K-ras mutant mouse model of lung cancer used in this study, CCSP^Cre/LSL-K-ras^G12D (CC-LR), was created by crossing mice harboring LSL-K-ras^G12D allele with mice containing Cre recombinase inserted into the Club Cell Secretory Protein (CCSP) locus., This mouse model has K-ras mutant tumor initiation exclusively in lung epithelium as we previously described (8). Then CC-LR mice were crossed with TLR2, TLR4, and TLR9 knockout (KO) mice separately to generate CC-LR/TLR2KO, CC-LR/TLR4KO, and CC-LR/TLR9KO mice respectively. TLR2 (22) and TLR4 KO (23) mice were purchased from Jackson Laboratory, and TLR9 KO was a generous gift from Dr. Shizuo Akira (Osaka University, Japan) (24). These mutant mice globally lack their respective TLR while expressing a mutant K-ras in the lung epithelium. LR/MyD88^Δ/Δ and LR/IKKβ^Δ/Δ mice, with lung epithelial specific loss of MyD88 and NF-κB activity, were generated by crossing CC-LR with MyD88^fl/fl and IKKβ ^fl/fl conditional KO mice respectively. IKKβ ^fl/fl mouse was kindly provided by Dr. Michael Karin (University of California, San Diego, CA) (25). MyD88^fl/fl mouse was purchased from the Jackson Laboratory (26). All the mice used in this study were housed under specific pathogen-free conditions and handled in accordance with the institutional animal care and use committee (IACUC) of the University of Texas MD Anderson Cancer Center. Mice were also monitored daily for evidence of disease or death.

A lysate of NTHi strain 12 (27) was prepared as previously described (28). This was prepared at a protein concentration of 2.5 mg/ml in phosphate-buffered saline (PBS) and stored in 7ml frozen aliquots at −80°C. When delivering the lysate to mice, a thawed aliquot was placed in an AeroMist CA-209 nebulizer (CIS-US, Bedford, MA) driven by 10 liters/minute of room air mixed with 5% CO₂ for 20 minutes. Mice in all studies were exposed to the lysate once weekly at six weeks of age for a duration of eight weeks, with a final readout at fourteen weeks of age.

In order to assess lung tumor burden and inflammation, fourteen-week-old mice were anesthetized. Their tracheas were then exposed, cannulated, and sutured into place. If visible, lung surface tumor numbers were counted, then the lungs were perfused with PBS through the right ventricle. For half of the mice in this study, lungs were inflated with 10% buffered formalin (Sigma) for 10 min, then collected and embedded in paraffin blocks for histopathological studies. Formalin fixed paraffin embedded (FFPE) blocks were sectioned at 5-mm thickness, placed on glass slides, dried, deparaffinized, and stained with hematoxylin and eosin. Tumor/lung area percentages were calculated using the respective formula; Tumor/Lung Area =∑i=ni=1TiL × 100%, as described previously (29). For the other half of the mice in this study, bronchoalveolar lavage fluid (BALF) was collected by sequentially instilling and extracting 2ml of PBS through a tracheostomy cannula. Afterward, the lungs were removed and snap-frozen to be stored for RNA and protein analysis. Total WBC counts in BALFs were calculated with the use of a hematocytometer, and differential cell populations (macrophages, neutrophils, and lymphocytes) were determined by cytocentrifugation of BALFs onto slides and Wright–Giemsa (Sigma) staining.

Previously sectioned lung tissue samples were immunohistochemically stained as previously described (29) for proliferation marker Ki-67 (1:200; ab16667; Abcam, MA), angiogenesis marker ETS-related gene (ERG) (1:200; ab92513; Abcam) (30), and leukocyte cell marker, CD45 (1:200; ab10558; Abcam). To calculate the percentages of positively stained cells, slides were analyzed using Nuclear v9 in ImageScope 12.4.3 (Leica Biosystems, Nussloch, Germany) and presented as the fraction of positive cells per total tumor cells per 10x field.

Total RNA from mouse whole lung was extracted using a Zymo Research RNA extraction kit. Reverse transcription PCR was performed using the qScript cDNA SuperMix (Quanta Biosciences, Gaithersburg, MD). We then performed qPCR by using SYBR Green FastMix (Quanta Bioscience) on CFX96 Touch™ Real-Time PCR Detection System (Bio-Rad, Hercules, CA). CD45 was used as our housekeeping gene and data are presented as fold changes of experimental groups versus controls (CC-LR) as indicated in figure legends. Primers used in this study are listed in Supplementary Table S9 .

