Retinoic acid suppresses intestinal mucus production and exacerbates experimental enterocolitis

Altered goblet cell physiology is a hallmark of inflammatory bowel disease (IBD) pathology (Kaser et al., 2010). Individuals with Crohn’s disease (CD) or ulcerative colitis (UC) have decreased numbers of goblet cells and reduced mucus thickness at presentation; however, goblet cells are replaced during active inflammation in individuals with CD but not in those with UC, suggesting an etiological role for goblet cell abnormalities in UC (Trabucchi et al., 1986; Gersemann et al., 2009; Zheng et al., 2011). The most commonly used chemically induced animal models of IBD utilize dextran sodium sulfate (DSS) or trinitrobenzene sulfonic acid (TNBS) to induce colitis (Wirtz et al., 2007). These chemically induced models share a conspicuous loss of goblet cells with a genetic model of spontaneous colitis, the Winnie mouse, demonstrating a conserved role for mucin production in maintaining intestinal barrier function (Heazlewood et al., 2008; Eri et al., 2011).

Treatment with retinoids increases the risk of developing IBD in humans (Crockett et al., 2010). The best-explored hypothesis to explain this effect is that retinoic acid (RA) in the intestine affects the production of regulatory dendritic and T cells, leading to immune dysregulation (Eksteen et al., 2009; Iliev et al., 2009; Westendorf et al., 2009). Administration of exogenous RA is known to stimulate the release of mucin in cells of the upper airways and the addition of vitamin A to growing rats inhibits the differentiation of intestinal goblet cells (Gadzhieva and Kon, 1984). However, to our knowledge, the effects of exogenous RA on intestinal mucus secretion in response to inflammation have not been reported.

Intestinal development and physiology is highly conserved between mammals and zebrafish. Larval zebrafish intestinal physiology is amenable to relatively simple chemical-genetic interrogation by immersion of larvae in solutions of drugs and microinjection of morpholinos (modified oligonucleotides to knock down gene function) (Hama et al., 2009). Of importance to the modeling of IBD, the antimicrobial roles of NOD1 and NOD2 are conserved in zebrafish (Oehlers et al., 2011c; Oehlers et al., 2011b), as are many aspects of the host response to microbial colonization, including intestinal alkaline phosphatase expression and NFκB activation (Bates et al., 2007; Kanther et al., 2011). Furthermore, zebrafish models of IBD-like enterocolitis have been recently described (Brugman et al., 2009; Fleming et al., 2010; Oehlers et al., 2011a). Because zebrafish adaptive immunity is only functional after 4–6 weeks of development (Lam et al., 2004), larvae in the first week of development were used to investigate the effects of exogenous RA on intestinal physiology in the absence of adaptive immunity.

A range of DSS doses were initially investigated to determine the maximum non-lethal dose that could be continuously tolerated by larvae at 3 days postfertilization (dpf). A dose of 0.5% (w/v) was established to be the highest concentration that did not cause significant mortality (data not shown). DSS-exposed larvae at 6 dpf manifested liver discoloration reminiscent of TNBS-exposed larvae but otherwise appeared morphologically normal (Fig. 1A).

Enumeration of bacteria from whole larval homogenates revealed a significant (P<0.0001, Student’s t-test, n=7) increase in culturable bacterial load in DSS-, but not TNBS-, treated larvae compared with control larvae (Fig. 1B).

A leukocyte response that can be monitored visually has been recognized as a useful readout of inflammation in neutrophilic inflammation models (Hall et al., 2009; d’Alencon et al., 2010; Oehlers et al., 2011a). The location and number of neutrophils in DSS-exposed larvae was characterized by live imaging of 6-dpf Tg(mpx:EGFP)ⁱ¹¹⁴ larvae and by fluorescence activated cell sorting (FACS). Exposure to DSS resulted in a microbiota-dependent distribution of neutrophils away from the caudal hematopoietic tissue towards the intestine and epidermis, and an overall increase in neutrophil numbers, reminiscent of TNBS-induced enterocolitis (Fig. 1C). Consistent with the previously described TNBS-induced enterocolitis model, the DSS-induced enterocolitis protocol was standardized to include addition of DSS at 3 dpf and assay of outcome at 6 dpf (3 days post-exposure).

