The Prosurvival IKK-Related Kinase IKKϵ Integrates LPS and IL17A Signaling Cascades to Promote Wnt-Dependent Tumor Development in the Intestine Free

Colorectal cancer results from multiple genetic mutations and inflammatory processes (1). Somatic mutations associated with 80% of colorectal cancer cases target the adenomatous polyposis coli (APC) tumor suppressor gene, which leads to β-catenin activation, followed by additional mutations in K-Ras, PI3K3CA, and TP53 among others as tumors develop (2, 3).

The majority of colorectal cancer cases have no preexisting inflammation but nevertheless displays tissue infiltration by inflammatory cells, which is referred to as “tumor-elicited inflammation” (4, 5). Those infiltrates include CD4⁺ T-helper 1 (Th1) and CD8⁺ cytotoxic T cells (CTL), tumor-associated macrophages (TAM), and T-helper interleukin 17 (IL17)–producing (Th17) cells. The tumor-promoting functions of TAMs and T lymphocytes are mediated through the secretion of cytokines. TAMs produce IL23, which enhances tumor-promoting inflammatory processes through IL17A synthesis by Th17 cells and also suppresses the adaptive immune surveillance by reducing CD8⁺ CTL cell infiltration in tumors (4, 6–8). In turn, IL17A triggers MAPKs and NF-κB activations in intestinal epithelial cells (IEC) to support early tumor growth (9).

The establishment of a tumor microenvironment relies on transcription factors such as NF-κB (10, 11). IκB-kinase (Ikk) β-dependent NF-κB activity in IECs promotes cell survival and drives the expression of proinflammatory cytokines in myeloid cells to link inflammation to cancer (12). In addition, NF-κB signaling in IECs also cooperates with β-catenin to facilitate the crypt stem cell expansion (13).

Both NF-κB and Stat3 transcription factors are activated by cytokines through parallel signaling pathways in solid tumors (14). Similar to NF-κB, Stat3 controls the expression of genes involved in cell survival, proliferation, and immunity. IL6, whose expression relies on NF-κB in lamina propria myeloid cells, protects premalignant IECs from apoptosis through Stat3 activation in a model of colitis-associated cancer (15, 16). IL23 signaling also promotes Stat3 phosphorylation in Apc-mutated IECs through IL17A production by Th17 cells (7).

Constitutive β-catenin activation and/or Apc loss in the intestinal epithelium cause the loss of epithelial barrier function, an early event in intestinal tumorigenesis (7). As a result, commensal bacteria infiltrate the stroma and lead to tumor-associated inflammation (17). Bacterial products are sensed by Toll-like receptors (TLR) such as TLR4, which promotes colitis-associated cancer (18).

TLR signaling triggers IKKβ/NF-κB activation, leading to synthesis of proinflammatory cytokines and the phosphorylation of IKK-related kinases TBK1 and IKKϵ to induce type I interferons synthesis through IRF3 (19–21). IKKϵ is believed to play key roles in cancer by targeting multiple substrates, many of which act in NF-κB–dependent pathways (22–25). Both TBK1 and IKKϵ also directly phosphorylate AKT/protein kinase B (26, 27). So far, it remains to be demonstrated that IKKϵ acts as an oncogenic kinase in vivo.

Here we report that LPS and IL17A-dependent signaling pathways converge to Ikkϵ to promote Wnt-dependent tumor development in IECs in vivo. These pathways drive the expression of proinflammatory cytokines, anti-microbial peptides, and chemokines, the latter recruiting macrophages to support IL23 and IL17A synthesis and subsequent Stat3 activation in transformed IECs. Ikkϵ also promotes cell survival in these cells by limiting TNF- and caspase-8–dependent apoptosis. The establishment of an inflammatory tumor-promoting microenvironment by Ikkϵ thus relies on the activation of signaling pathways distinct from the NF-κB–dependent cascades.

