NOD2 suppresses colorectal tumorigenesis via downregulation of the TLR pathways

SUMMARY

Although NOD2 is the major inflammatory bowel disease susceptibility gene, its role in colorectal tumorigenesis is poorly defined. Here, we show that Nod2-deficient mice are highly susceptible to experimental colorectal tumorigenesis independent of gut microbial dysbiosis. Interestingly, the expression of inflammatory genes and the activation of inflammatory pathways including NF-κB, ERK, and STAT3 are significantly higher in Nod2^−/− mouse colons during colitis and colorectal tumorigenesis, but not at homeostasis. Consistent with higher inflammation, there is greater proliferation of epithelial cells in hyperplastic regions of Nod2^−/− colons. In vitro studies demonstrate that while NOD2 activates the NF-κB and MAPK pathways in response to MDP, it inhibits TLR-mediated activation of NF-κB and MAPK. Notably, NOD2-mediated downregulation of NF-κB and MAPK is associated with the induction of IRF4. Taken together, NOD2 plays a critical role in the suppression of inflammation and tumorigenesis in the colon via downregulation of the TLR signaling pathways.

INTRODUCTION

Colorectal cancer (CRC) is one of the most common malignancies worldwide and the second leading cause of cancer-related death in Western countries (Jemal et al., 2006). The pathogenesis of CRC is a complex process that involves the accumulation of genetic alterations in the intestinal epithelium, transformation of normal epithelium into the neoplastic epithelium, proliferation of the neoplastic cells, and potentially metastasis. While genetic alterations in oncogenes or cancer suppressor genes are the primary trigger for the induction of colorectal tumorigenesis, inflammation plays a driving role in every step of the tumorigenesis process (Balkwill and Mantovani, 2001; Grivennikov et al., 2010). Therefore, patients with inflammatory bowel diseases (IBD), such as Crohn’s disease (CD) and ulcerative colitis (UC), are at elevated risk of developing CRC (Eaden et al., 2001; Jess et al., 2005; Lutgens et al., 2013).

Inflammation is regulated by multiple cell signaling pathways including the NF-κB and MAP Kinase (P38, ERK and JNK) signaling. The critical contribution of the NF-κB and MAPK pathways in colorectal tumorigenesis has been documented in multiple studies (Ben-Neriah and Karin, 2011; Greten et al., 2004; Lee et al., 2010). NF-κB and MAPK are primarily activated by the Toll-like receptors (TLRs) that recognize pathogen- or danger-associated molecular patterns (PAMPs and DAMPs) at the cell surface or in the endosomal compartment. Dysregulation of the expression and function of TLRs is therefore associated with the pathogenesis of IBD and CRC (Li et al., 2014). For example, TLR4 was shown to be overexpressed in human and murine inflammation-associated colorectal carcinogenesis, and TLR4-deficient mice were protected against colon carcinogenesis (Fukata et al., 2007; Tang et al., 2010). Deletion of MyD88, a downstream adaptor of most of the TLRs, prevented tumorigenesis in APC^min/+ mice (Rakoff-Nahoum and Medzhitov, 2007). On the other hand, MyD88^−/− mice were shown to be highly susceptible to chemical injury-induced colitis and colorectal tumorigenesis (Salcedo et al., 2010). Therefore, a balanced activation of the NF-κB and MAPK signaling pathways is essential for immune homeostasis and protection against CRC.

In addition to TLRs, many cytosolic pattern recognition receptors (PRRs) participate in the regulation of inflammation. Recently the cytosolic NOD-like receptors (NLRs) have emerged as critical players in intestinal inflammation and tumorigenesis (Chen et al., 2011; Chen et al., 2008; Hugot et al., 2001; Watanabe et al., 2005; Zaki et al., 2010; Zaki et al., 2011b). The NLR family member NOD2 is a major IBD susceptibility gene with 15–20% of IBD patients carrying mutations in NOD2 (Hugot et al., 2001; Ogura et al., 2001). Not surprisingly, many population-based studies have attempted to explore the association of NOD2 mutations with the pathogenesis of CRC. In 2004, Kurzawski, et al. first reported an association of the Nod2 3020insC single nucleotide polymorphism with the risk of CRC (Kurzawski et al., 2004). This observation was later supported by other clinical studies (Papaconstantinou et al., 2005; Roberts et al., 2006). Recently, two meta-analyses studies also suggested that polymorphisms in NOD2 are linked with CRC (Liu et al., 2014; Tian et al., 2010). However, experimental evidence of the role of NOD2 in CRC is limited and mechanistic understanding of the functional role of NOD2 in carcinogenesis is currently lacking.

NOD2 has been characterized as a sensor for MDP, a component of peptidoglycan present in both Gram-positive and Gram-negative bacteria (Girardin et al., 2003). It is composed of two N-terminal CARD domains, a central nucleotide binding domain, and a C-terminal leucine-rich repeat. MDP-stimulated NOD2 undergoes conformational changes leading to the interaction with RIP2 kinase through CARD-CARD homotypic interaction. Activated RIP2 subsequently stimulates NF-κB and MAPK (Strober et al., 2006), suggesting a critical role for NOD2/RIP2 signaling in innate host responses against many bacterial pathogens (Kobayashi et al., 2005; Philpott et al., 2014; Strober and Watanabe, 2011). However, it is increasingly evident that the physiological function of NOD2 likely extends beyond its response to MDP. For example, NOD2 was shown to downregulate TLR2 signaling (Watanabe et al., 2004; Watanabe et al., 2005). Sabbah et al. showed that NOD2 senses viral ssRNA leading to the activation of MAVS, which in turn activates NF-κB and induces IFNβ (Sabbah et al., 2009). NOD2 may also play a role in host defense via the regulation of autophagy and antimicrobial peptide production by Paneth cells (Cooney et al., 2010; Ogura et al., 2003; Petnicki-Ocwieja et al., 2009; Strober and Watanabe, 2011). Thus, the precise mechanism of NOD2-mediated regulation of intestinal inflammation and tumorigenesis is yet to be defined.