Zebrafish larvae produce nitric oxide in response to inflammatory stress, and this can be live-imaged in larvae incubated with a 4-amino-5-methylamino-2′,7′-difluorofluorescein diacetate (DAF-FM-DA) probe (Grimes et al., 2006; Lepiller et al., 2007). Larvae exposed to DSS manifested identical patterns and intensity of nitric oxide to control larvae, whereas TNBS exposure induced excess nitric oxide production in the region of the cleithrum and notochord (Fig. 1D). To confirm the fidelity of the nitric oxide readout of TNBS-induced inflammation, nitric oxide production was assessed in TNBS-exposed larvae in which inflammation had been modified by the administration of antibiotics and anti-inflammatory drugs (Oehlers et al., 2011a). The production of nitric oxide in these larvae was identical to control samples (supplementary material Fig. S1). Interestingly, the production of nitric oxide was increased in larvae treated with the non-steroidal anti-inflammatory drug 5-aminosalicylic acid. This observation suggests that administration of this drug alone induces a nitric oxide response in zebrafish larvae while not affecting other measures of inflammation (Oehlers et al., 2011a).

To examine other measures of inflammation, quantitative PCR (qPCR) was undertaken. Exposure to DSS caused an appreciable upregulation of ccl20, il1b, il23, il8, mmp9 and tnfa transcription (Fig. 1E). Interestingly, the analysis of proliferating cell nuclear antigen (pcna) transcription revealed significantly reduced transcription of this marker of cellular proliferation (P<0.01).

Further analysis of cellular proliferation was performed by adapting the resazurin reduction assay to enable fluorescent whole-mount live quantification, with confirmation by bromodeoxyuridine (BrdU) labeling of proliferative cells. Resazurin staining was most prominent in the otic vesicle and gut of 6-dpf larvae (Fig. 1F). Comparison of fluorescence from control and DSS-exposed samples indicated reduced proliferation in DSS-exposed samples. This observation was confirmed by manual counting of BrdU-positive cells in the gut and trunk of control and DSS-exposed samples. There was no evidence of induction of cell death, as assessed by acridine orange or terminal deoxynucleotidyl transferase-mediated deoxyuridinetriphosphate nick end-labeling (TUNEL) staining (data not shown).

To examine changes to intestinal goblet cells in the DSS exposure model, alcian blue staining for acidic mucins was performed on whole-mount specimens. A striking feature of DSS-exposed samples was the appearance of alcian blue staining in the intestinal bulb (Fig. 2A). We termed this accumulation of acidic mucins in the intestinal bulb the ‘mucosecretory phenotype’. Sectioning revealed the stained substance to be in the lumen of the intestinal bulb, with no evidence of staining in intestinal bulb epithelial cells (Fig. 2B). Further examination of transverse intestinal sections did not reveal overt histological changes at 6 dpf (Fig. 2C,D).

Because there was no evidence to suggest transdifferentiation of intestinal bulb cells and a similar number of alcian-blue-stained goblet cells were observed in the mid intestine of untreated, DSS- and TNBS-exposed larvae at 6 dpf (Fig. 2E), it was hypothesized that the stained substance was mucus produced by esophageal goblet cells. However, examination of alcian-blue-stained transverse esophageal sections did not reveal any changes in alcian blue staining within the esophageal goblet cells or in the thickness of mucus lining the esophageal–intestinal-bulb junction of DSS-exposed larvae when compared with controls (supplementary material Fig. S2).

To examine the effect of DSS or TNBS treatment on mucin gene expression, quantitative PCR analysis of muc (a putative zebrafish ortholog of the human esophagus-expressed MUC5 gene family) expression was carried out on control, DSS- and TNBS-exposed larvae. Exposure to TNBS, but not to DSS, increased the expression of muc (Fig. 2F).

