Immune Chaperone gp96 Drives the Contributions of Macrophages to Inflammatory Colon Tumorigenesis

The multifaceted roles of inflammation in cancer have been increasingly recognized (1, 2). One of the best pieces of evidence to support the link between unresolving inflammation and cancer is the strong clinical association between inflammatory bowel disease and increased risk of colon cancer (3, 4). Using a colitis-associated colon tumorigenesis model in mice, it was found, in general, that proinflammatory responses via direct activation of epithelial cells is important for gut homeostasis and colon cancer progression (5–9). However, the roles and mechanisms of the host immune system, such as macrophages, in colon cancer development are far from clear (10). Depending on the model system examined, inflammation could be both friend and foe for oncogenesis. MyD88-mediated signaling drives colon cancer in the APC^min mouse model (8), but is protective against colitis-associated colon cancer induced by azoxymethane (AOM) and dextran sodium sulfate (DSS; ref. 11). In addition, although microbial stimulation is important for inflammation-associated colon cancer (12), both cellular and molecular mediators in the process are yet to be defined. For example, it is unclear what specific roles macrophages play in the initiation of colon cancer, despite the well-recognized attributes of tumor-associated macrophages in promoting cancer invasion and metastasis in other models (13–16). In the case of innate microbial sensors, such as Toll-like receptors (TLR), it has also been a challenge to tease out the specific contribution by macrophage-intrinsic TLRs in colonic oncogenesis, because TLRs are expressed by both gut epithelial cells and the host immune cells. The selective deletion of MyD88 in myeloid cells is not an ideal model (17) to address this question, inasmuch as MyD88 is the common adaptor for TLRs as well as the receptors for interleukin (IL)-1 and IL-18. Moreover, the MyD88-independent pathway may also contribute to TLR-mediated oncogenesis, an idea that has not been explored.

Glycoprotein 96 (gp96), also known as glucose-regulated protein 94 (grp94) or HSP90b1, is a molecular chaperone in the lumen of the endoplasmic reticulum. Although constitutively expressed in all cell types, gp96 is strongly induced in tumor-associated macrophages with unknown functional significance (18). As a major effector chaperone downstream of the unfolded protein response (UPR), gp96 surprisingly and selectively catalyzes the folding of a majority of TLRs except TLR3 (19). Deletion of gp96 genetically leads to categorical loss of responsiveness to all TLR ligands except dsRNA (20). We previously demonstrated that LysM-Cre–mediated deletion of gp96 is highly macrophage selective, without affecting other cell populations such as dendritic cells, B cells, or T cells (21). Because of a relatively long half-life of gp96, cre-mediated deletion of its gene in neutrophils did not affect the protein expression level of gp96 during the life cycle of neutrophils. Only tissue macrophages in these mice were defective of gp96, and hence in TLR assembly. TLR signaling particularly through TLR4 was further compromised in gp96-null cells because of loss of β2 integrins that play important roles in cross-talking with TLRs through membrane recruitment of TIRAP and MyD88 (22). In this study, we took advantage of these unique macrophage-specific gp96 knockout (KO) mice (referred to as Mac96^KO mice) to address the contribution of macrophages to colitis and colon tumorigenesis. Using the AOM/DSS-induced colon tumorigenesis model, we found that macrophages promote colitis and colitis-associated colon tumorigenesis in a gp96-dependent manner, which correlated with increased levels of IL-17, IL-23, IL-12, IL-6, and IFN-γ. Moreover, we discovered that macrophage-mediated inflammation promotes genotoxic insult resulting in the activation of the oncogenic machinery in the colon. Our study has thus demonstrated the importance of macrophage-intrinsic gp96 and its clientele in promoting both colitis and inflammation-induced colon tumorigenesis; it has revealed the critical roles of macrophages in directly initiating oncogenesis within the gut epithelium during chronic inflammation.

LysM^CreHsp90b1^flox/flox mice were previously described (21). All animal experiments involving control or wild-type (WT; LysM^CreHsp90b1^flox/WT) and knockout (LysM^CreHsp90b1^flox/flox) mice were approved by our Institutional Animal Care and Use Committee, and guidelines were strictly followed. Wild-type and knockout mice were cohoused, and age-matched mice were used for all experiments.

