNF, ADAM8, and IL-23A in Kras-CL and Kras^IKKαL ADC cells, analyzed by RT-PCR. Data represent mean ± SEM (three repeats). ***P < 0.001; **P < 0.01; Student’s t test. (B) Correlation of CHUK (IKKα) expression levels (using quartiles’ analysis to divide IKKα expression groups) with the expression levels of IL-23A or ADAM8 genes in human lung ADCs by analyzing the TCGA data (PanCancer Atlas, cBioPortal). H-CHUK, high CHUK; L-CHUK, low CHUK; P value at the Top of panels, Student’s t test; n, ADC numbers. (C–D) Expression of CHUK (IKKα), IL-23A (IL-23a), CCL22, CSF1, and TNF in human A549 (C) and mouse Kras-CL (D) lung ADC cells treated with Si-control (Si-Cont) or Si-IKKα RNA, analyzed by RT-PCR. Data represent mean ± SEM (three repeats). *P < 0.05; **P < 0.01; ***P < 0.001; Student’s t test. (E–F) ChIP analyses for binding of IKKα to the promoter regions of TNF (Tnf), CSF1 (Csf1), CCL22 (Ccl22), and IL-23A (Il-23a) genes by using an anti-IKKα antibody for immunoprecipitation, followed by PCR with primers for these genes in A549 cells (E) and in Kras-CL cells (F) treated with Si-control (Si-Cont) or Si-IKKα RNA. Data represent mean ± SEM (three repeats). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; Student’s t test. (G) ChIP-seq analyses for IKKα enrichment on the TNF gene in A549 cells. The enrichment peaks are marked by red lines. bp, nucleotide base pair; red arrow, 6,066 bp; Ab, antibody for immunoprecipitation. TNF gene: thickest lines, exons; second-thickest lines, untranslated sequences; thin lines, introns; red line, consensus sequences containing eight nucleotides. (H) A working model showing that ADC-IKKα reduction up-regulates the expression of TNF, CSF1, CCL-22, and IL-23A and macrophage recruitment, which generate a TNF/TNFRSF1B/c-Rel pathway for CD4 T cells to stimulate Treg-cell induction, accelerating Kras^G12D ADC development. Arrows, promotion; lines, inhibition; Mϕ, monocyte-derived macrophage.

To determine how IKKα modulates cytokine expression, we silenced IKKα in either human or mouse lung ADC cells and found it to increase expression of IL23A, CCL22, CSF1, and TNF mRNAs (Fig. 5 C and D). In addition to being a cytoplasmic protein kinase, IKKα was reported to act in the nucleus (30–32). We therefore hypothesized that IKKα may directly or indirectly suppress expression of these cytokine genes. We conducted chromatin immunoprecipitation (ChIP) assays with an IKKα antibody and found enrichment for IKKα on the IL23A, CCL22, CSF1, and TNF promoter regions in both human A549 and mouse Kras-CL cells, and this was attenuated by IKKα silencing (Fig. 5 E and F and SI Appendix, Fig. S5C). Of note, the correlation between IKKα enrichment at these cytokine genes and their low expression in IKKα-deficient cells, suggests that IKKα act as a transcriptional suppressor of these particular genes. Because an interaction between IKKα and SMAD3/4 regulates expression of certain genes in keratinocytes (33–35), we tested whether IKKα suppresses the above cytokine gene through SMAD transcription factors. We performed a ChIP assay with SMAD3 or SMAD4 antibodies and could not detect SMAD3/4 enrichment on the IKKα-bound genes (SI Appendix, Fig. S5D). We further performed unbiased ChIP-sequencing (ChIP-seq) experiments with an IKKα antibody (Fig. 5G). Sequence analysis of the regions at which IKKα was enriched revealed an 8-bp-long consensus sequence on the TNF, CSF1, CCL22, and IL-23A genes (Fig. 5G and SI Appendix, Fig. S5E). Future studies will probe the regulatory importance of this consensus sequence and whether it is recognized by IKKα or another transcription factor or chromatin protein with which IKKα interacts.

