Malignant pleural effusion (MPE) is a frequent metastatic manifestation of human cancers. While we previously identified KRAS mutations as molecular culprits of MPE formation, the underlying mechanism remained unknown. Here, we determine that non-canonical IKKα-RelB pathway activation of KRAS-mutant tumor cells mediates MPE development and this is fueled by host-provided interleukin IL-1β. Indeed, IKKα is required for the MPE-competence of KRAS-mutant tumor cells by activating non-canonical NF-κB signaling. IL-1β fuels addiction of mutant KRAS to IKKα resulting in increased CXCL1 secretion that fosters MPE-associated inflammation. Importantly, IL-1β-mediated NF-κB induction in KRAS-mutant tumor cells, as well as their resulting MPE-competence, can only be blocked by co-inhibition of both KRAS and IKKα, a strategy that overcomes drug resistance to individual treatments. Hence we show that mutant KRAS facilitates IKKα-mediated responsiveness of tumor cells to host IL-1β, thereby establishing a host-to-tumor signaling circuit that culminates in inflammatory MPE development and drug resistance.
Malignant pleural effusion (MPE) is one of the most challenging cancer-related disorders. It ranks among the top prevalent metastatic manifestations of tumors of the lungs, breast, pleura, gastrointestinal tract, urogenital tract, and hematopoietic tissues, killing an estimated two million patients worldwide every year and causing 126,825 admissions in U.S. hospitals in 2012 alone1,2. The presence of a MPE at diagnosis is an independent negative prognostic factor in patients with lung cancer and mesothelioma3,4. In addition, current therapies are non-etiologic and often ineffective, may cause further morbidity and mortality, and have not yielded significant improvements in survival5,6.
To meet the pressing need for mechanistic insights into the pathobiology of MPE, we developed immunocompetent mouse models of the condition that unveiled inflammatory tumor-to-host signaling networks causing active plasma extravasation into the pleural space7. Nuclear factor (NF)-κB activity in tumor cells was pivotal for MPE formation in preclinical models, driving pro-inflammatory gene expression and promoting pleural tumor cell survival8,9,10. However, the mechanism of oncogenic NF-κB activation of MPE-competent pleural tumor cells remained unknown. In parallel, we recently pinned mutant KRAS as a molecular determinant of the propensity of pleural-metastasized tumor cells for MPE formation: mutant KRAS delivered its pro-MPE effects by directly promoting C-C chemokine motif ligand 2 (CCL2) secretion by pleural tumor cells, resulting in pleural accumulation of MPE-fostering myeloid cells11. However, a unifying mechanism linking KRAS mutations with oncogenic NF-κB activation and MPE competence of pleural tumor cells was missing.
KRAS mutations have been previously linked to elevated or aberrant NF-κB activity via cell-autonomous and paracrine mechanisms. KRAS-mutant tumors, including lung and pancreatic adenocarcinomas, require active NF-κB signaling12,13,14 and NF-κB inhibition blocks KRAS-induced tumor growth14,15,16. In turn, NF-κB activation of KRAS-mutant tumor cells has been associated with enhanced RAS signaling, drug resistance, and stemness17,18. Despite significant research efforts, the NF-κB-activating kinases (ΙκΒ kinases, IKK) and pathways (canonical, involving ΙκΒα, ΙΚΚβ, and RelA/P50, versus non-canonical, comprising IκΒβ, ΙΚΚα, and RelB/P52) that mediate this oncogenic addiction between mutant KRAS and NF-κB signaling are still elusive and diverse, and different studies indicate that IKKα, IKKβ, IKKγ, IKKε, and/or TANK-binding kinase 1 (TBK1) are key for this17,18,19,20,21,22,23,24.
Here we use immunocompetent mouse models of MPE to show that mutant KRAS determines the responsiveness of pleural tumor cells to host-delivered interleukin (IL)-1β signals by directly regulating IL-1 receptor 1 (IL1R1) expression. IKKα is further shown to critically mediate IL-1β signaling in KRAS-mutant tumor cells, culminating in marked MPE-promoting effects delivered by C-X-C chemokine motif ligand 1 (CXCL1), and in oncogenic addiction with mutant KRAS evident as drug resistance. Importantly, simultaneous inhibition of IKKα and KRAS is effective in annihilating mutant KRAS-IKKα addiction in MPE.
Non-canonical NF-κB signaling of KRAS-mutant cancer cells
We first evaluated resting-state NF-κB activity of five mouse cancer cell lines with defined KRAS mutations and MPE capabilities in syngeneic C57BL/6 mice11: Lewis lung carcinoma (LLC; MPE-competent; KrasG12C), MC38 colon adenocarcinoma (MPE-competent; KrasG13R), AE17 malignant pleural mesothelioma (MPE-competent; KrasG12C), B16F10 skin melanoma, and PANO2 pancreatic adenocarcinoma (both MPE-incompetent and KrasWT) cells. Parallel transient transfection of these cell lines with reporter plasmids encoding Photinus Pyralis LUC under control of either a constitutive (pCAG.LUC) or a NF-κB-dependent (pNF-κB.GFP.LUC; pNGL) promoter8,9,10,11,25 (Fig. 1a) revealed that unstimulated NF-κB activity did not segregate by KRAS mutation status (Fig. 1b). However, when PANO2 cells, a cell line with relatively low NF-κB activity, were transiently transfected with pKrasG12C, their NF-κΒ expression levels were elevated (Fig. 1c). Moreover, KRAS mutant (MUT) cells displayed elevated DNA-binding activity of non-canonical NF-κB subunits P52 and RelB by functional NF-κΒ enzyme-linked immunosorbent assay (ELISA) and enhanced nuclear immunofluorescent localization of RelB compared with KRASWT cells (Fig. 1d, e). Immunoblotting of cytoplasmic and nuclear extracts revealed that KRASMUT cells had increased levels of cytoplasmic RelA and IκΒα and of nuclear RelB, ΙκΒβ, and ΙΚΚα compared with KrasWT cells (Fig. 1f). These results suggest that KRASMUT cancer cells exhibit non-canonical endogenous NF-κB activity.
