Introduction

About 150 years ago, Rudolf Virchow described the presence of leukocytes in tumors, and hypothesized that the origin of cancer was at the site of chronic inflammation. It has been reported that chronic infections are associated with 15–20% of malignant cancers [1, 2]. The principal infectious agents are Helicobacter pylori, hepatitis B and C viruses, and the human papilloma virus, which are closely associated with gastric cancer, hepatocellular carcinoma, and cervical cancer, respectively. Moreover, about 30% of all cancers have been attributed to smoking and 20% to obesity [3], and it has been shown that both tobacco smoke and obesity can trigger inflammatory responses in the lungs and liver, respectively, which promote tumorigenesis [4, 5]. These results, together with those of other recent studies (reviewed in [6–8]), indicate that inflammation plays an important role in promoting cancer development, and “tumor-promoting inflammation” is now included in the next generation of the criteria considered to be “hallmarks of cancer” [9].

Epidemiological studies have revealed that regular use of non-steroidal anti-inflammatory drugs (NSAIDs) lowers the mortality rate from cancers in the gastrointestinal tract [10, 11]. The target molecules of NSAIDs are cyclooxygenase (COX)-1 and COX-2, and accumulating evidence has indicated that COX-2 and its downstream product, prostaglandin E2 (PGE2), play an important role in cancer development [12, 13]. On the other hand, proinflammatory cytokines are expressed in the tumor microenvironment, and such cytokine signaling is also important for cancer development through the activation of downstream transcription factors [6]. Among them, tumor necrosis factor (TNF)-α and interleukin (IL)-6 activate nuclear factor (NF) -κB and Stat3, respectively, and both the TNF-α/NF-κB and IL-6/Stat3 pathways have been shown to be important for the development of inflammation-associated intestinal tumorigenesis [7, 8]. Moreover, NF-κB induces the expression of COX-2, IL-6, and TNF-α. Accordingly, these signaling pathways construct an inflammatory network in the tumor microenvironment, which plays an important role in tumor promotion (Fig. 1). In this review, we discuss the roles of these inflammatory pathways in gastrointestinal tumorigenesis, which have been identified by a number of mouse model studies, as listed in Table 1.

Fig. 1
figure 1

The interaction of the cyclooxygenase-2 (COX-2)/prostaglandin E2 (PGE 2 ), tumor necrosis factor-α (TNF-α)/nuclear factor-κB (NF-κB), and interleukin-6 (IL-6)/gp130/Stat3 pathways in the inflammatory environment. cAMP cyclic AMP, IKK inhibitor of κB kinase, IκB inhibitor of κB, NSAIDs non-steroidal anti-inflammatory drugs

Full size image
Table 1 Mouse model studies performed to examine inflammatory networks in gastrointestinal tumorigenesis
Full size table

Mouse models of gastrointestinal cancer

The roles of inflammatory responses in gastrointestinal cancers have been studied using several tumor mouse models (Table 1). Apc Δ716 knockout mice and Apc Min mice carry heterozygous truncation mutations at codons 716 and 850 of the mouse Apc gene, respectively, and somatic deletion of the wild-type Apc gene results in the activation of Wnt/β-catenin signaling, which causes tumor development in the entire intestinal tract [14, 15]. Approximately 80% of colorectal cancers harbor APC gene mutations and half of the remainder have mutations in Ctnnb1 gene encoding β-catenin, both of which activate Wnt/β-catenin signaling [16–18]. Thus, Apc Δ716 and Apc Min mice recapitulate the molecular mechanism of sporadic colon cancer development. Several other types of Apc mutant mice also develop intestinal polyposis (as described in this review and Table 1).