In this study, we examined the physiological function of NOD2 using the azoxymethane (AOM) plus dextran sodium sulfate (DSS) model of colorectal tumorigenesis. We observed that Nod2^−/− mice are susceptible to colorectal tumorigenesis with increased tumor burden and inflammatory responses in the colon. The susceptibility of Nod2^−/− mice to CRC was not associated with altered microbiota, but greater activation of major inflammatory signaling pathways including NF-κB, ERK, and STAT3. Further biochemical studies revealed that NOD2 inhibits TLR-mediated activation of the NF-κB and MAPK pathways. Notably, NOD2-mediated suppression of the NF-κB and MAPK pathways during TLR activation was associated the induction of IRF4, which was previously shown as a negative regulator of TLR activation (Negishi et al., 2005). These data provide experimental evidence of a genetic association of NOD2 in CRC and offer a mechanistic view of NOD2-mediated regulation of intestinal inflammation and tumorigenesis.

RESULTS

NOD2 deficiency predisposes mice to colitis-associated colorectal cancer

While mutations in NOD2 are genetically linked to IBD and CRC (Liu et al., 2014; Ogura et al., 2001; Papaconstantinou et al., 2005; Roberts et al., 2006), the underlying mechanism remains unknown. Therefore, we induced colorectal tumorigenesis in wild-type (WT) and Nod2^−/− mice with AOM/DSS (Figure S1 A), a widely used model for studying colitis-associated CRC. In the AOM/DSS model, repeated cycles of DSS administration results in chronic inflammation that supports the growth of neoplastic epithelium induced by AOM. We measured colitis phenotype by monitoring body weight changes and observed that Nod2^−/− mice lost a greater percentage of body weight than WT mice (Figure S1B). Mice were sacrificed 80 days after AOM injection to quantify the colonic tumor burden. Tumor burden in Nod2^−/− mouse colons was significantly higher than that of WT mice (Figure 1A and 1B), although tumor sizes of Nod2^−/− mice were not remarkably different (Figure 1C). Histopathological analysis of tumor-bearing colons showed that both WT and Nod2^−/− mice developed low-grade and high-grade dysplasia. However, the occurrence of high-grade dysplasia was significantly more in Nod2^−/− mice (Figure 1D and 1E). Notably, 30% of Nod2^−/− mice developed invasive carcinoma, whereas no invasive carcinoma was observed in WT mice (Figure 1D and 1F), suggesting that Nod2 deficiency leads to faster progression of tumorigenesis. WT and Nod2^−/− mice treated with a single dose of AOM didn’t develop any polyps (Figure S1C). Similarly, no visible polyps were identified in WT and Nod2^−/− mice treated with 3 cycles of DSS (without AOM). However, significantly shorter colons and increased inflammation were noted in Nod2^−/− compared to WT mice during chronic colitis (Figure S1D and S1E). To determine whether NOD2-mediated regulation of CRC is RIP2 dependent or not, we induced colorectal tumorigenesis in Rip2^−/− mice as well. Similar to the tumorigenic phenotype in Nod2^−/− mice, there was a significantly higher number of tumors in Rip2^−/− mouse colons as compared to WT (Figure S1F–S1H). These data indicate that NOD2/RIP2 signaling pathway plays a protective role against CRC.

Figure 1. Nod2-deficient mice are susceptible to colorectal tumorigenesis.

WT (n=19) and Nod2^−/− (n=19) mice were treated with AOM (10mg/kg). After 5 days of DSS administration, mice were treated with 2.5% DSS in drinking water for 5 days. DSS cycle was repeated two more times. Mice were sacrificed on day 80 after tumor induction. (A) Representative image of tumor-bearing colons from WT and Nod2^−/− mice. (B) The number of tumors per colon was counted. (C) Size of the tumor was measured. Colon tissues collected at day 80 after tumor induction were stained with H&E (D) Representative pictures of H&E staining of tumor-bearing colons. Arrow indicates invasive carcinoma. (E) Histopathological analyses for tumor grading of H&E-stained colon tissues sections. (F) Distribution of animals having low grade dysplasia, high grade dysplasia and invasive carcinoma based on histological analysis. Data represent means ± SEM; *p < 0.05, ***p < 0.0001. See also Figure S1.

Higher CRC susceptibility of Nod2-deficient mice is independent of gut microbial composition

Increasing evidence points to a critical role of gut microbiota in CRC pathogenesis (Zackular et al., 2013). A previous study showed that susceptibility to colitis and colorectal tumorigenesis in Nod2^−/− mice is associated with altered gut microbiota characterized by a higher abundance of Bacteroides (Couturier-Maillard et al., 2013). Microbial composition of laboratory animals varies depending on the laboratory environment as well as mouse handling practices. We therefore sought to verify the association of Bacteroides with the disease phenotype in Nod2^−/− mice in our laboratory settings by measuring the relative abundance of Bacteroides sp, Bacteroides vulgatus, Bacteroides fragilis, and mouse intestinal Bacteroides (MIB) in WT and Nod2^−/− mouse feces with real-time PCR analysis of 16S rDNA. Our results showed no remarkable difference in the level of different species of Bacteroides between WT and Nod2^−/− mice (Figure 2A). We also analyzed the frequency of several other commensal bacteria including Bifidobacterium sp, Prevotellaceae, segmented filamentous bacteria (SFB), TM7, Clostridium cluster IV, E. coli, and Enterobacteriaceae. We observed a modest reduction of Clostridium cluster IV and a slight increase of E. coli in Nod2^−/− mice (Figure S2A). The relative abundance of some other commensal bacteria such as Bifidobacterium, SFB, TM7, Prevotellaceae, Streptococcus, and Staphylococcus was very similar in WT and Nod2^−/− mice (Figure S2A).