To determine whether DSS-induced changes in goblet cell function were restricted to the esophagus, the fin fold mucus producing cells marked in Tg(kita:GAL4, UAS:EGFP) larvae were enumerated by live imaging (Feng et al., 2010; Santoriello et al., 2010). Exposure to DSS, but not to TNBS, was found to cause a small but statistically significant increase in the number of fin fold mucus-secreting cells (Fig. 2G).

Because Kit expression has been observed in zebrafish mast cells (Dobson et al., 2008), crosses with the Tg(lyzC:dsRed)^nz50 line were performed to determine whether the Tg(kita:GAL4-EGFP) construct drives expression in mast cells, because expression of lyzC has also been observed in zebrafish mast cells (Hall et al., 2007; Dobson et al., 2008). Live imaging of untreated and TNBS-challenged compound transgenic larvae by confocal microscopy failed to detect any transgene coexpression, indicating that the kita:GAL4-EGFP double transgenic line does not label mast cells (supplementary material Fig. S3).

To investigate the possibility that DSS-induced mucus accumulation in the intestinal bulb is caused by reduced peristalsis, peristalsis in untreated, DSS- and TNBS-exposed larvae was observed. However, the rate of peristalsis did not differ between control (mean 3.577, n=26) and DSS-exposed (mean 3.448, n=29) larvae (95% CI of difference −0.6085 to 0.8658, determined by ANOVA), or between control and TNBS-exposed (mean 3.571, n=21) larvae (95% CI of difference −0.7953 to 0.8063, determined by ANOVA).

To investigate the role of inflammation in the mucosecretory phenotype, larvae were co-treated with DSS and the anti-inflammatory glucocorticoid dexamethasone to suppress inflammation. Dexamethasone co-treatment at a dose of 50 μg/ml prevented DSS-induced neutrophilic inflammation in Tg(mpx:EGFP)ⁱ¹¹⁴ larvae (Fig. 3A). However, dexamethasone treatment had no effect on the mucosecretory phenotype (Fig. 3B). These data suggested that the DSS-induced mucosecretory phenotype was not dependent on inflammatory signaling.

The role of the microbiota as a trigger for the mucosecretory phenotype was investigated by co-administration of broad-spectrum antibiotics with DSS (Oehlers et al., 2011a). Depletion of the microbiota prevented the appearance of the DSS-induced mucosecretory phenotype (Fig. 3C).

Because increased mucin secretion protects against TNBS-induced colitis in mice (Krimi et al., 2008), it was hypothesized that DSS-induced intestinal bulb mucin accumulation might protect larvae against TNBS challenge. To control for possible direct interaction between DSS and TNBS, exposure to 0.25% dextran was used as an additional control for this experiment (Fig. 4A). Larvae that were coexposed to DSS and TNBS had less neutrophilic inflammation than larvae exposed to dextran and TNBS or TNBS alone (Fig. 4B,C).

To complement this experiment, larvae were exposed to either dextran or DSS for 3 days and then challenged with a high dose of TNBS (Fig. 4D). Analysis of larval survival and neutrophilic inflammation following high-dose TNBS treatment confirmed a protective effect of DSS pretreatment against TNBS-induced mortality and neutrophilic inflammation, respectively (Fig. 4E,F).

RA plays an important role in the epithelial transformation associated with Barrett’s esophagus, whereby esophageal epithelial cells adopt intestinal phenotypes (Chang et al., 2007; Cooke et al., 2008). Preliminary investigations by our group have indicated exogenous RA can induce in vivo transformation of zebrafish esophageal epithelial cells into intestinal bulb cells (M.V.F., unpublished data). RA is known to be a conserved modulator of intestinal epithelial cell differentiation in zebrafish through activation of Hoxc8 (Nadauld et al., 2004). Initial experiments suggested that coexposure to RA could suppress the DSS-induced mucosecretory phenotype (data not shown). Titration of micromolar concentrations of RA established a dose-dependent relationship between exogenous RA and suppression of the DSS-induced mucosecretory phenotype (Fig. 5A and supplementary material Fig. S4). Co-administration of 1 μM RA to 3-dpf larvae was found to profoundly suppress the DSS-induced mucosecretory phenotype, while maintaining a normal morphology.

