Dysregulated phosphatidylinositol signaling promotes endoplasmic-reticulum-stress-mediated intestinal mucosal injury and inflammation in zebrafish

Intestinal epithelial cells (IECs) play a major role in mucosal homeostasis, barrier function and immunity in addition to their digestive functions. Physiological stress to the IECs affects intestinal mucosal integrity, making the host susceptible to various gastrointestinal (GI) diseases. Epithelial disruption is a hallmark pathological feature of GI inflammatory disorders, particularly inflammatory bowel diseases (IBD) and necrotizing enterocolitis (NEC) (Xavier and Podolsky, 2007; Abraham and Cho, 2009; Henry and Moss, 2009).

Endoplasmic reticulum (ER) stress leading to epithelial dysfunction is believed to contribute to GI inflammation (Kaser and Blumberg, 2009; Hotamisligil, 2010; Kaser and Blumberg, 2010). ER stress results from perturbation of ER homeostasis through a multitude of factors and can trigger a conserved adaptive response, termed the unfolded protein response (UPR) or ER stress response (ERSR) (Ron and Walter, 2007). The UPR helps in protein folding capacity or ER-associated degradation (ERAD) of misfolded proteins to resolve the ER stress. However, chronic or unresolved ER stress, causing prolonged activation of UPR, can exacerbate the pathology of various human diseases (Lin et al., 2008). Recent studies on animal models of IBD have provided links between ERSR factors and GI inflammation. Mice lacking a crucial ERSR mediator, XBP-1, show cell-specific ER stress in the epithelium and develop spontaneous enteritis (Kaser et al., 2008). In humans, polymorphisms within the XBP1 locus confer an increased risk for both Crohn’s disease (CD) and ulcerative colitis (UC) (Kaser et al., 2008). In murine models, ER-stress-mediated goblet cell (GC) depletion is implicated in the pathogenesis of UC (Heazlewood et al., 2008). Differential expression of the proximal ER stress sensor HSPA5 is reported in human IBD tissues (Bogaert et al., 2011), and ER stress is hypothesized to activate pro-inflammatory signals through multiple mechanisms (Deng et al., 2004; Hu et al., 2006; Yamazaki et al., 2009). However, the precise molecular pathways leading to ER stress and the pathophysiological roles of various ERSR components in mucosal inflammation are largely unknown, necessitating the development of novel animal models to unravel these mechanisms. Furthermore, only a limited percentage of the estimated genetic heritability of IBD is explained by known genetic determinants identified by genome-wide association studies (Franke et al., 2010; Anderson et al., 2011), indicating the importance of finding new animal models to delineate specific genes and pathways that might contribute to IBD pathogenesis.

The zebrafish, Danio rerio, has been an effective tool in deciphering mechanisms of human GI diseases, due to the similarity in basic GI tissue structure, function and gene expression profiles (Stuckenholz et al., 2004; Stuckenholz et al., 2009). The zebrafish intestine is fully developed and becomes functional by 5 days post-fertilization (dpf), displaying a villous architecture with easily identifiable enterocytes, enteroendocrine cells, and the mucin-secreting GCs (Wallace and Pack, 2003; Ng et al., 2005; Stuckenholz et al., 2004). Zebrafish IECs secrete defensins and other antimicrobial peptides, and the IBD susceptibility genes nod1 and nod2 have been shown to maintain conserved antimicrobial roles in the zebrafish intestine (Oehlers et al., 2011a; Oehlers et al., 2011b). Chemical enterocolitis models in the zebrafish have shown that features of IBD seen in murine models can be rapidly recapitulated in larval zebrafish, emphasizing their utility for the study of IBD pathogenesis (Fleming et al., 2010). In addition, larval zebrafish models are being utilized to analyze interactions between the commensal microbiota and host innate immunity, providing insights into the role of bacteria and inflammation in human IBD (Kanther et al., 2011; Roeselers et al., 2011).

Phosphatidylinositol (PI) signaling has been linked to a variety of human diseases and cancer. PI is a crucial phospholipid synthesized in the ER and in highly dynamic ER-derived compartments. PI is rapidly metabolized and its levels are tightly controlled in the cell to exert its spatiotemporal intracellular signaling functions (Balla et al., 2009; Kim et al., 2011). Phosphorylated PIs (PIPs) are believed to be the regulators of vesicular transport and secretory pathways. We have previously shown by transcriptome profiling that inositol metabolism and phosphoinositide 3-kinase (PI3-K) pathways are enriched during zebrafish GI development and that inhibition of PI3-K signaling results in GI developmental defects (Stuckenholz et al., 2009). To further define the pathophysiological significance of intracellular PI, we identified and characterized the zebrafish insertional mutant cdipt^hi559 (hi559), which is defective in PI synthesis. Cdipt (CDP-diacylglycerol–inositol 3-phosphatidyltransferase) is a highly conserved enzyme with its active site on the cytoplasmic face of the ER and is responsible for synthesis of intracellular PI from myo-inositol and CDP-diacylglycerol. PI synthesis has been suggested to occur in a dynamic domain of the ER positioned at the leading edge of the ER tubules (English and Voeltz, 2013). Our earlier studies characterizing the hi559 mutation showed that the lack of de novo PI synthesis leads to ER stress and hepatic steatosis (Thakur et al., 2011). Despite the ubiquitous need for PI, the hi559 mutation does not cause a broad, general developmental defect. The enrichment of Cdipt expression in the larval GI tissues accounts for the strong GI defects in hi559 larvae due to functional loss of Cdipt and consequent deficiency of de novo PI synthesis, in spite of maternally deposited PI in the yolk. Given the transient roles and dynamics of PI metabolism, it is conceivable that Cdipt-controlled de novo PI synthesis is crucial for intracellular availability of PIPs such as phosphatidylinositol 4-phosphate [PI(4)P] and phosphatidylinositol (4,5)-bisphosphate [PI(4,5)P2] to exert their secretory functions.