Based on the above findings, we propose a working model that explains how IKKα deficiency in lung ADC generates iTME that dismantles immune surveillance and accelerates tumor progression. Key to this model is the up-regulation of TNF, CSF1, CCL22, and IL-23A in IKKα-deficient lung ADC, which enhances macrophage recruitment and Treg differentiation (Fig. 5H).

Discussion

Human lung ADC and SCC, which originate from different cell types, show distinct histological features and genomic alterations (cBioPortal, The Cancer Genome Atlas [TCGA], refs. 6 and 36, PanCancer Atlas). KRAS mutations are common in human lung ADC but rare in lung SCC. Previous studies conducted by other investigators had focused on the role of canonical, IKKβ-dependent, NF-κB signaling in mouse Kras-initiated lung ADC (37, 38). We, on the other hand, had focused on the role of IKKα, which unlike IKKβ, activates noncanonical NF-κB signaling (39) and has several other, NF-κB unrelated, functions (31, 32, 40, 41). Previously, we reported that Ikkα ablation in mice promotes Kras^G12D-initiated lung ADC development and results in spontaneous SCC formation associated with increased macrophage infiltration (4, 17, 42). In the present study we followed on the significant correlation between monocyte-derived macrophages and Treg scores and reduced IKKα expression in human lung ADC cohorts. Low IKKα expression also correlated with elevated expression of cytokines and chemokines (CSF1R, CSF1, TNFRSF1B, TNF, CCL2, FOXP3, and CCL22) that regulate macrophage recruitment and Treg-cell development (24, 25, 43–45). Similar correlations were observed in the mouse lung ADC model we have investigated.

Although macrophage-produced ROS support Treg-cell differentiation in vitro (22), the mechanism by which macrophage ROS production stimulates tumorigenesis is unknown. Our results suggest that ROS are required for maintaining high macrophage TNF expression, which is needed for stimulation of Treg-cell differentiation via a TNF/TNFR2/c-Rel signaling cascade. Ablation of TNF or the ROS-producing enzyme Nox2 inhibited lung ADC pathogenesis and reduced tumoral monocyte-derived macrophage and Treg infiltration. Moreover, in human lung ADC cohorts (TCGA, ref. 6 and PanCancer Atlas, cBioPortal), TNFR2 expression is significantly correlated with CD4 expression. Also, Tnfrsf1b^−/− CD4⁺ T cells gave rise to fewer Treg cells than WT CD4⁺ T cells when cocultured with macrophages. Correspondingly, lung ADC burden and intratumoral Treg-cell numbers were reduced in Kras^G12D;Ikkα^ΔLU;Tnfrsf1b^−/− mice compared to Kras^G12D; Ikkα^ΔLU mice. However, TNF still led to residual Treg induction in cocultures of Tnfrsf1b^−/− CD4⁺ T cells and macrophages, suggesting that in the absence of TNFR2, TNF may stimulate Treg differentiation via TNFR1, even though TNFR1 expression is not significantly correlated with CD4 expression in human lung ADC. Even in the complete absence of TNF, we still observed a basal level of Treg induction in the coculture system, suggesting that other macrophage-generated stimuli can drive Treg differentiation. Regardless of the underlying mechanism, our results demonstrate that Treg cells are key components of the iTME that accompanies IKKα-deficient lung ADC.