On the other hand, inflammatory bowel diseases (IBDs), including ulcerative colitis (UC) and Crohn’s disease (CD), are also risk factors for colorectal cancer [19, 20]. IBD-related colon cancers are associated with a severe inflammatory response, which is an important microenvironment required for inflammation-associated tumor development. The treatment of mice with a genotoxic chemical carcinogen, azoxymethane (AOM), followed by a non-genotoxic agent, dextran sodium sulfate (DSS), induces the development of colitis-associated colon cancer (CAC) [21]. Mutations in Ctnnb1 gene induced by AOM result in the activation of the Wnt/β-catenin pathway, which is thought to trigger tumor initiation [22]. On the other hand, DSS induces colonic inflammation in rodents, which is required for tumor promotion. Accordingly, the AOM/DSS model is a well-established and widely used mouse model for CAC development, and it mimics IBD-related colon cancer (Table 1). Rag2 gene-deficient mice lack functional lymphocytes, and are susceptible to infection-induced inflammation in the colon. When Rag2−/− mice are infected with the enteric bacterial pathogen, Helicobacter hepaticus, the mice rapidly develop CAC [23, 24]. This model system is also used for IBD-related colon cancer (Table 1).

Activation of Wnt/β-catenin signaling is found in approximately 30–50% of gastric cancers, suggesting a causal role of Wnt signaling in a subpopulation of gastric cancers [25, 26]. On the other hand, Helicobacter pylori infection is an important risk factor for gastric cancer [27]. In H. pylori-associated gastritis, COX-2 expression is induced significantly, whereas its level is decreased by H. pylori eradication [28, 29]. Two transgenic mouse strains, K19-Wnt1 mice expressing Wnt1 in the stomach, and K19-C2mE mice expressing Ptgs2 and Ptges that encode COX-2 and microsomal PGE synthase-1 (mPGES-1), mimic Wnt activation and H. pylori-induced inflammation, respectively, in the stomach [26, 29–31]. Gan mice, which are compound transgenic mice with K19-Wnt1 and K19-C2mE, express Wnt1, Ptgs2, and Ptges simultaneously in the gastric mucosa, resulting in the activation of both Wnt signaling and the COX-2/PGE2 pathways, as found in human gastric cancer. Gan mice are thus used as a model of inflammation-associated gastric tumors [26, 30, 31] (Table 1).

The COX-2/PGE2/EP2 pathway in gastrointestinal tumorigenesis

It has been demonstrated that treatment of familial adenomatous polyposis (FAP) patients with NSAIDs results in significant regression of colon polyps [32]. Moreover, a large number of animal experiments have shown that treatment with NSAIDs suppressed chemical carcinogen-induced colon tumorigenesis [33]. As a target molecule of NSAIDs, the inducible enzyme COX-2 plays an important role in inflammation and cancer, while COX-1 is expressed constitutively and functions as a house-keeping gene. Importantly, disruption of Ptgs2 gene in Apc Δ716 mice and Apc Min mice resulted in a significant suppression of intestinal tumorigenesis, thus indicating an essential role of COX-2 in intestinal polyp development [34, 35]. Interestingly, disruption of Ptgs1 gene encoding COX-1 also suppressed intestinal tumorigenesis [35]. It is possible that COX-1-derived prostaglandins are required for tumor cell proliferation during the initial stage when COX-2 expression is not yet induced [33]. Other lines of genetic evidence also support the role of the COX-2 pathway in intestinal tumorigenesis. Prostaglandins are catalyzed and inactivated by 15-hydroxyprostaglandin dehydrogenase (15-PGDH). The number of colon polyps in Apc Min mice was markedly increased when Hpgd gene encoding 15-PGDH was disrupted, suggesting that PGE2 has a role in colon tumorigenesis [36]. Moreover, the transgenic expression of Ptgs2 in the mouse intestine accelerated chemical carcinogen-induced tumorigenesis [37].

