IκB Kinase α Is Required for Development and Progression of KRAS-Mutant Lung Adenocarcinoma

Tumors harboring mutations in the V-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) are notoriously resistant to current treatments (1). Lung adenocarcinoma (LADC), the number one cancer killer worldwide (2), harbors KRAS mutations in up to 30% to 40% of the cases diagnosed in Europe and North America (3). A cardinal mechanism of KRAS mutation–associated drug resistance appears to be the oncogene's addiction to transcriptional programs that facilitate sustained tumor-initiated cell survival, such as NFκB (4). To this end, mutant KRAS was recently shown to interact with NFκB–activating kinases [inhibitor of NFκB (IκB) kinases, IKK] to promote cancer cell survival, stemness, and drug resistance (5, 6).

NFκB is activated via the canonical (involving IκBα, IKKβ, and RelA/P50) and noncanonical (comprising IκBβ, IKKα, and RelB/P52) pathways (7). We and others previously documented NFκB activation in murine and human LADC (8–10). However, the IKKs responsible for this remain elusive, and most studies focused on IKKβ, IKKϵ, and TANK-binding kinase 1 (TBK1; refs. 11–14). IKKα participates in both canonical and noncanonical NFκB pathways and cooperates with IKKβ for tumor cell growth in vitro (11, 15), but its role in LADC development in vivo is uncharted.

We deployed NFκB reporter and conditional IKKα- and IKKβ-deleted mice to decipher the timing of NFκB activation and the mutual impact of IKKα and IKKβ on LADC development. In mouse models of tobacco carcinogen- and oncogenic KRAS^G12D-triggered LADC, IKKα was cardinal for disease initiation and progression. Moreover, IKKα selectively fostered cellular proliferation in the context of mutant KRAS and was also highly expressed in human LADC. Importantly, dual IKKα /IKKβ inhibition yielded promising results against KRAS-driven LADC, lending hope for translational applications of our findings.

Additional Methods are described in the Online Supplement.

C57BL/6J (#000664), FVB/NJ (#001800), B6.129(Cg)-Gt(ROSA)26Sor^{tm4(ACTB-tdTomato,-EGFP)Luo}/J (mT/mG; #007676; ref. 16), B6.129S4-Kras^tm4Tyj/J (LSL.KRAS^G12D; #008179; ref. 17), NOD.CB17-Prkdc^<scid>/J (NOD/SCID; #001303), and FVB.129S6(B6)-Gt(ROSA)26Sor^tm1(Luc)Kael/J (LSL.R26.Luc; #005125; ref. 18) mice were from The Jackson Laboratory. NFκB reporter mice (NGL; NF-κB.GFP.Luciferase), B6.B4B6-Chuk^<tm1Mpa>/Cgn (Chukf/f), and B6.B4B6-Ikbkb^<tm2.1Mpa>/Cgn (Ikbkbf/f), B6;CBA-Tg(Scgb1a1-cre)1Vart/Flmg (Scgb1a1.Cre), and Tg(Sftpc-cre)1Blh (Sftpc.Cre) mice have been described (8, 19–21). Mice were bred >F9 to the C57BL/6 and/or FVB backgrounds at the University of Patras Center for Animal Models of Disease. The number of mice used for these studies (n = 542) is detailed in Supplementary Table S1.

Urethane (CAS#51-79-6) and 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay were from Sigma, adenoviruses from the Vector Development Lab of the Baylor College of Medicine (Houston, TX), d-luciferin from Gold Biotechnology, HEK293T cells from ATCC, and Lewis lung carcinoma (LLC) and A549 lung adenocarcinoma cells from the NCI Tumor Repository (Frederick, MD). Primers and antibodies are listed in Supplementary Tables S2 and S3. Lentiviral shRNA pools (Santa Cruz Biotechnology) are described in Supplementary Table S4.