To examine the reversibility of the DSS-induced mucosecretory phenotype, larvae were exposed to DSS from 3 dpf and administered 1 μM RA at 3, 4 or 5 dpf. DSS-induced mucosecretion was assayed at 6 dpf, revealing an exposure-duration-dependent relationship between exogenous RA and suppression of the DSS-induced mucosecretory phenotype (Fig. 5B).

To examine the effect of RA on esophageal muc expression, whole-mount in situ hybridization was carried out. Although the intestinal bulb staining for muc expression varied widely within untreated and RA-treated groups, esophageal expression of muc was profoundly reduced in larvae treated with RA (Fig. 5C).

Histological examination of alcian-blue-stained esophageal sections of RA co-treated larvae revealed a decrease in the intensity of acidic mucin staining in the anterior esophagus in larvae that were co-treated with DSS and RA (Fig. 5D). Examination of posterior esophagus sections also revealed a striking RA-induced loss of mucin staining in the pneumatic duct. However, the thickness of the mucus lining the esophageal–intestinal-bulb junction did not seem to be altered following administration of RA.

To further examine the effects of RA treatment on intestinal glycobiology, isolectin GS-IB4 was used to stain the esophagus–intestinal-bulb junction. The staining pattern of isolectin GS-IB4 was unchanged following exposure to RA, confirming the results from the examination of alcian-blue-stained esophageal–intestinal-bulb junction sections (supplementary material Fig. S5A). Furthermore, expression of RFP in the Tg(ifabp:RFP)^as200 transgenic line and neutral red uptake was unchanged, indicating normal intestinal bulb and mid intestinal specification (supplementary material Fig. S5B and S5C, respectively) (Flores et al., 2008; Oehlers et al., 2011a). Together, these data indicate that the dose and exposure parameters of RA used in this study selectively suppressed mucin production by esophageal cells but did not alter esophageal cell fate.

To determine whether RA administration affected neutrophilic inflammation in the DSS model, live imaging of Tg(mpx:EGFP)ⁱ¹¹⁴ larvae was performed (supplementary material Fig. S5D). Administration of RA did not suppress DSS-induced neutrophilic inflammation.

Because TNBS-induced enterocolitis increased the expression of muc, it was hypothesized that suppression of muc expression by exogenous RA would exacerbate TNBS-induced enterocolitis. To test this hypothesis, larvae were coexposed to RA and TNBS from 3 dpf (Fig. 6A). Analysis of larval survival revealed a marked decrease in the survival of larvae coexposed to RA and TNBS compared with larvae exposed to TNBS alone (Fig. 6B). Furthermore, neutrophilic inflammation was increased in larvae that were coexposed with RA and TNBS compared with control or mono-exposed larvae (Fig. 6C).

Because pre-exposure to DSS caused mucus secretion that seemed to protect against TNBS-induced mortality and neutrophilic inflammation, it was hypothesized that suppression of DSS-induced mucosecretion by exogenous RA would negate the protective effects of DSS pre-exposure. To test this hypothesis, larvae were coexposed to combinations of RA and DSS from 3 dpf and subsequently challenged with TNBS at 6 dpf (Fig. 6D). Analysis of larval survival following TNBS challenge showed that, although coexposure to RA did not affect the survival of untreated or dextran-exposed larvae, the protective effect of DSS pre-exposure was suppressed by coexposure with RA (Fig. 6E). Furthermore, coexposure to RA increased neutrophilic inflammation in larvae that were pre-exposed to DSS compared with larvae that were exposed to only DSS (Fig. 6F).

This paper presents a novel model of enterocolitis with increased mucin production. This model adds a useful tool for investigating inflammatory processes in zebrafish larvae. Furthermore, by using larval zebrafish, in which key cell types involved in enterocolitis can be studied in isolation from the adaptive immune system, exogenous RA has been shown to both suppress intestinal mucin production and exacerbate experimental inflammation.