COX-2 catalyzes the synthesis of prostaglandin (PG) H2, which is then converted to PGE2. There are four G protein-coupled receptors for PGE2; EP1, EP2, EP3, and EP4. Notably, disruption of Ptger2 gene encoding EP2 caused significant suppression of intestinal polyposis in Apc Δ716 mice, whereas suppression of EP1 or EP3 signaling did not affect tumorigenesis [38]. EP2 signaling increases the expression of vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (bFGF), which enhances angiogenesis in intestinal tumors [39]. Moreover, treatment of Apc Min mice with PGE2 increased the development of intestinal tumors through the activation of peroxisome proliferators-activated receptor (PPAR) δ, which promotes the survival of tumor cells [40]. Furthermore, PGE2 signaling through the EP2 receptor has been shown to activate Wnt/β-catenin signaling directly in colon cancer cells by the suppression of β-catenin phosphorylation [41]. Accordingly, it is possible that the COX-2/PGE2 pathway contributes to intestinal tumorigenesis through a variety of PGE2 functions [12, 13] (Fig. 2).

Fig. 2
figure 2

A schematic diagram of the inflammatory microenvironment in gastrointestinal tumor tissues. Bacterial infection or endogenous ligand activates the Toll-like receptor (TLR)/MyD88 pathway in epithelial cells, which further activate stromal macrophages inducing CCL2. In the activated macrophages, NF-κB induces the expression of COX-2, IL-6, and TNF-α itself. The COX-2/PGE2 pathway is important for the construction and maintenance of the inflammatory network, and PGE2 accelerates angiogenesis, cell survival, and Wnt activation. TNF-α activates NF-κB in both epithelial cells and stromal macrophages, whereas IL-6 activates Stat3 in epithelial cells through gp130. NF-κB and Stat3 play an important role in the promotion of tumorigenesis through suppression of apoptosis and acceleration of the cell cycle. mPGES-1 Microsomal PGE synthase-1

Full size image

Microsomal PGE synthase-1 (mPGES-1) is an enzyme that converts PGH2 to PGE2, and expression of mPGES-1 is induced in gastric and colon cancers similar to COX-2 [42, 43]. Notably, disruption of Ptges gene encoding mPGES-1 in Apc Δ14 mice or AOM-treated mice resulted in a marked decrease in the PGE2 level in the intestinal mucosa, which led to a further significant suppression of intestinal tumorigenesis [44, 45]. Taken together, these results indicate that the simultaneous expression of COX-2 and mPGES-1 is required for intestinal tumorigenesis through the induction of PGE2 signaling.

In gastric cancer tissues, induction of COX-2 is found in approximately 70% of cases [46, 47], and mPGES-1 expression is also induced [42], suggesting that the COX-2/PGE2 pathway is also important for gastric tumorigenesis. In the K19-C2mE transgenic mice, an increased PGE2 level causes inflammatory infiltration and metaplastic hyperplasia in the gastric mucosa [29]. Although Wnt activation alone is not sufficient for tumor development in K19-Wnt1 mice, Gan mice (compound mutants of K19-Wnt1 and K19-C2mE mice) develop inflammation-associated gastric tumors with 100% incidence [26], indicating that cooperation of the Wnt and PGE2 pathways can lead to gastric tumorigenesis. Moreover, gastric inflammation and tumorigenesis were significantly suppressed in Gan mice when the mice were treated with a COX-2 inhibitor, celecoxib, or an EP4 receptor inhibitor [48, 49]. These results indicate that COX-2/PGE2/EP4-induced inflammation is involved in the development of gastric cancer. Signaling through both EP2 and EP4 stimulates the intracellular cyclic AMP signaling pathway. It is therefore possible that either EP2 or EP4 receptor signaling plays an important role in gastrointestinal tumorigenesis (Fig. 1).

The Toll-like receptor (TLR)/MyD88 pathway for COX-2 induction in gastrointestinal tumors

Expression of COX-2 and mPGES-1 is detected predominantly in stromal cells, including macrophages and fibroblasts, but not in the epithelial cells of mouse intestinal polyps, and the same is true for human colon polyps [36, 50–52]. Heterozygous mutations in the Lkb1, Smad4, or Cdx2 gene lead to the development of gastrointestinal hamartomas, which show histological characteristics distinct from those of dysplastic adenomas developed in Apc mutant mice [52–55]. Notably, expression of COX-2 and mPGES-1 is detected in the stromal cells, but not in tumor epithelial cells, in these models [56], thus indicating that the COX-2/PGE2 pathway is induced in the tumor stroma by a common mechanism, regardless of the type of tumor.