Mice were administered 3% DSS dissolved in water and fed ad libitum for 5 days. Fresh DSS was provided on day 3. One cohort of mice was euthanized at day 5. Another cohort of mice was given regular drinking water from day 5 until day 8, at which point they were euthanized. Five mice per group were used. Mice were weighed, and their stool was scored daily. Colitis pathology score was obtained as follows: 0, normal epithelium; 1, minimal leukocyte infiltration; 2, moderate leukocyte infiltration; 3, severe leukocyte infiltration with no major architecture destruction; and 4, severe leukocyte infiltration with destruction of normal architecture. Stool consistency was scored as follows: 0, well-formed stool; 1, semiformed stool; 2, semiformed stool that adheres to the anus; 3, liquid stool/diarrhea. Colon length was measured ex vivo from cecum to rectum.

A total of 10 to 17 mice per group were given 12.5 mg/kg body weight AOM via intraperitoneal injection on day 1. DSS added to drinking water was given at 2.5% for 5 days on weeks 2 and 5, and then at 2% for 4 days on week 8. Mice were weighed and monitored weekly for overall health, and bled every other week for serum analysis. Mice were euthanized at 19 to 21 weeks, at which point tissue was harvested. Colons were fixed in 10% paraformaldehyde overnight, and then stained with 0.2% methylene blue to facilitate the visualization of adenomas. Tumors were counted and measured under a dissecting microscope. Colons were then frozen as in a fashion of “Swiss rolls” in OCT for later histologic analysis.

Tissue was fixed in 4% formalin at least overnight, and then switched to 30% sucrose-PBS overnight. Tissue was then frozen in OCT medium and kept at −80°C. Five micrometers of sections were cut on a Shandon Cryotome and mounted on charged slides (Fisher). Slides were processed by hematoxylin and eosin (H&E) staining and then mounted. H&E slides were scored blindly by a surgical pathologist, based on the percentage of high-grade dysplasia in lesions: 1, 0%; 2, 1% to 25%; 3, 26% to 50%; 4, 51% to 75%; and 5, 76% to 100%. For immunohistochemistry (IHC), tissue was paraffin embedded.

Immunoblot were performed as described previously (23, 24). In brief, cells or colonic tissues were lysed and homogenized respectively in radioimmunoprecipitation assay lysis buffer along with a cocktail of protease inhibitors. Total lysates were then resolved on 10% SDS-PAGE, and the proteins were transferred from the gel onto Immobilon-P membranes. The membrane was blocked with 5% nonfat milk, probed with different antibodies, followed by incubation with horseradish peroxidase–conjugated secondary antibody. Protein bands were visualized by using enhanced chemiluminescent substrate (Pierce). Intensity of the band on the scanned films was analyzed by Image J program. β-Catenin antibody was purchased from Cell Signaling Technology. Antibodies against MSH6 (ab92471) and MLH1 (ab92312) were obtained from Abcam.

Paraffin sections were cut and mounted on positively charged slides. Paraffin was removed by incubating the slides at 55°C for 30 minutes. Slides were dipped in xylene and then rehydrated through a series of ethanol and then water and PBS. For F4/80 IHC, antigen retrieval was performed by applying 20 μg/mL proteinase K to the samples. For β-catenin IHC, antigen retrieval was carried out by boiling in sodium citrate buffer. Slides were washed and then incubated in 0.3% H₂O₂ + 0.3% normal goat serum diluted in PBS. Slides were again washed and then blocked with 2% BSA plus 10% goat serum diluted in PBS for 2 hours. Slides were then stained with anti-F4/80 antibody (Santa Cruz) or rat immunoglobulin G (IgG)2b isotype control antibody, or with β-catenin antibody or rabbit isotype control antibody, at 4°C overnight. Slides were washed, incubated with biotinylated anti-rat or -rabbit antibody plus 2% goat serum diluted in PBS, washed again, and then streptavidin antibody diluted in PBS was applied (peroxidase rat IgG Vectastain ABC Kit). Slides were again washed and then developed with DAB Substrate Kit (Vector Labs). After staining, slides were counterstained with hematoxylin, washed, and then dehydrated in a series of ethanols followed by xylenes, and then mounted.