Chemical-induced LADC was induced in FVB and C57BL/6 mice, respectively, by a single or by 10 consecutive weekly intraperitoneal exposures to 1 g/kg urethane (8, 22–24). KRAS^G12D-driven LADC was induced via intratracheal injections of 5 × 10⁸ plaque-forming units (PFU) adenovirus type 5 encoding CRE recombinase (Ad-Cre) to LSL.KRAS^G12D mice on the C57BL/6 background (9, 17). NOD/SCID and C57BL/6 mice were anesthetized by isoflurane and received 2 × 10⁶ HEK293T and 0.5 × 10⁶ tumor cells into the rear flank, and vertical tumor diameters (δ) were measured and mice were imaged for bioluminescent detection of cell mass weekly thereafter. Cell spot volume (V) was calculated as V = π × (δ1 × δ2 × δ3)/6, and mice were killed after 6 weeks. Flank tumors were harvested and fixed with 4% paraformaldehyde or processed for immunoblotting.

LSL.KrasG12D;LSL.R26.Luc mice received 5 × 10⁸ PFU intratracheal Ad-Cre, followed by daily intraperitoneal injections of 100 μL saline or 0.5 mg/Kg TPCA-1 or 17-DMAG in 100 μL saline at days 14 to 28 or 84 to 112 after Ad-Cre. Mice were imaged for bioluminescent detection of LADC burden at 0, 14, 28, 84, and 112 days after Ad-Cre. Mice were sacrificed and lungs were harvested at 112 days after Ad-Cre.

Mouse primary lung adenocarcinoma cells and airway epithelial cells were derived from the lungs of urethane (single dose 1 g/kg) or saline-treated FVB or C57BL/6 mice by simple tumor or large airway dissection or epithelial stripping, respectively, and 5-month or 5-day culture, respectively, as described elsewhere (25). These cell lines were named XYLA#, with X signifying the mouse strain (F, FVB; C, C57BL/6), Y the carcinogen used (U, urethane), LA lung adenocarcinoma, and # their serial number by derivation date. Cells were cultured at 37°C in 5% CO₂–95% air using DMEM supplemented with 10% FBS, 2 mmol/L l-glutamine, 1 mmol/L pyruvate, 100 U/mL penicillin, and 100 mg/mL streptomycin. Cells were tested biannually for identity (by the short tandem repeat method) and for Mycoplasma spp. (by PCR). For experiments, frozen cells were reconstituted and were passaged 2 to 5 times for less than 2 weeks. In vitro cancer cell proliferation was determined using the MTT assay. For this, 2 × 10⁴ cells/well were plated onto 96-well plates. Daily thereafter, 15 μL of 5 mmol/L MTT working solution in PBS was added to wells to be measured that day. The plate was incubated for 4 hours at 37°C in a 5% CO₂ humidified incubator, followed by the addition of 100 μL acidified isopropanol per well for sediment solubilization and absorbance measurement at 492 nm on an MR-96A photometer (Mindray). For soft-agar colony formation assay, 7.5 × 10³ cells were plated in 60-mm culture vessels in semisolid 0.7% agarose in full culture medium and were incubated for 30 days at 37°C in a 5% CO₂ humidified incubator. Fresh culture medium (2 mL) was added to each vessel biweekly. After incubation, 500 μL MTT working solution was added to each vessel and plates were dried, inverted, photographed, and colonies were counted, as described elsewhere (25).

Matched tumor and normal lung tissue RNA and sections of 23 and 35, respectively, previously reported patients with LADC from Institution 3 were used for microarray and immunohistochemistry for IKKα and IKKβ (26). Human studies were approved a priori by the ethics committee of the University of Lübeck, Germany (approval #AZ 12-220) and were conducted according to the Declaration of Helsinki. Written informed consent was obtained from all patients. IKK score was 0, 1, 2, or 3 for no, cytoplasmic only, cytoplasmic and nuclear, and nuclear only immunoreactivity, respectively (modified from ref. 11).

Sample size (n; always biological) was determined using G*power (http://www.gpower.hhu.de/), assuming α = 0.05, β = 0.05, and d = 1.5. Data were acquired by two blinded readers, reevaluated if >20% deviant (no data were excluded), examined for normality by the Kolmogorov–Smirnov test and presented as median (interquartile range) or mean ± SEM. Differences in frequencies were examined by Fisher exact/χ² tests, in means of normally distributed variables by t test or one-way ANOVA/Bonferroni posttests, and in medians of nonnormally distributed variables by Mann–Whitney test or Kruskal–Wallis/Dunn posttests. Survival and flank tumor volumes were examined by Kaplan–Meier estimates/log-rank tests and two-way ANOVA/Bonferroni posttests. Probability (P) is two-tailed; P < 0.05 was considered statistically significant. Statistics and plots were done on Prism v5.0 (GraphPad).