Figure 2. The CRC susceptibility of Nod2-deficient mice is not microbiota dependent.

(A) Genomic DNA was isolated from stools of separately housed WT (n=20) and Nod2^−/− (n=20) mice. 16S rDNA of the indicated bacteria was analyzed by real-time PCR and normalized to Universal bacterial 16S rDNA. (B–C) WT (n=8) and Nod2^−/− (n=7) mice were co-housed for 4 weeks before treatment with AOM/DSS as described in the methods. Mice were sacrificed on day 80 after tumor induction. (B) Representative image of colon after tumor induction. (C) The number of tumors per colon was counted. (D–F) Littermate Nod2^+/+ (n=5) and Nod2^−/− (n=5) mice were treated with AOM/DSS at 8 weeks after birth and sacrificed at day 80 following AOM injection. (D) Representative image of the colon after tumor induction. (E) The number of tumors per colon was counted. (F) Representative H&E staining of tumor-bearing colons. Arrow indicates invasive carcinoma. (G–J) GF mice were co-housed with either conventionally raised WT (GF-WT) or Nod2^−/− mice (GF-Nod2) for 7 days. 2 weeks after colonization, mice were induced colorectal tumorigenesis with AOM/DSS regimen. (G)Body weigh changes were monitored for 10 days following 2.5% DSS administration. Clinical scores for stool consistency (H) and occult bleeding (I) were measured at days 3, 4 and 5 after DSS treatment. (J) Tumor numbers were counted at 80 days after tumor induction. Data represent means ± SEM; *p < 0.05. and **p < 0.001. See also Figure S2 and Table S1.

It remained possible that some other species of bacteria contributed to CRC susceptibility of Nod2^−/− mice. Indeed, microbiota plays an essential role in colitis and colorectal tumorigenesis as WT and Nod2^−/− mice treated with antibiotics prior to and during AOM/DSS treatment were resistant to developing colitis (Figure S2B–S2D). To elucidate the possible role of dysbiosis in tumor susceptibility of Nod2^−/− mice, we co-housed WT and Nod2^−/− mice for 4 weeks in a specific pathogen-free facility (SPF). Under co-housed conditions, the microbiota is shared between WT and Nod2^−/− mice. Tumorigenesis was induced in those co-housed mice with AOM/DSS and tumor burden was measured 10 weeks thereafter. Tumor burden in co-housed Nod2^−/− mice was significantly higher than that of WT mice (Figure 2B and 2C). Similarly, Nod2^+/+ and Nod2^−/− littermates that were housed separately after weaning showed similar Bacteroides sp. levels (data not shown), but maintained differential tumor burdens (Figure 2D and 2E). Tumor progression was also seemed faster in Nod2^−/− littermate mice with the evidence of invasion of tumor epithelium in the submucosa (Figure 2F). Furthermore, transferring Nod2^−/− microbiota into germ-free wild-type (GF) mice did not confer colitis or CRC susceptibility as GF mice cohoused with WT mice (GF-WT flora) developed similar tumor burden as in GF mice cohoused with Nod2^−/− mice (GF-Nod2^−/− flora) (Figure 2G–2J). Thus, our data suggest that altered microbial composition is not the major player of higher CRC susceptibility of Nod2^−/− mice.

NOD2 attenuates inflammatory responses in the colon during tumorigenesis

The AOM/DSS model of colorectal tumorigenesis features human colitis-associated CRC that is primarily driven by inflammation (Lee et al., 2010; Zaki et al., 2011b). To understand the role of NOD2 in inflammatory responses during tumorigenesis, we measured the expression of inflammatory mediators, such as IL-6, KC, MIP2, and iNOS, and Cox2, in tumor-bearing whole colons by real-time qPCR. Consistent with the inflammatory responses, the expression of the proinflammatory genes was significantly higher in Nod2^−/− colons as compared to WT mice (Figure 3A). Higher production of IL-6 and KC in colon tissue extracts of Nod2^−/− mice was confirmed by ELISA (Figure S3). IL-6 has been implicated in colorectal tumorigenesis via STAT3 activation (Grivennikov et al., 2009; Lee et al., 2010). Chemokines, such as KC and MIP2, contribute by recruiting neutrophils and macrophages into the tumor microenvironment. Nitric oxide produced by iNOS and prostaglandin synthesized by COX2 are also considered tumor-promoting mediators. Therefore, enhanced expression of these inflammatory mediators in Nod2^−/− mice likely contributes to higher CRC pathogenesis.

Figure 3. NOD2 deficiency induces higher inflammatory responses in the colon during tumorigenesis.

Colorectal tumorigenesis was induced in WT, Nod2^−/− and Rip2^−/− mice with AOM/DSS treatment as described in methods. Mice were sacrificed on day 80 after tumor induction. (A) RNA was isolated from whole colons and expression of the indicated genes was analyzed by real-time PCR. (B) Homogenates of whole colon from WT and Nod2^−/− mice were analyzed for the activation of ERK, NF-κB, and STAT3 by western blotting. (C) Densitometric analysis of band intensity of western blots as shown in panel B. (D) Homogenates of tumor from WT and Nod2^−/− mice were analyzed for the activation of ERK, NF-κB and STAT3 by western blotting. (E) Densitometric analysis of band intensity of western blots as shown in panel D. (F) Homogenates of whole colon from WT and Rip2^−/− mice were analyzed for the activation of ERK, NF-κB, and STAT3 by western blotting. (G) Densitometric analysis of band intensity of western blots as shown in panel F. Data represent mean ± SEM; *p < 0.05, **p < 0.001, ***p < 0.0001. See also Figure S3 and Table S2.