Small and large intestines were removed and flushed of content with PBS. Peyers patches were removed, the intestine was cut open longitudinally, and then cut into ∼1 cm pieces. Samples were washed 3 times in cold Hanks balanced salt solution (HBSS) + 2% FBS and vortexed for 15 seconds to help remove the mucus layer. Samples were then spun at 100 rpm on a stir plate in HBSS + 5% FBS + 1 μmol/L dithiothreitol (DTT) for 20 minutes at 37°C, and then vortexed. Samples were next spun as before in 1.3 mmol/L of EDTA + PBS at 37°C for 20 to 30 minutes, and then vortexed. This step was repeated to fully remove epithelium. Samples were washed twice in HBSS + 2% FBS and then once in RPMI, vortexing after each wash. Samples were further cut into smaller pieces, and then incubated twice with collagenase IV + 0.1 mg/mL DNAse I + 5% FBS + RPMI, and spun as above for 20 minutes. The solution was then passed through a 19-gauge syringe and a cell strainer, and washed with RPMI. Total lamina propria isolates were then analyzed for myeloid cells, or enriched for lymphocytes via Percoll centrifugation (44% and 66% 1× Percoll diluted in RPMI) at 2,800 rpm for 20 minutes without braking. The interface was collected and washed in RPMI. Live cells were counted using a hemacytometer after Trypan blue exclusion.

Leukocytes isolated from the lamina propria were stimulated with phorbol 12-myristate 13-acetate (PMA) (300 ng/mL) and ionomycin (1 μg/mL) for 4 hours in RPMI complete media. Flow cytometry was performed as previously described (25).

RNA was extracted from colon tissue by TriZol reagent, and quantified with a Nanodrop. mRNA was then reverse transcribed into cDNA using Superscript II H. Reverse Transcriptase. cDNA was then diluted so that 5 ng of RNA was used per reaction. Quantitative real-time PCR (qRT-PCR) was carried out with SYBR Green Supermix (Bio-Rad) using a Bio-Rad iCycler. Data were exported into Excel for analysis using the ΔCt (the number of PCR cycles to reach the threshold of the product detection) method and normalized to β-actin as an internal control. Plots represent fold change in mRNA.

Bone marrow cells were isolated from femurs and tibias. Bone marrow cells were cultured in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal calf serum (FCS), 100 U/mL of penicillin, 100 mg/mL of streptomycin, 2 mmol/L l-glutamine, 20% L929 conditioned medium for 6 days. Nonadherent cells were washed away. The adherent cells were harvested after trypsin digestion, which were mostly bone marrow–derived macrophages (BMDM) by phenotypic analysis.

Wild-type and gp96 knockout BMDMs were stimulated with 200 ng/mL lipopolysaccharide (LPS; 055:B5) from Sigma for 6 hours. Total RNA was extracted from wild-type and gp96 knockout BMDMs using an RNeasy Mini Kit (Qiagen). Fluorescent antisense RNA (aRNA) target preparation was performed using an Eberwine-based amplification method with Amino Allyl MessageAmp II aRNA Amplification Kit. Cy5-labeled RNA targets were hybridized to Mouse Whole Genome OneArray v2 (Phalanx Biotech Group), and slides were scanned by the Axon 4000 scanner (Molecular Devices). The Cy5 fluorescent intensity of each spot was analyzed by Genepix 4.1 software (Molecular Devices). Heat map was generated with the intensity value of each spot over the mean of the intensity of each gene from wild-type BMDMs in triplicates. Genes that were downregulated in knockout macrophages were classified using KEGG Pathway Database and those involved in Wnt signaling and cancer were highlighted. Heat map was generated demonstrating differential expression of genes in the wild-type and knockout cells, using Gene Pattern program provided by the Broad Institute.

Genes were amplified by PCR from colon cDNA samples. PCR reactions were purified with Qiagen kits and then sent for sequencing by Genewiz. Sequencing results were then aligned using Vector NTI software (Invitrogen). The sequences of β-catenin primers were: 300F gct gac ctg atg gag ttg ga, 526R gct act tgc tct tgc gtg aa (5).

Using a cesium irradiator, wild-type recipient C57BL/6 male mice were lethally irradiated at a dose of 1,100 cGy, split into 2 sessions (550 cGy each) 4 hours apart. Twenty-four hours after the first irradiation session, mice were injected with 2 × 10⁶ bone marrow cells from donor wild-type or Mac96^KO mice via the lateral tail vein. Donor bone marrow cells were obtained under sterile conditions by flushing the tibias and femurs with PBS, lysing red blood cells with ACK buffer, and suspending the cells at a concentration of 10⁷ cells/mL. Twelve weeks after bone marrow transplant, mice were treated with the same AOM/DSS course as above, except that DSS was administered in weeks 2, 8, and 11 because of increased DSS susceptibility at a different animal facility. After 16 weeks, mice were sacrificed.