All animal experiments were approved a priori by the Veterinary Administration of Western Greece according to a full and detailed protocol (approval #276134/14873/2). Male and female mice were sex-, weight (20–25 g)-, and age (6–12 weeks)-matched. Human studies were approved a priori by the ethics committee of the University of Lübeck, Germany (approval #AZ 12-220).

To map pulmonary NFκB activity during KRAS-driven neoplasia, NFκB reporter mice (NGL) on the carcinogen-susceptible FVB background expressing NFκB–driven Photinus Pyralis luciferase (LUC) in-frame with EGFP (8, 23) received a single intraperitoneal injection of saline or the tobacco carcinogen urethane (1 g/kg) and were serially imaged for bioluminescence. Urethane causes respiratory epithelial Kras^G12V/Q61R mutations and progressive inflammation, hyperplasias, adenomas, and adenocarcinomas in FVB mice (22–25) that in this experiment also expressed the NGL reporter (Fig. 1A and B). In addition to the baseline signals of these mice, markedly increased light emission from the chest was exclusively detected in urethane-treated mice at early and late time points corresponding to carcinogen-induced inflammation and LADC, respectively (8). Enhanced NFκB activation indicated by the EGFP reporter emanated exclusively from LADC (Fig. 1C–F). In a second approach, NGL mice were intercrossed with mice carrying a conditional loxP-STOP-loxP.KRAS^G12D allele (LSL.KRAS^G12D; ref. 17), and NGL and NGL;LSL.KRAS^G12D offspring (all C57BL/6 background) received intratracheal Ad-Cre and were longitudinally imaged. In LSL.KRAS^G12D mice, progressive inflammation, hyperplasia, adenomas, and adenocarcinomas carrying the KRAS mutation are inflicted by Ad-Cre (9, 17). To titrate Ad-Cre, mT/mG CRE reporters that switch from membranous Tomato (mT) to EGFP (mG) fluorescence upon CRE recombination (16) received 0, 5 × 10⁷, 5 × 10⁸, or 5 × 10⁹ PFU intratracheal Ad-Luc or Ad-Cre and were killed upon subsidence of transient Ad-mediated transgene expression at 2 weeks after injection (27). The low, intermediate, and high Ad-Cre titers, respectively, caused infrequent, stochastic, and ubiquitous respiratory epithelial recombination (Supplementary Figs. S1A and S1B). We selected 5 × 10⁸ PFU Ad-Cre to stochastically induce recombination into the respiratory epithelium of NGL and NGL;LSL.KRAS^G12D mice (Supplementary Fig. S1C). Similar to the urethane model, two phases of enhanced chest light emission by NGL;LSL.KRAS^G12D but not NGL mice were observed, coinciding with early inflammation and late LADC development (8, 22–24, 28). Again, NFκB–dependent EGFP expression was confined to LADC (Supplementary Fig. S1D–S1G). These data demonstrate biphasic pulmonary NFκB activation during KRAS-driven LADC development.

