Selective PGE₂ Suppression Inhibits Colon Carcinogenesis and Modifies Local Mucosal Immunity

Selective inhibition of prostaglandin E₂ (PGE₂) synthesis via pharmacologic targeting of mPGES-1 may provide chemopreventive efficacy while limiting the toxicity that is often associated with long-term use of COX-2 inhibitors (1). mPGES-1 is an inducible terminal synthase with only moderate expression under normal physiologic conditions (2). However, coordinated induction of mPGES-1 and COX-2 is often observed within a variety of cancer types (3). Our laboratory recently reported that genetic deletion of mPGES-1 in Apc mutant mice significantly suppressed intestinal tumorigenesis, a protective effect that occurred in the absence of significant metabolic shunting of the COX-2 product, PGH₂, to other prostaglandins (4).

Despite the profound suppression in tumor formation that was observed in this earlier study, the effect of mPGES-1 deletion in the colon was less apparent due to the propensity of Apc mutant mice to develop tumors in the small intestine (5). To address the issue of tissue specificity, we introduced the mPGES-1 gene deletion onto strain A mice, a line that is highly sensitive to chemical-induced colon cancer (6–8). In the following study, we examined the impact of mPGES-1 genetic status on colon carcinogenesis induced by azoxymethane (AOM), an organospecific carcinogen that causes multiple adenomas in the distal colon (8). Genetic deletion of mPGES-1 affords significant protection against AOM-induced colon cancer. Associated with the striking cancer protection, we identified a critical role for PGE₂ in the control of immunoregulatory cell expansion [FoxP3-positive regulatory T cells (Treg)] within the colon-draining mesenteric lymph nodes (MLN), providing a potential mechanism by which inhibition of mPGES-1 and inducible PGE₂ synthesis may protect against colorectal cancers (CRC).

Male A/J mice were purchased from The Jackson Laboratory and crossed with female mPGES-1 knockout (KO) mice (C57BL/6; ref. 9). mPGES-1 heterozygous mice were backcrossed onto A/J mice for 9 additional generations (N10). N10 heterozygous mice were intercrossed to generate A/J:mPGES-1 KO mice. Genotyping was carried out by tail biopsy. Mice were maintained in a temperature-controlled, light-cycled room and allowed free access to drinking water and standard diet (LM-485; Harlan Teklad).

Six-week-old wild-type (WT) and KO mice were injected intraperitoneally with AOM (10 mg/kg of body weight; Sigma-Aldrich) or vehicle control (0.9% NaCl) once a week for a total of 6 weeks. Twenty weeks after the last injection, mice were sacrificed and blood, spleen, MLNs, and colon were harvested for further analysis. Colons were flushed immediately with ice-cold PBS and excised longitudinally. Specimens were fixed flat in 10% neutral buffered formalin for 4 hours and stored in 70% ethanol. Animal experiments were conducted after approval by the Animal Care Committee (ACC/IACUS) at the University of Connecticut Health Center.

Whole-mount colons were stained with 0.2% methylene blue and the number and size of aberrant crypt foci (ACF) and tumors were scored under a dissecting microscope. Colon tumor load per mouse was determined using tumor diameter to calculate the spherical tumor volume (V = 4/3 π r³). The amount of ulcerated tissue was determined as the percentage across the entire length of colon (n = 10 per group).

Colons were paraffin embedded and sectioned at 5-μm thickness. Sections were treated with 1% to 3% hydrogen peroxide, blocked, and incubated with anti-APC (1:800; Millipore), anti-PCNA (1:150; Novocastra Laboratories Ltd.), anti-PECAM1 (1:500; Santa Cruz Biotechnologies, Inc.), anti-β-catenin (1:2,000; Sigma-Aldrich), anti-cleaved caspase-3 (1:200; Cell Signaling Technology Inc.), anti-cyclinD1 (1:20; Novus Biologicals), anti-mPGES-1 (1:5,000; Abnova), and anti-Ki-67 (1:50; Dako North America, Inc.). Sections were incubated with biotinylated secondary antibody, followed by ABC reagent (Vector Laboratories Inc.). Signal was detected using 3,3′-diaminobenzidine (DAB) solution (Vector Laboratories). Tissues were counterstained with hematoxylin.

