Loss of intestinal O-glycans promotes spontaneous duodenal tumors

Abstract

Mucin-type O-glycans, primarily core 1- and core 3-derived O-glycans, are the major mucus barrier components throughout the gastrointestinal tract. Previous reports identified the biological role of O-glycans in the stomach and colon. However, the biological function of O-glycans in the small intestine remains unknown. Using mice lacking intestinal core 1- and core 3-derived O-glycans [intestinal epithelial cell C1galt1^−/−;C3GnT^−/− or double knockout (DKO)], we found that loss of O-glycans predisposes DKO mice to spontaneous duodenal tumorigenesis by ∼1 yr of age. Tumor incidence did not increase with age; however, tumors advanced in aggressiveness by 20 mo. O-glycan deficiency was associated with reduced luminal mucus in DKO mice before tumor development. Altered intestinal epithelial homeostasis with enhanced baseline crypt proliferation characterizes these phenotypes as assayed by Ki67 staining. In addition, fluorescence in situ hybridization analysis reveals a significantly lower bacterial burden in the duodenum compared with the large intestine. This phenotype is not reduced with antibiotic treatment, implying O-glycosylation defects, rather than bacterial-induced inflammation, which causes spontaneous duodenal tumorigenesis. Moreover, inflammatory responses in DKO duodenal mucosa are mild as assayed with histology, quantitative PCR for inflammation-associated cytokines, and immunostaining for immune cells. Importantly, inducible deletion of intestinal O-glycans in adult mice leads to analogous spontaneous duodenal tumors, although with higher incidence and heightened severity compared with mice with O-glycans constitutive deletion. In conclusion, these studies reveal O-glycans within the small intestine are critical determinants of duodenal cancer risk. Future studies will provide insights into the pathogenesis in the general population and those at risk for this rare but deadly cancer.

mucins are divided into either gel-forming polymeric forms secreted onto external surfaces of mucosal sites to protect from insult, or membrane-bound forms for signaling functions (35). The intestine is a major site of mucin expression. Intestinal epithelial cells produce both mucin forms; however, goblet cells of both the small and large bowel in mice and humans produce abundant amounts of MUC2 (mouse, Muc2), the major gel-forming mucin of the intestinal tract (1, 39, 40). The colon harbors a dense microbiota (10 trillion/ml) (43), where Muc2 plays an essential homeostatic role through forming an inner tissue-adherent layer as a barrier to the microbiota, and an outer layer colonized by microbiota (18). Interestingly, only a single mucus layer exists in the small intestine (19). The biological function of this small intestine mucus layer remains unclear. Evidence suggests small intestinal mucus has an immunoregulatory rather than barrier role by capturing epithelial-derived antimicrobial peptides (26) and by interacting with antigen-presenting cells to induce immune tolerance to luminal antigens (34).

Mucin-type O-linked oligosaccharides (O-glycans) make up about four-fifths of the molecular mass of mucins and contribute to their functions; although only recently have these functions been characterized (3, 3a, 4, 12, 13, 38). Previously, our laboratory demonstrated O-glycans are essential for the barrier capacity of Muc2 in the colon by preserving its ability to form a functional mucus layer and preventing spontaneous colitis and colitis-associated cancer (3a, 4, 13). O-glycans are predominantly core 1 and core 3 derived, with both synthesized on a common substrate, GalNAcα-O-Ser/Thr (also called Tn antigen) (5). Core 1 O-glycan biosynthesis is mediated by the enzyme core 1 β1,3-galactosyltransferase (C1GalT1), ubiquitously expressed in most tissues (13). C1GalT1 catalyzes the addition of galactose to Tn to form the core 1 structure Galβ1,3GalNAcα-O-Ser/Thr (also called T antigen) (20, 42). Core 3-derived O-glycans show a restricted expression pattern, mainly in salivary glands and intestine (3). It is generated by core 3 β1,3-N-acetylglucosaminyltransferase (C3GnT) in the same sites (3, 17). Both core 1 and 3 structures can be branched to form core 2 and 4 structures, respectively, and further modified by other transferases in the Golgi to form more complex structures. Interestingly, O-glycosylation defects are associated with intestinal diseases, including colon cancer (37), as well as gastric cancer and defective glycosylation (22). However, whether O-glycans influence small intestinal cancer pathogenesis remains undetermined.