The inflammatory parameters of the DSS-induced enterocolitis model seem to be largely similar to the TNBS-induced enterocolitis model (see Table 1). However, the difference in nitric oxide production between DSS- and TNBS-exposed larvae is striking because increased nitric oxide production has been previously observed as a core zebrafish inflammatory response (Lepiller et al., 2007). We speculate that the stark difference in the cellular proliferation response following exposure to TNBS (increased cellular proliferation) and DSS (decreased cellular proliferation) is a manifestation of the opposing stimulatory effects of TNBS haptenization and erosive effects of DSS on epithelial surfaces (Oehlers et al., 2011a). Further study of the zebrafish DSS- and TNBS-induced enterocolitis models will lead to insights into innate-immunity-mediated intestinal inflammation.

Comparison of alterations to mucus physiology between animal models of intestinal inflammation reveals a striking difference between phenotypes, with a reduced number of goblet cells observed in adult mice and an increased number of goblet cells or accumulation of mucus observed in zebrafish larvae exposed to a high dose of TNBS or DSS, respectively (see Table 1). From the limited number of reports using zebrafish models of intestinal inflammation, the dichotomy in responses seems to be a function of the age of the organism. Whereas goblet cells seem to be susceptible to chemically induced cell death in adult animals, goblet cells in the larval zebrafish intestine seem to both be resistant to chemically induced cell death and increase their function in response to chemically induced intestinal damage. We speculate that the heightened resilience of goblet cells in the developing intestine is due to higher levels of growth factors associated with organogenesis and maturation.

Comparison by electron microscopy of the ultrastructural morphology of the intestinal epithelium in chemically induced enterocolitis zebrafish models is necessary to complete our understanding of these models (Fleming et al., 2010). Further comparison between adult and larval zebrafish models could further our understanding of the intestinal response to damage at different stages of development.

From the results of the microbiota depletion experiments, it has been hypothesized that DSS-induced mucin is produced in response to bacterial products penetrating the intestinal epithelium as a result of the erosive effects of DSS (Iger et al., 1994). Two zebrafish mutants with an accumulation of intestinal mucus, stuffed and stuffy, were characterized in a gynogenetic screen (Mohideen et al., 2003). It is possible that these mutants might phenocopy aspects of the DSS-induced mucosecretory phenotype, and further characterization of these mutants could yield insight into intestinal mucosecretory diseases such as cystic-fibrosis-associated small intestinal bacterial overgrowth (De Lisle et al., 2006).

The recent characterization of Fusobacterium nucleatum-stimulated MUC2 expression in humans (Dharmani et al., 2011) has demonstrated that specific components of the microbiota stimulate intestinal mucus production. Because the increase in total bacterial load following exposure to DSS could either be a cause or effect of the mucosecretory phenotype, it will be intriguing to study the compositional changes to the microbiota of inflamed larval zebrafish with molecular techniques. Additionally, because the larval zebrafish is amenable to gnotobiotic techniques (Pham et al., 2008), it provides a tractable model for the study of microbiota-induced mucin production.

Important limitations of chemically induced zebrafish larval models of intestinal inflammation for the modeling of IBD are the extraintestinal effects of immersion exposure on the epidermis and the lack of T cell involvement or a chronic disease stage with IBD-like histopathology (see Table 1). Conversely, the zebrafish larval platform offers the ability to perform high resolution live imaging, rapid genetic manipulation including the targeting of IBD susceptibility genes, and a level of experimental complexity and throughput necessary for performing high content disease modifier screens that are relevant to IBD. We postulate that the most important factors that influence an IBD model are that inflammation is microbiota-dependent and responsive to existing IBD therapies.

Quantification of intestinal neutrophilic inflammation by stereomicroscopy as defined in our methodology is an attractive readout for chemical and/or genetic disease modifier screening. It is important to acknowledge that this readout is complicated by the recruitment of neutrophils to the epidermis, including the epidermis of the larval intestinal tract, following exposure to either DSS or TNBS. However, as evidenced by our previous work and this paper, there is a strong correlation between the level of inflammation [as determined by cytokine expression, neutrophil number, nitric oxide production (in the case of TNBS model) and bona fide intestinal epithelium-associated neutrophils (as determined by confocal microscopy)] and the number of intestinal neutrophils as defined by our methodology (Oehlers et al., 2011a).