The cdipt mutant zebrafish develop consistent GI defects during late larval stages after tissue differentiation, exhibiting a complex pathology in the intestine: abnormal IEC proliferation and apoptosis, villous atrophy, GC depletion, bacterial overgrowth and inflammation, all of which are hallmarks of human IBD. The hi559 IECs show disruption of ER architecture followed by mitochondrial defects and increased autophagy and cell death, consistent with ER-stress-induced cytopathology. Pharmacological induction of ER stress in wild-type larvae results in similar inflammatory pathologies, suggesting a contributory role of aberrant PI synthesis in ER-stress-mediated GI inflammation. In addition, akin to IBD treatment strategies, the mutant phenotype is partially ameliorated by antibiotic and anti-inflammatory drugs. This highlights the utility of this system as a tool for studying the pathogenesis of ER stress and mucosal inflammation. These studies facilitate novel insights into the mechanistic relationships between intracellular PI signaling, ER stress and GI pathophysiology in a whole-organism in vivo setting.

The hi559 homozygous mutant lacks Cdipt expression due to a retroviral insertion within the cdipt gene. Larvae develop normally until 5 dpf, when they begin to exhibit hepatic defects, including hepatomegaly and steatosis (Thakur et al., 2011). Another striking feature of the mutant is a smaller intestine by 5 dpf (Fig. 1A). Analyses of hi559 larvae expressing green fluorescent protein (GFP) in the gut [hi559Tg(gut:gfp)] confirmed that the hi559 mutation is fully penetrant, consistently presenting with a significantly smaller intestine at 5 dpf (P<0.001; Fig. 1E). Homozygous mutant larvae die at larval stages between 6.5 and 7 dpf. To prove that loss of Cdipt and its PI synthesis function underlies the mutant GI phenotype, we showed that knockdown of Cdipt by morpholino injection into wild-type embryos or chemical inhibition of PI synthesis in wild-type larvae by δ-hexachlorocyclohexane (δ-HCH) replicates the hi559 phenotype (supplementary material Fig. S1E-G) (Thakur et al., 2011). In addition, larvae with a weaker Cdipt mutant allele, cdipt^lop (lop) (Murphy et al., 2011), which carries a point mutation in the phosphatidylinositol synthase (PIS) domain, also replicated the hi559 GI phenotype (supplementary material Fig. S1C), but exhibited a milder, delayed phenotype. They developed normally until 7 dpf, subsequently exhibiting similar gross and histological intestinal abnormalities as seen in hi559 larvae and dying at about 10 dpf. Both hi559 and lop failed to rescue each other in a complementation assay (supplementary material Fig. S1D), supporting the conclusion that lop is a hypomorphic allele of hi559. In homozygous hi559 and lop mutants, Cdipt function is eliminated, resulting in abrogation of de novo PI synthesis (Murphy et al., 2011; Thakur et al., 2011). The similarity of the intestinal abnormalities in both lop and hi559 larvae suggest that the mutant phenotype is not primarily an early developmental defect, but reflects a requirement for de novo PI in intestinal function at later larval stages. Because hi559 mutants offer the advantage of an earlier and consistently stronger GI phenotype, we used them to elucidate the role of PI in GI tissues.

The hypomorphic nature of hi559 intestine is evident from whole-mount staining and histology (Fig. 1A-G; supplementary material Fig. S1). Incorporation of Nile Red was diminished in the mutants, indicating decreased luminal volume (Fig. 1C; supplementary material Fig. S1). The intestinal epithelial structure was reduced in size, as demonstrated by Cy3-streptavadin (Cy3-SA) immunostaining (Fig. 1D), suggesting that the hypomorphic intestinal manifests itself in both smaller lumen and thinner epithelium. To analyze developmental abnormalities in the intestine, we characterized the hi559 larvae by whole-mount in situ hybridization (ISH) using RNA probes against the intestine-specific markers fabp2, vil1 and anxa2b (Fig. 1B; supplementary material Fig. S1A). The observed decrease in marker gene expression suggests loss of structural and functional components of the hi559 intestine by 5 dpf. There was no difference in expression of intestinal markers or Nile Red staining until 4 dpf in hi559 compared with wild type (data not shown), suggesting no gross physical defects during early intestinal development. Intestinal expression of cdipt in wild-type larvae (Thakur et al., 2011) and the intestinal defects of hi559 and lop larvae (Fig. 1; supplementary material Fig. S1) implicate an important role of PIS in intestinal integrity and function.