Villin-Cre-ER^T2 Ctnnb1^+/lox(ex3) (β-cat^c.a.) mice were previously described (28, 29). Villin-Cre-ER^T2Ctnnb1^+/lox(ex3)mice were gavaged 5 consecutive days with 1 mg tamoxifen (Sigma) to induce β-catenin activation in enterocytes as described previously (30). Both Apc^+/min and Ikkϵ^KO mouse strains were from Jackson Laboratories (Bar Harbor, ME). For antibiotics treatments, 0.5 g ciprofloxacin, 1 g ampicillin, and 0.5 g metronidazole per liter were added in the drinking water 1 week before tamoxifen administration. All mice used were 8 to 16 weeks old when started with experiments (except for the Apc^+/min survival experiments) and littermate controls were used. All procedures were approved by the local Ethical Committee of the University of Liege.

Bone marrow transplantation and bone marrow cell isolation were done as described previously (30). Minor changes are described in the Supplementary data section.

Intestinal crypts from Apc^+/min-Ikkϵ^−/− and Apc^+/min-Ikkϵ^+/+ mice were isolated and cultured as described (31). Stimulations of ex vivo organoid cultures with IL17A and LPS were carried out as described (9).

Mice were injected intraperitoneal with 100 mg/kg BrdU (Sigma) 90 minutes before their sacrifice and paraffin sections of duodenum tissues were stained using anti-BrdU antibody (RPN201; Amersham Biosciences/GE Healthcare) to quantify proliferating nuclei. Proliferative rates were determined by the ratio of average of positive cells in 10 crypts or by the ratio of positive cells to total cells in three proliferative cryptic area (where individual crypts could no longer be identified) per sample. Apoptotic cells in a given tissue section were determined histologically by TUNEL assay using an ApoAlert DNA Fragmentation Assay Kit (BD Biosciences Clontech).

SW480, HCT116, and HT-29 cells were obtained from ATCC in 2009. These cells were characterized by ATCC, using a comprehensive database of short tendem repeat (STR) DNA profiles. Frozen aliquots of freshly cultured cells were generated and experiments were done with resuscitated cells cultured for less than 6 months. Cell culture reagents, cytokines, and kinase inhibitors are described in the Supplementary data.

Infections of Lenti-X 293T cells (Clontech) using lentiviral constructs described in the Supplementary data were carried out as previously described (32).

Isolation of enterocytes and Western blot analyses were performed as described previously (30). Paraffin sections (4 μm) and Western blots were stained using antibodies described in the Supplementary data section. For Immunoprecipitations, anti-TANK, -NAP1 and -IgG (negative control) antibodies were coupled covalently to a mixture of Protein A/G-Sepharose (see the Supplementary data for details). Immunoprecipitations were done as previously described (33).

Total RNAs were extracted and subjected to real-time PCR analyses as described (32). Primer sequences are available on request. Gene expression profiling of tumor tissues was carried out by RNA-Seq analysis. Both sample preparations and sequencing were performed at the GIGA transcriptomic facility (GIGA, University of Liege, Liege, Belgium). Methods to check total RNAs integrity, to carry out RNA-Seq expression analyses and for data analysis are described in the Supplementary Data.

Sample tissues were fixed with the standard procedures using 4% PFA (1 hour) and sucrose (15% 6 hours; 30% o/n) at 4°C and frozen in OCT freezing medium by the use of supercooled isopropanol-dry ice mixture and stored at −80°C up to 6 month. Frozen samples were cut 5 to 10 μm with a cryostat microtome at −20°C on superfrost slides. In situ hybridization was carried out using the protocol provided by the manufacturer (RNAscope Multiplex Assay System; Advanced Cell Diagnostics Inc.).

Control or IKKϵ-depleted SW480 cells were pretreated or not with the pan-caspase inhibitor Z-VAD-FMK (Promega; 20 μmol/L) for 1 hour and subsequently untreated or stimulated with TNF (100 ng/mL)/cycloheximide (CHX; 50 μg/mL) for up to 8 hours. The quantification of apoptosis was done as previously described (32).