Resistance of KRAS-mutant cancer cells to IKKβ inhibition
We next examined the effects of small molecule inhibitors of the proteasome (bortezomib26), of IKKβ (IMD-035427), or of heat shock protein 90 (HSP90) (17-dimethylaminoethylamino-17-demethoxygeldanamycin (17-DMAG)28) that display significant inhibitory activity against IKKβ and/or IKKα (of note, a specific IKKα inhibitor does not exist) on NF-κB reporter activity and cellular proliferation of our murine cancer cell lines (Fig. 1g, h; Supplementary Table 1). Bortezomib, an indirect inhibitor of IKKβ via cytoplasmic accumulation of non-degraded ΙκΒα16,26, attenuated endogenous NF-κB activity of KrasWT cells but paradoxically activated NF-κB in KRASMUT cells, at the same time more effectively killing KRASWT than KRASMUT cells in vitro. Similarly, IKKβ-selective IMD-035427 blocked NF-κB activity and cellular proliferation of unstimulated KRASWT cells but not of KRASMUT cells. Interestingly, the HSP90 and dual IKKα/IKKβ inhibitor 17-DMAG29 was equally effective in limiting NF-κB activity and cellular proliferation of all cell lines irrespective of KRAS mutation status. These results suggest the existence of endogenous resistance of KRAS-mutant cells to IKKβ inhibition, which can be overcome by combined HSP90/IKKα/IKKβ inhibition.
IL-1-inducible NF-κB activation of KRAS-mutant cancer cells
We next studied NF-κB activation patterns of our murine cancer cells in response to exogenous stimuli. For this, cells were stably transfected with pNGL, were pretreated with saline or bortezomib (1 μM ~5–10-fold the 50% NF-κB inhibitory concentration obtained from KrasWT cells; Supplementary Table 1), were exposed to 60 different candidate NF-κB-pathway ligands at 1 nM concentration30, and were longitudinally monitored for NF-κB-dependent LUC activity by bioluminescence imaging of live cells in vitro (Fig. 2a, b; Supplementary Table 2). Incubation with lipopolysaccharide (LPS) and tumor necrosis factor (TNF) resulted in markedly increased NF-κB activity in all cells irrespective of KRAS status, while lymphotoxin β activated NF-κB in all but PANO2 cells, effects that peaked by 4–8 h of incubation and subsided by 16–24 h. Uniquely, IL-1α and IL-β induced NF-κB exclusively in KRASMUT cells. In addition, bortezomib exaggerated endogenous and inducible NF-κB activation of KRASMUT cells, in contrast to KRASWT cells that displayed efficient NF-κΒ blockade by bortezomib. In line with the above, Il1r1 (encoding IL1R1, cognate to IL-1α/β) expression, but not Tnfrsf1a/Tnfrsf1b (encoding TNF receptors) or Il1a/Il1b expression (that was undetectable in all cell lines), was exclusively restricted to KRASMUT MPE-proficient tumor cells (Fig. 2c, d). We subsequently tested whether inducible NF-κB activation occurs in tumor cells entering the pleural space in vivo, simulating incipient pleural carcinomatosis4,7. For this, naive C57BL/6 mice were pulsed with a million intrapleural pNGL-expressing tumor cells and were serially imaged for NF-κB-dependent bioluminescence. Amazingly, KRASMUT MPE-competent cells responded to the pleural environment with markedly escalated NF-κB activity within 4 h after injection, while KRASWT MPE-incompetent cells showed diminishing NF-κB signals (Fig. 3a). Interestingly, this in vivo NF-κB response of KRASMUT cells was abolished in IL-1β-deficient (Il1b−/−31), but not in TNF-deficient (Tnf−/−32), mice (Fig. 3b), indicating that KRASMUT tumor cells selectively respond to IL-1β of the pleural environment by activating NF-κB.