Young adult murine colon (YAMC) cells (26) were obtained from Dr. Philip Howe (Medical University of South Carolina, Charleston, SC) in December 2012. The normal gut epithelial cell morphology was verified microscopically and the expression of E-cadherin was confirmed by immunoblot. The temperature-sensitive growth at 25°C because of the expression of a temperature-sensitive mutant of the SV0-40 large T antigen was confirmed. The cells were grown in DMEM (Sigma) supplemented with 10% FBS and 1% penicillin and streptomycin. Cells were grown at 33°C with 5% CO₂. For stimulation experiments, 0.3 × 10⁶ cells were seeded in each well of a 12-well plate and left to adhere and grow overnight. Cells were then left untreated, or treated with a mixture of recombinant murine cytokines (5 ng/mL IL-1β, 10 ng/mL IL-17A, and 1 ng/mL TNF-α, all from Shenandoah) for 48 hours. Invasion assay was performed using Transwell system by seeding cells in the upper chamber of a 1% gelatin-coated Transwell membrane (Corning). At 15 hours, cells were fixed in 90% ethanol for 10 minutes and stained with 1% crystal violet for 10 minutes. Cells in the lower chamber were eluted with 10% acetic acid for 10 minutes and cell number was determined by optical density (OD) at 595 nm. epithelial–messenchymal transition (EMT) markers were analyzed by immunoblot for E-cadherin and vimentin of whole cell lysates.

Serum IL-1α and IL-10 for the bone marrow transplantation (BMT) experiment were quantified using cytokine bead arrays from Eve Technologies, Inc. Quantification for IL-6 and IL-12 (p40) was performed by ELISA from R&D Systems. IL-17A ELISA antibodies were purchased from BD Pharmingen.

Two-sample Student t test was performed. Error bars represent either SD or SE, as indicated. For mouse weight in the BMT experiment, two-way ANOVA was used for statistical analysis.

We reported previously that gp96 deletion is more pronounced in tissue macrophages in Mac96^KO mice because of the long half-life of gp96 (∼7 days; ref. 21). To verify the loss of gp96 in intestinal macrophages, lamina propria cells of wild-type and knockout mice were isolated from both the small intestine and colon, and F4/80⁺ macrophages were analyzed via flow cytometry. Similar percentages of F4/80⁺ cells in the lamina propria were demonstrated, indicating no defect in recruitment of knockout macrophages into the lamina propria. Consistent with our previous findings, F4/80⁺ lamina propria macrophages from both the small and large intestine of Mac96^KO mice showed significant reduction of gp96 expression by intracellular stain using a gp96 monoclonal antibody (Fig. 1A).

To address the role of macrophage-specific gp96 in inflammation, we induced colitis by adding DSS in the drinking water. After treating wild-type and Mac96^KO mice with 3% DSS for 5 days and then analyzing at either day 5 or day 8, we saw weight loss in both groups of mice. Histologic examination, however, demonstrated significantly more inflammation in the colon of wild-type mice compared with knockout mice. This was evidenced by shortening of villi, infiltration of the colon mucosa with mononuclear cells including neutrophils, and destruction of crypt–villus structure (Fig. 1B and C). As a further indication of inflammation, wild-type colons were significantly more shortened than knockout colons on day 5 (Fig. 1D). Wild-type mice also had lower stool consistency score than knockout mice (Fig. 1E). By qRT-PCR on colon samples, we found a general trend of increased cytokine mRNA levels in wild-type colons compared with knockout ones, such as IL-1β, IL-6, IL-17A, IL-17F, IL-22, IL-23, and IFN-γ (Fig. 1F). However, macrophage density in the inflamed colon is similar between wild-type and knockout colons as indicated by equal levels of F4/80 mRNA (Fig. 1F). Collectively, deletion of gp96 from macrophages renders mice less susceptible to DSS-induced colitis.