Several studies have suggested the commensal flora to play a role in the homeostasis of the intestinal mucosa. Toll-like receptors (TLRs) are a family of pattern-recognition receptors that detect the molecular products of microorganisms. MyD88 is an adaptor molecule for the TLR-mediated induction of inflammatory cytokines. It has been shown that a disruption of Tlr2/4 or Myd88 genes encoding TLR2/4 or MyD88, respectively, in mice results in the impaired mucosal repair of DSS-induced ulcers [57], suggesting that infectious stimulation through the TLR/MyD88 pathway is important for the regeneration process of injured mucosa. Moreover, stromal macrophages are also required for the mucosal repair of DSS-induced ulcers in the colon [58]. Accordingly, it is possible that stromal macrophages are activated by the TLR/MyD88 pathway. In the case of intestinal tumorigenesis, cancer cells may use such a TLR/MyD88-induced regeneration system to increase their proliferation.

The treatment of mice with DSS induces COX-2 expression and PGE2 production, predominantly in macrophages of the inflamed colon mucosa. However, such COX-2 induction is not found in the mice lacking the TLR4/MyD88 pathway [59], thus indicating the role of bacterial infection in inducing COX-2 expression in colitis tissues. Importantly, AOM/DSS treatment-induced colon tumor development was dramatically suppressed in Tlr4−/− mice [60], while exogenous administration of PGE2 promoted CAC development in the AOM/DSS-treated Tlr4−/− mice [61]. Moreover, bone marrow chimera experiments have indicated that TLR4 expression in the intestinal epithelial cells, but not in myeloid cells, is required for CAC development [62]. Taken together, these results suggest that bacterial infection stimulates the TLR/MyD88 pathway in epithelial cells, which leads to activation of stromal macrophages, thus resulting in the induction of the COX-2/PGE2 pathway in the tumor stroma (Fig. 2).

In contrast to the AOM/DSS model, intestinal polyposis in the Apc mutant mice is not associated with UC. It is therefore possible that COX-2 expression is induced by a different mechanism in sporadic colon cancer compared to that in IBD-related tumors. However, Apc Min Myd88−/− mice showed significant suppression of intestinal polyposis with dramatically decreased mortality compared with control Apc Min Myd88+/+ mice [63, 64]. Moreover, the induction of COX-2 expression was also suppressed in Apc Min Myd88−/− mouse intestinal tumors [63]. Bone marrow chimera experiments also indicated that MyD88 expression in the epithelial cells was important for intestinal tumorigenesis [64]. These results have demonstrated that activation of the TLR/MyD88 pathway in epithelial cells is also important for COX-2 expression in non-IBD-related colon cancer (Fig. 2).

If bacterial infection is also required for COX-2 induction in the stomach, a low bacterial count due to the acidic environment in the stomach may protect against COX-2 expression in tumorous lesions. Therefore, it is possible that COX-2 induction through the TLR/MyD88 pathway is one of the important mechanisms by which Helicobacter pylori infection promotes gastric cancer.

It has also been shown that TLRs can be stimulated with endogenous ligands, including heat shock proteins and various products of the extracellular matrix [65]. Accordingly, it is possible that tumor cell proliferation causes tissue damage, releasing endogenous ligands for TLRs, thus resulting in their activation, which induces COX-2 expression in stromal macrophages. Such “cancer-induced inflammation” may also be one of the mechanisms responsible for the generation of an inflammatory microenvironment, especially in cancers that are not associated with infection [65].