Histological examination showed that the villous and luminal architecture was disrupted in the hi559 intestine (Fig. 1F,G). In wild-type larvae, columnar IECs are well polarized, forming a continuous epithelial monolayer with villi. In contrast, hi559 IECs were disorganized, less columnar with incomplete cytoplasmic maturity, and sporadically detached from the epithelium into the lumen (Fig. 1F,G and Fig. 2A). The intermittently detached IECs had nuclear condensation, suggesting apoptosis (supplementary material Fig. S2A). TEM analysis of the 5.5-dpf intestinal mucosa demonstrated that the wild-type IECs exhibit a highly elaborate apical brush border with microvilli projecting into the lumen, whereas the hi559 IECs had enlarged cytoplasmic vesicles, abnormal brush border, reduced terminal web and microvillus atrophy (Fig. 2B). The hi559 intestinal lumen was consistently filled with basophilic plaques (Fig. 2A), which TEM and colony formation assays confirmed to be largely due to increased bacterial colonization (P=0.0027; Fig. 2C).

Mucin-secreting GCs in the esophageal and mid-intestinal regions are typically evident in zebrafish larvae by 5 dpf. In hi559, these cells appeared abnormal with pyknotic or fragmented nuclei, suggesting apoptosis (Fig. 2A). We used PAS staining and TEM to analyze these GCs. In wild-type intestine, a thick secreted mucinous layer was consistently seen covering the apical border of the epithelium, which was diminished in the hi559 GI tract, suggesting alteration of GC physiology and their secretory function (Fig. 2D,E). Ultrastructurally, the 5-dpf wild-type GCs showed mature theca containing large mucinous vacuoles. In hi559, these appeared immature and degenerated (supplementary material Fig. S2B). Interestingly, there was no difference in numbers of GCs at 5 dpf (P=0.667), supporting normal IEC differentiation. However, the population of esophageal GCs declined by 6 dpf (P=0.0456; Fig. 2D–F) due to apoptosis and detachment. Mid-intestinal GCs and mucus secretion were similarly depleted in hi559 (P=0.0489; supplementary material Fig. S2C–E). Concomitant with GC depletion, nearly all hi559 larvae (>90%) showed histological features of bacterial overgrowth in the intestine by 6 dpf (P=0.002; Fig. 2F).

We studied the fate of the IECs by analyzing their cell-cycle status by BrdU incorporation and TUNEL assays. In the wild-type intestine, BrdU-positive cells occur frequently, typically at the base of the epithelial villi; TUNEL-positive cells are rare (Fig. 3A,B). Although BrdU labeling and TUNEL assays did not reveal abnormal cell proliferation or apoptosis in the hi559 GI tract prior to 5 dpf, IEC proliferation was significantly reduced and disorganized at 5 dpf (P=0.006; Fig. 3A,C) and the frequency of apoptotic cells increased as the hi559 intestinal pathology worsened by 6 dpf (P=0.0003; Fig. 3B,C), resulting in focal ulceration of the intestinal epithelium. TUNEL assays confirmed that the GCs with pyknotic nuclei observed in the hi559 GI tract were predominantly undergoing apoptosis, which might account for their depletion (Fig. 3B,C). In the disorganized epithelial region, IECs are often largely vacuolated (supplementary material Fig. S3A). These vacuoles did not show PAS or Oil-Red-O staining, suggesting that they were neither mucinous nor steatotic (data not shown). Taking the results together, we conclude that PI deficiency impedes proliferation and induces apoptosis of IECs, causing villous atrophy and intestinal hypoplasia.

During larval development, the hi559 intestine exhibited increasing bacterial overgrowth coinciding with depletion of the GCs (Fig. 2F). Because loss of mucinous secretions and aberrant bacterial growth in the gut can cause spontaneous inflammation, as seen in IBD, we assayed intestinal inflammation in hi559 larvae. Onset of an inflammatory response was evident at 5.5-6 dpf, as ISH demonstrated infiltration of the epithelium with mpo-positive neutrophils and spi1-positive macrophages (Fig. 3D). Necro-inflammatory injury was evident histologically by 6 dpf (supplementary material Fig. S3A). To quantify leukocyte infiltration, we utilized the leukocyte-specific reporter line Tg(lyzc:egfp). Analysis of hi559 Tg(lyzc:egfp) mutant larvae revealed significantly higher leukocyte aggregation in the 6-dpf intestine compared with wild-type Tg(lyzc:egfp) (P=0.008; Fig. 3E). Interestingly, pharmacological inhibition of PI synthesis by the chemical inhibitor δ-HCH in wild-type larvae replicated the increased intestinal leukocyte aggregation seen in hi559 (P=0.031; supplementary material Fig. S3B), further substantiating the hypothesis that deficient PI synthesis leads to mucosal inflammation.