We next analyzed the cellular components of the lamina propria from wild-type and knockout DSS-treated mice and examined the level of proinflammatory cytokines at the protein level. CD8⁺ cells in knockout mice demonstrated a trend toward lower production of IFN-γ (Fig. 1G and H). In the CD4 compartment, knockout colons contained lower percentages of Th17 (Ror-γt⁺ and IL-17A⁺) and Th1 (Tbet⁺ and IFN-γ⁺) helper T cells (Fig. 1G and I–O). Intuitively, macrophages from knockout mice secreted lower amounts of TNF-α (Fig. 1G and P), IL-6 (Fig. 1G and Q), and IL-17A (Fig. 1G and R).

We next studied the role of macrophage-intrinsic gp96 in the development of colitis-associated colon tumorigenesis by injecting wild-type and Mac96^KO mice with the carcinogen AOM, followed by chronic induction of colitis using 3 cycles of DSS. We found that wild-type mice suffered from increased tumor burden (number of tumors multiplied by the average tumor size) comparing with Mac96^KO mice (Fig. 2). At week 20, wild-type mice had increased colonic tumors by gross pathologic examination (Fig. 2A). Histologic analysis demonstrated that the colons were enriched with dysplasia and high-grade adenomas in the wild-type mice, whereas Mac96^KO mice had low-grade adenomas (Fig. 2B and C). Importantly, knockout mice had fewer (Fig. 2D) and smaller tumors, especially fewer large tumors >2 mm in diameter (Fig. 2E). The decrease in both tumor number and size in Mac96^KO mice translated into a significant reduction in tumor burden (Fig. 2F). Therefore, we conclude that deletion of gp96 in macrophages renders mice resistant against colitis-associated colon cancer.

The Wnt pathway, important in cell growth and embryogenesis, has been definitely linked to the development of colon cancer (27). By analyzing the colon samples from AOM- and DSS-treated mice, we found that Wnt target molecules, including Axin, β-catenin, c-Myc, Cox-2, cyclin D1, Fzd1, and p53, were significantly increased compared with untreated mice (Fig. 3A). Of note, colon from Mac96^KO mice expressed significantly less mRNA for the Wnt receptor Fzd1 and the downstream target gene p53. Interestingly, we also found significantly less β-catenin mRNA levels basally in the colon of knockout mice, indicating that loss of gp96 in macrophages might have an effect on baseline mucosal biology. To determine if inflammation accelerates AOM-induced mutations in β-catenin, thus driving oncogenesis, we sequenced exon 3 of β-catenin, which contains GSK3β phosphorylation sites. Mutations were absent in both untreated wild-type (n = 3) and Mac96^KO (n = 3) mice. However, whereas 67% of wild-type mice (6 of 9 mice) treated with AOM/DSS harbored mutations in exon 3, we found no mutations in AOM/DSS-treated knockout mice (0 of 14 mice; Fig. 3B). When we closely examined the mutation pattern in each mouse, we found that wild-type mice contained anywhere from 0 to 6 mutations each (Fig. 3C). These mutations in exon 3 are indeed clustered around the GSK3β phosphorylation sites (capitalized and underlined bases, respectively; Fig. 3D). By IHC, β-catenin nuclear translocation was observed (Fig. 3E). We therefore conclude that gp96 deletion from macrophages alone without alteration of any other cellular compartment protects colonic epithelium from β-catenin mutation and the activation of the canonical Wnt pathway.

In addition to the increased β-catenin mutation rate in wild-type colons, we also found that inflammation per se, without an exogenous mutagen (AOM) being present, is capable of inducing β-catenin expression (Fig. 3F). Furthermore, the higher prevalence of β-catenin mutation in wild-type colons suggests that macrophages and their mediators exert genotoxic pressure during the process of oncogenesis. If this surmise were true, one would expect to see the induction of components of the DNA repair pathway during inflammation. To examine this possibility, we looked at the mismatch repair (MMR) and base excision repair (BER) pathways, both of which have been implicated in the pathogenesis of colon cancer in human and mouse (28–30). We used qRT-PCR to examine the mRNA levels of DNA repair molecules during acute inflammation induced by DSS alone and during chronic inflammation induced by both AOM and DSS. We found that DSS alone indeed turned on the MMR pathway, evidenced by increased mRNA levels of MLH1 and MSH6 (Fig. 3G). We also found that AOM/DSS caused significant upregulation of AAG mRNA, a key component of the BER pathway. On the protein level, acute inflammation induced MLH1 and MSH6 expression in both wild-type and knockout mice (Fig. 3H and I). However, in the absence of an acute inflammatory insult; 2 months after the latest DSS exposure and 5 months after AOM exposure, knockout mice maintained higher levels of both MLH1 and MSH6 than wild-type mice (Fig. 3H and I), which could potentially contribute to the lower prevalence of β-catenin mutation rates as well as the lower tumor burden in knockout mice.