Since NF-κB, MAPK, and STAT3 pathways drive proinflammatory gene expression (Grivennikov et al., 2009; Zaki et al., 2011b), we analyzed their activation in the colons of WT and Nod2^−/− mice by western blot analyses. Consistent with higher inflammatory responses, the levels of phospho-p65, phospho-IκBα, phospho-ERK, and phospho-STAT3 were higher in Nod2^−/− mouse colons as compared to those of WT mice (Figure 3B and 3C). Greater activation of the NF-κB, ERK, and STAT3 pathways was also seen in Nod2^−/− tumor tissue (Figure 3D and 3E). Furthermore, similar to Nod2^−/− mice, there was significantly higher activation of the NF-κB, ERK, and STAT3 pathways in tumor-bearing colons of Rip2^−/− mice compared to WT mice (Figure 3F and 3G). These observations indicate an association of higher tumorigenesis in Nod2^−/− and Rip2^−/− mice with increased activation of the NF-κB, ERK, and STAT3 signaling pathways.

Defects in NOD2 leads to increased inflammation and activation of inflammatory signaling pathways during colitis

Since NOD2 is known for its function in activating the NF-κB and MAPK signaling pathways in response to MDP, the above findings of increased activation of NF-κB, ERK, and STAT3 in Nod2^−/− mouse colons is intriguing. To verify whether increased activation of inflammatory signaling pathways in tumor-bearing Nod2^−/− mouse colons is secondary to higher induction of tumorigenesis, we examined the activation level of those pathways in the colons during colitis. We therefore measured the activation of NF-κB, ERK, and STAT3 in the colons collected at day 3, day 5, and day 10 following DSS administration. We observed that the NF-κB, ERK and STAT3 signaling pathways remained less activated in the Nod2^−/−mouse colons at the beginning of colitis (day 3) (Figure 4A and 4B). However, after day 5, there was increased activation of NF-κB, ERK and STAT3 in Nod2^−/− mouse colons relative to WT mice. Similar results were observed in Rip2^−/− mouse colons at day 10 following DSS administration (Figure S4A). Consistently, the expression of IL-6, KC, iNOS, and Cox2 was seen significantly higher in Nod2^−/− mice at day 10 post DSS (Figure 4C). Excessive inflammation often damages mucosal tissue. Indeed, histopathological analyses of colons showed a remarkable difference in tissue architecture between WT and Nod2^−/− mice at day 10 following DSS administration with increased ulceration, crypt loss, and infiltration of immune cells (Figure 4D and 4E). Nod2^−/− mice also showed increased intensity of colitis with higher body weight loss, diarrhea, rectal bleeding and shortening of the colon length (Figure S4B–S4F). Infiltrating immune cells in the lamina propria were analyzed by flow cytometry. There were increased numbers of myeloid cells (CD11b⁺) and macrophages (F4/80⁺), and the percentage of IL-6-positive myeloid cells was significantly higher in Nod2^−/− mice (Figure 4F and 4G). Previous studies suggest that Nod2 regulates intestinal inflammation via multiple mechanisms, including induction of α-defensin (Kobayashi et al., 2005). However, we could not detect any difference in the expression of α-defensin between WT and Nod2^−/− colons at homeostasis or day 10 after DSS administration (Figure S4G). Taken together, dysregulated activation of NF-κB and other proinflammatory signaling pathways drives increased inflammation and ultimately tumorigenesis in Nod2^−/− mice. Moreover, it appears that Nod2 regulates the inflammatory responses in a biphasic manner with initial NF-κB activation at the onset of the colitis but later suppression during acute and chronic colitis.

Figure 4. NOD2 suppresses inflammatory signaling pathways during colitis.

WT and Nod2^−/− mice were allowed to drink 3% DSS for 5 days. Mouse colons were collected at days 3, 5 and 10 after DSS administration. Colons collected from healthy mice are designated as day 0. (A) Whole colon homogenates were analyzed for the activation of ERK, NF-κB and STAT3 by western blotting. (B) Densitometric analyses of P-p65, P-ERK, and P-STAT3 bands relative to β-actin. (C) mRNA isolated from whole colons was analyzed for inflammatory and tumorigenic mediators by real-time PCR. (D) Representative images of H&E staining of colon sections collected at day 10 following DSS administration. Images were captured at 40X magnification. (E) Histopathological analysis of H&E-stained colon sections collected at day 10 following DSS administration. (F–G) Lamina propria cells were isolated from colons of WT and Nod2^−/− mice at day 10 after DSS administration and stained for CD11b, F4/80 and IL-6. (F) Representative image of flow cytometry analysis of lamina propria cells. (G) Frequency of CD11b⁺F4//80⁺ and CD11b⁺IL-6⁺ cells in the lamina propria of WT and Nod2^−/− mice. Data represent mean ± SEM; *p < 0.05, **p < 0.001, ***p < 0.0001. See also Figure S4 and Table S2.