Pathway analysis of hi559 gene expression identified acute phase response (APR) signaling as the most significantly upregulated canonical pathway, suggesting activated transcription of pro-inflammatory factors (P=0.002; Fig. 3F). In addition to the complement cascade pro-inflammatory factors, interleukin (IL-6, IL-8 and IL-17) signaling and NF-κB signaling were among the most significantly dysregulated gene sets, suggesting that pro-inflammatory activity might be mediated via these pathways (P≤0.01; Fig. 3F and supplementary material Fig. S4A,B).

Because our microarray-based pathway analyses revealed enrichment of ERSR gene sets in hi559 larvae (Thakur et al., 2011), we further investigated the nature of ER stress pathology in the GI tract. Unresolved ER stress is typically marked by persistent UPR induction and subsequent disruption of ER architecture and function. HSPA5 (also known as GRP78) is a heat shock protein that chaperones proteins in the ER lumen and is upregulated in response to ER stress (Marciniak and Ron, 2006), and XBP1 mRNA splicing is a key marker of ER stress and UPR activation (Calfon et al., 2002). The expression of proximal ERSR sensors hspa5 and xbp1 was robustly elevated in the hi559 GI tract as seen by ISH (supplementary material Fig. S5A,B) (Thakur et al., 2011). In addition, splicing of xbp1 was evident in the micro-dissected GI tissue of hi559 larvae by RT-PCR (supplementary material Fig. S5C). The elevation of both unspliced and spliced transcripts of xbp1 indicated that there was ongoing ER stress in the hi559 GI tissues. To further clarify the ER stress within tissues, we performed immunohistochemistry to detect active Hspa5 protein. Robust enrichment of Hspa5 protein was seen within hi559 GI tissues, specifically in the mucin secreting GCs and subsequently in the IECs of the intestinal mucosa (Fig. 4A).

Because these ER stress UPR factors are associated with molecular pathogenesis of IBD (Kaser et al., 2008; Bogaert et al., 2011), we wanted to further analyze the temporal ultrastructural defects within hi559 GI tissues to dissect the sequence of ER-stress-mediated pathology at cellular level. We performed extensive TEM analysis of the intestinal mucosa at different stages of the phenotype (Fig. 4B,D). At 5 dpf, the most striking defect in hi559 IECs was a disruption of the ER-Golgi architecture, without any overt changes in other cellular components (Fig. 4B). Large double-membrane macroautophagic bodies (autophagosomes) causing focal cytoplasmic necrosis were evident by 5.5 dpf (Fig. 4B). At 6 dpf, the IECs showed extensive mitophagy, depletion of mitochondria, and large multilamellar autophagosomes containing cytoplasmic organelles (Fig. 4C). The lumens of the distended ER-Golgi compartments in the hi559 IECs were often filled with aggregates of variable electron density, suggesting protein accumulation. Significantly increased autophagy and loss of mitochondria could account for the large cytoplasmic vesicles of 6-dpf IECs (P≤0.003; Fig. 4E). ER-stress-associated cytopathology and autophagic vesicles were also evident in the secretory enteroendocrine cells (Fig. 4D). ER expansion and autophagy occurred in both the pancreatic endocrine cells and the zymogen-rich acinar cells by 6 dpf, suggesting that the ER-stress-induced pathology is subsequently propagated in the majority of the secretory cells of the digestive system (supplementary material Fig. S5D,E). Disrupted ER architecture, grossly expanded ER lumens and vacuolization, consequent mitochondrial damage, and autophagy are consistent with ER-stress-induced cytopathology. These results demonstrate that unresolved ER stress in the highly secretory GI cells is the major etiology of the hi559 phenotype, implying that the lack of de novo PI impedes secretory function, leading to pathological ER-stress-induced GI defects in cdipt mutants.

Our analysis of hi559 mutants has not yet addressed the sequence of events leading to the overt phenotype: is the ER stress a direct consequence of functional loss of Cdipt, or induced by another unrecognized process such as inflammation? To help distinguish between these possibilities, we tested whether tunicamycin, a compound known to induce ER stress by inhibition of N-glycosylation, can cause GI inflammation. Chronic treatment of wild-type zebrafish with 1 μM tunicamycin from 3.5 dpf through to 6 dpf resulted in a smaller intestine (P=0.002) and defects in intestinal architecture (Fig. 5A–C and supplementary material Fig. S6A). Interestingly, tunicamycin-treated larvae exhibited increased bacterial growth (P=0.0006), GC depletion (P≤0.02), and increased intestinal macrophage and neutrophil infiltration (P≤0.006; Fig. 5C–E and supplementary material Fig. S6B,D). Necro-inflammatory lesions containing large numbers of vacuolated or autophagic IECs in tunicamycin-treated Tg(lc3-gfp) larval intestines were clearly evident (P=0.005; Fig. 5F and supplementary material Fig. S6C). These results clearly suggest that chronic ER stress is sufficient to trigger necro-inflammatory injuries in the zebrafish intestine, leading to hi559-like GI pathology.