To investigate the NFκB pathway at play during KRAS-mutant inflammation, hyperproliferation, and LADC formation, the immunoreactivity of nuclear and cytoplasmic protein extracts of whole lungs of urethane-treated FVB mice and of Ad-Cre–treated LSL.KRAS^G12D mice for NFκB subunits, kinases, and inhibitors were probed longitudinally (Fig. 2A–D). In the urethane model, marked RelB and RelA immunoreactivity was detected in nuclear extracts and enhanced IκBα, IκBβ, IKKα, IKKβ, and TBK1 immunoreactivity in cytoplasmic extracts of the neoplastic stage. Some immunoreactivity was also present in early stages but their expression peaked in tumor-bearing lungs, while no IKKϵ signal was evident at any time point. In the LSL.KRAS^G12D model, enhanced nuclear RelB and P52 and modest RelA immunoreactivity was detected in nuclear extracts of tumor-bearing lungs, together with some cytoplasmic immunoreactivity for IκBβ, IKKα, and TBK1 (120 days). In addition, some RelA, RelB, P52, IκBα, IκBβ, and TBK1 immunoreactivity was evident in same-day–treated lungs (0 days), some RelA, RelB, P52, IKKα, and TBK1 immunoreactivity in inflammatory and proliferative lungs (30 and 60 days), and no IKKβ and IKKϵ signal at any stage. IKK expression patterns were corroborated using immunofluorescent detection of IKKα/IKKβ on lung sections of urethane-treated FVB and Ad-Cre–treated LSL.KRAS^G12D mice at 240 and 120 days after treatment, respectively. In both models, IKKα was expressed by a significant proportion of LADC cells, while minimal IKKβ expression was detectable (Fig. 2E and F). To further characterize NFκB activity of LADC, LADC cells were derived from the lungs of saline and urethane-treated FVB mice, according to established methods (Supplementary Fig. S2A and S2B; refs. 25, 29). LADC cells exhibited enhanced nuclear RelB (but not RelA) localization and activity compared with saline- and urethane-treated lungs (Supplementary Fig. S2C and S2D). Taken together, these results indicate coactivation of the canonical and noncanonical NFκB pathways in LADC.

We next functionally assessed the role of IKKα and IKKβ in LADC development, utilizing conditional IKKα and IKKβ gene–deleted mice (Chukf/f and Ikbkbf/f) that feature loxP-flanked alleles excised upon CRE expression (19). In a first line of experiments, mT/mG CRE reporter (control), Chukf/f, and Ikbkbf/f mice on the urethane-resistant C57BL/6 background (8) received 5 × 10⁹ PFU intratracheal Ad-Cre (a titer causing recombination in ∼75% of the respiratory epithelium within 2 weeks; Supplementary Fig. S1A and S1B), and were started 2 weeks thereafter on 10 weekly doses of 1 g/kg intraperitoneal urethane, a regimen that reproducibly induces LADC in C57BL/6 mice (23, 29). In this multihit model, stochastic KRAS mutations, inflammation, apoptosis, and regeneration were repeatedly inflicted across IKK-deleted and nondeleted respiratory epithelium (Fig. 3A). Interestingly, Ikbkbf/f mice displayed decreased survival during repeated urethane exposures, suggesting a role for IKKβ in epithelial repair (Fig. 3B). However, at 6 months after urethane start, IKKα-deleted mice had markedly decreased LADC incidence, multiplicity, and burden per lung compared with controls, while IKKβ-deleted mice displayed only minor reductions in tumor multiplicity but not burden (Fig. 3C–G). These experiments were replicated on Chukf/f and Ikbkbf/f mice back-crossed >F9 to the single-hit FVB model that recapitulates the mutation spectrum of human LADC and allows separate insights into the effects of IKK deletion on tumor initiation and progression via observations on LADC number and size after 6 months (8, 22–24). For this, WT control, Chukf/f, and Ikbkbf/f mice (all FVB) received 5 × 10⁹ PFU intratracheal Ad-Cre, followed by a single intraperitoneal exposure to 1 g/kg urethane (Supplementary Fig. S3A). All genotypes comparably survived single-hit urethane (Supplementary Fig. S3B). Again, Chukf/f mice developed fewer and smaller LADC compared with controls, indicating marked tumor-initiating and promoting effects of IKKα, but Ikbkbf/f mice displayed tumor incidence, number, size, and load closely resembling WT littermates, suggesting that the minor tumor-promoting effects of IKKβ require repetitive carcinogen challenge to become evident (Supplementary Fig. S3C–S3G). Chukf/f and Ikbkbf/f mice were also intercrossed with Scgb1a1.Cre (20) and Sftpc.Cre (21) CRE drivers (all C57BL/6) and their offspring received 10 consecutive weekly intraperitoneal injections of 1 g/kg urethane starting at 6 weeks of age (Supplementary Fig. S4A). Interestingly, both Scgb1a1.Cre- and Sftpc.Cre-driven IKKα-deletion was equally protective against LADC, while IKKβ-deletion had no effect (hence pooled Scgb1a1.Cre and Sftpc.Cre data are presented; Supplementary Fig. S4B–S4E). To solidify the link between IKKα and mutant KRAS and to discriminate between cell-autonomous and paracrine IKKα effects, Chukf/f and Ikbkbf/f mice were intercrossed with LSL.KRAS^G12D mice (all C57BL/6) and their offspring received 5 × 10⁸ PFU intratracheal Ad-Cre, a model where KRAS^G12D expression and IKK deletion coincide (Fig. 4A). Lung morphometry (30) at 4 months after Ad-Cre showed that IKKα-deleted mice had markedly decreased LADC burden compared with controls, while IKKβ-deleted mice displayed an intermediate phenotype (Fig. 4B–E). Collectively, these findings show that IKKα promotes KRAS-mutant LADC in a cell-autonomous fashion, independent from and more pronounced than IKKβ.