The paradox of the COX-2 pathway in colitis-associated cancer

Although COX-2 inhibition causes suppression of gastrointestinal tumorigenesis, treatment with NSAIDs or a COX-2 inhibitor exacerbates DSS-induced colon injury in rodent models [66]. Consistently, Ptgs2 gene knockout mice exposed to DSS showed exacerbated phenotypes, such as more severe inflammation, compared with wild-type mice [67]. One of the functions of PGE2 is to protect the gastrointestinal mucosa. Therefore, COX-2 is important for mucosal protection and regeneration in the DSS-treated mouse colon. Notably, treatment of Ptgs2 knockout mice with AOM/DSS resulted in a significant increase in tumors, with severe inflammatory responses [68, 69], which appears to be contradictory to the results showing that Ptgs2 gene disruption causes significant suppression of intestinal polyposis in Apc mutant mice [34, 35]. It is possible that in the case of severe UC, cytokine signaling is highly activated and may be sufficient for tumor promotion if the COX-2/PGE2 pathway is blocked (Fig. 2).

The TNF-α/NF-κB pathway in gastrointestinal tumorigenesis

TNF-α is one of key regulators of inflammatory responses. Although TNF was originally recognized as a tumor-necrotizing factor, accumulating evidence has indicated that TNF-α has tumor-promoting functions [70]. TNF-α signaling activates NF-κB, which further induces the expression of inflammatory factors including COX-2, IL-6, IL-8, and TNF-α itself (Fig. 1). Several genetic studies have demonstrated a link between the TNF-α/NF-κB pathway and cancer development [71].

Conditional disruption of Ikbkb gene encoding IKKβ in myeloid cells, which results in specific inhibition of NF-κB, caused significant suppression of the tumor incidence in AOM/DSS-treated mice, and was associated with decreased expression of cytokines and COX-2 [72]. Conditional deletion of Ikbkb in epithelial cells also suppressed AOM/DSS-induced tumorigenesis [72]. Similar results were found in TNF-α receptor gene knockout mice [73]. AOM/DSS treatment in mice lacking Tnfrsf1a gene encoding TNF receptor p55 (TNF-Rp55) resulted in attenuated tumor formation, with reduced inflammatory cell infiltration compared with findings in wild-type mice. Moreover, wild-type mice transplanted with Tnfrsf1a-deficient bone marrow developed significantly fewer tumors after AOM/DSS treatment [73], indicating that TNF-α stimulation of myeloid cells is important for tumorigenesis. Accordingly, it is possible that the TNF-α-induced NF-κB activation in myeloid cells is important for CAC development, and NF-κB activation in epithelial cells also contributes to tumor formation (Fig. 2).

It has also been shown that a disruption of Ccr2 gene encoding a CCL2-specific receptor, CCR2, led to significantly decreased macrophage infiltration and lower tumor numbers when mice were treated with AOM/DSS [74]. CCL2 is a chemokine that is chemotactic for monocytes and macrophages [75]. Taken together, these findings indicate that the activation of NF-κB in activated macrophages by TNF-α in an autocrine or paracrine manner is important for the promotion of intestinal tumorigenesis (Fig. 2). The inhibition of TNF-α in Apc mutant mice by treatment with an anti-TNF-α antibody also suppressed intestinal polyposis, with the suppression of angiogenesis [76]. Accordingly, it is possible that TNF-α/NF-κB activation is important in both IBD-related and sporadic colon carcinogenesis.

An important role for the TNF-α/NF-κB pathway was also discovered in Mdr2 knockout mice that develop inflammation-associated hepatocellular carcinoma (HCC) [77]. In this mouse model, NF-κB is activated in the liver by TNF-α signaling, and the inhibition of NF-κB significantly suppressed the development of HCC after 7 months of age [77]. On the other hand, preneoplastic dysplastic lesions in younger mice were not affected by NF-κB inhibition. It is possible that the TNF-α/NF-κB pathway is not required for the initiation step, but it does play a role in the promotion step of HCC development.