Here, we report a model of spontaneous duodenal tumorigenesis induced by O-glycan deficiency in the small intestinal epithelium. These tumors develop when O-glycans are deleted in either a constitutive or inducible manner, with a tumor subset progressing to adenocarcinomas. The primary defect correlates with a loss of duodenal mucus in the lumen and an altered homeostasis in the duodenal mucosa. These findings provide insight into duodenal cancer development.

METHODS

Mice.

Intestinal core 1 O-glycan-deficient mice [C1galt1^f/f;VillinCre or intestinal epithelial cell (IEC) C1galt1^−/−] were generated as previously described (13). Mice lacking both core 1 and core 3 O-glycans [double knockout (DKO)] were generated through crossing IEC C1galt1^−/− with mice lacking core 3 O-glycans (C3GnT^−/−), as described (3a, 4). For deletion of intestinal O-glycans in adult mice, tamoxifen (TM)-inducible intestine-specific Cre system (C1galt1^f/f;VillinCreER^T2) were used as described (13), along with C3GnT^−/−;C1galt1^f/f;VillinCreERT2 mice to generate TM-DKO mice, as described (3a). Mice are on a mixed C57BL/6J and 129Sv background used with littermate controls. Animals were fed standard chow (PicoLab Rodent Diet 20). Mice were housed in a specific pathogen-free facility. All animal procedures were approved by the institutional animal care and use committee.

Tissue preservation and histology.

For histology, duodenal tissues were harvested from mice immediately after euthanization and fixed with 10% neutral buffered formalin or in Carnoy's fixative (60% methanol, 30% chloroform, and 10% acetic acid) before paraffin embedding. Paraffin-embedded sections (5 μm) were stained with hematoxylin and eosin. Pictures were taken with a Nikon Eclipse E600 microscope equipped with a Nikon DS-2Mv camera, using the software NIS Freeware 2.10.

Histological staining of mucins.

Carnoy's fixed sections were cut (5 μm), deparaffinized, and hydrated. Sections were then immersed with Alcian blue (AB) pH 2.5 (Newcomer Supply), which stains acidic glycans (13) for 20 min, and then thoroughly rinsed in tap water. Sections were then stained with periodic acid Schiff staining (PAS) by treating samples with 0.1% periodic acid for 10 min, then Schiff reagent for 10 min, followed by thorough rinsing. Sections were dehydrated, mounted with Permount, and imaged.

Histological scoring for inflammation and dysplasia.

Small intestinal inflammation scoring was determined using published protocols (10). Briefly, the total scores were based on two features: 1) inflammatory cell infiltrate, and 2) intestinal architecture. For inflammatory cell infiltrate, a score of 0 was given for no changes; 1 for mild infiltration mainly in the mucosa; 2 for moderate infiltration into the mucosa and sometimes submucosa; 3 for marked infiltration throughout the mucosa and submucosa, and sometimes transmural; and 4 for marked infiltration that was completely transmural. For intestinal architecture, a score of 0 was given for no overt changes; 1 for mild villus blunting; 2 for moderate villus blunting and mild epithelial hyperplasia; 3 for moderate hyperplasia, loss of mature goblet cells, and blunting and broadening of villi; and 4 for extensive hyperplasia and atrophy of villi and ulcerations. The total score is the sum of the two scores. Scoring was done by a clinician blinded to the experimental conditions.

Dysplasia scoring was performed using a published scheme (11). A score of 0 was given for normal tissue; 1 for hyperplasia and mild dysplastic changes; 2 for extensive low-grade dysplasia; 3 for high grade dysplasia; and 4 for invasive carcinoma. The grade of dysplasia was determined as described previously (6, 15). Cellular features characterizing low-grade dysplasia included crypt branching and elongation; low nuclear-to-cytoplasm ratio; loss of mature cells on luminal surface; and elongated, crowded, and sometimes stratified nuclei, although cells still retain polarization. Cellular features of high-grade dysplasia include highly irregular glands, high nuclear-to-cytoplasm ratio, extensive hyperchromasia, presence of ovoid/round stratified nuclei, loss of cell polarization, and dramatic nuclear atypia (irregular shapes, sizes). Invasive carcinoma was determined by the presence of dysplastic glands (high or low grade) in the submucosa or outer muscle surrounded by desmoplastic stroma. Mild dysplastic changes were determined by focal regions of low-grade dysplasia amid a primarily hyperplastic epithelium. Scoring was performed by a clinician blinded to the experimental conditions.