NOD2 negatively regulates the TLR signaling pathways

The data above demonstrating hyperactivation of NF-κB pathway in Nod2^−/− mouse colons is similar to that seen in CD. However, our data are at odds with the consensus role of NOD2-mediated activation of the NF-κB and MAPK pathways in response to MDP. To resolve this inconsistency, we postulated that Nod2 may suppress TLR-mediated activation of NF-κB and MAPK. To test this hypothesis, we stimulated WT and Nod2^−/− bone marrow-derived macrophages (BMDM) with NOD2, TLR4, and TLR3 agonists MDP, LPS and Poly I:C respectively. Interestingly, activation of the NF-κB, ERK, and STAT3 pathways were significantly higher in Nod2^−/− BMDM during stimulation with LPS and Poly I:C, but not MDP (Figure 5A–5C). Increased activation of NF-κB in Nod2^−/− macrophages was further confirmed by increased nuclear translocation of P-p65 during stimulation with LPS as compared to that in WT macrophages (Figure 5D). Consistently, the expression of downstream proinflammatory cytokines IL-1β, IL-6, and IL-12p40 was significantly higher in LPS-stimulated Nod2^−/− macrophages (Figure 5E). Compared to WT, stimulation of Nod2^−/− macrophages with TLR2 ligand PGN and TLR5 ligand flagellin also yielded higher production of proinflammatory cytokines IL-1β, IL-6, IL-12p40 (Figure S5A), suggesting that NOD2 is a suppressor of multiple TLR signaling pathways. We further examined the role of NOD2 in suppressing the TLR pathways in dendritic cells using bone marrow-derived dendritic cells (BMDC). Nod2^−/− BMDC showed increased activation of NF-κB, ERK, and STAT3, and induction of proinflammatory cytokines in response to LPS and Poly I:C as compared to WT BMDC (Figure 5F and 5G). Rip2^−/− BMDM showed similar changes as those from Nod2^−/− mice upon stimulation with LPS and Poly I:C. There was significantly greater activation of NF-κB, ERK, and STAT3, and significantly enhanced expression of IL-1β, IL-6, and IL-12 in Rip2^−/− macrophages as compared to WT (Figure 5H and S5B).

Figure 5. NOD2 negatively regulates TLR signaling pathways.

(A–E) BMDM from WT and Nod2^−/− mice were stimulated with MDP (10 μg/ml), LPS (1 μg/ml), or poly I:C (1 μg/ml). (A) Cell lysates collected at indicated times were analyzed for P-IκB, IκB, P-ERK, ERK, P-p65, and P-STAT3 by western blotting. (B–C) Densitometric analyses of P-IκB and P-ERK bands relative to IκB and ERK respectively in LPS and poly I:C-treated cells. Data represent mean ± SD of 3 independent experiments; *p < 0.05, **p < 0.001, ***p < 0.0001. (D) Cytoplasmic and nuclear fraction was isolated from LPS-treated BMDM cell lysates and was analyzed for P-p65 by western blotting. α-tubulin and Lamin b1 were used as loading controls of cytoplasmic and nuclear fraction respectively. (E) mRNA was isolated from LPS-stimulated WT and Nod2^−/− macrophages at indicated times and analyzed for the expression of IL-1β, IL-6, and IL-12p40. (F–G) WT and Nod2^−/− BMDC were stimulated with MDP (10 μg/ml), LPS (1 μg/ml) or poly I:C (1 μg/ml). (F) Western blot analysis of P-IκB, IκB, P-ERK, ERK, P-p65, and P-STAT3 in MDP, LPS, and Poly I:C-stimulated BMDC. (G) mRNA was isolated at 0, 2 and 4h after stimulation with LPS and analyzed for the expression of IL-1β, IL-6, and IL-12p40 by real-time PCR. Data represent mean ± SEM; *p < 0.05, **p < 0.001, ***p < 0.0001. (H) WT and Rip2^−/− BMDM were stimulated with LPS (1 μg/ml) and the cell lysates were analyzed for P-IκB, IκB, P-ERK, ERK, P-p65, and P-STAT3 by western blotting. See also Figure S5 and Table S2.

A hallmark feature of colitis is increased infiltration of gut commensal bacteria into the lamina propria, leading to the recruitment and activation of macrophages and dendritic cells (Zaki et al., 2011a). The activation of NF-κB and MAPK in macrophages (RAW264.7) following exposure to bacterial cell wall components is much stronger than that seen in epithelial cells (Figure S5C). Therefore, interaction of bacteria with myeloid cells in the lamina propria is central to the initiation and propagation of intestinal inflammation. We observed an increased invasion of commensal bacteria into the lamina propria of both WT and Nod2^−/− mice during colitis (Figure S5D). To understand the role of NOD2 in inflammatory responses against bacteria, we infected WT and Nod2^−/− BMDM with E. coli, an IBD-associated bacterium (Darfeuille-Michaud et al., 2004). E. coli infection induced greater activation of the NF-κB and ERK signaling pathways in Nod2^−/− and Rip2^−/− macrophages than WT (Figure S5E). Since LPS is a more potent activator of the NF-κB pathways than MDP (Figure S5C), this result implies that NOD2-mediated negative regulation of TLR pathways is physiologically more relevant and prominent than its function in activating NF-κB in response MDP during bacterial infection. Overall, these data support the notion that NOD2 exerts an anti-inflammatory effect in myeloid cells during interactions with bacteria or exposure to TLR ligands.

