The protein disulfide isomerase AGR2 is essential for production of intestinal mucus

AGR2 Is Present Within the ER of Intestinal Secretory Epithelial Cells.

We first examined the localization of AGR2 in the intestine. In the small intestine, AGR2 was found within a subset of cells near the base of the crypt, where Paneth cells and pregoblet cells are found (13), and in cells with goblet cell morphology along the villi (Fig. 1A). AGR2-stained regions also stained with antibodies recognizing the ER proteins GRP78 and GRP94, indicating that AGR2 was present within the ER of these intestinal epithelial cells. The majority of epithelial cells lining the intestinal villi are absorptive enterocytes, and the lack of AGR2 staining in most of the epithelium is consistent with prior studies showing that AGR2 is not present in these nonsecretory cells (8, 14). Immunohistochemical staining of colon goblet cells showed AGR2 concentrated in areas surrounding Alcian blue-stained mucous granules, but little, if any, AGR2 staining was seen within these granules (Fig. 1B). Consistent with a previous report (12), an epitope-tagged AGR2 construct colocalized with the archetypical family member PDI (also known as P4HB) in transfected cells (Fig. 1C), providing further evidence that AGR2, like other PDI family members, is present within the ER.

AGR2 Is Essential for Mucus Production in the Intestine.

To investigate the role of AGR2 in vivo, we produced mice with a null mutation in the Agr2 gene. Agr2 mRNA and AGR2 protein were expressed in Agr2^+/+ control mice but could not be detected in Agr2^−/− mice (Fig. 2 A and B and Fig. S1). Periodic acid-Schiff, which stains heavily glycosylated proteins, including mucins, revealed mucus-containing goblet cells within the small intestine and colon of control mice, but staining was markedly reduced in Agr2^−/− mice (Fig. 2 C and D). Furthermore, MUC2 core protein was abundant in control mice but undetectable in Agr2^−/− mice (Fig. 2 E and F). Pulse–chase analysis revealed synthesis of MUC2 in colon explants from control mice, but there was little or no MUC2 synthesis detected in Agr2^−/− explants, even in the presence of a proteasome inhibitor (Fig. S2). Taken together, these results demonstrate that AGR2 has an essential role in mucin and mucus production in the intestine.

AGR2 Is Not Required for Establishment of Intestinal Secretory Epithelial Cell Lineages.

Intestinal epithelial stem cells near the base of the crypt give rise to absorptive enterocytes and to a common precursor for all 3 secretory cell types (goblet cells, Paneth cells, and enteroendocrine cells) (15). Because AGR2 is found in all 3 secretory cell types in the small intestine, it has been suggested that AGR2 might have a general role in establishing secretory cell fate (14). However, analyses of proteins and mRNA transcripts expressed selectively in the Paneth and enteroendocrine cells showed that these lineages were still present in Agr2^−/− mice (Fig. S3). To determine whether loss of mucus and MUC2 resulted from loss of the goblet cell lineage, we analyzed expression of goblet cell-specific proteins and mRNAs. MUC2 protein was undetectable by either immunohistochemistry (Fig. 2 E and F) or immunoblotting (Fig. 3A). In contrast, Muc2 mRNA was easily detectable in control and Agr2^−/− mice (Fig. 3B), although quantitative RT-PCR demonstrated that levels were 3-fold lower in Agr2^−/− mice (Fig. 3C). Tff3, the mRNA encoding the goblet cell secretory product intestinal trefoil factor (16), was also present but reduced 3-fold in Agr2^−/− mice. TFF3 protein was detectable in Agr2^−/− mice, although the number of TFF3-stained cells and the size of the TFF3-containing granules were substantially reduced (Fig. 3 D and E). A reduction in TFF3-containing granule size has also been described in mice that lack MUC2 because of a null mutation in the Muc2 gene (17). In summary, we found that the goblet cell lineage and Muc2 mRNA were present but MUC2 protein absent in Agr2^−/− mice.

A Cysteine Residue in the Thioredoxin-Like Domain of AGR2 Forms Mixed Disulfide Bonds with MUC2.

We next considered whether AGR2 might have a direct role in processing of MUC2, which contains 215 cysteine residues (Fig. 4A). PDIs form mixed disulfide bonds with client proteins during processing in the ER. To determine whether AGR2 associates with MUC2, we performed coimmunoprecipitation experiments. We found that MUC2 from a human intestinal cell line coimmunoprecipitated with AGR2 (Fig. 4B). To analyze the AGR2–MUC2 interaction further, we coexpressed wild-type or mutant AGR2 with cysteine-rich regions of MUC2. Wild-type AGR2 coimmunoprecipitated with a 1,397-aa, cysteine-rich region from the amino terminus of MUC2 (Fig. 4C). This association was disrupted by treating cells with reducing agent. AGR2 contains a single thioredoxin-like domain with a CXXS motif, and mutation of the cysteine residue in that motif abolished the association of AGR2 with the amino terminus of MUC2 (Fig. 4C). A 981-aa, cysteine-rich region from the carboxyl terminus of MUC2 also associated with wild-type AGR2 but not with mutant AGR2 lacking the cysteine residue. These results show that the single cysteine residue in the AGR2 thioredoxin-like domain can form mixed disulfide bonds with cysteines in the amino-terminal and carboxyl-terminal portions of MUC2.