Immunostaining.

Formalin or Carnoy's fixed paraffin-embedded sections were deparaffinized and rehydrated. For some antibodies (e.g., Ki67), antigen retrieval was performed by heating sections between 95 and 100°C for 22 min in antigen unmasking buffer (Vector Labs), according to manufacturer's directions. For streptavidin-biotin-based staining, sections were blocked with biotin blocking solution (Vector Labs). For immunohistochemistry (IHC), endogenous peroxidase blocking was performed by incubation with 3% H₂O₂ for 10 min at room temperature (RT). For immune-epifluorescence or IHC, serum-free protein block (DAKO) was used to block nonspecific antibody binding (10 min, RT). Following blocking, sections were incubated with rabbit anti-MUC2 (Santa-Cruz H300, sc15334, 2.0 μg/ml), biotinylated mouse anti-Tn-antigen IgM (11), or rat anti-CD45 (Abcam no. 25386, 2.5 μg/ml), or rabbit anti-Ki67 (ThermoScientific RM-9106-S0, 1:100) in a humidified chamber overnight at 4°C or 2 h at ambient temperature. For CD45 detection, sections were stained with DL488-labeled goat anti-rat IgG (1:50; Jackson ImmunoResearch). For MUC2 immunofluorescent staining, sections were incubated with biotinylated goat anti-rabbit IgG (1.5 μg/ml) and then with DyLight 594-labeled atreptavidin (Thermo Scientific) (5 μg/ml) for 30 min at RT. Then sections were counterstained with 25 ng/ml 4,6-diamidino-2-phenylindole in dH₂O for 3 min, washed with dH₂O, and mounted using PermaFluor antifade (ThermoScientific). For anti-Tn-antigen IHC, sections were incubated with horseradish peroxidase-conjugated streptavidin (2 μg/ml; Jackson ImmunoResearch) for 40 min at RT, detected through diaminobenzidine staining kit (Vector Labs), counterstained with hematoxylin, dehydrated, and mounted. For epifluorescent imaging, specimens were analyzed with a Nikon Eclipse 80i microscope equipped with a Nikon DS-Qi1MC camera operating through NIS Elements AR software (version 3.0). For imaging IHC staining, sections were imaged as above. Species-specific, isotype control antibodies or nonprimary antibody-treated sections were used for negative controls.

Fluorescence in situ hybridization.

Stool pellet containing cecal or duodenal sections fixed with Carnoy's fixative were incubated with Texas red-conjugated universal bacterial probe EUB338 (5′-GCTGCCTCCCGTAGGAGT-3′; bp 337–354 in bacteria EU622773), or with a nonspecific probe (NON338 5′-ACTCCTACGGGAGGCAGC-3′) as a negative control (MWG/Operon) in hybridization buffer (20 mM Tris·HCl, pH 7.4; 0.9 M NaCl; 0.1% SDS) at 37°C overnight. The sections were rinsed in wash buffer (20 mM Tris·HCl, pH 7.4; 0.9 M NaCl) at 37°C for 15 min.

Calculation of villus spacing.

To determine villus spacing on AB-PAS-stained sections, 20 measurements along the length of neighboring villi were taken per pair, and 5–10 pairs were analyzed. Data were acquired using Image J software (version 1.46r) with a precalibrated scale and analyzed using Microsoft Excel and GraphPad Prism software.

Antibiotic treatment.

DKO mice were given filter sterilized (0.22 um) tap water containing 1 g/l each of metronidazole, neomycin sulfate, and ampicillin, 0.5 g/l of Vancomycin (Sigma Aldrich) (13), and 0.1% sucrose. Control mice were given water with 0.1% sucrose. After 3 wk, mice were euthanized. For bacterial plating, preweighed stool contents were put into 1-ml sterile PBS with 0.5 ml Zirconia-Silica beads and homogenized using a mini-bead-beater at 25 Hz for 1 min. Ten-fold serial dilutions were made, and 20-μl aliquots of each dilution were plated onto sheep's blood agar under aerobic or anaerobic (GasPAK) conditions, at 37°C. Total colonies were enumerated the next day and normalized to the weight of the stool to determine colony-forming units per milligram of stool contents.