NOD2-mediated suppression of NF-κB is associated with the induction of IRF4

There are many regulatory molecules, including SOCS, A20, IRF4, IRAK-M, ST2, etc., that dampen excessive activation of the TLR/NF-κB signaling pathways. Previous studies have shown that MDP-stimulated NOD2 induces IRF4 which suppresses TLR-induced activation of NF-κB via interaction with MyD88 and TRAF6 (Watanabe et al., 2014; Watanabe et al., 2008). We therefore examined whether NOD2 downregulates TLR pathways via IRF4. In agreement, MDP was shown to induce IRF4 via NOD2 in BMDM (Figure 6A–C). Interestingly, LPS also induced IRF4 which was markedly reduced in Nod2^−/− BMDM and BMDC (Figure 6D–6I), suggesting a possible role of NOD2 in expressing IRF4 in an MDP-independent manner. To further confirm that NOD2 regulates IRF4 expression during TLR activation which contributes to the downregulation of NF-κB activation, we overexpressed Nod2 in the murine macrophage cell line RAW264.7 and stimulated the mock (GFP)-transfected or Nod2-transfected cells with LPS (Figure 6J). While the activation of NF-κB and ERK was decreased in Nod2-transfected cells upon stimulation with LPS, there was increased expression of IRF4 as compared to mock-transfected cells (Figure 6J–6L). Consistently, the expression of proinflammatory molecule IL-6 was suppressed with induction of IRF4 during stimulation with LPS (Figure 6L). Furthermore, knocking down IRF4 with siRNA led to upregulation of IL-6 (Figure S6A). Notably, NOD2-dependent induction of IRF4 during LPS stimulation requires TLR4 (Figure S6B). Finally, we investigated in vivo relevance of NOD2-mediated induction of IRF4 in the regulation of inflammatory response by measuring the expression of IRF4 in the colons collected from healthy and colitic WT and Nod2^−/− mice. The levels of IRF4 were significantly reduced in Nod2^−/− colons at day 3 and day 10 following colitis induction compared to those in WT mice (Figure 6M–6O). All these results suggest that NOD2-mediated induction of IRF4 is involved, at least partially, in the suppression of the TLR pathways.

Figure 6. NOD2 downregulates TLR-induced NF-κB activation via induction of IRF4.

(A–F) WT and Nod2^−/− BMDM were stimulated with MDP (10 μg/ml) or LPS (1 μg/ml). Expression of IRF4 was measured by real-time RT-PCR (A and D) and western blotting (B and E) at indicated times. (C and F) IRF4 band intensity relative to β-actin was analyzed. Data represent mean ± SD of 3 independent experiments; **p < 0.001, ***p < 0.0001. (G–I) WT and Nod2^−/− BMDC were stimulated with LPS (1 μg/ml). Expression of IRF4 was measured by real-time RT-PCR (G) and western blotting (H) at indicated times. (I) Densitometric analysis of IRF4 band intensity relative to β-actin. Data represent mean ± SD of 3 independent experiments; *p < 0.05, **p < 0.001. (J–L) Nod2 was cloned into pcDNA4/TO vector and transfected into RAW264.7 cells. Nod2- or GFP (mock)-transfected cells were stimulated with LPS. (J) The cell lysates collected at the indicated times were analyzed for P-ERK, ERK, P-p65, and IRF4 by western blotting. Expression of Nod2 was measured by anti-Nod2 and anti-Flag antibodies. (K) Densitometric analysis of IRF4 and P-p65 band intensities relative to β-actin. Data represent mean ± SD of 3 independent experiments; *p < 0.05, **p < 0.001. (L) Nod2 or GFP-transfected RAW264.7 cells were stimulated with LPS (1 μg/ml) or MDP (10 μg/ml) for 4 h. mRNA was isolated and analyzed for the expression of IRF4, IL-6, Nod2 by real-time PCR. Data represent means ± SD of triplicate wells; *p < 0.05, **p < 0.01, ***p < 0.001. (M) Expression of IRF4 was measured in the colons of WT and Nod2^−/− mice at different days after DSS (3%) administration. (N–O) Colon lysates collected at day 0, day 3 and day 10 following DSS administration were analyzed for IRF4 by western blotting. (O) Densitometric analysis of IRF4 band relative to β-actin. Data represent mean ± SEM; *p < 0.05, **p < 0.001, ***p < 0.0001. See also Figure S6 and Table S2.

Increased IL-6 production in Nod2-deficient mouse colon leads to hyperproliferation of tumor epithelium

Based on our findings, it is apparent that NOD2 deficiency enhances inflammation in the colon. The association of inflammation with tumorigenesis is linked by its ability to induce cellular proliferation (Grivennikov et al., 2010). The proliferative responses in WT and Nod2^−/− mouse epithelia during tumorigenesis were assessed by the expression of Ki67 in tumor-bearing colons at day 80 after AOM/DSS treatment. A significantly higher number of Ki67-positive cells were seen exclusively in tumor tissue of Nod2^−/− mice compared to WT mice (Figure 7A and 7B). Ki67 mRNA expression was also elevated in tumor tissues but not in unaffected colon tissue of Nod2^−/− mice as compared to WT mice (Figure 7C). To further examine the association of inflammation with the proliferative responses, we analyzed Ki67 in the colons from healthy or colitic mice at day 10 following DSS administration. In agreement with the above observation, there was increased proliferation in the inflamed areas, but not in the healthy or non-inflamed areas of the colon of Nod2^−/− mice as compared to those in WT mice (Figure S7A and S7B). One of the cytokines regulated by the NF-κB pathway and linked with tumorigenesis is IL-6 (Grivennikov et al., 2009). IL-6 production was increased in Nod2^−/− mouse colons during colitis and colorectal tumorigenesis (Figure 3A and 4C), suggesting its role in tumorigenesis phenotype in Nod2^−/− mice. Indeed, stimulation of mouse intestinal epithelial cell line MODE-K cells with IL-6 resulted in induction of Ki67 (Figure 7D). IL-6 regulates cellular proliferation via activation of STAT3. As expected, STAT3 was strongly activated by IL-6 in MODE-K cells (Figure 7E). Therefore, we propose that elevated IL-6 production in Nod2-deficient mouse colons due to higher activation of the NF-κB and MAPK pathways leads to enhanced epithelial proliferation and ultimately increased tumorigenesis (Figure 7F).