We next stably transfected HEK293T benign human embryonic kidney cells with vectors encoding control random sequence (pC), RFP (pRFP), EGFP (peGFP), wild-type (peGFP.Kras^WT), or mutant (peGFP.Kras^G12C) murine Kras in-frame with EGFP, and murine IKKα (pChuk) or IKKβ (pIkbkb) in various combinations. After transgene expression was validated, RFP-expressing control cells and EGFP-expressing intervention cells cotransfected with various combinations of peGFP.Kras^WT/peGFP.Kras^G12C and pChuk/pIkbkb were mixed at equal ratios and cocultured for 1 week, followed by quantification by fluorescent microscopy and flow cytometry (Supplementary Fig. S5A and S5B). Of note, as opposed to successful pIkbkb coexpression with peGFP.Kras^WT, pIkbkb coexpression with peGFP.Kras^G12C was repeatedly impossible (n = 5), indicating mutual repulsion of mutant Kras^G12C and IKKβ, similar to what was previously observed with other RAS/IκB-like GTPases called κB-RAS (Supplementary Fig. S5B; ref. 31). Despite this, IKKβ provided a proliferative advantage to HEK293T cells expressing Kras^WT, whereas Kras^G12C-expressing HEK293T cells proliferated more efficiently upon IKKα overexpression (Supplementary Fig. S5C and S5D). Subsequently, HEK293T cells were stably transfected with pCAG.Luc, followed by various combinations of pC, peGFP.Kras^WT, peGFP.Kras^G12C, pChuk, and/or pIkbkb were validated, and two million cells were injected at different dorsal skin spots of NOD/SCID mice, followed by serial spot volume assessment and bioluminescence imaging. Again, IKKβ boosted in vivo growth of HEK293T cells expressing Kras^WT, while Kras^G12C-expressing HEK293T cells were rendered more tumorigenic upon IKKα overexpression (Fig. 5A–C). Interestingly, none of 8 NOD/SCID mice bearing subcutaneous Kras^WT cells developed pulmonary lesions, while 5 of 8 mice with subcutaneous Kras^G12C cells developed lung metastases (P = 0.0256, χ² test; Fig. 5D). We next examined HEK293T spots that had grown into tumors for NFκB pathway component immunoreactivity. By immunoblotting, we observed nuclear localization of IKKα but not IKKβ in control tumors expressing Kras^WT that was further enhanced by coexpression of IKKβ. Kras^G12C tumors showed both IKKα and IKKβ nuclear immunoreactivity, while Kras^G12C-IKKα-expressing tumors had enhanced IKKα and diminished IKKβ nuclear signals, and Kras^G12C-IKKβ-expressing tumors displayed loss of both nuclear signals (Fig. 5E). The nuclear localization of IKKα in Kras^G12C-IKKα coexpressing tumors was also evident on tissue sections by immunofluorescence (Fig. 5F). In addition to KRAS-IKK coexpression in benign cells, we stably expressed shRNAs specifically targeting IKKα and IKKβ transcripts (Chuk and Ikbkb, respectively) in different lung adenocarcinoma cell lines [LLC, Lewis lung adenocarcinoma cells; and primary lung adenocarcinoma cells derived from urethane-induced lung tumors of FVB (FULA) and C57BL/6 (CULA) mice] bearing wild-type Kras^WT (CULA cells), Kras^G12C (LLC cells), Kras^Q61R (FULA1 and FULA3 cells), or silenced Kras^Q61R (FULA3 cells stably expressing shKras; refs. 25, 29). Interestingly, IKKα silencing resulted in decreased clonogenic capacity in vitro and decreased tumor growth in vivo specifically of KRAS-mutant tumor cells (Supplementary Fig. 6A–C). Moreover, this effect was not obvious in vitro, in line with recent observations on the in vivo–restricted effects of the oncogene (32). Collectively, these results indicated selective addiction of mutant KRAS to IKKα during carcinogenesis, possibly via nuclear IKKα functions reported elsewhere (33, 34).