Statistics.

All data are presented as mean ± SD, unless otherwise indicated. Student's t-test was used to analyze the differences between two groups, and one-way ANOVA followed by Bonferroni posttest was used to analyze the difference between three or more groups, unless stated otherwise. A P value < 0.05 was considered to be significant. Data were plotted and analyzed using GraphPad Prism version 5.0.

RESULTS

Absence of intestinal mucin-type O-glycans leads to spontaneous duodenal tumorigenesis.

Core 1- and core 3-derived O-glycans are essential for colon homeostasis (3, 13). To determine whether this was true in the small intestine, we crossed our previously generated C3GnT^−/− mice (3) with intestinal epithelial cell-specific C1galt1^−/− mice (IEC C1galt1^−/−) (13) to generate mice lacking both O-glycan types throughout the intestinal tract (DKO) (3a, 4).

Although uncommon, the risk of developing small intestinal carcinoma increases with age (2). Therefore, we looked at the effect of O-glycan deficiency on the small intestine in DKO mice ∼1 yr of age. In the small intestine between about 12 and 20 mo, we found moderate thickening in different regions of the intestinal tract in a subset of DKO mice but not wild-type (WT) mice. This thickening was most pronounced in the proximal duodenum, immediately after the pyloric sphincter (Fig. 1A, arrow). Numerous raised round lesions within the first 2 cm of the duodenum were found in DKO mice, but not age-matched WT controls (Fig. 1B, arrows). Histological examination showed these regions to be composed of hyperplastic duodenal mucosa overlaying submucosal Brunner's glands (Fig. 1C). These legions were devoid of obvious villous structures and contained dysplastic epithelium, which suggest adenomatous polyps (Fig. 1, C and D). In contrast, WT mucosa appeared normal, with clear regular villous structures observed (Fig. 1C). Similar hyperplastic mucosa could be seen in regions about 1 cm distally when Brunner's glands were no longer seen, indicating this phenotype was independent of these glands (data not shown). These lesions appeared similar in mice >20 mo of age, although they exhibited a more aggressive phenotype, including high-grade dysplasia and carcinoma (Fig. 1, C and D). Overall, ∼27% of DKO mice showed evidence of spontaneous duodenal tumors, whereas no WT mice exhibited tumor development in this region (Fig. 1E). On average, four lesions were observed per mouse, with sizes ranging from 3 to 6 mm in diameter (Fig. 1, F and G). Thirty percent of DKO mice exhibited thickening in the terminal ileum, although histology revealed no tumor development in this region (data not shown). Immunohistochemical staining of the proliferation marker Ki67 (24) revealed aggressive growth characteristics of a duodenal tumor (Fig. 1H). Ki67 staining was primarily in DKO duodenal mucosal epithelial nuclei compared with cells within Brunner's glands, further revealing the epithelial origin of these tumors (Fig. 1H). These results indicate mucin-type O-glycan loss in the small intestine promotes spontaneous duodenal tumorigenesis.

Fig. 1.

Defective intestinal mucin-type O-glycans promotes spontaneous duodenal tumorigenesis. A: gross morphology of stomach and proximal duodenum of WT (19 mo old) and DKO (12 mo old) mice (n = 4–5 mice/group). Arrow, thickened region. B: luminal view of duodenum. Arrows, tumors. C: hematoxylin and eosin (H&E) staining of duodenal sections. D: dysplasia score (n = 4/time point/group). E: tumor incidence (%mice with tumors, n = 5 mice/group). F: average tumor diameter from mice up to 24 mo of age. Each data point represents 1 tumor. G: average tumor burden in <20-mo-old mice (n = 5–11 mice/group). H, left and middle: immunohistochemical (IHC) staining for Ki67 (brown) in 12-mo-old mice. Right: quantitation of Ki67⁺ cells per crypt (n = 4 mice/group). *P < 0.05, two-way ANOVA followed by Bonferroni post test. Scale bars: 200 μm (C and H). B.G., Brunner's glands.