Taken together, we conclude that inflammation is genotoxic, and suggest that deletion of gp96 in macrophages protects mice from inflammation-associated colon tumorigenesis partially because of a reduced genotoxic pressure, as well as an increased DNA repair mechanism.

gp96 deletion did not significantly alter the level of tissue macrophages as evidenced by the comparable density of F4/80⁺ cells in wild-type and knockout colonic tumors by IHC (Fig. 4A). This was also supported by similar levels of F4/80 on the mRNA level (Fig. 4B). We next focused on the potential difference between wild-type and knockout macrophages in producing oncogenic inflammatory mediators that might contribute to the distinct susceptibility to colon tumorigenesis in wild-type and knockout mice. After AOM and DSS treatment, we found a significant induction of many proinflammatory cytokines in the colon tissues in both wild-type and knockout mice, including IL-1β, IL-6, IL-17A, IL-17F, IL-18, IL-22, IL-23, IFN-γ, and TNF-α (Fig. 4B). When the induction of cytokine levels was compared between wild-type and knockout mice, the levels of selective cytokines, such as IL-17A, IL-17F, IL-23, and TNF-α, were significantly reduced in knockout mice.

To further characterize the impact of loss of macrophage gp96 in gut immunity during colon tumorigenesis, we isolated lamina propria cells and performed immunophenotypic analysis to probe cytokine production at the protein level. No differences in CD4 and CD8 distribution were observed between wild-type and knockout mice (Fig. 5A and B). However, there were fewer IL-17A–producing CD4⁺ cells in the LP of untreated knockout mice, in response to stimulation by PMA and ionomycin (Fig. 7C and D). Knockout mice also possessed lower levels of IFN-γ⁺ CD4 cells at baseline and after AOM/DSS treatment (Fig. 5C and E). Consistent with this observation, serum from knockout mice at week 20 (upon sacrifice) after AOM + DSS treatment demonstrated lower levels of multiple cytokines, including GM-CSF, TNF-α, IL-2, IL-7, IL-12, IFN-γ, IL-10, as well as chemokines CCL3 and CCL4 (Fig. 5F). Importantly, the systemic levels of IL-17A, TNF-α, and IL-12p40 were significantly reduced in the knockout mice comparing with the wild-type mice during the induction phase of the colon cancer (on Day 70; Fig. 5G). Taken together, deletion of gp96 in macrophages causes selective reduction of a number of cytokines including IL-23, IL-17, and IFN-γ. Given the established oncogenic roles of IL-17 in the AOM/DSS-induced colon cancer model (31), our data strongly favor the idea that macrophages can promote colon tumorigenesis via a gp96–TLR–IL-23–IL-17 axis.

So far we have used LysM-Cre–mediated gp96 deletion to address the role of macrophage-intrinsic gp96 in colitis and colon tumorigenesis. This strategy is expected to delete gp96 from other tissue-resident antigen-presenting cells (APC) such as radio-resistant Langerhans cells (32). There was also a possibility of cre-expression in other nonhematopoietic parenchymal tissues to complicate the interpretation of our data. To address this possibility, we next reconstituted lethally irradiated wild-type mice with bone marrow cells from wild-type or Mac96^KO mice. This strategy resulted in greater than 90% donor chimerism in the gut-associated lymphoid tissues. We then treated these chimeric mice with AOM and DSS. Twenty weeks later, mice were sacrificed for analysis. As expected, we found that F4/80⁺ macrophages in the peritoneal exudate cells from knockout → wild-type bone marrow chimeric mice had nearly 100% knockdown of gp96 by intracellular stain (Fig. 6A). Similar to the straight Mac96^KO mice, the knockout chimeric mice had lower tumor burden compared with wild-type → wild-type mice (15 mm² in knockout compared with 33 mm² in wild-type mice, P = 0.04; Fig. 6B). During the course of tumor induction, wild-type mice lost significantly more weight than knockout mice in the AOM/DSS-treated group (Fig. 6C). In addition, we observed significantly higher levels of serum IL-1α and lower levels of IL-10 in knockout mice compared with wild-type mice (Fig. 6D). Immunophenotypic analysis of the gut-associated lymphoid tissues also demonstrated decreased CD4⁺ Foxp3⁺ Treg cells in knockout mice (21% of CD4⁺ cells) compared with wild-type mice (25% of CD4⁺ cells, P = 0.03; Fig. 6E). As expected, macrophage density in AOM/DSS-treated mice was similar in wild-type and knockout tumors (Fig. 6F). This experiment provided further evidence to support the important role of macrophage-intrinsic gp96 in promoting colitis-associated colon tumorigenesis.