Mutant KRAS promotes non-canonical NF-κB signaling
To define the role of mutant KRAS in the aberrant NF-κB activation patterns of KRASMUT tumor cells, including non-canonical endogenous NF-κB activity, resistance to IKKβ inhibition, and IL-1β-inducibility, we undertook short hairpin RNA (shRNA)-mediated KRAS silencing (shKras) and plasmid-mediated overexpression of a mutant dominant-negative form of ΙκΒα (pΙκΒαDN; inhibits canonical NF-κΒ signaling) in KRASMUT cell lines, as well as plasmid-mediated overexpression of mutant KRAS (pKrasG12C) in KRASWT cell lines11,33. Stable pΙκΒαDN expression in MC38 cells (KrasG13R) resulted in decreased RelA and sustained RelB nuclear-binding activity, while shKras did not affect RelA but abolished RelB nuclear-binding activity (Fig. 4a). shKras also eliminated nuclear RelB localization in these cells without affecting RelA (Fig. 4b) and abolished nuclear IKKα immunoreactivity of LLC (KrasG12C) and MC38 cells (Fig. 4c). shKras expression reversed the endogenous resistance of MC38 cells to bortezomib and IMD-0354, rendering them as sensitive as KRASWT cells (Fig. 4d, e). In addition, shKras annihilated IL-1β-induced NF-κB transcriptional activity of pNGL-expressing LLC, MC38, and AE17 (KrasG12C) cells (Fig. 4f), and pKrasG12C transmitted this phenotype to KrasWT PANO2 cells (Fig. 4g). Importantly, shKras abrogated the in vivo NF-κB response of pleural-inoculated MC38 cells, which was reinstated in PANO2 cells by stable pKrasG12C expression (Fig. 4h, i). In parallel, KRAS silencing in KRASMUT cells significantly decreased, whereas pKrasG12C overexpression in KRASWT cells significantly increased Il1r1 expression, as well as resting-state and IL-1β-inducible nuclear immunoreactivity for RelB, IκΒβ, and ΙΚΚα (Fig. 4j–l). Collectively, these data indicate that mutant KRAS induces non-canonical NF-κΒ signaling of cancer cells in unstimulated and IL-1β-stimulated conditions.
IKKα in mutant KRAS-dependent MPE
To define the NF-κB-activating kinase responsible for aberrant NF-κΒ signaling of KRASMUT cancer cells, we stably expressed shRNAs specifically targeting IKKα, IKKβ, IKKε, and TBK1 transcripts (Chuk, Ikbkb, Ikbke, and Tbk1, respectively) in our pNGL-expressing cell lines and validated them (Fig. 5a). In addition, we cloned these murine transcripts into an eukaryotic expression vector and generated stable transfectants of our cell lines. Interestingly, resting-state NF-κB transcriptional activity across KRASMUT cells was markedly suppressed by shChuk but not by shIkbkb or shTbk1, while shIkbke yielded minor NF-κΒ inhibition in MC38 and AE17 cells. On the contrary, endogenous NF-κΒ-mediated transcription of B16F10 cells was exclusively silenced by shIkbkb and, to a lesser extent, shIkbke, and of PANO2 cells by no shRNA (Fig. 5b). In a reverse approach, overexpression of any kinase resulted in enhanced NF-κΒ activity in all KRASMUT cells, of IKKβ only in B16F10 cells, and of no kinase in PANO2 cells (Fig. 5c). In addition to intrinsic, IKKα also mediated IL-1β-inducible NF-κΒ activity of KRASMUT tumor cells, since shChuk but not shIkbkb abolished IL-1β-induced NF-κB activity across KRASMUT cell lines (Fig. 5d). In line with the above, shChuk abolished the immunoreactivity of MC38 cell nuclear extracts for RelB, ΙκΒβ, and IKKα, both at resting and IL-1β-stimulated states (Fig. 5e). Taken together, these data suggest that KRAS-mutant cancer cells respond to pleural IL-1β via IKKα-mediated non-canonical NF-κB activation. Based on these results and our previous identification of the importance of KRAS mutations and NF-κΒ signaling in MPE development8,9,10,11, we hypothesized that IKKα is required for sustained NF-κΒ activation and MPE induction by pleural-homed KRASMUT cancer cells. To test this, we injected IKK-silenced pNGL-expressing LLC cells (KrasG12C; MPE-competent) into the pleural space of C57BL/6 mice. Indeed, recipients of IKKα-silenced LLC cells displayed significant reductions in MPE incidence and volume, pleural inflammatory cell influx, and pleural tumor NF-κΒ activity and prolonged survival. IKKε silencing delivered more modest and equivocal beneficial effects, while IKKβ and TBK1 silencing had no impact (Fig. 6a–c; Supplementary Table 3). These experiments were repeated with IKKα- and IKKβ-silenced MC38 cells (KrasG13R; MPE-competent) stably expressing pNGL, confirming that IKKα is cardinal for oncogenic NF-κΒ activation and MPE precipitation by pleural-metastatic KRASMUT tumor cells (Fig. 6d–f; Supplementary Table 3). However, standalone overexpression of IKKα or IKKβ did not confer MPE competence to KRASWT PANO2 cells, as opposed to pKrasG12C (Fig. 6g–i; Supplementary Table 3), in accord with our previous observations11. Collectively, these results suggest that mutant KRAS-potentiated IL-1β signaling results in KRASMUT addiction to IKKα activity, which is required but not sufficient for oncogenic NF-κΒ activation and MPE formation.
Myeloid IL-1β fosters mutant KRAS-IKKα addiction in MPE
To study the importance of host-delivered IL-1β in the proposed KRASMUT-IKKα addiction culminating in MPE, we delivered pNGL-expressing KRASMUT LLC and MC38 cells into the pleural space of Il1b−/−, Tnf−/−, and WT C57BL/6 mice. Interestingly, Il1b−/− but not Tnf−/− mice displayed decreased MPE incidence, volume, inflammatory cell influx, and oncogenic NF-κB activation (Fig. 7a–d; Supplementary Table 3). Host-provided IL-1β was of myeloid origin, since bone marrow (BM) transplants34,35 from C57BL/6 and Tnf−/−, but not Il1b−/− donors, to lethally irradiated Il1b−/− recipients unable to foster MPE rendered LLC cells MPE proficient (Fig. 7e, f; Supplementary Table 3). To define which myeloid cells provide the bulk of IL-1β to fuel tumor cell NF-κΒ activity, we isolated BM cells from C57BL/6 mice and drove them toward monocyte and neutrophil differentiation by macrophage colony-stimulating factor (M-CSF) and granulocyte-colony-stimulating factor (G-CSF culture, respectively. Both BM-derived monocytes and neutrophils secreted IL-1β upon 24-hour treatment with cell-free LLC supernatants as measured by ELISA, but monocytes secreted ~200 times higher cytokine levels than undifferentiated BM cells and neutrophils (Fig. 7g). These data clearly show that the main source of IL-1β in the pleural space during MPE development likely are recruited myeloid monocyte cells.