O-glycan deficiency impairs maintenance of luminal mucus in the duodenum.

Previously, our laboratory demonstrated that core 1- and 3-derived O-glycans are critical for maintenance of the mucus barrier and protection from colon inflammation (3, 3a, 13). However, whether O-glycans contribute to mucus function in the small intestine remains undetermined. We chose 5-mo-old mice, an age before duodenal tumorigenesis. We first confirmed the functional defects in glycosylation by Tn antigen staining, which is normally not expressed extracellularly, in our DKO mice (3a, 4, 13). As expected, Tn staining was observed only in villous epithelia of DKO mice, both in enterocyte cytoplasm and goblet cell thecae; in contrast, weak staining was observed in WT mice localized to subcellular regions corresponding to Golgi apparatus where immature glycans are processed (Fig. 2, A and B). Next, we analyzed glycosylation status of the mucus through AB staining, which dyes acidic glycans blue, combined with PAS, which stains mainly neutral and acidic glycans a deep magenta (25). Consistent with our laboratory's previous work (3a, 4), goblet cells and secreted mucus stained purple in WT duodenal mucosa, indicating the presence of both acidic and neutral glycans in goblet cell mucin (Fig. 2C). In contrast, DKO goblet cells were devoid of any AB staining, showing only PAS-positive cells (Fig. 2C). This indicates a loss of acidic glycans, such as sialic acid and sulfated mucin, and presence of neutral structures, such as Tn antigen. Strikingly, little secreted mucus was observed in luminal regions or between villi in DKO mice (Fig. 2C). We also observed a strong association of O-glycosylation status with villus spacing, as the villi in WT mice were significantly more spaced apart compared with DKO villi, which were usually adherent to each other (Fig. 2C, black double arrow vs. black arrow; and Fig. 2D). To determine the relationship of this phenotype to Muc2 secretion, we performed Muc2 immunostaining and found robust staining in the lumen between villi in WT mice, but almost no luminal staining in DKO mice (Fig. 2E). There was no difference in staining between goblet cells (Fig. 2E) or in Muc2 gene expression to account for these staining differences (Fig. 2F). A similar analysis was done in dysplastic tissues of older mice (11–12 mo old), showing that no further changes to mucins in control mice occurred with age, or during the course of transformation (data not shown). These results support an association between altered mucus secretion and glycan composition in DKO mice and duodenal tumor development.

Fig. 2.

O-glycan deficiency leads to impaired luminal mucus layer in the duodenum and reduced intervillus spacing. A: representative IHC staining for Tn antigen (brown) in 5-mo-old mice. Arrowheads, goblet cells. B: quantitation of Tn staining in goblet cells. C: representative combined Alcian blue-PAS (AB-PAS) staining in 5-mo-old mice. Double arrows show intervillus spacing; single arrow, lack of spacing. D: graph of intervillus distances. Each data point represents 1 measurement. Results are pooled from n = 5 mice/group. E: representative epifluorescent staining for Muc2 in Carnoy's fixed duodenal sections. White arrows indicate luminal mucus. F: quantitative PCR (qPCR) analysis for Muc2 gene expression in duodenal sections of 5-mo-old mice. All results are representative of at least 2 experiments, 5 mice/group. *P < 0.05. Scale bars: 50 μm (A, E) and 20 μm (C).

O-glycan deficiency results in mild spontaneous duodenal inflammation.

Loss of O-glycans leads to severe spontaneous bacterial-dependent colitis (13). To assess whether lack of core 1- and 3-derived O-glycans results in duodenal inflammation, we examined histological sections of 5-mo-old DKO mice. We found the duodenal mucosa was slightly thicker and the villi wider, while the overall structure was intact compared with WT mice (Fig. 3A, top). High-power magnification revealed a modest influx of polymorphonuclear cells in the DKO mucosa (Fig. 3A, bottom). Immunostaining for the pan leukocyte marker, CD45, showed a mild increase in DKO mice at 5 mo relative to WT controls (Fig. 3, B and C). Quantitative PCR showed a minor, but significant, increase in tnfa as well as cxcl1 but no change in il18 in DKO compared with WT mice (Fig. 3D). Overall, histological scoring of duodenal mucosa did not reveal a major impact of O-glycan deficiency on duodenal tissue architecture, showing only a slight increase vs. WT mice (Fig. 3E). These results suggest that O-glycan deficiency is associated with only a modest spontaneous duodenal inflammation, unlike the severe inflammation observed in the DKO colon (4).