Recruited macrophages and neutrophils are potent sources of cytokines and tissue destructive enzymes, contributing to necro-inflammatory injury by loss of IEC integrity. Anti-inflammatory agents, such as 5-aminosalicylic acid (5-ASA) and prednisolone, are thus widely used therapies for alleviating human inflammatory disorders, and remain an important option for treating patients presenting with moderate to severe IBD. In recent years, co-administration of antibiotics or probiotics with anti-inflammatory drugs is also being prescribed as an effective regimen for IBD treatment (Perencevich and Burakoff, 2006). To illustrate the similarity of cdipt mutants to human inflammatory states and as an in vivo system in which to assay potential suppressors of the inflammatory pathology, we assessed the response of hi559 mutants to antibiotics and anti-inflammatory drugs.

Treatment with antibiotics or anti-inflammatory drugs alone failed to rescue the hi559 intestinal phenotype and did not increase gut size (P=0.72 and 0.39, respectively), although leukocyte infiltration was effectively reduced (P≤0.01; Fig. 6E; supplementary material Fig. S7A–C). Co-administration of anti-inflammatory drugs 5-ASA and prednisolone together with antibiotics from 3.5 to 6 dpf resulted in marginal alleviation of hi559 intestinal size, as seen by ISH with fabp2, showing a minor increase in gut size compared with the mutant (P=0.051; Fig. 6A,E), clearing of luminal bacterial plaques, and reduced intestinal leukocyte infiltration as seen by decreased mpo expression and lyzc:egfp punctates (P=0.0006; Fig. 6B–E). The similar response to anti-inflammatory agents in hi559 larvae and to anti-inflammatory treatment in humans in reducing leukocyte infiltration suggests a conserved mode of action. Intriguingly, the antibiotic and anti-inflammatory treatments did not significantly alleviate GCs, autophagy and apoptosis of the hi559 intestinal mucosa (P=0.527, 0.59, 0.298, respectively; Fig. 6C,E), suggesting that ER stress is a precursor to bacterial overgrowth and inflammation. Subsequently, inflammation might then further exacerbate GI pathology. Hence, we hypothesize that inflammation is a downstream pathological event of ER stress and that relieving ER stress in hi559 larvae would rescue the GI inflammation.

Chemical chaperones enhance the protein folding and adaptive capacity of the ER and thus act as a potent suppressor of ER stress. As a proof of principle, we investigated whether the hi559 larvae are responsive to chemical chaperones that might alleviate ER stress. Phenylbutyric acid (PBA) is a small chemical chaperone and a well-established drug proven to reduce ER stress in both in vivo and in vitro studies (Ozcan et al., 2006). Prolonged exposure of hi559 larvae to 4-PBA ameliorated the intestinal phenotype, showing significant increase in gut size (P=0.0006; Fig. 7A,D), improvement in intestinal villous architecture (Fig. 7B), mitigation of inflammation with a significant reduction of intestinal leukocyte infiltration (P=0.0037; Fig. 7C,D), and increased survival of GCs (P=0.0031; Fig. 7D). Exposure of 4-PBA in δ-HCH-treated Tg(lc3:gfp) larvae showed reduction of IEC autophagosomes (P=0.0028; Fig. 7D), suggesting that alleviating ER stress can potentially reduce autophagy in PI-deficient larvae. Notably, the alleviation of hi559 GI phenotype by 4-PBA also resulted in improved survival of the mutant larvae (P=0.0008; Fig. 7E). Taken together, these results strongly support a feedback model in which unresolved ER stress initially triggers inflammation, which in turn can further worsen ER stress, exacerbating the GI pathology.

Using zebrafish genetic and pharmacological models, this study demonstrates that intracellular PI synthesis plays a vital role in maintaining physiological homeostasis and integrity of intestinal mucosa. Analysis of hi559 mutants has proven that loss of Cdipt function abrogates de novo PI synthesis, resulting in multiple GI pathologies. The cdipt mutant intestine fails to maintain its integrity and villous architecture as IECs exhibit reduced proliferation, reduced columnar shape, and degeneration at later larval stages. PI signaling has been linked to both cytokinesis and specification of apicobasal polarity in epithelial organs (Janetopoulos et al., 2005; Janetopoulos and Devreotes, 2006; Comer and Parent, 2007). Because IECs are rapidly proliferating at larval stages (Fig. 3A), they might require increased levels of de novo PI synthesis. Thus, despite maternal deposits of PI within the embryo itself (Thakur et al., 2011), these PI pools might not satisfy the specific needs of IECs in the hi559 larval intestine, leading to aberrant cytological architecture and a decrease in proliferation.