Mutant KRAS-IKKα addiction promotes MPE via CXCL1 secretion
To identify the MPE effectors and transcriptional signatures of IL-1β/KRAS/IKKα-addicted tumor cells, we subjected KRAS-silenced, IKKα-silenced, and IL-1β-challenged LLC and MC38 cells to microarray analyses, seeking for transcripts altered heterodirectionally by silencing/challenge. Thirty transcripts fulfilled these criteria in LLC (including Ppbp, encoding pro-platelet basic protein, PPBP, and Cxcl1, encoding CXCL1) and 20 in MC38 (including Cxcl1) cells, with Cxcl1 being the only common gene of these two signatures (Fig. 8a, b; Supplementary Tables 4, 5). Cxcl1 microarray results were validated by quantitative PCR (qPCR) and ELISA (Fig. 8c–e). Furthermore, chromatin immunoprecipitation (ChIP) was performed in LLC cells treated with phosphate-buffered saline (PBS) or IL-1β in order to specify whether and which NF-κB component directly binds to the promoter region of Cxcl1. The data indicate that only RelB and IKKα bind to the NF-κB element in the Cxcl1 promoter and that IL-1β significantly strengthens this binding (Fig. 8f). These findings are consistent with the enhanced transcriptional induction of Cxcl1. Moreover, Cxcl1 and Ppbp expression was pivotal for MPE induction by IL-1β/KRAS/IKKα-addicted LLC cells, since these were MPE incompetent in both C-X-C chemokine motif receptor 1 (CXCR1) and CXCR2 gene-deficient mice36,37 that lack the genes encoding CXCL1/PPBP-cognate CXCR1 and CXCR2 receptors38 (Fig. 8g; Supplementary Table 3). Notably, in MPEs from CXCR1 and CXCR2 gene-deficient mice the predominant cell population was monocytes, whereas in MPEs from CCR2 gene-deficient mice11 the prevalent cell type was neutrophils. This result was not unexpected since the majority of myeloid cells recruited in the pleural space during MPE development in C57BL/6 mice consist of both neutrophils and monocytes (Fig. 8h). Of note, the monocyte population is the most prevalent during MPE development.
Combined targeting of KRAS/IKKα is effective against MPE
To explore the therapeutic implications of the proposed mechanism, we examined potential synergy of the KRAS inhibitor deltarasin39 with the IKKβ-specific inhibitor IMD-0354 or the HSP90/IKKα/IKKβ inhibitor 17-DMAG using TNF- or IL-1β-stimulated LLC murine and A549 human lung adenocarcinoma cells expressing pNGL (Fig. 9a, b). Interestingly, all inhibitors alone or in combination failed to block TNF-inducible NF-κΒ activation in both cell lines. In addition, all standalone drugs failed to inhibit IL-1β-inducible NF-κΒ activation in both cell lines, except from partial effects observed in A549 cells by 17-DMAG. However, deltarasin/17-DMAG but not deltrarasin/IMD-0354 combination treatment completely abolished IL-1β-induced NF-κB activation in both cell types to unstimulated levels (Fig. 9a, b), indicating that drugging the KRAS/IKKα axis can halt IL-1β responsiveness. To determine the potential efficacy of this approach against MPE, standalone or combined deltarasin, and 17-DMAG treatments (both 15 mg/Kg) were delivered to mice with established pleural tumors. For this, C57BL/6 mice received pleural LLC cells and treatments commenced after 5 days to allow initial pleural tumor implantation11. At day 13 post-tumor cells, standalone deltarasin and 17-DMAG-treated mice had significantly decreased MPE volume compared with saline-treated controls (40% reductions for both groups; P < 0.05; one-way analysis of variance (ANOVA) with Bonferroni post-tests). However, combination-treated mice were markedly protected from MPE development (57% incidence) and progression (65% volume reduction; P < 0.001; one-way ANOVA with Bonferroni post-tests) (Fig. 9c; Supplementary Table 3). Hence combined targeting of mutant KRAS and IKKα is effective in halting oncogenic NF-κΒ activation and MPE in mice.
IL-1β-inducible NF-κΒ activity in human KRAS-mutant cells
To assess whether our findings are relevant to human cancer, we screened nine human cancer cell lines of known KRAS mutation status40 for Rel-binding activity of nuclear extracts. In accord with murine data, KRASMUT cells displayed enhanced nuclear RelB compared with RelA binding (Fig. 10a). In addition, A549 (KRASG12S) and NCI-H23 (KRASG12C) cells displayed IL-1β-induced NF-κΒ activation, as opposed to HT-29 and SKMEL2 cells (both KRASWT). Importantly, stable pKrasG12C expression in SKMEL2 cells rendered them responsive to IL-1β (Fig. 10b, c).