We subsequently evaluated the therapeutic potential of our findings using cellular and animal systems tailored to noninvasively monitor tumor growth and NFκB activity. For this, three KRAS-mutant LADC cell lines (mouse primary LADC, Kras^Q61R; murine LLC, Kras^G12C; A549 human LADC, KRAS^G12S) were stably transfected with constitutive (pCAG.Luc) and NFκB–dependent (pNGL) LUC reporters, inducibility of the NFκB reporter was validated, and cells were treated with increasing concentrations of the selective IKKβ inhibitor TPCA-1 {2-[(aminocarbonyl)amino]-5-[4-fluorophenyl]-3-thiophenecarboxamide; ref. 35) or the heat shock protein 90 inhibitor 17-DMAG (alvespimycin; 17-dimethylaminoethylamino-17-demethoxygeldanamycin; refs. 36, 37) that blocks, among other targets, IKKα and IKKβ function; bioluminescence imaging of live pCAG.Luc cells after 48-hour treatments was used to determine cell killing and bioluminescence imaging of live pNGL cells after 4-hour treatments was used to determine NFκB inhibition. Intriguingly, 17-DMAG displayed superior efficacy in halting cell proliferation and NFκB activity in all three cell lines compared with TPCA-1, as evident by 4- to 5-fold lower 50% inhibitory concentrations of pCAG.Luc activity (mean ± SD: 28 ± 12 μmol/L for 17-DMAG and 114 ± 30 for TPCA-1) and 200- to 1,000-fold lower 50% inhibitory concentrations of pNGL activity (mean ± SD: 0.133 ± 0.068 μmol/L for 17-DMAG and 62 ± 30 for TPCA-1; Supplementary Fig. S7A–S7E). Based on these results and the data from NGL mice with KRAS^G12D tumors (Supplementary Fig. S1), we designed an in vivo study where mice with KRAS^G12D-mutant LADCs received low doses of either drug tailored to inhibit NFκB activity rather than cell proliferation in both preventive and curative modes. To enable repetitive noninvasive bioluminescent quantification of tumor burden in vivo, mice harboring a conditional loxP-STOP-loxP.R26.Luc allele (LSL.R26.Luc; ref. 18) were intercrossed with LSL.KRAS^G12D mice (all C57BL/6; ref. 17), yielding a model where CRE-recombination leads to simultaneous KRAS^G12D and LUC expression (Fig. 6A; ref. 38). LSL.KRAS^G12D;LSL.R26.Luc mice (n = 30) received 5 × 10⁸ intratracheal PFU Ad-Cre and were allocated to drug treatments during the two distinct phases of NFκB activation identified from LSL.KRAS^G12D;NGL mice (Supplementary Fig. S1): between days 14 and 28 after Ad-Cre (prevention trial) or between days 84 and 112 after Ad-Cre (regression trial; Fig. 6B). Treatments consisted of 100 μL daily intraperitoneal saline, TPCA-1, or 17-DMAG, both at 0.5 mg/Kg in 100 μL saline, equivalent by body volume extrapolation to maximal in vivo concentrations of 1.79 μmol/L for TPCA-1 and of 0.77 μmol/L for 17-DMAG, far inferior to cytotoxic concentrations (Supplementary Fig. S7). Bioluminescent detection of developing LADC revealed that TPCA-1 had no effect, while 17-DMAG prevention and regression regimens efficiently blocked tumor development compared with controls (Fig. 6C and D). Collectively, these results indicate that 17-DMAG exerts beneficial effects against KRAS-mutant LADCs in vitro and in vivo, even at low doses tailored to inhibit NFκB. On the contrary, a specific IKKβ inhibitor failed to show any effect, further supporting a druggable addiction of IKKα with mutant KRAS in LADC.