Data are expressed as mean ± SEM. Differences were analyzed by Student t-test or log-rank test (for Kaplan–Meier survival graphs of animal models) using Prism5 (GraphPad Software). The P values ≤ 0.05 (covering 95% confidence intervals) were considered significant (30).

We investigated whether Ikkϵ inactivation impacts on tumor formation in the Apc^+/min mouse model, which spontaneously develops adenocarcinomas due to constitutive Wnt signaling (34). Inactivating Ikkϵ in Apc^+/min mice significantly enhanced survival (226 days vs. 143 days, P < 0.001 in Apc^+/min-Ikkϵ^−/− and Apc^+/min-Ikkϵ^+/+ mice, respectively) due to a decreased tumor incidence in distinct parts of the intestine (Fig. 1A–D). As a result, Apc^+/min-Ikkϵ^−/− mice did not suffer from anemia and splenomegaly was less dramatic (Fig. 1E and F, respectively). Ikkϵ deletion slightly impaired cell proliferation in tumors but not in normal intestinal crypts in Apc^+/min mice (Fig. 1G). Consistently, pErk1/2 levels and, to some extent, cell proliferation as assessed by BrdU staining, were decreased in Apc^+/min-Ikkϵ^−/− mice (Fig. 1H). Ikkϵ did not control cell proliferation in a cell-autonomous manner as ex vivo organoïds generated with intestinal crypts from Apc^+/min-Ikkϵ^−/− and Apc^+/min-Ikkϵ^+/+ mice showed similar cell growth (Supplementary Fig. S1A). Ikkϵ phosphorylation on serine 172 was higher in intestinal tumors than in normal adjacent tissues from Apc^+/min mice, as were protein levels of Tank, one of the Ikkϵ scaffold proteins (Supplementary Fig. S1B).

We took advantage of the tamoxifen-inducible β-cat^c.a. mouse model, which expresses truncated and stabilized β-catenin protein in IECs (28). Intestinal crypts rapidly expand because of constitutive Wnt signaling and loss of differentiated IECs, with β-cat^c.a. mice succumbing to disease within 4 weeks of age because of continuous adenoma formation (13). Ikkϵ mRNA expression was detected by in situ hybridization both in transformed IECs and in inflammatory cells (Fig. 2A). Ikkϵ inactivation in β-cat^c.a. mice also extended their survival (37 vs. 30 and 28.5 days, P = 0.0044 in β-cat^c.a.-Ikkϵ^−/−, β-cat^c.a.-Ikkϵ^+/−, and β-cat^c.a.-Ikkϵ^+/+ mice, respectively; Fig. 2B). Ikkϵ was essential for Akt, Mek1/2, Erk1/2, and Msk1 activation and for Creb1 (a Msk1 substrate) and Stat3 phosphorylation but not for Wnt-dependent Pdk1 phosphorylation (Fig. 2C and Supplementary Fig. S2, respectively). Tank expression was also higher upon constitutive Wnt signaling (Fig. 2C). Enhanced pAkt and pErk1/2 levels in tumors from β-cat^c.a.-Ikkϵ^+/+versus β-cat^c.a.-Ikkϵ^−/− mice were confirmed by Immunohistochemistry (Fig. 2D). Thus, Ikkϵ promotes the activation of multiple oncogenic pathways in transformed IECs to support tumor development.