In summary, KRAS mutations alter NF-κΒ signaling in tumor cells. KRASWT cells preferentially utilize intrinsically or exogenously (i.e., by LPS, TNF) stimulated IKKβ-mediated NF-κΒ signaling, display sensitivity to IKKβ inhibition, poor CXCR1/2 ligand secretion, and MPE incompetence. KRASMUT cells predominantly use IKKα-mediated non-canonical NF-κΒ signaling at resting state and in response to myeloid IL-1β, display enhanced CXCR1/2 ligand secretion and MPE proficiency, and are addicted to sustained IKKα activity evident as resistance to IKKβ inhibitors.
We provide a novel paradigm of how an oncogene can co-opt the host environment to foster addiction with a perturbed signaling pathway. KRAS-mutant cancer cells are shown to respond to host provided IL-1β in the pleural space by increasing non-canonical IKKα-RelB pathway activity. The co-existence of mutant KRAS and elevated IKKα-mediated non-canonical NF-κB signaling in the cancer cell, relentlessly driven by host IL-1β, leads to two important consequences. First, to enhanced transcription of CXCL1/PPBP chemokines, recruitment of CXCR1+ and CXCR2+ myeloid cells, and frank escalation of inflammatory MPE development. Second, to oncogenic addiction between mutant KRAS and IKKα that culminates in drug resistance. Using immunocompetent mouse models of MPE, we show how IL-1β, mutant KRAS, and IKKα interplay to mediate non-canonical NF-κB activation, resistance to proteasome and IKKβ inhibitors, CXCL1/PPBP secretion, and MPE. Finally, we show that this partnership can be annihilated by combined inhibition of KRAS together with IKKα but not alone.
Although cell-autonomous pro-tumorigenic functions of mutant KRAS are well charted41,42,43, mechanisms utilized by the oncogene to co-opt host cells from the tumor microenvironment in order to favor tumor progression have only recently begun to be elucidated. In this regard, mutant KRAS was first shown to promote chemokine secretion by tumor-initiated cells, thereby promoting tumor-associated inflammation44,45. Along similar lines, we recently showed that the oncogene is responsible for CCL2 secretion by pleural metastatic cancer cells, fostering inflammatory MPE formation11. Our present findings expand the paradigm of how mutant KRAS impacts tumor–host interactions: it renders tumor cells capable of sensing inflammatory IL-1β signals originating from the CCL2-recruited monocytes. The increased Il1r1 expression in these cells could be a result of IL-1β -induced phosphorylation by nuclear IKKa of Ser10 in histone H3 that could be especially important for subsequent modifications in a variety of genes, including Il1r1. Moreover, integrated by IKKα-mediated non-canonical NF-κΒ activity, IL-1β signaling culminates in enhanced CXCL1/PPBP expression and secretion that function to escalate tumor-associated inflammation required for MPE. Hence, in addition to directly promoting chemokine expression, mutant KRAS is shown here to amplify host-originated inflammatory signals in order to escalate MPE-promoting inflammation.
Mutant KRAS is known to enhance oncogenic NF-κB activity; however, it was mainly linked to IKKβ, ΙΚΚε, and TBK1 function12,14,17,18,21,23,24,43, and only two studies identified IKKα as an accessory to IKKβ in KRAS-mutant lung adenocarcinoma22 and epidermal growth factor receptor-driven head and neck cancers19. Here we show for the first time that KRAS-mutant cancer cells display altered NF-κΒ utilization in resting and stimulated states, a phenomenon previously identified in pancreatic β cells46. Indeed, KRAS-mutant cancer cells displayed non-canonical endogenous NF-κΒ activity evident by enhanced nuclear localization and/or DNA-binding activity of RelB, ΙκΒβ, and ΙΚΚα, which was further inducible by exogenous IL-1β. Importantly, non-canonical NF-κΒ utilization by KRAS-mutant cancer cells was IKΚα driven, involved RelB activation, and was required for MPE. Nuclear IKΚα functions have been identified previously, including histone 3 modifications augmenting TNF and receptor activator of NF-κB ligand-induced gene expression and repression of maspin, a metastasis gate-keeper47,48,49. Our work links IKKα function with IL-1β-induced RelB activation and CXCL1/PPBP transcription. Moreover, we provide novel evidence that mutant KRAS is indirectly responsible for non-canonical NF-κB activation, which is IKKa and RelB based, via sensitization of cancer cells to host IL-1β. Finally, IKKα is found to be responsible for MPE, an important metastatic manifestation of various cancers. The findings concur with previous reports of a combined requirement for IKKα and IKKβ for oncogenic NF-κB activation19,22, as well as with human observations of predominant non-canonical NF-κB activity of tumors with high incidence of KRAS mutations, such as lung adenocarcinoma50. However, we demonstrate an isolated requirement for IKKα in KRAS-driven MPE, an important cancer phenotype.