Following antigen retrieval, sections were blocked and incubated with anti-mPGES-1 (1:5,000; Abnova) and then incubated with secondary antibody conjugated with Cy5 (1:500; Millipore). Nuclei were stained with Sytox Orange (1:10,000; Invitrogen). Staining was visualized by confocal microscopy using a Zeiss LSM 510/Confocor II and images were analyzed by LSM image browser software.

RBC-depleted spleens and MLN cells were resuspended with staining buffer (balanced salt solution, 3% FBS and 0.1% sodium azide), followed by blocking solution containing normal mouse serum, anti-Fc receptor supernatant from the 2.4 G2 hybridoma (10), and human γ-globulin. Cells were incubated with labeled primary antibodies (CD4, CD8, CD11b, Gr-1 from eBioscience; Ki-67 from BD Biosciences) and analyzed by flow cytometry as described before (11). Intracellular staining of FoxP3 was conducted according to manufacturer's protocol (eBioscience).

Spleens and MLNs were harvested from 20- to 25-week-old untreated female A/J mice, and 1 × 10⁶ cells were stimulated with phorbol-12-myristate-13-acetate (PMA; 50 ng/mL; Sigma) and ionomycin (1 μg/mL; Sigma) for 4 hours in the presence of Brefeldin A (10 μg/mL; Sigma). Intracellular staining was conducted for interleukin (IL) 10, IFN-γ, and Foxp3 (eBioscience,) as described earlier (n = 6 per group).

Colon histology was evaluated on Swiss-rolled hematoxylin and eosin (H&E) sections by a board-certified pathologist (T.T.). Colonic crypt dysplasia was diagnosed according to established criteria (12), and colon tumors were further diagnosed using histopathologic descriptions (13).

Serum was prepared from blood and samples were purified using a PGE₂ affinity column (Cayman Chemical). PGE₂ concentrations were determined by ELISA (Cayman Chemical; n = 13 per group). Serum cytokine levels were determined using a Bio-Plex Pro assay (Bio-Rad Laboratories, Inc.; n = 5 per group).

Quantification of the proliferative zone was assessed on proliferating cell nuclear antigen (PCNA)-stained tissues, randomly selecting 25 crypts per mouse for analysis (n = 5 per group). The ratio of positive cells to total cells within the crypt was calculated for each colon, and an average of the ratios were compared between WT and mPGES-1 KO mice.

Statistical analyses of tumor size and multiplicity, as well as a comparison of PGE₂ levels and the frequency of immune cells using fluorescence-activated cell-sorting (FACS) analysis, were conducted by Student's t test. A value of P < 0.05 was considered statistically significant.

mPGES-1 KO mice and WT littermates were injected with AOM and colons were harvested for analysis 20 weeks after the last injection. In response to AOM treatment, WT mice developed multiple, large, and highly vascularized tumors, primarily confined to the distal colon (Fig. 1A, arrows). In mPGES-1 KO mice, however, only several colons had macroscopically visible tumors (Fig. 1A). Tumor enumeration revealed a remarkable suppression (up to 85%) in mPGES-1 KO mice (33.6 ± 2.0 vs. 5.0 ± 0.7 in WT and KO, respectively; P < 0.0001; Fig. 1B), whereas tumor load was reduced by up to 90% (264.0 ± 29.0 vs. 24.3 ± 4.3 in WT and KO, respectively; P < 0.0001; Fig. 1C). Despite the virtually complete protection against tumor formation, the total number of ACF, a preneoplastic lesion common to the distal colon, was reduced by less than 40% (32.4 ± 2.5 vs. 20.7 ± 2.3 in WT and KO, respectively; P < 0.002; Fig. 1D). The most effective protection occurred in ACF of intermediate size, including those with 2 to 3 (16.0 ± 1.4 vs. 10.0 ± 1.4 in WT and KO, respectively; P < 0.005) and 4 to 6 crypts per focus (10.4 ± 1.3 vs. 6.3 ± 0.9 in WT and KO, respectively; P < 0.01).