Loss of de novo PI synthesis might impede various cellular functions; however, our data suggest a primary role for secretory pathways in the development of the hi559 phenotype. Highly secretory cells of the digestive system of Cdipt-deficient larvae exhibit ER-stress-associated cytopathology, such as extensive disruption of the ER-Golgi complex and aggregates of electron-dense granules within the ER-Golgi lumen, suggesting protein accumulation consistent with pathological ER stress (Fig. 4). Furthermore, the observed diminution of mucus secretory functions of GCs and the aberrant ER-Golgi architecture of enteroendocrine and pancreatic cells clearly implicate a secretory defect in the hi559 digestive system. A mutation in sec13, which is primarily involved in protein trafficking from ER to Golgi, causes similar ER disruption, cell-cycle arrest, and cell death phenotypes in zebrafish (Niu et al., 2012), thus substantiating the hypothesis that a secretory defect might be linked to the persistent ER stress pathology. Given the role of PI metabolism and its phosphorylated derivatives (PIPs) in vesicular trafficking at ER exit sites and within the secretory pathway (Hama et al., 1999; Blumental-Perry et al., 2006; Yakir-Tamang and Gerst, 2009), it is likely that the tissue-specific expression of cdipt reflects a requirement for de novo intracellular PI pools by IECs to support the secretory function of intestinal cells, which could make them particularly vulnerable to ER stress. Therefore, specific intracellular PI signaling components might be intrinsic regulators of ER stress and UPR (Jesch et al., 2006), and we are pursuing their identification.

We observed abnormal cell death in the mutant intestine, predominantly of GCs by apoptosis. Interestingly, loss of GCs has been linked to human IBD pathogenesis (Xavier and Podolsky, 2007). Chronic ER stress and UPR can lead to cell death through different pathways, such as upregulation of apoptotic factors, including casp3 or ddit3/chop, which are found to be upregulated in hi559 larvae (Thakur et al., 2011), or via release of Ca²⁺ from the ER, which perturbs mitochondria and triggers oxidative-stress-induced cell death and inflammation (Kim et al., 2008; Lin et al., 2008). Additionally, ER stress is linked to increased autophagy (Yorimitsu et al., 2006), which is clearly evident in hi559 GI cells. These mechanisms could plausibly explain the increased apoptosis, mitochondrial damage and autophagy in hi559 GI cells, which collectively manifest in GC loss and intestinal dysfunction.

Increased inflammation in the hi559 intestine could be the result of at least two complementary mechanisms. Dysfunction and apoptosis of GCs and other IECs might result in reduced secretion of antimicrobial mucus and peptides. This loss would facilitate bacterial overgrowth, resulting in an inflammatory response to the increased intraluminal microbial load. Separately, in hi559 larvae, we observed the reported ER-UPR dependent dysregulation of NF-κB and pro-inflammatory interleukin signaling (Fig. 3 and supplementary material Fig. S4), which probably contributes to inflammation (Yamazaki et al., 2009; Kaser and Blumberg, 2010). This could set up a positive feedback loop in which bacterial overgrowth causes upregulation in the synthesis of secreted antimicrobial peptides and mucus, thus adding to the stress of the ER-Golgi secretory complex and further straining the intestinal mucosa. Additional studies using a recently developed NF-κB reporter line (Kanther et al., 2011), reared in a gnotobiotic environment, could help to dissect the complex interplay of gut microbiota, NF-κB signaling and ER stress in intestinal inflammation.

Although inflammation itself might contribute to ER stress, multiple lines of evidence in our study support a model in which ER stress initially triggers the development of necro-inflammatory injury. Temporal analysis of the hi559 ultrastructural pathology showed that hallmarks of ER stress occur prior to the onset of gross intestinal inflammatory pathology. Furthermore, pharmacological induction of ER stress in wild-type larvae using tunicamycin resulted in apoptosis of intestinal GCs, bacterial overgrowth and increased inflammation similar to that seen in hi559 larvae. Lastly, co-administration of antibiotics and anti-inflammatory drugs suppressed bacterial overgrowth and mucosal inflammation, but failed to alleviate the ER-stress-associated autophagy and necro-apoptosis (Fig. 6). Interestingly, treatment with the chemical chaperone 4-PBA, a known alleviator of ER stress, resulted in amelioration of the mucosal inflammation and increased survival of hi559 larvae (Fig. 7). Collectively, these results strongly suggest that bacterial overgrowth and inflammation do not directly cause ER stress, but result from it and exacerbate the ER-stress-induced pathology in our model.