Fig. 3.

DKO mice exhibit modest spontaneous duodenal inflammation. A: representative H&E-stained images of duodenal tissue from 5-mo-old mice. Arrows, polymorphonuclear cells. B: representative immune-epifluorescence (IF) staining images of CD45 staining. C: quantitation of CD45⁺ cells in duodenal mucosa. D: qPCR analysis of cytokine expression in duodenal tissue from 5-mo-old mice. E: inflammation score for duodenal tissue. F: representative epifluorescent staining of EUB338-positive bacteria on Carnoy's fixed sections by fluorescence in situ hybridization. Lu, lumen. Arrows, bacteria. Results are representative of at least 2 experiments, 4–5 mice/group. *P < 0.05. Scale bars: 200 μm (A, top), 20 μm (A, bottom), 50 μm (B), and 25 μm (F).

We predicted that the difference in inflammation susceptibility in the colon and small intestine was associated with bacterial load differences, which has the highest levels in the colon (32). We assayed bacterial load with fluorescence in situ hybridization using the universal bacterial probe EUB338, to compare luminal staining between large intestine (cecum) and duodenal sections (3a). Fluorescence in situ hybridization staining showed a dense mass of bacteria in the cecal lumen compared with the duodenal lumen in both strains (Fig. 3F).

To more definitively ascertain the role of the intestinal microbiota in the duodenal phenotype in DKO mice, we administered broad-spectrum antibiotics (ABX) (neomycin, metronidazole, vancomycin, and ampicillin) to 5-mo-old WT and DKO mice in their drinking water for 3 wk and then terminated the experiment to study duodenal responses. Serial plating of colon stool homogenates revealed the efficacy of the ABX treatment on aerobic and anaerobic bacteria (Fig. 4A). However, pathological scoring of duodenal tissue did not reveal a significant difference in duodenal pathology in ABX vs. vehicle-treated mice (Fig. 4B). We next determined if ABX treatment could rescue the luminal duodenal mucus, as recently described in the colon (3a), by immune-epifluorescence staining for Muc2 on Carnoy's fixed tissue. We found ABX treatment did not increase luminal mucus in the duodenum (Fig. 4C). Correspondingly, we found ABX treatment did not increase villus spacing in DKO mice (Fig. 4D). These studies collectively show the duodenal pathology associated with spontaneous duodenal tumorigenesis is independent of the microbiota.

Fig. 4.

Broad-spectrum antibiotics do not ameliorate duodenal pathology in DKO mice. A: enumeration of culturable bacteria in stool content from 5- to 6-mo-old mice to confirm efficacy of antibiotics. B, left: H&E staining of duodenal tissue of 5- to 6-mo-old mice. Right: inflammation score (mean ± SE). C: epifluorescent staining for Muc2 in Carnoy's fixed duodenal sections. White arrows, luminal mucus. D: graph of intervillus distances. Each data point represents 1 measurement. Results are pooled from representative sections of n = 3–5 mice/group. All results are representative of 2 independent experiments, n = 3–5 mice/group. *P < 0.05, Student's t-test (B), or one-way ANOVA (Bonferroni post test) (C). Scale bars: 100 μm (B) and 50 μm (C). B.D., below detection.

O-glycan deficiency in the duodenum leads to altered intestinal homeostasis by enhancing baseline crypt proliferation.

The mucus defects, villus spacing, and duodenal morphology suggest O-glycan deletion alters duodenal homeostasis. Epithelial proliferative status is a reliable measure of mucosal homeostasis (29). Ki67 staining of duodenal tissues in 4-mo-old DKO and WT mice revealed an expansion in the proliferative zone of the duodenal crypt that reached to the base of villi in DKO mice, whereas Ki67-positive cells were restricted to the crypt in WT mice (Fig. 5A). Similar results were observed at 8 mo of age (Fig. 5B). These results link defects in mucus production to altered duodenal homeostasis. These differences may indicate a risk factor for tumorigenesis.