The relevance of our findings with human LADC was subsequently addressed. For this, tumor and adjacent normal-appearing lung tissues of 23 patients with LADC were analyzed for CHUK and IKBKB expression by microarray and of another 35 patients from the same series for IKKα and IKKβ by immune labeling (26). CHUK mRNA was overrepresented in normal-appearing and LADC tissue compared with IKBKB mRNA, while the levels of both were not different between normal-appearing and tumor tissue (Fig. 7A). However, using a modified NFκB scoring system that examines staining intensity and localization (10), IKKα protein was significantly overexpressed in LADC compared with both normal-appearing tissues and with IKKβ (Fig. 7B and C), suggesting its possible involvement in the pathogenesis of human LADC.

We report an actionable requirement for IKKα in KRAS-mutant LADC. Using chemical and transgenic delivery of KRAS mutations to the respiratory tract in combination with NFκB reporter and conditional IKK-deleted mice, we map the patterns of NFκB activation in the lungs and identify the critical role of IKKα. We show that IKKα drives LADC through cell-autonomous effects that are specifically exerted in the cellular context of mutant KRAS. These findings have implications for human disease, because IKKα is overexpressed in human LADC and oncogenic KRAS-IKKα addiction was annihilated by treatment with 17-DMAG.

The findings are novel and important on various counts. First, NFκB activity of KRAS-mutant LADC is charted in living mice and is shown to be activated early after KRAS mutation induction and late in established LADC. This pattern is in line with observations from smokers at risk for LADC that feature airway epithelial NFκB activation (39) and from patients with established LADC that display oncogenic NFκB activation (10). The results are consistent with the hypothesis that NFκB activation occurs together with field mutagenesis in the respiratory tract, persists in mutated cells, and reappears during clinical manifestation of late disease (40), bearing implications for NFκB–based therapy and prevention (Fig. 7D).

Second, noncanonical together with canonical NFκB pathway components are shown to be activated in KRAS-driven LADC. Canonical NFκB signaling is long known to be important in human and experimental LADC (8–10), but activity of the alternative pathway has not been described. This finding is in accord with our previous observations of enhanced RelB activity of tumor cells in human LADC (10) and suggests important roles for alternative NFκB signaling in KRAS-driven LADC.

Importantly, IKKα is identified as the critical kinase for oncogenic NFκB activation of KRAS-mutant LADC. IKKα deletion provided beneficial effects in four different mouse models of combined KRAS-driven carcinogenesis and IKK depletion from the respiratory epithelium. In addition, 17-DMAG protected mice from KRAS^G12D-driven LADC when given early (preventive treatment) or late (regression trial), while the IKKβ blocker TPCA-1 did not. Although 17-DMAG likely suppresses a spectrum of targets broader than IKKα and IKKβ (41), inclusively targeting IKKα, utilizing even this nonspecific approach provided superior overall effects in reducing tumor burden compared with IKKβ-specific inhibition (Fig. 7E). We were the first to identify that indirect IKKβ blockade via overexpression of dominant-negative IκBα protects mice from urethane-induced LADC (8), a finding thereafter recapitulated in KRAS^G12D-mutant (9), and tobacco-smoke–induced (14) LADC. Urethane-triggered LADCs were recently genomically characterized and shown to harbor Kras^Q61R/Kras^G12V mutations (22), similar to human LADC (42). Based on these findings and results from other tumor types, research and drug discovery focused on IKKβ yielding proteasome and IKKβ inhibitors (35, 43). However, these provide poor outcomes in human LADC (44) and cause resistance or paradoxical tumor promotion in animal LADC models via myeloid NFκB inhibition, secondary mutation development, and/or enhanced neutrophil-provided IL1β (23, 45, 46). In addition, recent evidence indicates that IKKβ might not be the only kinase responsible for oncogenic NFκB activation of KRAS-mutant LADC (47). To this end, TBK1 emerged as a KRAS addiction partner and was found to mediate EGFR-inhibitor resistance (5, 6), while IKKϵ promoted tumorigenesis together with TBK1 (13). Only one study addressed the role of IKKα depletion together with IKKβ in lung cancer cells in vitro and found both kinases to be important (11). Our results identify for the first time the pivotal role of IKKα in de novo development of KRAS-mutant LADC in vivo and position the kinase as a marked therapeutic target.