The role of the TNF-α/NF-κB pathway in gastric tumorigenesis in Gan mice has not yet been examined. However, Tnf−/− K19-C2mE mice showed significant suppression of gastritis and hyperplasia compared with control K19-C2mE mice, although the COX-2/PGE2 pathway was still activated by the transgenic expression of COX-2 and mPGES-1 [78]. It is thus conceivable that the TNF-α/NF-κB pathway is activated in PGE2-associated gastritis, and that this contributes to inflammation-associated gastric tumorigenesis.

The IL-6/gp130/Stat3 pathway in gastrointestinal tumorigenesis

One of the NF-κB-inducible cytokines is IL-6, which is important for immune responses, cell survival, apoptosis, and proliferation [79]. The expression of IL-6 is often upregulated in tumor tissues and in the sera of humans and mice with cancers, including colon cancer [80]. The IL-6 cytokine family signals through a common receptor, gp130, which activates Stat3 (Fig. 1). Stat3 plays an important role in the development of a variety of cancers, including CAC [81–84]. AOM/DSS-induced CAC development was significantly suppressed in conditional Stat3-knockout mice that lacked Stat3 in the intestinal epithelial cells and also in Il6−/− mice [81, 82]. Survival and proliferation of tumor cells were suppressed in these mutant mice. On the other hand, the number and size of AOM/DSS-induced colon tumors increased significantly in gp130757F/F mice, in which gp130-dependent Stat signaling is constitutively activated [81]. These results indicate that Stat3 activated in epithelial cells plays an important role in the promotion of intestinal tumorigenesis (Fig. 2).

It has been shown that gp130757F/F mice develop gastric tumors with abundant infiltration of inflammatory cells [85]. Moreover, heterozygous mutations of the Stat3 gene in gp130757F/F mice reduced the incidence and multiplicity of gastric tumors, with the suppression of inflammatory responses [86, 87]. These results indicate that Stat3 is also an important tumor-promoting factor in gastric tumorigenesis, and that it is activated by the inflammatory network of the tumor microenvironment. Transforming growth factor (TGF)-β signaling promotes epithelial differentiation, and thus, suppression of the TGF-β signaling pathway has been thought to promote gastrointestinal tumorigenesis. Notably, the activation of Stat3 in gp130757F/F cells desensitizes them to TGF-β by inducing inhibitory Smad7. This may be one of the mechanisms by which Stat3 promotes tumor formation [86]. IL-11 is another member of the IL-6 cytokine family and also signals through gp130. Interestingly, disruption of the IL-11 co-receptor in 130757F/F mice significantly ablated gastric tumorigenesis [88]. Because IL-11 is upregulated in human and mouse gastric tumors, these results suggest that the IL-11/Stat3 pathway, together with IL-6/Stat3, promotes gastric tumorigenesis.

Inflammatory cells that promote or suppress tumorigenesis

The major source of inflammatory cytokines and prostaglandins in tumor tissues is macrophages. Tumor-associated macrophages (TAMs) have been shown to promote the progression and metastasis of cancer [89]. TAMs can be classified into several distinct groups by their functions, such as inducing inflammation, invasion, angiogenesis, or metastasis. Macrophages are polarized to either the classical M1 type or the alternative M2 type, and it has been suggested that TAMs are polarized to M2 or M2-like types [90]. It has been shown that CD4+ T cells regulate the polarization of macrophages to the M2 type in mammary tumors [91]. It has recently been shown that COX-2 is important for the M2-polarization of TAMs in Apc Min mouse tumors [92]. Accordingly, it is possible that the COX-2/PGE2 pathway-induced inflammatory network is important for the education of macrophages into the pro-tumorigenic M2 or M2-like types. Importantly, depletion of functional macrophages in Apc Δ716 mice by crossing them with op/op mutant mice resulted in significant suppression of intestinal polyposis [93]. Moreover, inhibition of macrophage recruitment in AOM/DSS-treated mice by Ccl2 gene disruption resulted in the suppression of colon tumor development [74]. Accordingly, macrophages play an important role in both sporadic and IBD-related intestinal tumorigenesis (Fig. 2). In Gan mouse gastric tumors, macrophage depletion caused atrophic changes of the tumor cells and apoptosis of stromal cells [49]. Accordingly, it is possible that macrophages play a role in the maintenance of both tumor epithelial cells and stromal cells.