Fig. 5.

DKO mice have altered intestinal homeostasis by enhancing baseline crypt proliferation in the duodenum. A: representative IF for Ki67 in 4-mo-old mice. B: quantitation of Ki67-positive nuclei. Results show the mean of 10 crypt/villus structures from 3 mice/group. *P < 0.05, Student's t-test. Scale bar: 50 μm.

Inducible loss of intestinal mucin-type O-glycans predisposes adult mice to duodenal tumorigenesis.

These results demonstrate spontaneous duodenal tumorigenesis in DKO mice. We next sought to exclude developmental effects on duodenal programming during embryonic stages when O-glycans are conditionally deleted. Moreover, defects in O-glycosylation typically occur in adults (21). To circumvent these potential problems, we utilized our TM-inducible model of O-glycan deletion, which ablates O-glycosylation in adult mice (3a, 13, 27). Thus we injected 1.5- to 4-mo-old IEC C1galt ^f/f;VillinCre-ER^T2;C3GnT^−/− mice with 1 mg TM to generate TM-DKO mice. Consistent with the results from constitutive O-glycan deletion, we found TM-DKO mice developed spontaneous duodenal tumorigenesis between 12 and 24 mo (Fig. 6, A–C), exhibiting mostly proliferative epithelium, as shown by Ki67 staining (Fig. 6D). Nontumor tissues of TM-DKO mice showed only a mild phenotype (Fig. 6, C and D). Notably, the incidence of these tumors appeared to be higher in the TM-DKO (inducible deletion of O-glycans) compared with DKO mice (constitutive deletion of O-glycans, Fig. 6E vs. Fig. 1E). The number and size of the tumors were similar to those observed in DKO mice (Fig. 6, F and G). However, the tumors in TM-DKO mice appeared more aggressive, advancing to adenocarcinoma (Fig. 6B, arrow). As expected, TM-DKO showed high levels of Tn antigen staining, which indicates the efficiency in the induced deletion of mucin-type O-glycans (Fig. 6H). In parallel, we conducted similar studies in TM-IEC C1galt1^−/− mice (IEC C1galt^f/f;VillinCre-ER^T2) (3a, 13), which demonstrated a similar phenotype, indicating a critical role for C1galt1, not C3GnT, in duodenal tumorigenesis (data not shown). These results indicate that loss of O-glycans directly predisposes adult mice to duodenal adenocarcinoma.

Fig. 6.

Inducible loss of intestinal mucin-type O-glycans predisposes adult mice to duodenal tumorigenesis. A: gross morphology of duodenal tissues of 17-mo-old WT and TM-DKO mice, 15-mo posttamoxifen (TM) treatment. Arrows, tumors. B: H&E-stained images of duodenal tissues. Arrow, carcinoma. C: dysplasia score. NT, nontumor; T, tumor. D, left: representative IF for Ki67 in 17-mo-old mice (15-mo post-TM treatment). Right: quantitation of Ki67⁺ cells per crypt (n = 3–5/group). E: tumor incidence. F: average tumor diameter from mice up to 24 mo of age. Each data point represents 1 tumor. G: average tumor burden. H: IHC for Tn antigen. Results are representative of n = 5 of WT and 8 of TM-DKO mice/group. *P < 0.05 vs. WT. Scale bars: 200 μm (B) and 50 μm (D and H).

DISCUSSION

Our results show that core 1- and 3-derived O-glycans deletion leads to spontaneous duodenal tumor development. Because the small intestinal tumors were found in regions of low bacterial abundance, tumors resulted from low-grade tumor-promoting inflammation due to excessive exposure of mucosal epithelium to luminal insults, such as digestive juices. These findings underscore the importance of mucin-type O-glycans in intestinal mucosal protection.