As Wnt-driven tumor development was impaired upon Ikkϵ deficiency, we next explored whether this resulted from enhanced cell death. Ikkϵ inactivation in Apc^+/min mice enhanced the number of TUNEL⁺ cells in small intestinal tumors (Fig. 3A). Consistently, IKKϵ-deficient and p53-mutated colon cancer SW480 cells were sensitized to TNF + CHX-dependent cell death, as judged by FACS analysis (Fig. 3B). IKKϵ-deficient SW480 cells were dying of apoptosis as the caspase inhibitor Z-VAD-FMK blocked TNF/CHX-dependent cell death (Fig. 3B). Consistently, IKKϵ-deficient SW480 cells subjected to TNF/CHX stimulation showed increased levels of cleaved forms of caspases 3/8 and RIPK1, a caspase-8 substrate (Fig. 3C). Cell death in IKKϵ-deficient SW480 cells did not result from decreased NF-κB activity as the TNF-dependent IκBα degradation and p65 phosphorylation were unchanged (Supplementary Fig. S3). The TNF-dependent activation the other IKK-related kinase TBK1 was potentiated upon IKKϵ deficiency, suggesting a compensatory mechanism (Supplementary Fig. S3). Enhanced cell apoptosis was also seen upon TNF/CHX stimulation in other IKKϵ-deficient colon cancer cell lines showing constitutive Wnt signaling, namely in p53-mutated HT29 and in p53-proficient HCT116 cells (Supplementary Fig. S4A–S4C). Therefore, IKKϵ protects from TNF-dependent apoptosis through p53- and NF-κB–independent mechanisms in transformed IECs.

We next characterized the IKKϵ-dependent pathways in colon cancer cells. Constitutive phosphorylation of ERK1/2 relied on IKKϵ in differentiated HT-29 cells (Supplementary Fig. S5A). Moreover, Lipopolysaccharide (LPS)-induced phosphorylation of ERK1/2 was defective in IKKϵ-depleted SW480 cells (Supplementary Fig. S5B). Thus, our data link IKKϵ to ERK1/2 activation in transformed IECs.

IL17A signals in transformed IECs and IKKϵ is activated by IL-17A in airway epithelial cells (9, 35, 36). Therefore, we assessed if IL17A promotes Wnt-dependent tumor development through IKKϵ. IL-17A alone or in combination with LPS triggered IKKϵ phosphorylation in ex-vivo organoid cultures of transformed IECs (Supplementary Figs. S6A and 4A, respectively). IKKϵ deficiency in ex-vivo organoïd cultures from Apc^+/min mice as well as in SW480 cells impaired AKT, MEK1, p38 and ERK1/2 activation upon stimulation with both LPS and IL17A (Fig. 4A and B, respectively). IKKϵ constitutively bound TANK but not with NAP1, another scaffold protein, in unstimulated or IL17A-treated SW480 cells (Supplementary Fig. S6B). TANK deficiency also severely impaired AKT, MEK1/2, ERK1/2 and p38 activation in cells stimulated with both LPS and IL17A (Supplementary Fig. S7). Therefore, the IKKϵ-TANK complex integrates LPS- and IL17A-dependent cascades to activate multiple kinases.

To identify target genes induced through Ikkϵ, RNA-Seq analysis was done using total RNAs from duodenal samples of β-cat^c.a.-Ikkϵ^+/+ and β-cat^c.a.-Ikkϵ^−/− mice, 0 and 22 days following tamoxifen administration and WebGestalt GSAT enrichment analysis was carried out. A remarkable number of Ikkϵ-regulated genes were involved in the immune response (Supplementary Figs. S8A and S8B). GSEA analysis further highlighted defective interferon and innate immune responses in β-cat^c.a.-Ikkϵ^−/− mice 22 days after tamoxifen administration (Fig. 5A). Consistently, a heatmap representation of gene expression demonstrated the lack of upregulation of immune response genes in duodenal extracts from β-cat^c.a.-Ikkϵ^−/− mice compared to β-cat^c.a.-Ikkϵ^+/+ mice (Fig. 5A). These candidates included Fcamr (Fc receptor, IgA, IgM, high affinity), Aicda (Activation-induced cytidine deaminase), Fcrl5 (Fc Receptor-like 5), Reg3β/γ (regenerating islet-derived protein 3-beta/gamma), IL23A, and IL17A (Fig. 5A). Among the 236 downregulated transcripts in β-cat^c.a.-Ikkϵ^−/− tissue, many were proinflammatory genes (Supplementary Fig. S9A). The most prominent candidates were Ly6a/c, Retnlb, Cxcl9, C3, and Nlrc5 (Supplementary Fig. S9A). In addition, RNA-Seq data revealed numerous chemokines whose expression required Ikkϵ upon constitutive Wnt activation in the intestine. Indeed, mRNA levels of Cxcl12 (also referred as to SDF-1α), Cxcl5, and Cxcl1 were severely downregulated upon Ikkϵ inactivation (Fig. 5B). A chemokine/cytokine protein array confirmed the decreased expression of chemokines, including Cxcl9, Cxcl11, Cxcl12, G-Csf, and cytokines (IL7 and IL17A) in whole duodenal extracts from β-cat^c.a.-Ikkϵ^−/− mice (Fig. 5C). In contrast, IL1rα was upregulated in these conditions (Fig. 5C). Ikkϵ deficiency impairs Stat3 phosphorylation in transformed IECs (Fig. 2C), possibly because of an impaired IL17A production rather than from an intrinsic signaling defect. IL17A controls STAT3 phosphorylation through IL6. IKKϵ was dispensable for IL6-dependent STAT3 activation in SW480 cells, which demonstrates that Ikkϵ controls Stat3 phosphorylation through a paracrine mechanism involving IL17A production in β-cat^c.a.mice (Supplementary Fig. S9B).