In the Apc Δ468 mouse intestinal tumors, mast cells are preferentially enriched in the polyp tissues, and these mast cells express TNF-α [76]. Notably, the depletion of mast cells in Apc Δ468 mice by bone marrow transplantation from Kit Wsh/Wsh mice caused significant suppression of intestinal polyposis, with a decreased level of TNF-α expression. It is therefore possible that both macrophages and mast cells are important components of the inflammatory network in the tumor microenvironment.

Bone marrow transplantation from Rag2−/− mice did not affect the intestinal tumorigenesis in Apc Δ468 mice [76]. Consistently, the gastritis phenotype was not altered in the K19-C2mE Rag2−/− mouse stomach [78]. These results indicate that lymphocytes are not required for the construction of the inflammatory network and the promotion of tumorigenesis in the gastrointestinal tract. Importantly, however, the adaptive transfer of CD4+ CD25+ regulatory T cells to Apc Min mice dramatically reduced the number of intestinal polyps, with the induction of necrosis of tumor cells [94, 95]. Moreover, such regression of intestinal polyps was not found when CD4+ CD25+ cells were prepared from Il10−/− mice.

The transfer of CD4+ CD25+ T cells to Helicobacter hepaticus-infected Rag2−/− mice, another IBD-related colon cancer model, resulted in suppression of colitis and tumor development, but Il10-disrupted CD4+ CD25+ T cells could not suppress the development of CAC [23, 24]. IL-10 suppresses inflammatory responses, thus indicating that regulatory T cells expressing IL-10 suppress intestinal tumorigenesis by inhibiting the formation of the inflammatory network. Although CD25+ Foxp3+ T cells are found in Apc Min mouse polyp stroma, they no longer express IL-10, and instead switch to the production of IL-17 [95]. It is therefore possible that anti-inflammatory regulatory T cells (Foxp3+ IL-10+ IL-17−) shift to pro-inflammatory T cells (Foxp3+ IL-10− IL-17+) in polyp tissues [96]. Moreover, the ablation of Il17a gene in Apc Min mice significantly suppressed the development of intestinal polyps and inflammatory cytokine expression, indicating that T-cell-derived IL-17 plays an important role in intestinal tumorigenesis [97].

Concluding remarks

The development of the inflammatory network in tumor tissues and its possible roles are summarized in Fig. 2. Chronic infection or endogenous ligands derived from tumor cells stimulate the TLRs of epithelial cells, leading to the activation of MyD88. The activation of the epithelial TLR/MyD88 pathway further induces COX-2 expression and PGE2 production in stromal macrophages through the TNF-α/NF-κB pathway. Epidemiological studies and genetic experiments have demonstrated that COX-2 and its downstream product, PGE2, play an important role in gastrointestinal tumorigenesis. NF-κB is activated by TNF-α in TAMs, which further induces the expression of TNF-α, IL-6, and COX-2. TNF-α in turn stimulates both tumor epithelial cells and stromal cells, activating NF-κB in these cells, which promotes tumorigenesis. On the other hand, IL-6 activates Stat3 through gp130 in epithelial cells, thus leading to an increase in cell cycling and a decrease of apoptosis. The induction of the COX-2/PGE2 pathway is important for the development of such an inflammatory tumor microenvironment. When the TNF-α/NF-κB and/or IL-6/Stat3 pathways are activated beyond a threshold, they can promote tumorigenesis if the COX-2/PGE2 pathway is inhibited. In the inflammatory microenvironment, not only TAMs, but also mast cells and IL-17-expressing T cells, infiltrate and contribute to tumor development. Therefore, targeting the inflammatory network in tumor tissues by the inhibition of PGE2, NF-κB, Stat3, or downstream pathways may provide an effective preventive or therapeutic strategy against gastrointestinal cancer.