Using in vitro and in vivo competition studies, we determine that IKKα selectively fosters the survival of KRAS-mutant cells and is therefore addicted to the oncogene, while IKKβ promotes the survival and maintenance of nonmutated cells. We hypothesize that in a stochastically KRAS-mutated respiratory field, this opposing addiction of IKKα and IKKβ to mutant and wild-type KRAS, respectively, would lead over time via clonal selection to KRAS-mutant LADCs with enhanced IKKα activity (Fig. 7D). This cell-autonomous model is supported by the results from KRAS^G12D mice (where IKKα was selectively deleted in KRAS-mutant cells) and from HEK293T cells (where IKK/KRAS combinations functioned similarly in vitro and in vivo), notwithstanding the possibility for autocrine IKKα-triggered cytokine networks identified elsewhere (48, 49). To this end, IKKα localized to the nucleus of our murine LADCs, a phenomenon that could enhance gene transcription or repress oncogenes (33, 34). Nuclear IKKα was also present in human LADC, which displayed enhanced nuclear IKKα immunoreactivity. The proposed IKKα function to site independently foster KRAS-mutant cells also emanates from tissue-restricted IKK-deletion studies where IKKα was critical in both airway and alveolar cells, a result of importance given the cellular and histologic diversity of human LADC (50).

Finally, a feasible approach for translation of the findings is explored. Treatment with 17-DMAG was tailored to target NFκB activation of KRAS-mutant LADC in vitro and was translated to a preclinical study, where it was well tolerated and effective against LADC in vivo, both preventively and therapeutically. The efficacy of 17-DMAG and the inefficacy of TPCA-1 strengthen the proposed link between mutant KRAS and IKKα and open up new avenues for therapy/prevention of KRAS-mutant LADC (1). In summary, we report a requirement for IKKα in KRAS-driven LADC, implicate IKKα as a KRAS nononcogene addiction partner, and show that targeting IKKα may confer beneficial effects against a currently untreatable disease that is the number one cancer killer in the world.

No potential conflicts of interest were disclosed.

Conception and design: M. Papageorgopoulou, T.S. Blackwell, A. Marazioti, G.T. Stathopoulos

Development of methodology: M. Vreka, I. Lilis, M. Papageorgopoulou, G.A. Giotopoulou, M. Lianou, I. Giopanou, M. Spella, T. Agalioti, G.T. Stathopoulos

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Vreka, M. Papageorgopoulou, N.I. Kanellakis, M. Spella, T. Agalioti, V. Armenis, S. Marwitz, F.E. Yull, G.T. Stathopoulos

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Vreka, I. Lilis, M. Papageorgopoulou, G.A. Giotopoulou, M. Lianou, I. Giopanou, N.I. Kanellakis, T. Agalioti, T. Goldmann, S. Marwitz, G.T. Stathopoulos

Writing, review, and/or revision of the manuscript: M. Vreka, M. Papageorgopoulou, T. Goldmann, S. Marwitz, F.E. Yull, A. Marazioti, G.T. Stathopoulos

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Vreka, M. Papageorgopoulou, V. Armenis

Study supervision: M. Vreka, T. Goldmann, A. Marazioti, G.T. Stathopoulos

Other (provided mouse models used in the study): M. Pasparakis

This work was supported by European Research Council 2010 Starting Independent Investigator and 2015 Proof of Concept Grants (260524 and 679345, respectively, awarded to G.T. Stathopoulos), as well as a Research Award by the Hellenic Thoracic Society (awarded to M. Vreka).

The authors thank the University of Patras Center for Animal Models of Disease and Advanced Light Microscopy Cores for experimental support.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.