In recent years, inflammation was established as a conditional tumor promoter51. IL-1α/β are important components of the tumor microenvironment that stimulate tumor invasiveness and angiogenesis52. Myeloid-derived IL-1β is implicated in the resistance to NF-κΒ inhibitors and IL-1β antagonism yielded beneficial effects in a mouse model of KRAS-mutant pancreatic cancer53,54. We found previously that IL-1α/β are present in human and experimental MPE and that MPE-competent adenocarcinomas trigger myeloid cells to secrete IL-1β35. Here the mechanism of pleural IL-1β function in MPE promotion is elucidated: CCL2-attracted monocyte-released IL-1β fosters NF-κΒ activation of MPE-prone KRAS-mutant carcinomas by potentiating non-canonical NF-κB signaling via IKKα. Undoubtedly, IL-1β is not the sole NF-κΒ ligand expressed in the malignancy-affected pleural space: TNF, a known stimulator of canonical NF-κB signaling, is present in MPE and promotes disease progression9. However, TNF likely originates from tumor cells in MPE9 and non-specifically triggers NF-κB activation in any tumor type irrespective of its KRAS status and MPE competence, suggesting it functions as an autocrine growth factor across tumor types. On the contrary, IL-1α/β selectively fostered MPE competence of KRAS-mutant carcinomas, in agreement with previous reports of IL-1β-induced NF-κB activation independent from IKKβ55. Our findings explain how the tumor microenvironment fuels tumor NF-κΒ activity56 and link the pro-tumorigenic functions of IL-1β with KRAS mutations, setting a rationale for genotype-stratified future investigations on IL-1β functions and therapies in cancer.
Unbiased analyses identified cancer-elaborated CXCL1/PPBP, potent myeloid cell chemoattractants that drive inflammation and metastasis via CXCR1/CXCR2 on host cells57,58, as the transcriptional targets of IL-1β-fostered KRAS-IKKα addiction. Indeed, Cxcl1 expression was downregulated by Kras or Chuk silencing and IL-1β induced Cxcl1 expression by two different cancer cell lines and Ppbp by LLC cells (MC38 cells do not express Ppbp25). Our experiments using CXCR1- and CXCR2-deficient mice support that pleural tumor cell-secreted CXCL1/PPBP is cardinal for MPE and are in line with a previous study demonstrating increased production of CXCL1 by tumor cells during human MPE development that mobilizes regulatory T cells59.
In addition to the mechanistic insights into host environment-fostered co-option of IKKα activity by mutant KRAS, our data bear therapeutic implications for KRAS inhibitors39. KRAS is notoriously undruggable, and proteasome and IKKβ inhibitors have yielded suboptimal results in mice and men with cancer. Focusing on lung cancer, a tumor with high KRAS mutation frequency60, bortezomib has shown poor efficacy in clinical trials61. In animal models of lung cancer, bortezomib and IKKβ inhibitors caused resistance or paradoxical tumor promotion via development of secondary mutations, NF-κB inhibition in myeloid cells, or enhanced IL-1β secretion by tumor-associated neutrophils through an unknown mechanism15,16,53. We show how KRAS-mutant cancer cells utilize myeloid-IL-1β in order to activate IKKα and alternative NF-κB signaling and to by-pass IKKβ canonical NF-κB dependence. We provide proof-of-concept data that KRAS-mutant cancer cells can be targeted by combined inhibition of KRAS and HSP90/IKKα/IKKβ signaling, a strategy that blocks IL-1β-inducible oncogenic NF-κB activation and in vivo MPE development, a cancer phenotype that requires mutant KRAS-potentiated, IL-1β-induced IKKα activity. These results challenge the prevailing focus on IKKβ for the development of anti-tumor drugs and establish IL-1β and IKKα as important targets in KRAS-mutant tumors.
In conclusion, we show that KRAS-mutant cancer cells use host IL-1β to sustain IKKα-mediated non-canonical NF-κB activity responsible for MPE development and primary drug resistance. We identify CXCL1/PPBP as effectors of MPE downstream of KRAS/IKKα addiction. Finally, we provide proof-of-concept data suggesting that KRAS/IKKα addiction may occur in human cancers and may be targeted by combined KRAS/IKKα inhibition.
All mouse experiments were prospectively approved by the Veterinary Administration of Western Greece (approval # 276134/14873/2) and were conducted according to Directive 2010/63/EU (http://eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex%3A32010L0063).
D-Luciferin was from Gold Biotechnology (St. Louis, MO); lentiviral shRNA and puromycin from Santa Cruz (Dallas, TX); 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay and Hoechst 33528 from Sigma-Aldrich (St. Louis, MO); mouse gene ST2.0 microarrays and relevant reagents from Affymetrix (Santa Clara, CA); recombinant cytokines and growth factors from Immunotools (Friesoythe, Germany); NF-κB-binding ELISA from Active Motif (La Hulpe, Belgium); bortezomib, IMD-0354, 17-DMAG, and deltarasin from Selleckchem (Houston, TX); G418 from Applichem (Darmstadt, Germany); IL-1β and CXCL1 ELISA from Peprotech (London, UK); and primers from VBC Biotech (Vienna, Austria). Primers, antibodies, and lentiviral shRNA pools are listed in Supplementary Tables 6–8.
LLC, B16F10, PANO2, and A549 cells were from the National Cancer Institute Tumor Repository (Frederick, MD); MC38 cells were a gift from Dr. Barbara Fingleton (Vanderbilt University, Nashville, TN)34,35, and AE17 cells from Dr. YC Gary Lee (University of Western Australia, Perth, Australia)11,25. All cell lines were cultured at 37 °C in 5% CO2–95% air using Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal bovine serum, 2 mM L-glutamine, 1 mM pyruvate, 100 U/mL penicillin, and 100 mg/mL streptomycin. Cell lines were tested annually for identity by short tandem repeats and for MycoplasmaSpp. by PCR. For in vivo injections, cells were harvested using trypsin, incubated with Trypan blue, counted in a hemocytometer, and 95% viable cells were injected intrapleurally8,11,34,35.