A strong causative correlation exists between inflammation and increased susceptibility to cancer. Of interest, activation of wild-type but not gp96 knockout macrophages by TLR ligands (LPS) upregulates not only the proinflammatory genes but also genes that control cell growth and survival, such as Wnt receptor, phosphoinositide 3-kinase-AKT, and NF-κB (Fig. 7A). Given the differences in IL-1β, IL-17A, and TNF-α between wild-type and Mac^96KO mice, we asked if inflammatory cytokines could directly induce cancer pathways and promote oncogenesis. To this end, we treated normal colonic murine epithelial cells (YAMC; ref. 26) with a mixture of the 3 cytokines mentioned above. We found that the cytokine mixture can induce significant EMT and invasion, but not the expression of DNA mismatch repair genes (Fig. 7B and C). Exposure of normal epithelial cells to this cytokine cocktail as short as 48 hours led to reduction of E-cadherin and induction of vimentin, which was accompanied by cell-cycle arrest but increased invasion. Testing those cytokines individually showed that TNF-α and IL-1β could independently induce YAMC cell invasion, but IL-17A could not (Fig. 7C).

The macrophage possesses excellent phagocytic and cytotoxic functions (33); it expresses multiple pattern recognition receptors such as TLRs, scavenger receptors, cytosolic NOD, and NOD-like receptors (34); it is armed with both antigen-presenting (35) and costimulatory molecules; and it produces proinflammatory cytokines during infection. Many functions of macrophage are regulated by TLRs, including cytokine production, bactericidal activity, phagocytosis, and even self discrimination versus nonself discrimination in that TLR ligand-“tainted” antigen is preferentially presented to MHC class II (36). However, the roles of macrophage in cancer have not been completely resolved (1, 37). gp96 is a master chaperone for TLRs and integrins (19–21). Loss of gp96 from macrophages results in profound disturbance of their function (21). Thus, by selectively deleting gp96 from macrophages, we were finally able to tease out the roles of this important cell type in inflammation and cancer. This work is also significant because gp96 is strongly induced in tumor-associated macrophages with unknown functional consequences (18). In this study, we specifically focused on the roles of macrophages in colitis and colitis-associated colon tumorigenesis. We demonstrated the critical roles of macrophages via gp96 and its clientele in conferring intestinal inflammation, inducing genetic instability in the colonic epithelium, and promoting colitis-associated colon tumorigenesis. Because oncogenesis occurs in the wild-type epithelium, our study has conclusively demonstrated for the first time that macrophage-intrinsic gp96 plays critical roles in trans in promoting inflammation-associated cancer in the gut mucosa.

One fundamental question is the mechanism by which macrophage-derived inflammatory mediators drive oncogenic program. Our study suggests that the upregulated Wnt activity is the key missing link based on the following considerations. First, consistent with findings that the AOM/DSS model is associated with mutations within exon 3 of β-catenin (5), we found a high rate of mutation in colon tissue in the AOM/DSS-treated wild-type mice. Strikingly, however, no mutation was found in that of knockout mice. Because these mutations are in and around the GSK3β phosphorylation sites, β-catenin is no longer phosphorylated in wild-type mice and will thus be liberated from the β-catenin–destruction complex, contributing to the increased Wnt activity in wild-type mice. Second, looking at other members of the Wnt pathway, we further see a reduction of the Wnt receptor Fzd1 and the downstream target p53 in knockout mice. This decreased Wnt pathway activation in knockout mice is consistent with less tumor burden, as Wnt activation is also seen in patients with colon cancer (38). Third, both IL-1β and TNF-α have been linked to Wnt activation in a variety of cancer models (39–42). These 2 cytokines were significantly reduced in Mac96^KO mice. Taken together, our data favor a model to support the roles of macrophage-intrinsic gp96 in promoting cancer via activating Wnt pathway in colonic epithelial cells.