Increased expression of MUC5AC is observed in IBD and other forms of colonic inflammation (Shaoul et al., 2004; Fyderek et al., 2009). Both the DSS and TNBS exposure models of zebrafish enterocolitis investigated in this report demonstrated increased mucus production, albeit by different mechanisms. Our finding that the increased mucus secretion caused by exposure to DSS protects larvae from TNBS-induced pathologies suggests a cytoprotective role for DSS-induced intestinal mucus accumulation.

Because intestinal mucus is known to be an important scaffold for antimicrobial activity in the mammalian intestine (Meyer-Hoffert et al., 2008), it is likely that zebrafish intestinal mucus also has an antimicrobial function. The zebrafish larval intestinal damage models presented in this paper and elsewhere should be tractable platforms to explore intestinal mucus function (Oehlers et al., 2011b; Oehlers et al., 2011a).

The association between retinoid treatment for dermatological conditions and the development of IBD is stronger for UC than it is for CD (Popescu and Popescu, 2011). Interestingly, UC, but not CD, is characterized by a deficient goblet cell phenotype (Gersemann et al., 2009). Our observations of RA-induced exacerbation of TNBS-induced enterocolitis, as analyzed by survival and neutrophilic inflammation, was surprising because RA has been largely demonstrated to have an anti-inflammatory effect by altering dendritic cell function (Eksteen et al., 2009; Iliev et al., 2009; Westendorf et al., 2009). We hypothesize that this unexpected result is largely due to RA-induced suppression of intestinal mucus production rather than an effect on dendritic cells, because few dendritic cells are present in larval zebrafish (Lugo-Villarino et al., 2010; Wittamer et al., 2011).

Retinoid-containing drugs are primarily prescribed to adolescents for the treatment of acne; however, few studies have examined the effect of retinoid treatment on the developing intestine. A sole study on young rats has shown that hypervitaminosis A reduces the number of small intestine goblet cells and induces an inflammatory cell infiltrate (Gadzhieva and Kon, 1984). Our study adds to these findings by demonstrating that exogenous RA suppresses protective DSS-induced mucus secretion, subsequently leading to increased mortality and neutrophilic inflammation upon further enterocolitic challenge. Collectively, these data suggest that retinoids deliver a ‘first hit’ in a multistep etiology by suppressing intestinal MUC expression in adolescents. In a subset of already susceptible individuals, reduced mucus production and impaired mucus layer restoration could then combine with other genetic and environmental factors to trigger UC. Further study is warranted to examine the manifestation of these effects in patients undergoing treatment with retinoids.

Zebrafish (Danio rerio) embryos were obtained from natural spawnings and raised until 1 dpf at 28.5°C in E3 embryo medium supplemented with methylene blue. E3 embryo medium used after 1 dpf did not contain methylene blue. Induction of enterocolitis by DSS exposure was carried out essentially as described for TNBS (Oehlers et al., 2011a), with the substitution of DSS, or dextran as a control (MP Biomedicals), dissolved in E3 embryo medium for solutions of TNBS. Briefly, 10% (w/v) stock solutions of DSS or dextran were prepared in E3 at room temperature with gentle rocking. All analyses were performed at 6 dpf unless otherwise noted. Dexamethasone (Sigma-Aldrich) was dissolved in dimethyl sulfoxide (DMSO). All-trans RA (Sigma-Aldrich) was dissolved in DMSO and diluted to working concentrations with water.