AGR2-Deficient Mice Had an Increased Incidence of Rectal Prolapse and Were More Susceptible to Experimentally Induced Acute Colitis.

A total of 30 litters produced by breeding an Agr2^+/− male with an Agr2^+/− female included 240 pups that survived until weaning at 3–4 weeks of age, and 31% of these pups were Agr2^−/−, indicating that lack of AGR2 did not adversely affect viability during early life. Mice that lack intestinal mucin because of null mutations in Muc2 may develop spontaneous colonic inflammation and weight loss, depending on genetic background and/or environmental factors (17, 18). Despite a lack of intestinal mucus, young Agr2^−/− mice appeared healthy, did not have diarrhea, and had body weights similar to those of wild-type littermates until at least 4 months of age (Fig. 5A). Routine histological examination of the colon again revealed the lack of mucus-filled goblet cells but showed no apparent inflammation (Fig. S4A). DNA microarrays were used to further analyze the effects of AGR2 deficiency (Table S1). The transcript with the largest fold decrease encodes cytochrome P450 1A1, a xenobiotic-metabolizing enzyme that is regulated in response to toxins and certain dietary components (19). There were prominent increases in mast cell proteases in small intestine from Agr2^−/− mice. Toluidine blue staining of mast cells confirmed that there were increased numbers of these innate immune cells in the intestine in Agr2^−/− mice (Fig. S4B). We also found decreased expression of kallikrein 1 and several related peptidases. Goblet cells are a prominent source of kallikreins, and goblet cell kallikrein expression is substantially reduced in inflammatory bowel disease (20). Despite the lack of overt disease, colon from Agr2^−/− mice had increased expression of several proinflammatory cytokines (Table S2). In addition, expressions of Hsp90b1 (an ER chaperone) and Sec61a1 (which encodes an ER pore protein involved in the removal of misfolded proteins) were increased in Agr2^−/− colon. Similar changes in these genes were recently identified in mice with Muc2 mutations that cause MUC2 misfolding (21). Although we did not detect overt colitis in young mice, 5 of 8 Agr2^−/− mice that we followed for 1 year after birth developed rectal prolapse, a condition that is seen in various mouse models of intestinal inflammation (22, 23). None of the 7 Agr2^+/+ littermate controls developed rectal prolapse (P = 0.03 by Fisher's exact test). In summary, Agr2^−/− mice had increases in mast cells, increased expression of some genes associated with inflammation and MUC2 misfolding, and an increased incidence of rectal prolapse.

To determine whether Agr2^−/− mice had altered susceptibility to experimentally induced colitis, we used a widely used model involving oral administration of dextran sodium sulfate (DSS) that mimics some features of human inflammatory bowel disease (24, 25). Administration of DSS in the drinking water for 7–8 days produces a dose-dependent acute colitis characterized by diarrhea, weight loss, and colonic mucosal ulceration. We exposed Agr2^+/+ controls and Agr2^−/− mice to a concentration of DSS (1.5%) that is lower than the dose usually used to induce colitis. All 6 Agr2^+/+ control mice maintained a normal body weight, and none had bloody stool (Fig. 5 B and C). In contrast, all 6 DSS-exposed Agr2^−/− mice had bloody stool and lost weight. DSS exposure resulted in obvious colonic epithelial damage in Agr2^−/− mice but not in the controls (Fig. 5D). These findings indicate that Agr2^−/− mice had increased susceptibility to DSS-induced colitis.

Immunodetection of Proteins in Mouse Intestine.

Antisera against AGR2 (ab22208; Abcam) and MUC2 (sc-15334 from Santa Cruz Biotechnology and HPA006197 from Sigma) and antibodies against other proteins of interest were used for immunofluorescence, immunohistochemistry, immunoblotting, and immunoprecipitation (including pulse–chase analysis) as detailed in SI Methods.

Heterologous Expression of AGR2 and MUC2 Constructs.

Plasmids encoding EGFP-tagged and Myc-tagged amino-terminal (10) and carboxyl-terminal (11) portions of human MUC2 were gifts from G. C. Hansson (Göteburg University, Gothenburg, Sweden). The production of a control EGFP-myc construct and FLAG-tagged AGR2 constructs is described in SI Methods. Cos-7 and HEK-293T cells were grown in DMEM with 10% FBS and transfected by using Lipofectamine 2000 (Invitrogen) or calcium phosphate and were analyzed 36–48 h after transfection.

Generation of Agr2^−/− Mice.

The gene-targeting strategy is described in SI Methods and summarized in Fig. S1. Mice were housed in a specific pathogen-free facility. The University of California, San Francisco Committee on Animal Research approved the use of mice for these experiments.

Measurements of Intestinal mRNA Expression.

RNA isolates from small intestine and colon of Agr2^−/− and control Agr2^+/+ littermates were analyzed by using quantitative RT-PCR and Agilent oligonucleotide microarrays as described in detail in SI Methods.

Colitis Model.

Agr2^−/− mice and control Agr2^+/+ littermates were given drinking water containing 1.5% DSS (40 kDa; MP Biomedicals). Body weight was measured and stools were examined for the presence of gross blood each day. After 8 days, mice were killed for histological analysis.

Statistical Analysis.

Statistical analyses of microarray data are described in SI Methods. For other data, P values for comparisons between Agr2^+/+ and Agr2^−/− mice were obtained by using Student's t test unless otherwise indicated.