This finding has important implications for the treatment of many human GI diseases. It suggests that pharmacologic manipulation of the ER stress pathway might be a novel treatment paradigm for particular GI diseases (such as IBD) and other disparate diseases (including cancer) that have been linked with chronic inflammation. Phospholipids are believed to have potential anti-inflammatory roles and can suppress activation of pro-inflammatory cells in vivo. Phosphatidylserine has been shown to inhibit macrophage activation (Gilbreath et al., 1985), and administration of phosphatidylcholine prevented stricture formation in a rat model of colitis (Mourelle et al., 1996). Interestingly, the PI3-K subunit p110δ was shown to play a vital role in maintaining mucosal homeostasis. Its inactivation caused defects in B and T cell signaling and in bactericidal activity, resulting in chronic colitis in mice (Uno et al., 2010). Recently, PI itself has been shown to inhibit T cell proliferation and function, implicating it as a novel physiological immune suppressant (van Dieren et al., 2011). The μ-opioid receptor ligand DALDA, a compound that might activate PI3-K signaling, has been shown to protect glafenine-induced intestinal injury in zebrafish by ameliorating ER stress (Goldsmith et al., 2013). We hypothesize that PI exerts its anti-inflammatory function via its ability to alleviate pathological ER stress.

This study provides the first evidence linking PI synthase to ER-stress-mediated GI pathologies, including bacterial overgrowth, mucosal apoptosis and inflammation, that are reminiscent of human IBD. In addition to genes regulating the immune system, mutations in genes affecting epithelial ER stress and function have been associated with IBD risk factors (Khor et al., 2011). Because ER homeostasis and inflammatory pathways appear to be conserved between zebrafish and human, investigating the mechanisms of ER stress in the zebrafish might reveal novel markers for IBD treatment. Currently, metabolically stabilized PI-derivative analogs and ER-stress-modulating compounds are being tested for their physiological relevance (Kim et al., 2008; He et al., 2011). Within this context, the cdipt mutants could provide an excellent platform for preclinical in vivo whole-organism studies evaluating the therapeutic potential of such compounds in ameliorating epithelial injury and inflammation.

The zebrafish line cdipt^hi559 was isolated from a large-scale insertional mutagenesis screen (Amsterdam et al., 2004). Heterozygous and homozygous mutants were sorted by genotyping using PCR (Thakur et al., 2011). The cdipt^lop mutant was isolated from a chemical mutagenesis screen (Murphy et al., 2011). The cdipt^hi559/loptrans-heterozygotes were generated by crossing hi559 heterozygotes with cdipt^lop. Fish were maintained in accordance with the institutional animal care and use committee protocols.

We used the regulatory region of the zebrafish fatty acid binding protein 2, intestinal (fabp2) to generate a transgenic zebrafish line. A 1.2-kb upstream fragment of the fabp2 promoter was cloned into the plasmid vector pEGFP. The plasmids pEGFP-fabp2 and pTOL2 (pT2KxIG in) were double-digested with BamHI and XhoI. The linearized ~3.5-kb fragment from pEGFP-fabp2 and the 6.8-kb fragment from pTOL2 were ligated using the T4 DNA ligase. The linearized construct pTOL2-EGFP-fabp2 was micro-injected into single-cell zebrafish embryos to obtain a germ-line transgene integration of fabp2-TOL2-EGFP. The founder fish were screened for the stable integration of the transgene, and subsequent transgenic fish generations were maintained.

Double transgenic fish used in this study were generated by crossing hi559 heterozygotes with Tg(gut:gfp) and Tg(lyzc:egfp) lines. The Tg(gut:gfp) transgenic zebrafish line expresses GFP throughout the digestive system and is used as a tool to analyze development of the GI tract and digestive organs (Field et al., 2003). The Tg(lyzc:egfp) transgenic line expresses enhanced green fluorescent protein (EGFP) under the regulatory regions of the zebrafish lysozyme-C (lyzc) gene, and is used to study infiltration of myeloid-derived inflammatory cells, representing a subset of macrophages and granulocytes (Hall et al., 2007; Hall et al., 2009). The Tg(mpx:gfp) transgenic line expresses GFP under the neutrophil-specific myeloperoxidase (mpx, also known as mpo) promoter and is used effectively to analyze intravital inflammatory response in vivo in zebrafish larvae (Renshaw et al., 2006). The Tg(lc3:gfp) transgenic line expresses GFP-fused Lc3 (GFP-Lc3), which can be visualized in vivo to monitor autophagy as Lc3 specifically labels the growing phagophores and completed autophagosomes (Kabeya et al., 2000; He et al., 2009; He and Klionsky, 2010).

Live imaging of zebrafish larvae was done by brightfield or fluorescent microscopy (Leica or Zeiss Axiovert). Confocal imaging was performed using a laser scanning confocal microscope (Leica or Zeiss), and the acquired images were analyzed using ImageJ (NIH, Bethesda, MD). The GFP intensity and puncta were quantified to assess leukocytes and autophagosomes in the GI tract of the respective transgenic larvae (n≥12).