Figure 7. Increased IL-6 production in Nod2-deficient mouse colon leads to hyperproliferation of tumor epithelium.

(A–D) Colorectal tumorigenesis was induced in WT and Nod2^−/− mice with AOM/DSS treatment as described in methods. Colons collected at day 80 after tumor induction were processed for immunostaining of Ki67. (A) Representative images of Ki67 staining in non-tumor and tumor area of colon sections. (B) Ki67⁺ cells were counted in tumor and non-tumor tissue under 20X microscopic lenses (n=3 mice/group). (C) mRNA was isolated from whole colons and analyzed for Ki67 by real-time PCR (n=8/group). Data represent means ± SEM; *p < 0.05. and **p < 0.001. (D) MODE-K cells were stimulated with IL-6 (20 ng/ml) and the expression of Ki67 mRNA was analyzed by real-time PCR. Data represent means ± SD; *p < 0.05. and **p < 0.001. (E) The cell lysates from IL-6 stimulated MODE-K cells were analyzed for P-STAT3, P-p65, P-ERK, and ERK by western blotting. (F) Proposed pathway for NOD2-mediated regulation of inflammation and tumorigenesis in the colon. See also Figure S7 and Table S2.

DISCUSSION

IBD is a major risk factor for CRC. NOD2 polymorphisms have been strongly associated with IBD, suggesting they may also be risk factors for CRC (Liu et al., 2014; Tian et al., 2010). However, lack of experimental evidence establishing a causal relationship between NOD2 and CRC is still lacking. Here we demonstrated that deficiency of Nod2 leads to increased tumorigenesis in mice which is independent of dysbiosis. This observation is in contrast with another report showing that colitis and colorectal tumorigenesis in Nod2^−/− mice was driven by Bacteroides in the gut (Couturier-Maillard et al., 2013). Similar to our observation, no major difference was observed in the composition of gut microbiota between WT and Nod2^−/− mice in a separate study (Robertson et al., 2013). Due to the influence of environmental factors, animals from different laboratories show major differences in their gut microbiota. Thus, it appears that the altered composition of microbiota in Nod2^−/− mice is due to environmental factors, not the genotype. The findings of this study therefore represent the genetic role of NOD2 in the pathogenesis IBD of colorectal tumorigenesis.

The relevance of NOD2 with inflammatory disorders in the gut was demonstrated more than a decade ago (Hugot et al., 2001). Yet, the mechanism of NOD2-mediated regulation of inflammation and tumorigenesis remains poorly understood. Sensing MDP and subsequent activation of NF-κB and MAPK via RIP2 kinase is a well-characterized function of NOD2. Since CD-associated NOD2 mutations are usually missense (Bonen et al., 2003; Hugot et al., 2001; Ogura et al., 2001), how defective NOD2 function is associated with hyperinflammatory responses in the gut remains unclear. As CD patient’s colonic mucosa show increased but not decreased activation of NF-κB (Eckmann and Karin, 2005; Rogler et al., 1998), NOD2 must have additional functions that involve the suppression of the NF-κB pathway. This hypothesis is supported by other independent studies. Watanabe et al reported that TLR2-mediated activation of NF-κB in T cells was increased in the absence of NOD2 (Watanabe et al., 2004). Subsequent studies from the same research group demonstrated that MDP-stimulated NOD2 negatively regulates TLR-mediated activation of NF-κB in dendritic cells (Watanabe et al., 2008; Watanabe et al., 2006; Watanabe et al., 2005). However, these results are in contradiction with some other studies showing that MDP exerts a synergistic effect on TLR ligand-induced NF-κB activation (Kobayashi et al., 2005; Park et al., 2007; Tada et al., 2005). Such a synergistic relationship can be explained by the fact that TLR activation induces NOD2 which participates in MDP-mediated activation of NF-κB (Tsai et al., 2011). Given that MDP-mediated activation of NF-κB via NOD2 is much weaker than that mediated by LPS/TLR4 pathway, the expression level of NOD2 should exert modest impact on overall NF-κB activation during co-stimulation with LPS and MDP. Alternatively, in the context of co-stimulation of LPS and MDP, the interaction of NOD2 with MDP results in an attenuation of NOD2-mediated negative regulation of the TLR4/NF-κB pathway. Indeed, our data strongly supports the latter concept as we observed that NOD2 can downregulate TLR activation even in the absence of MDP. Since TLR-mediated activation of NF-κB is the central pathway for inflammatory responses, NOD2-mediated repression likely constitutes an important protective mechanism against IBD and CRC.

The findings of this study suggest that NOD2 maintains intestinal homeostasis via early activation but eventual downregulation of NF-κB. Under homeostatic conditions, NOD2 activates NF-κB and MAPK pathways. On the other hand, NOD2 suppresses these pathways during acute and chronic inflammation. The activation of the NF-κB and MAPK pathways in the intestinal epithelium by NOD2 at homeostasis or the early stage of inflammation is very critical since these pathways regulate the production of the antimicrobial peptides, cytokines, and pro-proliferative genes that are involved in host responses against infection and injury (Kaser et al., 2010). The expression of TLR4 and TLR2 is downregulated in epithelial cells of healthy intestines (Saleh and Trinchieri, 2011), suggesting that MDP-dependent activation of NOD2 plays a central role in immune homeostasis of the healthy gut. However, macrophages and dendritic cells in the lamina propria of inflamed gut provide robust inflammatory response via the TLR activation. Notably, NF-κB is more potently activated by bacterial TLR ligands than MDP. Thus, while MDP-dependent activation of NOD2 in epithelial and myeloid cells plays an important role in immune responses at homeostasis, the activation of TLRs in myeloid cells constitutes major trigger for the inflammatory disorders in the gut. Indeed, Tlr4^−/− mice are resistant to colitis and CRC (Fukata et al., 2007).