Whole-mount imaging was carried out as previously described using a Nikon SMZ1500 stereomicroscope equipped with a DS-U2/L2 camera or a Nikon D-Eclipse C1 confocal microscope (Oehlers et al., 2011a). Neutrophilic inflammation was analyzed as previously described (Oehlers et al., 2011a). Briefly, Tg(mpx:EGFP)ⁱ¹¹⁴ larvae were anesthetized in tricaine and mounted in 3% methylcellulose for imaging; ‘intestinal neutrophils’ were defined as EGFP-expressing cells within the two-dimensional boundary of the intestine from the intestinal bulb to the anus (see supplementary material Fig. S6 for illustration).

Enumeration of bacteria from larval zebrafish was carried out as previously described with the modification that overnight growth was carried out at 28.5°C to recover commensal microbiota (Oehlers et al., 2011b).

Cell sorting experiments were carried out as previously described (Oehlers et al., 2011a). Briefly, larvae were dissociated in 0.25% trypsin-EDTA (Invitrogen) and analyzed using a BD LSRII.

Live imaging detection of nitric oxide production was carried out as described (Lepiller et al., 2007). Briefly, larvae were incubated in a 10 μM solution of DAF-FM-DA (Molecular Probes) for 2 hours, rinsed in E3, anesthetized and imaged by epifluorescent microscopy.

The resazurin reduction assay was adapted from the manufacturer’s instructions. Briefly, larvae were incubated in a 1× solution of alamar blue (Invitrogen; final concentration 1 mg/ml resazurin) for 60 minutes, anesthetized and individually imaged by epifluorescent microscopy at fixed zoom and excitation intensity settings. Fluorescence intensity was quantified in ImageJ software Version 1.43 (National Institutes of Health).

Analysis of cell proliferation and death was performed as previously described using the In Situ Cell Proliferation Kit, FLUOS (Roche), the In Situ Cell Death Kit, TMR red (Roche) and acridine orange staining (Oehlers et al., 2011a). Gut proliferation was enumerated by manual counting along the length of the gut, and trunk proliferation was enumerated by manual counting of a single field of view immediately posterior to the swim bladder.

The Ambion MicroPoly(A)Purist kit was used to purify RNA from pools of whole larvae for reverse transcription (Kerr et al., 2011). Reverse transcription and quantitative PCR was carried out as previously described (Oehlers et al., 2011c). Additional primers for amplifying il23 were previously described (Holt et al., 2011). Primers to amplify muc (5′-TGTGGACCCGAGCAAAAATTA-3′ and 5′-AGCTAACTCGAATCCCACATCAC-3′) were designed in Primer Express (Applied Biosystems).

Larvae were fixed in paraformaldehyde, rinsed in acidic ethanol and stained with alcian blue. Unbound stain was removed by repeated rinsing with acidic ethanol prior to whole-mount imaging or preparation for histological sectioning. Specimens for sectioning were embedded in agarose, infiltrated with paraffin, cut into 7 μm sections, rehydrated and counterstained with nuclear fast red (Vector Labs). Histological imaging was carried out on a Leica DMR compound microscope with a DFC420C camera.

PCR with gene-specific primers (5′-CCTACACCTCCTCC-AGTTATCTGC-3′ and 5′-CACTCGCCTTTTTGTTCCACG-3′) for muc (CR854881.2-201; NCBI accession number: 100148804) were used to amplify a 908 bp product from larval cDNA. This PCR product was cloned and transcribed to create riboprobes. Whole-mount in situ hybridization was carried out as described (Thisse and Thisse, 2008), and no staining was observed in sense strand control riboprobes.

Lectin staining was carried out as described (Bates et al., 2006). Briefly, larvae were fixed for 2 hours at room temperature or overnight at 4°C, rinsed with PBST, blocked in PBST supplemented with 2% (w/v) blocking solution (Roche), incubated in PBST supplemented with 2% blocking solution and 10 μg/ml lectin, rinsed with PBST and imaged by epifluorescent microscopy.

All statistical tests were carried out with GraphPad Prism version 5.0a for Mac (GraphPad Software). Multiple comparisons within ANOVA tests were carried out by Tukey test.

We thank Alhad Mahagaonkar for expert running of the zebrafish facility; Kazuhide Okuda and Jonathan Astin for helpful discussions regarding phenotyping; and members of the zebrafish community for reagents.