Although the incidence of duodenal cancer is rare (0.01–0.3% in the general population), it accounts for one-half of small intestine malignancies (14). The most frequent type of duodenal tumor is adenocarcinoma (9, 28). The duodenal tumor type in DKO mice is epithelial cell derived and follows an expected hyperplasia to dysplasia to adenocarcinoma sequence over time. The high incidence of duodenal tumors in DKO mice suggests that O-glycans are critical determinants for duodenal tumor susceptibility. Interestingly, duodenal carcinoma incidence increases with specific risk factors, including familial adenomatous polyposis, celiac disease, and other disorders (8). A recent study showed duodenal tumors arise in mice with oncogenic Kras mutations fed a high-fat diet, and these tumors showed downregulated mucin production (33). These studies suggest a link between mucus barrier function and duodenal cancer susceptibility; however, how loss of mucin-type of O-glycans promotes spontaneous duodenal remains to be studied. Gel-forming mucins and their constituent O-glycans serve several key functions in the gastrointestinal tract. In the gastric mucosa, gel-forming MUC5AC and MUC6 contribute to the gastric mucus layer, which acts as a buffer between epithelium and low pH gastric secretions. Key glycans on MUC6 (i.e., α1,4-linked GlcNAc) have antimicrobial activity against Helicobacter pylori (23) and prevent the onset of spontaneous gastritis and gastritis-associated cancer, although this latter phenotype is independent of H. pylori status (22). Our recent studies show that deficiency of core 1 and 3-derived O-glycans impairs the barrier capacity of the colon mucus layer by reducing its stability to bacteria-derived proteases, leading to microbiota-dependent colitis (3a) and ultimately colitis-associated cancer (4). Similar to the colon, we observed impaired luminal expression of Muc2 normal and tumor tissues of the duodenum of DKO mice, indicating O-glycans similarly contribute to mucus layer maintenance in the small intestine. In contrast to the colon, duodenal inflammation in DKO mice was modest, yet tumors still formed. The mild form of inflammation itself was not surprising, since the bacterial burden in this region is among the lowest in the intestinal tract (32). Although a dysbiotic microbiota induced by high-fat diets (33), or conditional overexpression of innate immune receptors such as Toll-like receptor 4 (31), has implicated host-microbe interactions as potential promoters of duodenal tumorigenesis, our ABX depletion studies suggest not all phenotypes in DKO duodenum are due to aberrant host-microbe interactions. Other factors may, therefore, account for spontaneous tumor development, such as increased exposure to the cancer-associated carbohydrate Tn-antigen, which may alter cell-intrinsic behaviors (30). Alternatively, O-glycans on mucus may protect the duodenal mucosa from mucosal irritation by digestive juices, which cleave O-glycan-deficient mucus (41), and prevent chronic low-grade inflammation (or para-inflammation), which may promote a tumorigenic environment (7). Future studies should address this hypothesis.

While constitutive deletion of O-glycans promotes duodenal tumor development, the same also held true using induced deletion of O-glycans in adult mice. This is relevant as somatic mutations in the Cosmc gene, which encodes a chaperone protein, essential for C1galt1 activity, leads to core 1 O-glycans impairment (13, 21). In the murine small bowel, core 1 O-glycans predominate over core 3 O-glycans (16), and loss of C1galt1 activity exposes the carcinoma-associated antigen Tn (37) and increases susceptibility to duodenal tumors. Thus O-glycosylation defects may contribute to duodenal tumor development and/or serve as a putative biomarker for this patient population.

In conclusion, we report a novel mouse model of spontaneous duodenal tumorigenesis caused by loss of epithelial-derived, mucin-type O-glycans. These studies introduce a novel function of intestinal mucin-type O-glycans, which preserve homeostasis in the duodenal mucus barrier. Future studies will address the mechanisms underlying this phenotype and develop methodological advances to test novel therapeutic approaches for duodenal cancer.

GRANTS

This work was supported by National Institutes of Health Grants RR-018758 and R01-DK-085691; National Natural Science Foundation of China (81470825, 81172299, 81272737); Research Fellows Award 285148 from the Crohn's and Colitis Foundation of America; Oklahoma Center for Adult Stem Cell Research; and the Stephenson Cancer Center, University of Oklahoma Health Sciences Center.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

N.G., K.B., J.F., W.C., and L.X. conception and design of research; N.G., K.B., and B.X. performed experiments; N.G., K.B., J.F., and L.X. analyzed data; N.G., K.B., J.F., W.C., and L.X. interpreted results of experiments; N.G. and K.B. prepared figures; N.G. and K.B. drafted manuscript; K.B., W.C., and L.X. edited and revised manuscript; L.X. approved final version of manuscript.