Data are presented as mean ± SEM. GraphPad Prism software was used for statistical analysis, and Student’s t-test was used for comparison between every experimental group versus controls (CC-LR). Differences were considered significant for *p < 0.05 and labeled with an asterisk (*), figures with no significant differences were labeled with "ns". Some figures have additional symbols of “#” and “+” to denote different cellular populations as noted in the figure legend.

The exact role of TLRs in lung cancer promotion is not well established, therefore we sought to study the role of TLRs by crossing CC-LR mice with TLR2, TLR4, and TLR9 null mice separately to generate K-ras mutant mice lacking TLR2, TLR4, and TLR9, respectively (CC-LR/TLR2KO, CC-LR/TLR4KO, CC-LR/TLR9KO). These cohorts were then subjected to COPD-inducing conditions from the age of 6 weeks for 8 weeks as described above and previously (8) and their tumor burdens were compared with age and sex matched control CC-LR mice at the age of 14 weeks. The lack of TLRs in these three distinct mutant mouse models significantly inhibited lung cancer promotion by COPD-like airway inflammation in CC-LR mice. Lung surface tumor numbers decreased by 1.5-fold (72 vs. 112) in CC-LR/TLR2KO, 1.3-fold (84 vs. 112) in CC-LR/TLR4KO, and 1.3-fold (83 vs. 112) in CC-LR/TLR9KO mice ( Figure 1A ). Tumor area analysis of H&E-stained slides also showed a 1.5-fold (27% vs. 41%) reduction in CC-LR/TLR2KO and CC-LR/TLR4KO, while CC-LR/TLR9KO showed a 1.4-fold reduction (29% vs. 41%) in the lung area occupied by the tumors ( Figures 1B, C ). We also studied the role of TLRs in lung cancer promotion in mice lacking COPD-inducing conditions (Naïve). The naïve cohorts of CC-LR/TLR2, 4, and 9 mice also showed 1.7-fold (35 vs 60), 1.5-fold (40 vs. 60), and 2-fold (30 vs. 60) reductions in tumor numbers respectively compared to age and sex match naïve CC-LR mice ( Supplementary Figure S1 ). In these naïve cohorts, we also observed a significant 1.5-fold reduction (17% vs. 28%) across all three mouse models in the lung area occupied by the tumors ( Supplementary Figure S1 ). These findings collectively confirm the protumorigenic effects of these TLRs in K-ras mutant lung cancer development and promotion.

We aimed to further study the mechanism of tumor reduction in mice with a lack of TLRs and COPD-like airway inflammation. Lung tissues were examined via immunohistochemistry for proliferation, and angiogenesis markers, Ki-67, and ERG respectively. Ki-67 staining of the lung from CC-LR/TLR4KO and CC-LR/TLR9KO mutant mouse models showed a reduction in tumor cell proliferation by ~15%, while CC-LR/TLR2KO had a greater reduction of 50% in Ki-67 positive cells ( Figures 2A, B ). We also found that TLR knockout mice had reduced angiogenesis. CC-LR/TLR4KO and CC-LR/TLR9KO mutants showed a 30% reduction and CC-LR/TLR2KO a 75% reduction in ERG positive cells ( Figures 2C, D ). Our naïve cohort of mice displayed a low baseline of Ki-67/ERG positivity and showed significant reductions for both markers. TLR2, 4, and 9 displayed an 80% reduction in Ki-67 positivity, while TLR2 and 4 had a 65% reduction and TLR9 had a 35% reduction in ERG positivity ( Supplementary Figure S2 ). Indeed, these data hint at the promoting effects that TLRs may have on tumor cell proliferation and angiogenesis.