In WT mice, analysis of tumor histology identified primarily 2 forms of large adenomas, characterized by either a pedunculated or flat morphology (Fig. 2, WT). The tubular adenomas often exhibited an elongated structure with a stalk (Fig. 2, arrow). In the mPGES-1 KO mice, however, only 1 of 19 mice developed a large adenoma (>3 mm), which had either a flat or a slightly raised morphology (Fig. 2, KO). Microadenomas were found in the colons from either genotype, although carcinomas in situ were limited to the WT mice (Fig. 2, WT). The microadenomas showed slight, moderate, or severe nuclear atypia without evidence of expansion and invasive growth, characteristic of AOM-induced tumors occurring in this model (refs. 7, 12; Fig. 2). The morphology of the adenomas was similar between genotypes, comprised largely of microscopically tubular structures. However, tumors in WT mice were considerably larger, with frequent compression of surrounding normal crypts. In several cases, there was evidence for moderately differentiated tubular carcinomas in situ with invasive growth (Fig. 2, WT). Analysis of crypt dynamics revealed that within the normal colonic epithelium, the absence of mPGES-1 did not affect overall crypt length, nor alter the size of the proliferative compartment in comparison to the WT colons (Supplementary Fig. S1).

The profound suppression in tumor growth observed in the mPGES-1 KO mice raised the possibility that PGE₂ levels may directly affect cell turnover within the tumor epithelium. To evaluate this possibility, colon tumors were examined immunohistochemically for PCNA staining (proliferation), and cleaved caspase-3, which detects an earlier stage of apoptosis in cells that have not yet undergone major morphologic changes (14). A total of 10 colons from each genotype were selected for immunohistochemical analyses. Representative adenomas from a WT and mPGES-1 KO colon are shown in Figure 3A. Intense PCNA staining was present throughout the tumor epithelium, regardless of mPGES-1 genotype and independent of tumor size (Fig. 3A, PCNA). Cleaved caspase-3 immunostaining revealed few apoptotic cells within the normal crypts (data not shown), and their frequency was unaffected by mPGES-1 status, nor affected by tumor size (Fig. 3A, cleaved caspase-3). These observations suggest that mPGES-1 status does not influence cell turnover in the AOM colon tumor model.

In contrast to our previous findings, whereby deletion of mPGES-1 was found to disrupt neovessel growth within small intestinal tumors in Apc^Δ14/+ mice (4), mPGES-1 deletion did not directly affect PECAM-1 staining within and adjacent to AOM-induced colon tumors (Fig. 3A, PECAM-1). Even in the smallest colon lesions examined in the mPGES-1 KO mice, PECAM-1 staining within the tumor stroma and surrounding colonic mucosa showed the presence of well-formed vascular structures (arrows; Fig. 3A, PECAM-1). The difference in PECAM-1 staining between these 2 studies, however, may result from the dissimilar experimental systems employed including differing genetic backgrounds and distinct mechanisms of tumor initiation.

We next examined the possibility that disruption of PGE₂ formation may directly impact Wnt signaling, an effect that was shown earlier in adenomatous polyposis coli (APC)-deficient DLD-1 colon cancer cells (15). Surprisingly, loss of APC protein, with increased cytoplasmic staining and nuclear localization of β-catenin, was equivalent within tumors regardless of mPGES-1 genotype or tumor size (Fig. 3B). The loss of APC protein is consistent with previous findings showing that AOM-induced tumors do not express the full-length protein (16, 17). Taken together, these data suggest that the cancer suppression associated with reduced PGE₂ formation is not related to aberrant cell turnover or dysregulated β-catenin signaling in transformed epithelial cells.

To define the localization of mPGES-1, immunofluorescence imaging was done on colonic mucosa prepared from mice harboring tumors. In the WT colons, we identified increased expression of mPGES-1 localized primarily at the apical surface of the tumor stroma (Fig. 4). Positive staining was also found within the stroma immediately adjacent to the tumors (ref. 18; Fig. 4, arrows). Similar to that reported by Murakami and colleagues (19), mPGES-1 staining was confined to the perinuclear region of the cells (Fig. 4). In addition, the expression of mPGES-1 was observed within multiple cell types including macrophages and fibroblasts (ref. 20; Supplementary Fig. S2). The location of mPGES-1 indicates that the primary source of inducible PGE₂ originates within the tumor stroma (21). Therefore, an alternative possibility for tumor suppression is that genetic deletion of mPGES-1 interrupts localized production of PGE₂ within the tumor stroma, eliciting effects that extend into the epithelial compartment.