Together with IKKβ, IKKα is one of the two catalytic subunits of the IKK complex. Curiously, however, in human lung ADC, the expression patterns of IKKα and ΙΚΚβ associated with expression of macrophage- and Treg cell-regulating cytokines and chemokines are reciprocal. Ours and other animal studies are consistent with the findings in human lung ADC (4, 7, 38, 46, 47). However, a recent report showed that Ikkα, but not Ikkβ, deletion in lung epithelial cells attenuated Kras-initiated lung ADC development (48). The cause of this discrepancy is not clear, but it may be due to unknown microenvironmental conditions. Consistent with the results described herein, transgenic Tg-K5.IKKα and Tg-Lori.IKKα mice develop normally and are resistant to carcinogen-induced tumorigenesis and metastasis (15, 49, 50), whereas transgenic Tg-K5.IKKβ and Tg-EDL-2.IKKβ mice develop epidermal and esophageal hyperplasia and oral carcinomas (51–53). Our results are also consistent with previous studies showing that while IKKβ is the critical IKK catalytic subunit responsible for NF-κB activation, IKKα has numerous other functions. IKKα ablation in human and mouse lung ADC cells results in up-regulation of the CSF1, CCL22, TNF, and IL-23A genes, to whose promoter regions IKKα is recruited, suggesting that in lung ADC, IKKα works in the nucleus as a transcriptional repressor. However, ChIP assays cannot determine whether IKKα directly recognizes specific DNA sequences or interacts with transcription factors that bind to these sequences, a question that needs to be addressed in future studies. In past studies IKKα was reported to interact with SMAD transcription factors to determine expression of Myc antagonists (33–35) and was shown to modulate methylation of histone proteins (17, 31, 54). Whether these mechanisms apply to lung ADC cells remains to be determined.

Overall, Ikkα ablation promotes Kras^G12D-initiated lung ADC development in mice. In previous studies we showed that increased tumor-intrinsic ROS are associated with Kras^G12DIkkα^ΔLU ADC progression (4). We now extend these results to show that Ikkα-deficient lung ADC generates an iTME by up-regulating Treg-cell induction and that ROS produced by intratumoral macrophage facilitate Treg differentiation through the TNF/TNFR2/c-Rel pathway. Given the present findings, tumor cell–intrinsic ROS may also contribute to iTME generation through potentiation of TNF signaling.

Materials and Methods

Mice, Human Tissue Array, and Cell Lines.

All mice used in this study were cared for in accordance with the guidelines of the Institutional Animal Care and Use Committee (IACUC) of the NIH. All animal experiments (Protocols 17-051 and 17-052) were approved by the IACUC. The following mice were of a C57BL/6 background: Ikkα^f/f, Nox2^−/− (stock No. 002365; The Jackson Laboratory), Rag1^−/−, Kras^LSL-G12D (Kras^G12D, stock No. 008179; The Jackson Laboratory), Cd4^−/− (stock No. 002663; The Jackson Laboratory), Tnf^−/− (stock No. 003008; The Jackson Laboratory), Tnfr2^−/− (stock No. 002620; The Jackson Laboratory), CD45.1 B6 (stock No. 002014; The Jackson Laboratory), and B6.Cg-Foxp3 sf/J (Foxp3-DTR-gfp/DEREG, stock No. 32050; The Jackson Laboratory). Human lung ADC tissue arrays (BCS04017a and HLug03PG02) were purchased from US Biomax, Inc. We used an A549 human lung ADC cell line carrying a KRAS^G12S mutation, a Kras-CL mouse lung ADC cell line carrying a Kras^G12D mutation, and a Kras^IKKαL mouse lung ADC cell line carrying a Kras^G12D mutation, and multiple Ikkα mutations were used (4).

Antibodies.

For Western blotting and immunostaining, we used antibodies to: Lamin B (sc-6216), c-Rel (sc-6955), NOX2 (sc-5827), IKKα (sc-7183 and sc-7606), SMAD4 (sc-7154), TNFα (sc-1348), and normal mouse IgG (sc-2025) from Santa Cruz Biotechnology, Inc.; CD68 (M087629-2) from Dako Products; β-actin (A5441) from Sigma-Aldrich; α-tubulin (ab4074) and Foxp3 (ab10901) from Abcam; c-Rel (rabbit, No. 12707) and IKKβ (No. 2678) from Cell Signaling Technology; IKKα (No. NB100-56704) from Novus Biologicals; IKKγ (No. 559675) from BD Biosciences; and SMAD3 (ab28379) from Abcam.

Data Availability

Tumor-associated macrophage data have been deposited in National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) from mouse lung adenocarcinoma-associated macrophages (GSE114501). Anonymized tumor-associated CD4 T cells and ChIP-seq data have been deposited in GEO from mouse lung adenocarcinoma-associated CD4 cells (GSE122419 and GSE132460). All other study data are included in the article and/or supporting information.