We have examined the possibility of compromised DNA repair pathway as an alternative mechanism for macrophage-associated carcinogenesis. The MMR pathway of DNA repair has been associated with both human and murine colon cancer (28, 29, 43, 44). Even though the BER pathway is not as highly associated with spontaneous colon cancer as the MMR pathway, it has been shown that AAG^−/− mice are more susceptible to AOM/DSS-induced colon cancer (28, 30). Furthermore, the MMR pathway is responsible for fixing the O⁶-methylguanine adducts formed by AOM (43). Of interest, we saw that DSS alone without exposure to carcinogen induced the expression of MMR machinery, indicating indeed that inflammation imposes genotoxic pressure. Importantly, upon treatment with AOM and DSS, we found that both MLH1 and MSH6 were maintained at higher level in knockout comparing with wild-type colonic epithelium, raising an intriguing possibility that chronic inflammation destabilizes the DNA repair machinery.

The reduced tumor burden in Mac96^KO mice correlated significantly with decreased IL-17A, IL-17F, TNF-α, and IL-23. We also demonstrated that the treatment of normal colonic epithelial cells with a combination of IL-17, TNF-α, and IL-1β could cause EMT. To note, Langowski and colleagues demonstrated that IL-23, but not IL-12, was overexpressed in various human cancers, including colon cancer. They further showed that IL-23 promoted tumorigenesis in the 9,10-dimethyl-1,2-benzanthracene and 12-O-tetradecanoyl-phorbol acetate model of skin papillomas, whereas IL-12 proved protective (45). The tumor-promoting abilities of IL-23 have also been proven in the 3′-methylcholanthrene fibrosarcoma model (46). Besides cancer, IL-23 acting upon T cells proved important for promoting inflammatory colitis in the T-cell adoptive transfer model (47). IL-23 is also important clinically, where both the cytokine and its receptor are elevated in patients with ulcerative colitis and Crohn disease (48). A correlation is further seen between inflammatory bowel diseases and mutations in IL-23R (49). Even IL-17 has been shown to aid tumor growth by promoting angiogenesis (50), and its expression is elevated in patients with ulcerative colitis (48). Taken together, our data strongly suggest that macrophage-intrinsic gp96 promotes colon tumorigenesis via inducing IL-17 and IL-23 in the tumor microenvironment.

Few studies link macrophage-specific TLRs to cancer. Given that gp96 is a master chaperone for TLRs, our study indicate the roles of TLR sensing on myeloid cells in oncogenesis. However, it is important to point out that gp96 client network is not restricted to TLRs, but also include other important receptors such as integrins (19–21, 25, 51), insulin-like growth factors (52), and others. In addition, gp96 is a major effector chaperone of UPR and plays important roles in regulating endoplasmic reticulum homeostasis (53). Therefore, the immediate downstream mechanism of gp96 in regulating macrophage biology and cancer is likely complex. Nevertheless, our study has positioned gp96 as a strategically important chaperone in integrating stress and innate immunity, both of which could play important roles in macrophage biology and the tumor microenvironment to promote oncogenesis. Thus, macrophage-directed inhibition of gp96 may prove to be an attractive strategy for treatment of cancer in the future.

No potential conflicts of interest were disclosed.

Conception and design: C. Morales, S. Rachidi, E. Garret-Mayer, J. Wu, B. Liu, Z. Li

Development of methodology: C. Morales, S. Rachidi, X. Ouyang, Y. Zhang, Z. Li

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Morales, S. Rachidi, F. Hong, X. Ouyang, C. Wallace, B. Liu, Z. Li

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Morales, S. Rachidi, S. Sun, X. Ouyang, B. Liu, Z. Li

Writing, review, and/or revision of the manuscript: C. Morales, S. Rachidi, Z. Li

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Morales, F. Hong

Study supervision: Z. Li

This work was supported by NIH grants AI070603 and AI077283 (Z. Li), and by the Flow Cytometry Shared Resource, Hollings Cancer Center, Medical University of South Carolina (P30 CA138313). Z. Li is also the Abney Chair Remembering Sally Abney Rose in Stem Cell Biology & Therapy, and is supported by the SmartState Endowed Chair Program of South Carolina, USA.

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.