IL17A synergizes with TNF to induce Cxcl1 expression through IKKϵ in airway epithelial cells (36). The combination of LPS and IL17A failed to induce Cxcl1 expression in both ex vivo organoid cultures from Apc^+/min-Ikkϵ^−/− mice and in IKKϵ-depleted SW480 cells, as was the induction of Cxcl1 expression by LPS or IL17A alone (Fig. 5D and Supplementary Fig. S9C, respectively). Therefore, LPS and IL17 signals converge to IKKϵ to induce Cxcl1 expression in transformed IECs. AKT or ERK1/2 inhibitors (perifosine, GSK690693, and PD98059, respectively) interfered with the induction of Cxcl1 expression (Supplementary Fig. S9D). Therefore, IL17A and LPS signal through both AKT and ERK1/2 to induce Cxcl1 expression in colon cancer cells.

Consistent with a role of Ikkϵ in chemokines production, the number of macrophages infiltrating the tumor stroma of β-cat^c.a./Ikkϵ^−/− mice was significantly decreased, as evidenced by anti-F4/80 and CD163 immunofluorescence (IF) analysis (Fig. 5E and F). Reduced expression of both F4/80 and CD163 upon Ikkϵ deficiency was also revealed through real-time PCR analysis (Fig. 5G). Yet, Ikkϵ deletion did not impact on macrophages polarization as both M1 and M2 markers were similarly downregulated in duodena of β-cat^c.a./Ikkϵ^−/− mice (Fig. 5H). Cd4, Cd8a, and Cd68 mRNA levels were also downregulated in these samples (Fig. 5G). Therefore, Ikkϵ in IECs promotes the recruitment of macrophages to the tumor stroma through the expression of macrophage-attracting chemokines.

To assess whether Ikkϵ expression in hematopoietic cells also contributed to intestinal tumor development, bone marrow cells from β-cat^c.a.-Ikkϵ^+/+ or β-cat^c.a.-Ikkϵ^−/− mice were isolated and transplanted intravenously to irradiated β-cat^c.a.-Ikkϵ^−/− or β-cat^c.a.-Ikkϵ^+/+ mice. Mice were kept for a month for the regeneration of immune cells before tamoxifen administration (Supplementary Fig. S10). Irradiated β-cat^c.a.-Ikkϵ^−/− mice transplanted with bone marrow from Ikkϵ^−/− mice survived longer than β-cat^c.a.-Ikkϵ^+/+ mice transplanted with bone marrow from Ikkϵ^+/+ mice (33.5 days versus 27 days, respectively), which confirms the contribution of IKKϵ in Wnt-driven tumor development (Supplementary Fig. S10). β-cat^c.a.-Ikkϵ^+/+ mice transplanted with bone marrow from Ikkϵ^−/− mice also survived longer (34.6 days), suggesting a contribution of Ikkϵ expression in hematopoietic cells in the observed phenotype. Yet, irradiated β-cat^c.a.-Ikkϵ^−/− mice transplanted with bone marrow from Ikkϵ^−/− or Ikkϵ^+/+ mice also showed a similar survival advantage (34 and 33.5 days, respectively) compared to irradiated β-cat^c.a.-Ikkϵ^+/+ mice transplanted with bone marrow from Ikkϵ^+/+ mice, which also highlight the key contribution of Ikkϵ expression in transformed IECs (Supplementary Fig. S10). These data highlight the contribution of Ikkϵ in both IECs and hematopoietic cells (possibly through IL17A production through an Ikkϵ-dependent pathway in Th17 cells) to support Wnt-driven tumor development in the intestine.