Further evaluation of colon histology in the mPGES-1 KO mice revealed the presence of synchronous, localized mucosal ulcerations affecting up to 15% of the colonic epithelium. These cryptal lesions developed independently of AOM treatment and were characterized histologically by the presence of regenerative atypia (Fig. 5A). Mucosal ulcerations within crypt abscesses resembled the active phase of ulcerative colitis. Regenerative crypts adjacent to the ulcerated areas, as well as infiltrating immune cells, were positive for Ki-67, indicating active cell proliferation (Fig. 5A, Ki-67). However, these regenerative crypts did not share other molecular features typically associated with neoplasia. For example, APC expression was largely retained compared with the extensive loss of APC protein observed in dysplastic adenomatous crypts in the mPGES-1 KO mice (Fig. 5B, APC). Importantly, we found no evidence for β-catenin activation within these regenerative crypt lesions, with plasma membrane staining observed in all cases examined (Fig. 5B, β-catenin). Although cyclinD1, a key β-catenin target, showed intermittent nuclear staining within these epithelial structures (Fig. 5B, cyclinD1), the normal status of APC expression and β-catenin localization within these colonic structures support their nonneoplastic nature.

The mild and localized chronic inflammation observed within the colons of mPGES-1 KO mice was further substantiated by the presence of macroscopically inflamed MLNs, with a significant expansion of total lymphocytes (1.6 ± 0.3 vs. 7.4 ± 1.5 in WT and KO, respectively; P < 0.006;Fig. 6A). Total numbers of CD4⁺ and CD8⁺ cells were also markedly elevated (0.9 ± 0.2 vs. 2.9 ± 0.5 for CD4⁺ in WT and KO, respectively; P < 0.004 and 0.2 ± 0.04 vs. 0.8 ± 0.2 for CD8⁺ in WT and KO, respectively; P < 0.01, respectively; Fig. 6A), presumably a direct result of the ongoing localized inflammation. In the spleen, however, there were no significant differences in the total numbers of both CD4⁺ and CD8⁺ cells (Fig. 6A). Correspondingly, serum PGE₂ concentrations were moderately (P < 0.05) lower in the mPGES-1 KO than in WT mice (Supplementary Fig. S3A). Moreover, the concentration of a panel of pro- and anti-inflammatory cytokines in the serum was mostly unchanged between genotypes, confirming the localized effect associated with mPGES-1 deletion (Supplementary Fig. S3B). The only exception was a significant decrease in IL-1α in mPGES-1 KO mice, which was recently shown to be regulated by PGE₂ (22) and might be indicative of a stronger chronic inflammatory response.

We next investigated the immunoregulatory mechanisms that may underlie colonic inflammation in the mPGES-1 KO mice. PGE₂ has been shown in vitro to enhance the differentiation of naive CD4⁺ T cells into FoxP3-positive Tregs that have the potential to suppress effector T-cell function (23). Furthermore, Tregs play an important regulatory role in gastrointestinal immunity (24). As shown in Figure 6B, in the MLNs of mPGES-1 KO mice, the frequency of CD4 FoxP3 double-positive cells was reduced by 55% compared than in WT mice (21.1 ± 1.1 vs. 11.7 ± 1.3 in WT and KO, respectively; P < 0.0004). Importantly, this effect was not systemic, as the composition of Tregs within the spleen was unaffected by genotype (Fig. 6B).

We also examined the possibility that a population of myeloid-derived suppressor cells (MDSC), immunomodulatory cells that are often increased in tumor-bearing mice (25), may be expanded in mPGES-1 KO mice in response to the localized mucosal ulcerations. As anticipated, increased levels of GR-1 CD11b double-positive MDSCs were found in the MLN (0.03 ± 0.02 vs. 0.20 ± 0.03 in WT and KO, respectively; P < 0.00004) and spleen (2.08 ± 0.40 vs. 8.37 ± 2.76 in WT and KO, respectively; P < 0.01; Fig. 6C). However, the frequency of MDSCs was modest in comparison to changes found in other mouse tumor models. For example, in some cases, the spleen can harbor up to 40% of MDSCs within the T-cell population, depending of course on the underlying pathology (26). The present findings suggest that limited expansion of MDSCs in mPGES-1 KO mice, most likely attributed to reduced PGE₂ levels (27), contributes to the enhanced inflammatory state that is present within the colon and that may ultimately impede colon tumor progression.