Here, we sought to elucidate the intricate mechanisms of tumor suppression due to TLR knockouts in the setting of COPD-like airway inflammation, specifically regarding changes in the tumor microenvironment (TME) and tumor-infiltrating immune subsets. We accomplished this by analyzing the BALFs of TLR2, 4, and 9 KO/K-ras mutant mouse models in the presence of NTHi-induced COPD-like airway inflammation. We found that all three of these TLR KO mice had a significant reduction in lung neutrophilic infiltrates ( Figure 3A ). There were no significant changes in lymphocytes, although CC-LR/TLR4KO showed a significant decrease in macrophages, and CC-LR/TLR9KO showed a significant increase ( Figure 3A ). Additionally, mRNA expression analysis of whole lungs from these KO mice revealed significant reductions of neutrophil chemoattractant CXCL1, with a 2-fold decrease in TLR2 KO and a 2.5-fold decrease in TLR4 or 9 KO mice ( Figure 3B ). No significant changes in macrophage chemoattractant CCL2 were found ( Figure 3C ). These results were accompanied by significantly increased levels of IFNG, specifically a 15-fold increase in TLR2 KO and a 50-fold increase in TLR4 or 9 KO mice ( Figure 3D ). TNF also displayed significant increases of 7-fold in TLR2 or 4 KO and 20-fold in TLR9 KO mice ( Figure 3E ), and Gzmb showed a 2.5-fold increase in TLR2, a 2-fold increase in TLR4 and a 3-fold increase in TLR9KO mice ( Figure 3F ). IL1β revealed a 1.5-fold reduction in TLR2/4, and a 1.4-fold reduction in TLR9 KO mice ( Figure 3G ), while IL17 showed a homogenous 2.5-fold reduction across all TLR KO mice ( Figure 3H ). Lastly, protumorigenic and immunosuppressive macrophage polarization marker RETNLB (Fizz1) showed a significant 2-fold reduction in TLR2 KO, a 5-fold reduction in TLR4 KO, and a 10-fold reduction in TLR9 KO mice ( Figure 3I ). To further explore the unique immune cell changes across the different TLR KO mice in the context of NTHi-mediated inflammation, we studied the mRNA expression levels of different immune cell markers including total leukocytes (CD45), T-cells (CD3), neutrophils/granulocytes (Ly6G), cytotoxic T-cells (CD8) and monocytes (CD68). We also performed immunohistochemistry for CD45, which represents all immune cell types in our study. Results from mRNA expression of CD45, CD3, CD8, and CD68 showed no significant changes across TLR mutant mice, the only significant change was Ly6G which displayed a significant decrease in TLR2KO ( Supplementary Figure S3A-E ). Additionally, our immunohistochemistry studies showed a significant reduction in CD45 staining across all TLR KO mouse models, further supporting a decrease in pro-tumor inflammatory TME of these mice ( Supplementary Figure S3F, G ).

In comparison, our naïve cohort of mice displayed a distinct phenotype. There were no significant changes to lymphocytes or neutrophils, although all three mouse models showed a significant increase in macrophages. CXCL1 presented with a 1.6-fold reduction across all three mouse groups. No significant changes were seen for CCL2 or Gzmb. TNF showed a significant 4-fold increase for TLR2, and 4, and a 5-fold increase for TLR9. IFNG had a 3-fold increase for TLR2, a 9-fold increase for TLR4, and a 5-fold increase for TLR9. This was accompanied by significant reductions in IL1β, 2.5-fold reduction for TLR2, 1.6-fold reduction for TLR4, and 1.2-fold reduction for TLR9. IL17 results showed that all three groups shared a significant 1.6-fold reduction. Lastly, RETNLB displayed that TLR2 had a 1.6-fold reduction and TLR4 and 9 had a 2-fold reduction ( Supplementary Figure S4 ). In summary, these findings further support a shift to an antitumorigenic TME phenotype in response to the lack of TLRs.

To further dissect the cell type specific mechanism of tumor promotion by TLR activation in a COPD setting, we studied the signaling checkpoints shared among TLRs, specifically MyD88/NF-κB signaling. To do this, we developed CC-LR mice with airway specific deletion of MyD88 (LR/MyD88^Δ/Δ), and IKKβ (LR/IKKβ^Δ/Δ) separately, then exposed them to COPD-inducing conditions like TLR KO models. Interestingly, we found a robust reduction in tumor burden similar to the TLR KO models. LR/MyD88^Δ/Δ had a 1.4-fold reduction (87 vs. 124) and LR/IKKβ^Δ/Δ had a 1.3-fold (96 vs. 124) reduction in lung surface tumor numbers compared to age and sex match CC-LR control mice in the presence of COPD-like airway inflammation ( Figure 4A ). Tumor area analysis revealed a significant 1.5-fold (27% vs. 41%) and 1.3-fold (31% vs. 41%) reduction in LR/MyD88^Δ/Δ and LR/IKKβ^Δ/Δ mice, respectively ( Figures 4B, C ). Our naïve cohort of mice also presented with a significant 1.8-fold reduction (30 vs. 55) in tumor burden for both LR/MyD88^Δ/Δ and LR/IKKβ^Δ/Δ mice ( Supplementary Figure S5 ). These findings in naïve mice were accompanied by a 1.5-fold reduction (18% vs 28%) and an 85% reduction in both LR/MyD88^Δ/Δ and LR/IKKβ^Δ/Δ ( Supplementary Figure S5 ). These findings support our initial studies with TLR KO mice and suggest downstream signaling, specifically the MyD88/NF-κB pathway is essential in COPD-driven K-ras mutant lung cancer promotion.