Mouse models and drug treatments
C57BL/6 (#000664), B6.129P2-Cxcr1tm1Dgen/J (Cxcr1−/−; #00582036), B6.129 S2(C)-Cxcr2tm1Mwm/J (Cxcr2+/−; #00684837), B6;129S-Tnftm1Gkl/J (Tnf−/−; #00300832) (Jackson Laboratory, Bar Harbor, ME), and Il1btm1Yiw (Il1b−/−; MGI #215739631) mice were bred at the Center for Animal Models of Disease of the University of Patras. Male and female experimental mice and littermate controls were sex, weight (20–25 g), and age (6–12 weeks) matched. For MPE induction, mice received 150,000 cancer cells in 100 μL PBS intrapleurally. Mice were observed continuously till recovery and daily thereafter and were sacrificed when moribund (13–14 days post-tumor cells) for survival and pleural fluid analyses. Mice with pleural fluid volume ≥100 μL were judged to have a MPE and were subjected to pleural fluid aspiration, whereas animals with pleural fluid volume <100 μL were judged not to have a MPE and were subjected to pleural lavage. Injection, harvest, and sample handling are described elsewhere8,9,10,11,34,35. Drug treatments were initiated 5 days post-tumor cells and consisted of daily intraperitoneal injections of 100 μL PBS containing no drug, deltarasin39, 17-DMAG28, or both at 15 mg/kg.
pNGL, pΙκΒαDN, and pCAG.LUC (#74409) have been described elsewhere8,25,33. Lentiviral shRNA pools (Santa Cruz) are described in Supplementary Table 8. A pMIGR1-based (#27490) bicistronic retroviral expression vector was generated by replacing eGFP sequences with puromycin resistance gene (#58250). KrasG12C,Chuk, Ikbkb, Ikbke, and Tbk1 cDNAs were cloned via reverse transcriptase-PCR (RT-PCR) from LLC or MC38 RNA using specific primers (Supplementary Table 6) and were subcloned into peGFP-C1 (Takara, Mountain View, CA). eGFP, eGFP.KrasG12C, eGFP.Chuk, eGFP.Ikbkb, eGFP.Ikbke, and eGFP.Tbk1 cDNAs were subcloned into the new retroviral expression vector (#58249, #64372,# 87033, #58251, #87444, and #87443, respectively). Retroviral particles were obtained by co-transfecting HEK293T cells with retroviral vectors, pMD2.G (#12259), and pCMV-Gag-Pol (Cell Biolabs, San Diego, CA) at 1.5:1:1 stoichiometry using CaCl2/BES. After 2 days, culture media were collected and applied to cancer cells. After 48 h, media were replaced by selection medium containing 2–10 μg/mL puromycin. Stable clones were selected and subcultured11. For stable plasmid/shRNA transfection, 105 tumor cells in six-well culture vessels were transfected with 5 μg DNA using Xfect (Takara), and clones were selected by G418 (400–800 μg/mL) or puromycin (2–10 μg/mL).
In vitro cancer cell proliferation was determined using MTT assay. Nuclear extracts were assayed for RelA, RelB, c-Rel, P50, and P52 DNA-binding activity using a commercially available ELISA kit (Transam, Active Motif, Belgium). All cellular experiments were independently repeated at least thrice.
Living cells and mice were imaged 0, 4, 8, 24, and 48 h after cellular treatments and 0 h, 4 h, and 12–14 days after pleural delivery of pNGL-expressing cells on a Xenogen Lumina II (Perkin-Elmer, Waltham, MA) after addition of 300 μg/mL D-luciferin to culture media or isoflurane anesthesia and delivery of 1 mg intravenous D-luciferin to the retro-orbital veins8,9,10,11,16,25,34,35. Data were analyzed using Living Image v.4.2 (Perkin-Elmer).
qPCR and microarray
RNA was isolated using Trizol (Invitrogen, Carlsbad, CA) and RNAeasy (Qiagen, Hilden, Germany) was reverse transcribed using Superscript III (Invitrogen), and RT-PCR or qPCR was performed using SYBR Green Master Mix in a StepOnePlus (Applied Biosystems, Carlsbad, CA) and specific primers (Supplementary Table 6). Ct values from triplicate qPCR reactions were analyzed by the 2–ΔΔCT method62 relative to Gusb mRNA levels. For microarray, RNA was extracted from triplicate cultures of 106 cells. Five micrograms pooled total RNA were quality tested on an ABI 2000 (Agilent Technologies, Sta. Clara, CA), labeled, and hybridized to GeneChip Mouse Gene 2.0 ST arrays (Affymetrix, St. Clara, CA). For analysis of differential gene expression (ΔGE) and unsupervised hierarchical clustering, Affymetrix Expression and Transcriptome Analysis Consoles were used.
LLC cells were treated with PBS or 1 nM IL-1β, and 30 min later, cells were fixed sequentially with 2 mM di(N-succinimidyl) glutarate (Sigma) and 1% formaldehyde (Sigma) and quenched with 0.125 M glycine, followed by lysis with 1% sodium dodecyl sulfate (SDS), 10 mM EDTA, and 50 mM Tris pH 8. Sonication was performed in a Bioruptor (Diagenode) for 40 cycles (30 s on/off) power settings high), using 3 × 106 cells; 20 μg of chromatin was precipitated with 5 μg of RelA, RelB, IKKα, or IKKβ antibody or a mouse control immunoglobulin G (IgG). Immunoprecipitates were retrieved with 50 μl of magnetic Dynabeads conjugated to protein G (Invitrogen) and subjected to quantitative real-time PCR (Applied Biosystems StepOne), using the Kapa SYBR Fast qPCR Kit (KapaBiosystems, KK4605) for amplification of the Cxcl1 promoter or Gusb as control. The sequences of the primers used for Cxcl1 promoter are: 5′-ATACAGCAGGGTAGGGATGC, 3′-TTGCCAACTGTTTTTGTGG. The sequences of the primers used for Gusb are: 5′-TTACTTTAAGACGCTGATCACC, 3′-ACCTCCAAATGCCCATAGTC.