On the basis of the reduced levels of Tregs and the moderate effect on MDSCs, we postulated that additional immunoregulatory mechanisms might also contribute to tumor suppression. Thus to evaluate this possibility, we harvested cells from the spleens and MLNs of untreated WT and mPGES-1 KO mice. The total number of cells harvested from the MLN of the mPGES-1 KO mice was significantly higher in comparison to the WT mice, a result of the localized colonic ulcerations (Supplementary Fig. S4). In addition, the number of CD4⁺ and CD8⁺ cells was also markedly higher in the mPGES-1 KO MLN (Supplementary Fig. S4), consistent with our findings in tumor-bearing mice (Fig. 6A). Following a 4-hour stimulation with PMA/ionomycin, the ability of CD4⁺ cells to produce IL-10 or IFN-γ was slightly impaired in the mPGES-1 KO group (Supplementary Fig. S4). These observations suggest that mPGES-1 deficiency may affect the production of regulatory T type 1 cells (Tr1), another type of immunoregulatory cell present within the gut mucosa (28).

Elevated prostanoid production in the colon plays a key role in cancer pathogenesis and efforts have been made to suppress this pathway, primarily via inhibition of COX-2 activity. Although long-term COX-2 inhibition can be effective, it has also been associated with a number of adverse effects, notably cardiovascular and gastrointestinal (GI) toxicities (29). Evidence from several tumor models provides the rationale for the development of alternative chemoprevention targets within the arachidonic acid pathway including the terminal PGE₂ synthase, mPGES-1 (3). In the present study, we provide evidence that suppression of inducible PGE₂ production through genetic deletion of mPGES-1 effectively reduces colon cancer development. We also go on to show that suppressing inducible PGE₂ formation influences cancer development in part by promoting an effective immune response to the tumor.

Remarkably, tumor suppression was so effective that only 1 of 19 mPGES-1 KO mice (5.3%) developed a colon tumor exceeding 3 mm in size. Despite this protection afforded to the colon, however, mPGES-1 deletion did not influence the frequency of ACF to the same extent. This latter observation is consistent with the recent findings of Cho and colleagues (30) who reported that within a subset of patients on the Adenoma Prevention with Celecoxib trial, adenoma suppression by celecoxib treatment was not correlated with changes in the density of ACF within the distal colorectum. It is possible that at least for agents that target the COX-2 pathway, ACF do not provide a surrogate marker for colon cancer suppression.

PGE₂ is a pleiotropic molecule that is formed within a variety of cell types and can elicit effects that are both cell- and tissue-context dependent. The precise location of inducible PGE₂ formation, however, remains somewhat controversial. For example, it is broadly accepted that tumor cell–derived PGE₂ promotes tumor growth through an autocrine mechanism. Consistent with this mechanism, mPGES-1 expression has been identified directly within the epithelial cells of colon tumors (31–33). In the present study, however, we found that mPGES-1 was localized primarily within the tumor stroma, indicating that inducible formation of PGE₂ may impair tumor expansion via non–cell autonomous mechanisms. This finding is consistent with the results of Kamei and colleagues (32), who showed that the growth of Lewis lung carcinoma (LLC) tumor cells explanted into an mPGES-1–deficient host was markedly impaired in comparison to the growth observed in an mPGES-1–competent host. In addition, a number of studies have found COX-2 expression to be confined to the tumor stroma (reviewed in ref. 34).