Similar to TLR KO models, we sought to study the mechanism of epithelial specific knockouts of MyD88/NF-κB in K-ras mutant lung tumor promotion. Thus, we analyzed lung tissue for markers of cell proliferation and angiogenesis. LR/MyD88^Δ/Δ and LR/IKKβ^Δ/Δ mice showed a robust decrease in Ki-67 positive cells of about 80% ( Figures 5A, B ) and ERG positive cells of about 85% ( Figures 5C, D ). Upon analyzing our naïve cohort of mice, we found Ki-67 positive cells displayed a significant 55% reduction for LR/MyD88^Δ/Δ and an 85% reduction for LR/IKKβ^Δ/Δ mice. On the other hand, ERG had approximately a 30% reduction in LR/MyD88^Δ/Δ and an 80% reduction in LR/IKKβ^Δ/Δ mice ( Supplementary Figure S6 ). These results indicate that TLR downstream signaling pathways may play a role in regulating tumor cell proliferation and angiogenesis.

To gain insight into the immune cell phenotype of K-ras mutant MyD88 or IKKβ conditional knockout models, we studied the BALF inflammatory cells along with protumorigenic and anti-tumorigenic cytokine milieu. We found evidence of a similar reprogramming in the immune cell phenotype as TLR KOs, characterized by a reduction in total WBC counts driven mainly by a reduction in neutrophils followed by a significant reduction in the macrophage population ( Figure 6A ). Interestingly, neutrophil chemoattractant CXCL1 showed a 1.6-fold decrease in both LR/MyD88^Δ/Δ and LR/IKKβ^Δ/Δ mice ( Figure 6B ), however, macrophage chemoattractant CCL2 did not show significant changes ( Figure 6C ). IFNG displayed a significant 2-fold increase in LR/MyD88^Δ/Δ and a 3-fold increase in LR/IKKβ^Δ/Δ mice ( Figure 6D ). There was a 3.5-fold increase in TNF levels in both LR/MyD88^Δ/Δ and LR/IKKβ^Δ/Δ mice ( Figure 6E ), and Gzmb showed a 1.9-fold increase in LR/MyD88^Δ/Δ and 2.7-fold increase in LR/IKKβ^Δ/Δ mice ( Figure 6F ). IL-1β was unaffected ( Figure 6G ), while, IL17 revealed a 1.4-fold decrease in both KO mice ( Figure 6H ). Consistent with our findings in TLR KO mice, we also found a significant 2-fold decrease in protumorigenic/immunosuppressive marker RETNLB (Fizz1) in both LR/MyD88^Δ/Δ and LR/IKKβ^Δ/Δ mice ( Figure 6I ). In contrast to TLR KOs, CD45 mRNA expression in LR/MyD88^Δ/Δ and LR/IKKβ^Δ/Δ mice showed a significant decrease, which is consistent with the relatively lower number of leukocytes present in these mice compared to TLR KOs ( Supplementary Figure S7A ). T-cell marker CD3 and more specifically cytotoxic T-cell marker CD8 both showed a significant increase in LR/MyD88^Δ/Δ and LR/IKKβ^Δ/Δ ( Supplementary Figure S7D, E ), possibly due to the fact that MyD88 and IKKβ are downstream to multiple TLRs and blocking them leads to a signaling response akin to the promotion of T-cell cytotoxicity, which is nicely supported by the observed mRNA expression levels of Gzmb in these cohorts. CD68 expression showed no significant changes, while Ly6G was significantly decreased in LR/IKKβ^Δ/Δ mutant mice ( Supplementary Figures S7B, C ). Interestingly, IHC results were similar to TLR KO mice, supporting the previous notion of a less inflamed TME as we found a significant decrease of CD45-positive staining across all mutant mice models ( Supplementary Figures S7F, G ).

In the naïve cohorts of LR/MyD88^Δ/Δ and LR/IKKβ^Δ/Δ, like our naïve TLR KO cohort we found no significant changes in either neutrophils or lymphocytes. However, contrary to our naïve TLR KO cohort, targeting these downstream signaling pathways led to significant reductions in macrophages. Interestingly, CXCL1, CCL2, and Gzmb showed no significant changes. TNF levels in both LR/MyD88^Δ/Δ and LR/IKKβ^Δ/Δ naïve mice had a significant 3.5-fold increase. IFNG displayed a 2.2-fold increase for LR/MyD88^Δ/Δ and a 3-fold increase for LR/IKKβ^Δ/Δ. Expression of IL-1β displayed a 1.7-fold and 1.2-fold reduction in naïve LR/MyD88^Δ/Δ and LR/IKKβ^Δ/Δ respectively. Lastly, IL17 had a significant 1.4-fold reduction for both mouse models of downstream TLR signaling pathways and RETNLB had a 2-fold reduction as well ( Supplementary Figure S8 ). Ultimately, these findings in conjunction with our initial TLR KO models validate an antitumorigenic phenotype that the knockout of TLRs or epithelial specific knockout of MyD88/NF-κB could induce.