BM cell derivation and transfer
For adoptive BM replacement, Il1β−/− mice (C57BL/6 background) received 10 million BM cells flushed from the femurs and tibias of C57BL/6, Tnf−/−, or Il1β−/− donors (C57BL/6 background) intravenously 12 h after total-body irradiation (1100 Rad)11,25,34,35. One mouse in each experiment was not engrafted (sentinel) and was observed till moribund between days 5 and 15 post-irradiation. The mice were left to be engrafted for 1 month, when full BM reconstitution is complete, before experimental induction of pleural carcinomatosis via intrapleural injection of LLC cells. For BM cell retrieval, BM cells were flushed from C57BL/6 femurs and tibias using full DMEM and were simply cultured in full culture media (the same used for cancer cell line cultures), supplemented with 20 ng/ml M-CSF or G-CSF in order for cells to differentiate to monocytes or neutrophils, respectively. Supernatants and cytocentrifugal specimens were obtained at day 0 for undifferentiated cells, day 2 for neutrophils, and at day 6 for monocytes/macrophages.
Nuclear and cytoplasmic extracts were prepared using the NE-PER Extraction Kit (Thermo, Waltham, MA), separated by 12% SDS polyacrylamide gel electrophoresis, and electroblotted to polyvinylidene difluoride membranes (Merck Millipore, Darmstadt, Germany). Membranes were probed with specific antibodies (Supplementary Table 7) and were visualized by film exposure after incubation with enhanced chemiluminescence substrate (Merck Millipore, Darmstadt, Germany).
Electrophoretic mobility shift assay (EMSA)
Nuclear extracts were prepared using the NE-PER Extraction Kit. Proteins (10 μg) were incubated with NF-κB biotin-labeled probe using a commercially available non-radioactive EMSA Kit (Signosis Inc, Santa Clara, USA). DNA–protein complexes were electrophoresed in a prerinsed 6.5% polyacrylamide gel, transferred to a positively charged nylon membrane, and were visualized by film exposure after incubation with enhanced chemiluminescence substrate. For gel shift reactions, proteins were incubated with the specific antibody for 1 h at 4 °C before probe incubation. The antibodies used for observing the supershifted bands were RelA and RelB. IgG antibody served as negative control for super-shift assays.
For immunofluorescence, cells were fixed in 4% paraformaldehyde overnight at 4 °C and were labeled with the indicated primary antibodies (Supplementary Table 7) followed by incubation with fluorescent secondary antibodies (Invitrogen, Waltham, MA; Supplementary Table 7). Cells were then counterstained with Hoechst 33258 (Sigma-Aldrich, St. Louis, MO) and mounted with Mowiol 4-88 (Calbiochem, Gibbstown, NJ). For isotype control, the primary antibody was omitted. Fluorescent microscopy was carried out on an AxioObserver.D1 inverted microscope (Zeiss, Jena, Germany) connected to an AxioCam ERc 5 s camera (Zeiss), and digital images were processed with the Fiji academic imaging freeware63.
Sample size was calculated using G*power (http://www.gpower.hhu.de/)64 assuming α = 0.05, β = 0.05, and d = 1.5, tailored to detect 30% differences between means with 20–30% SD spans, yielding n = 13/group. Animals were allocated to groups by alternation (treatments or cells) or case–control-wise (transgenic animals). Data acquisition was blinded on samples coded by non-blinded investigators. No data were excluded. All data were examined for normality by Kolmogorov–Smirnof test and were normally distributed. Values are given as mean ± SD. Sample size (n) refers to biological replicates. Differences in means were examined by t-test and one-way or two-way ANOVA with Bonferroni post-tests, in frequencies by Fischer’s exact or χ2 tests, and in Kaplan–Meier survival estimates by log-rank test, as appropriate. P-values are two-tailed. P < 0.05 was considered significant. Analyses and plots were done on Prism v5.0 (GraphPad, La Jolla, CA).
All new plasmids have been deposited at the Addgene plasmid repository (https://www.addgene.org/search/advanced/?q=stathopoulos) and their IDs (#) are given in the text. Microarray data are available at the GEO (http://www.ncbi.nlm.nih.gov/geo/; Accession IDs: GSE93369 and GSE93370). The authors declare that all the other data supporting the findings of this study are available within the article and its supplementary information files and from the corresponding authors upon reasonable request.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work was supported by European Research Council 2010 Starting Independent Investigator and 2015 Proof of Concept Grants (260524 and 679345, to GTS), by European Respiratory Society 2013 Romain Pauwels Research Award (to G.T.S.), by a Hellenic Association for Molecular Cancer Research Award 2015 (to A.M.), by a Hellenic Thoracic Society Research Award 2014 (to M.V.), and by an Immunotools Award 2014 (to A.D.G.). The authors thank the University of Patras Center for Animal Models of Disease for experimental support.