While NOD2 emerges as a negative regulator of the TLR signaling pathways, the mechanism of such regulation is still poorly defined. Previous studies demonstrated that MDP-stimulated NOD2 downregulates TLR activation via induction of IRF4 (Watanabe et al., 2014; Watanabe et al., 2008). IRF4 was shown to interact with MyD88, TRAF6, and RICK, leading to the inhibition of TLR- and NOD2-mediated activation of the NF-κB pathways. Such a mechanism of counterbalancing NF-κB activation explains why MDP-stimulated cells are tolerized to subsequent TLR or NOD2 stimulation (Strober and Watanabe, 2011). Interestingly, our data illustrates a different context where NOD2 suppresses the NF-κB pathway via induction of IRF4 during activation of the TLR pathways without MDP prestimulation. This is an intriguing observation since the mechanism by which NOD2 is activated by LPS or other TLR ligands in the absence of MDP remains unknown. Given that LPS-mediated induction of IRF4 via NOD2 requires TLR4, a possible explanation would be that biochemical changes occur in NOD2 during the activation of the TLR pathways, leading to its activation and induction of IRF4. In fact, NOD2 in involved in multiple physiological functions in a context dependent manner (Strober and Watanabe, 2011). McDonald et al. showed that NOD2 binds with an array of cellular proteins that can either inhibit or enhance its function (McDonald et al., 2005). A recent proteomic study also suggests that NOD2 interacts with multiple proteins that are involved in the regulation of NF-κB in both positive or negative ways (Warner et al., 2013). Future studies should aim to explore the precise structural and biochemical changes in NOD2 during TLR activation.

In conclusion, here we have demonstrated that NOD2 is a physiological regulator of intestinal inflammation and tumorigenesis. It is increasingly evident that NOD2 possesses multiple biological functions and its relevance to the pathogenesis of IBD and CRC cannot be simply explained by its well-characterized function of MDP-mediated activation of NF-κB. This study demonstrates that NOD2-mediated protection against colorectal tumorigenesis reflects its unique function in the suppression of TLR-mediated activation of NF-κB pathways. Therefore, this study will help explain the underlying mechanism of IBD and CRC pathogenesis associated with NOD2 mutations, and suggest that NOD2-mediated negative regulation of TLR pathways may be a therapeutic target for IBD and CRC treatment.

MATERIALS AND METHODS

Mice

Nod2^−/−, Rip2^−/−, and wild-type (C57BL6/J) mice were purchased from Jackson Laboratory. All mice were bred and maintained in a specific pathogen free (SPF) facility at UT Southwestern Medical center. Unless otherwise stated, mice of different genetic backgrounds were housed in separate cages, maintained in same animal room, and used for in vivo and in vitro experiments. All studies were approved by the Institutional Animal Care and Use Committee (IACUC) and were conducted in accordance with the IACUC guidelines and the National Institutes of Health Guide for the Care and Use of Laboratory Animals. All experiments were conducted with sex and age-matched mice and both male and female mice were included.

Induction of colorectal tumorigenesis

Colorectal tumorigenesis was induced in mice using AOM plus DSS model as described previously (Zaki et al., 2011b). Briefly, mice were injected intraperitoneally (i.p.) with 10 mg/kg AOM (Sigma) and maintained on regular drinking water for 5 days, after which they were administered with 2.5% DSS (molecular mass 36–40 kDa; TdB Consultancy) in drinking water for 5 days. After omission of DSS, mice were allowed to drink regular water for two weeks. The DSS cycle was repeated for two more cycles. Mice were sacrificed at day 80 following AOM injection.

Histopathology and immunohistochemistry

Colon was washed with PBS, fixed in 4% paraformaldehyde and embedded in paraffin. Tissue sections of whole colon were stained with hematoxylin & eosin (H&E). Histological scoring was performed in a blinded fashion by a pathologist. Colitis scoring was made as a combined score for inflammatory cell infiltration (score 0–3), ulceration (score 0–3), hyperplasia (score 0–3) and area of crypt distortion (score 0–3) (Hu et al., 2015). Tumors were graded as low grade dysplasia, high grade dysplasia, and invasive adenocarcinoma. High grade dysplasia scoring (0–3) was done as 0 = no high grade dysplasia, 1 = high grade dysplasia in occasional area, 2 = high grade dysplasia in some area, 3 = high grade dysplasia in many area. For immunohistochemistry, 4% paraformaldehyde-fixed and paraffin-embedded colon tissues sections were de-paraffinized and hydrated through decreasing concentrations of ethanol. Heat-induced antigen retrieval was performed in 10 mM sodium citrate solution (pH 6.0) for 20 min at 95°C. Tissue sections were blocked with 5% goat serum for 30 min and stained for Ki67 using rabbit anti-Ki67 (ab16667; Abcam). After overnight incubation at 4°C, the tissue sections were washed three times and incubated with HRP-conjugated anti-rabbit antibody for 1 h at room temperature. The images were taken using a Zeiss microscope.

Statistical analysis

Data are presented as averages ± SD or ± SEM (as stated in figure legends). Statistical significance was determined by unpaired Student’s t test and p < 0.05 was considered statistically significant.