GSEA analysis also identified an enrichment of intestinal antimicrobial factors among Ikkϵ target genes in β-cat^c.a. mice (Fig. 6A). Indeed, both Reg3β/γ and Ang4, whose mRNA levels increased in transformed IECs from β-cat^c.a. mice, were downregulated upon Ikkϵ inactivation (Fig. 6A and B). The number of Paneth cells, the major source of antimicrobial factors, was intact in β-cat^c.a.-Ikkϵ^−/− mice (Supplementary Fig. S11). Yet, the number of visible antimicrobial factor releasing granules in each Paneth cell was severely decreased in intestinal crypts from β-cat^c.a.-Ikkϵ^−/− mice (Supplementary Fig. S11). Therefore, Ikkϵ deficiency interferes with Paneth cell differentiation. Of note, the goblet cell marker Muc1, whose expression increased upon constitutive Wnt signaling, was also decreased in Ikkϵ-deficient IECs from β-cat^c.a. mice (Fig. 6B). Moreover, mRNA levels of fucosyltransferase 2 (Fut2), an enzyme produced by innate lymphoid cells, which promotes epithelial fucosylation in the intestinal tract to protect from Salmonella typhirium infection (37), also severely increased upon Wnt activation but decreased in the absence of Ikkϵ (Fig. 6B). In addition, Ikkϵ was required for complement C3 expression in transformed IECs (Supplementary Fig. S12A). Therefore, Ikkϵ provides an inflammatory signature in IECs upon Wnt-dependent tumorigenesis in a cell-autonomous manner.

As LPS-dependent expression of complement C3 and activation of C/Ebpδ in mouse embryonic fibroblasts requires the transcriptional induction of Ikkϵ through NF-κB (38), we assessed C/Ebpδ expression in primary IECs from Ikkϵ^+/+ or Ikkϵ^−/− mice subjected or not to LPS stimulation. Complement C3 expression was strongly induced by LPS at the mRNA level and Ikkϵ deletion impaired its expression, especially after 4 and 8 hours of LPS stimulation in IECs (Supplementary Fig. S12B). Consistently, C/Ebpδ protein levels were also decreased in Ikkϵ-deficient IECs, with or without LPS stimulation and in IECs from β-cat^c.a.-Ikkϵ^−/− mice compared to β-cat^c.a.-Ikkϵ^+/+ mice (Supplementary Fig. S12C and S12D). Thus, Ikkϵ promotes C3 expression, by regulating C/Ebpδ levels in normal and transformed IECs.