The depletion of inducible PGE₂ formation is associated with the development of colonic mucosal abnormalities that are reminiscent of ulcerative colitis. The lesions are restricted to the large intestine, and the histologic features of these lesions consist of crypt erosion and an influx of inflammatory cells. Interestingly, Hara and colleagues (35) recently reported that mPGES-1 KO mice show enhanced susceptibility to dextran sodium sulfate (DSS)-induced ulcerative colitis, confirming a critical role for PGE₂ in maintaining colonic epithelial barrier function under conditions of chemical-induced stress. It is possible that the localized ulcerations present within the colons of the mPGES-1 KO mice induce a chronic inflammatory condition. We further speculate that this underlying inflammation may actually be a contributing factor in the suppression of colon tumors observed in the present study.

PGE₂ is among the most potent immunoregulatory molecules within the intestinal mucosa. In addition to its modulating effects on normal gut barrier function and mucosal response to pathogens (36), inducible formation of PGE₂ also plays a critical role in the immunosuppression associated with advanced neoplasia (37). Tregs, with the potential to suppress effector T-cell function (38, 39) have been shown to play an important immunomodulatory role within the GI tract (24, 40). Within the tumor microenvironment, PGE₂ has been reported to enhance Treg differentiation by inducing the expression of Foxp3 in naive CD4⁺ T cells (23, 41, 42). In CRC patients, increased levels of Tregs were identified within the tumors as well as in the regional lymph nodes (43). Moreover, the effect of Tregs on the production of proinflammatory cytokines was reversed by treatment with nonsteroidal anti-inflammatory drugs (NSAID), further evidence for Treg dependence on PGE₂ during tumor evolution (43). Given the direct influence that PGE₂ elicits on FoxP3 expression in T cells, it is entirely possible that mPGES-1 deficiency may result in a persistent overreactive immune response due to the loss of functional activation of Tregs.

Interestingly, the present study shows that the impaired immunoregulatory response in the mPGES-1 KO mice, reflected by the spontaneous development of localized ulcerations, was not entirely accounted for by attenuated Treg expansion. mPGES-1 deletion also modestly affected the levels of MDSCs and Tr1-like cells, suggesting that the absence of inducible PGE₂ formation may disrupt fundamental immunomodulatory mechanisms within the tumor microenvironment. One possibility is that mPGES-1 deficiency causes a shift in cytokine profiles during tumorigenesis. In fact, Monrad and colleagues (44) recently showed that mPGES-1–deficient bone marrow–derived dendritic cells (BMDC) had decreased production of IL-12 in response to lipopolysaccharide (LPS) stimulation. Furthermore, our preliminary data show that MLN cells harvested from the mPGES-1 KO mice produce higher levels of several cytokines when compared with similarly stimulated WT cells (data not shown). Because the mPGES-1 genotype did not affect systemic cytokine profiles with the exception of IL-1α, a more detailed analysis of tissue-specific cytokine profiles is warranted.

A number of studies suggest that chronic inflammation may play an important role in the pathogenesis of up to 20% of human cancers (45). In particular, long-standing inflammatory bowel disease (IBD) is considered a significant risk factor for CRC (46). Although the present data appear to contradict these earlier observations, we postulate that the underlying inflammation present in mPGES-1 KO mice, resulting from the mild and localized colonic ulceration, may actually confer protection against tumor progression by providing a mechanism for active clearance of cancer-initiating cells. Consistent with this hypothesis, an earlier study by Tanaka and colleagues (13) showed that mice administered DSS prior to a single injection of AOM failed to develop colon tumors 20 weeks later. In the present study, mPGES-1 KO mice did not exhibit clinical signs of severe ulcerative colitis, such as rectal bleeding and excessive weight loss (data not shown), despite the presence of these benign, localized mucosal ulcerations. Of course, the possibility exists that the mild inflammation that is present within the colons of the mPGES-1 KO mice may contribute to subsequent cancer risk. However, without additional genetic hits, these lesions may not have the capacity to progress to cancer (47). Therefore, additional studies to better define the mechanisms by which selective suppression of PGE₂ directly modulates antitumor immunity and contributes to colon cancer suppression are underway. These studies will ultimately enable the development of effective therapeutic strategies for targeting mPGES-1 for cancer prevention.

No potential conflicts of interest were disclosed.

The authors thank Dr. Robert Clark in Department of Immunology at University of Connecticut Health Center for his advice on designing experiments for Treg analyses.

This work was supported by NIH grant CA-125691 and CA-114635 (D.W. Rosenberg).

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.