Gut microbiota promotes tumor development in Apc^+/min mice (39). Moreover, commensal bacteria trigger TLR-dependent signaling pathways that converge on Ikkϵ (19). To assess whether bacterial products trigger TLRs-dependent Ikkϵ activation to provide the inflammatory tumor microenvironment, we subjected β-cat^c.a./Ikkϵ^+/+ mice to broad-spectrum antibiotics (Abx) to deplete commensal bacteria. Bacterial counts from feces of β-cat^c.a./Ikkϵ^+/+ mice validated the efficiency of antibiotics (Supplementary Fig. S13). Abx treatment prolonged mouse survival, probably by interfering with Ikkϵ, Akt, Msk1, and Stat3 activations (Fig. 7A and B). Therefore, bacterial products trigger the Ikkϵ-dependent activation of oncogenic pathways during Wnt-driven tumor development. We next assessed mRNA levels of pro-inflammatory cytokines and chemokines in whole duodenum from control versus Abx-treated β-cat^c.a./Ikkϵ^+/+ mice. Most candidate genes whose expression was decreased in β-cat^c.a./Ikkϵ^−/− mice also showed reduced expression in Abx-treated β-cat^c.a./Ikkϵ^+/+ mice (Fig. 7C). Also, similar to Ikkϵ deficiency, the expression of multiple Paneth cells markers significantly decreased upon Abx treatment in β-cat^c.a. mice (Fig. 7D). These data identified key Ikkϵ-dependent oncogenic pathways triggered by bacterial products that provide an inflammatory tumor microenvironment in the intestine showing constitutive Wnt signaling.

Here we define Ikkϵ as a LPS- and IL17A-activated kinase acting upstream of multiple pathways in transformed IECs, leading to the establishment of a proinflammatory environment in two mouse models of Wnt-driven intestinal tumorigenesis. Ikkϵ is also acting as a pro-survival kinase by limiting TNF- and caspase-8–dependent apoptosis in IECs showing constitutive Wnt signaling.

Ikkϵ counteracts TNF-dependent cell death, similarly to the prosurvival Ikkβ but through NF-κB-independent mechanisms as IκBα degradation and p65 phosphorylation by TNF remained intact in IKKϵ-deficient IECs. It is likely that the phosphorylation of multiple unknown IKKϵ substrates will provide prosurvival signals.

In addition to a prosurvival role, Ikkϵ acts as an oncogenic kinase by stimulating the recruitment of proinflammatory cells to support Wnt-driven tumorigenesis. Our bone marrow transplantation experiments highlight a dual function for Ikkϵ expression in both IECs and bone marrow-derived cells. This dual role is required to sustain a proinflammatory loop that supports tumor development, a loop initiated by Ikkϵ expression in transformed IECs (Supplementary Fig. S14). Th17 cells known to produce IL17A may critically rely on Ikkϵ to maintain this loop. Indeed, the key role of Ikkϵ in IL1β-driven Th17 maintenance supports this hypothesis (40). Removing Ikkϵ in transformed IECs or in bone marrow-derived cells disrupt this proinflammatory loop and tumor development is consequently delayed. Whether IKKϵ expression in cancer-associated fibroblasts also provide oncogenic signals deserves further investigation using conditional knockout mouse models.

Ikkβ is another proinflammatory molecule but mechanisms by which Ikkβ drives Wnt-dependent tumor initiation in the intestine are partially distinct. Ikkϵ is an Akt-activating kinase in Apc-mutated IECs whereas Ikkβ is not. Similarly, Erk1/2 is regulated by Ikkϵ but not by Ikkβ. Therefore, Ikkϵ provides a proinflammatory signature in transformed IECs, at least through some specific pathways distinct from those controlled by Ikkβ. Previous in vitro studies showed that Ikkϵ targets several substrates acting in NF-κB–activating cascades (23, 41). We show here that the oncogenic potential of Ikkϵ in transformed IECs mainly results from its capacity to provide a tumor microenvironment rather than from enhancing pro-proliferative cascades in a cell-autonomous manner.

Our data provide an in vivo demonstration that Ikkϵ promotes Akt activation in transformed IECs. The transcriptional program induced through the Ikkϵ–Akt pathway in Apc-mutated IECs remains unclear. One candidate could be Retn1b, which is upregulated in colon cancer, and protects against parasitic helminth infections by maintaining the colonic barrier function (42–44). Retn1b expression is induced through IL23 and Akt in intestinal goblet cells (45). Because Akt activation is Ikkϵ dependent in Apc-mutated IECs, Retn1b expression may be induced through this pathway. It is likely that CREB1, whose phosphorylation occurs through Akt and Msk1 (46), contributes to the induction of numerous Ikkϵ target genes. Similarly, C/Ebpδ is another transcription factor acting downstream of Ikkϵ that drives the expression of proinflammatory molecules such as complement C3.

Constitutive Stat3 activation cooperates with NF-κB to promote cell survival and proliferation in the intestine (14). The defective Stat3 phosphorylation profile seen upon Ikkϵ inactivation results from an impaired recruitment of macrophages in the tumor stroma rather than an epithelial cell-autonomous effect of Ikkϵ on Stat3. This defect causes decreased levels of Stat3-activating cytokines such as IL6 in whole duodenum from β-cat^c.a./Ikkϵ^−/− mice.

Multiple cytokines and chemokines show an Ikkϵ-dependent expression in our model of Wnt-driven tumor initiation. One of them is IL17A whose production was decreased upon Ikkϵ deficiency. Once synthesized, IL17A can establish a positive loop by re-activating Ikkϵ in transformed IECs. Consistently, IL17A or Ikkϵ deficiency in Apc^+/min mice similarly delays tumor development and also corrects splenomegaly (47). Therefore, signals from two distinct families of receptors, IL17RA and TLRs, converge to Ikkϵ to promote Wnt-dependent tumor development in the intestine. Few candidates such as IL1rα were upregulated in duodenum of β-cat^c.a./Ikkϵ^−/− mice, as similarly showed in a model of arthritis (48). As IL1rα antagonizes the function of IL1β, Ikkϵ may potentiate IL1β signaling by limiting IL1rα expression.

The recruitment of macrophages in the intestinal tumor stroma, but not their polarization, requires Ikkϵ. This is in sharp contrast with Ikkα whose kinase activity is required for Wnt-driven intestinal tumor development by negatively regulating the recruitment of Interferon γ (IFNγ)-producing M1-like myeloid cells (30). Therefore, Ikkϵ establishes an inflammatory signature to promote Wnt-driven tumor development through mechanisms distinct from those implying Ikkα and Ikkβ.

L.C. Heukamp has ownership interest in a NEO New Oncology and reports receiving a commercial research grant from Roche, Boehringer, and MSD. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S.I. Göktuna, K. Shostak, A. Ladang, A. Chariot

Development of methodology: S.I. Göktuna, H.-Q. Duong, A. Ladang, P. Close, I. Klevernic, A. Florin, F. Baron, S. Rahmouni, R. Büttner, A. Chariot

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.I. Göktuna, T.-L. Chau, L.C. Heukamp, B. Hennuy, H.-Q. Duong, A. Ladang, P. Close, F. Olivier, G. Ehx, M. Vandereyken, S. Rahmouni, R. Büttner, F. R. Greten, A. Chariot

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.I. Göktuna, K. Shostak, B. Hennuy, A. Ladang, G. van Loo, R. Büttner, A. Chariot

Writing, review, and/or revision of the manuscript: S.I. Göktuna, K. Shostak, I. Klevernic, L. Vereecke, R. Büttner, A. Chariot

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B. Hennuy, G. Ehx

Study supervision: A. Chariot

The authors thank Wouters Coppieters (GIGA Transcriptomic facility, ULG, Liege, Belgium) for RNA-Seq analyses, Sophie Dubois for her help in tail injections and the GIGA Imaging and Flow Cytometry Facility.

A. Chariot received grants from the Belgian National Funds for Scientific Research (F.N.R.S), TELEVIE, the Belgian Federation against cancer, the University of Liege (Concerted Research Action Program (BIO-ACET) and “Fonds Spéciaux” (C-11/03)), the “Centre Anti-Cancéreux”, the “Leon Fredericq” Fundation (ULg), and from the Walloon Excellence in Life Sciences and Biotechnology (WELBIO). P. Close and A. Chariot are Research Associate and Senior Research Associate at the F.N.R.S., respectively.

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.