For Nile Red staining, larvae were treated with 10 ng/ml Nile Red in E3 medium, starting at 3 dpf. The size and morphology of the gut lumen was assessed at different stages by observing Nile Red incorporation using fluorescent microscopy (Leica). Cy3-SA labeling and whole-mount in situ hybridization were performed as described previously (Sadler et al., 2005; Stuckenholz et al., 2009). Quantitative analyses of gut size were performed by ImageJ analyses of GFP-positive intestinal area using the Tg(gut:gfp) or Tg(fabp2-egfp) transgenic fish.

Histological sectioning and hematoxylin and eosin (H&E) staining were performed as described previously (Thakur et al., 2011). The IEC morphology, villous architecture and histological evidence of intraluminal bacteria were assessed by microscopic examination (Zeiss Axiovert) of at least ten alternate H&E-stained sagittal sections (5 μM), each representing larvae from wild type and mutants (n≥15) and larvae from DMSO control and drug treatment groups (n≥7). This allowed us to cover the analyses of histological features of the entire GI tract. For GC enumeration, PAS- and H&E-stained sections prepared at various time points during larval growth were imaged, and the total numbers of IECs and PAS-positive cells were determined for at least 12 alternate sections (4 μM) representing at least eight different larvae each from wild-type, mutant, DMSO and drug treatment groups. Phenotypically mature GCs were assessed according to the intensity of staining, the size of the apical region, the location in the intestinal epithelium and morphological appearance.

To assess the differential expression of Hspa5, sagittal cryosections (8 μM) through the entire GI tract of wild-type and mutant larvae were used for fluorescent immunohistochemistry using anti-HSPA5 primary antibody (Sigma) and FITC-conjugated anti-rabbit IgG secondary antibody (Sigma) and counterstained with DAPI (Sigma) for nuclear staining. Fluorescent image acquisition was performed using a confocal microscope (Zeiss) followed by analyses using ImageJ.

TEM was performed in the EM facility at the Center for Biological Imaging, University of Pittsburgh. For semithin sections, the epoxy resin-embedded larvae were transverse sectioned (350 nm) and stained with toluidine blue. At least ten ultrathin sections (70 nm) were collected, corresponding to the esophageal and intestinal region, for TEM staining and analyses. Sectioning depth from the beginning of the tissue as reference point and visualization of toluidine blue stained sections at regular intervals allowed us to select TEM sections from the same area of the tissue for wild type and mutants. The number of mitochondria, autophagosomes and lysosomes were counted from a set field of specific magnifications facilitating observation of GI cells and represented as numbers per field of observation. GI cells containing double or multi-membrane autophagosomes were considered positive and counted manually. Data were presented as percentage of IECs with autophagy in each field of observation. Mitochondrial counts were presented as total number of mitochondria per IEC.

Total RNA from the micro-dissected GI tissue (n=5) was isolated using RNA purification kit (Stratagene). cDNA was prepared by reverse transcription using Superscript II (Invitrogen). RT-PCR to detect xbp1 splicing was performed as described previously (Cinaroglu et al., 2011).

Quantification of intestinal bacterial density was adapted from previously described methods with applicable modifications (Oehlers et al., 2011b). Zebrafish larvae (6 dpf) were euthanized with tricaine (MS-222, Sigma) and washed three times with sterile PBS containing 0.1% Tween to remove non-adherent or loosely attached surface bacteria. Individual guts from each larva were micro-dissected using disposable sterile needles (n=3 for each genotype or treatment group) and the isolated gut tissues were washed three times with sterile PBS followed by homogenization with 500 μl PBS in sterile microfuge tubes with disposable microfuge pestles. Serial log₁₀ dilutions of the homogenates were plated on LB agar plates and incubated overnight at 28.5°C. Intestinal bacterial density was enumerated based on total colony forming units (cfu) per individual gut.

Cell proliferation was estimated by in vivo labeling with 5-bromo-2 -deoxy-uridine (BrdU, Roche) and apoptosis was quantified by TUNEL assay on histological sections using the ApopTag peroxidase kit (Chemicon). Larvae (4, 5 and 6 dpf) were incubated in E3 with 10 mM BrdU for 6 h at 28.5°C and fixed in 4% PFA overnight. Incorporated BrdU was detected with an anti-BrdU antibody (Amersham) and visualized with peroxidase substrate kit (Vector). Quantification of proliferating and apoptotic cells was represented in percentages of relative proportion of BrdU-positive and TUNEL-positive IECs to the total number of IECs in at least ten alternate sagittal sections (5 μm) representing the entire GI tract of at least eight different larvae.

Tunicamycin is known to prevent N-linked glycosylation of proteins, affecting their proper folding and accumulation in the ER, thus inducing ER stress in eukaryotic cells (Yoshida, 2007). The tunicamycin (Calbiochem, EMD Biosciences) treatment assay in larval zebrafish was optimized using the methods described in our previous study (Thakur et al., 2011).

δ-Hexachlorocyclohexane (δ-HCH), a compound with similar configuration to myo-inositol, inhibits PI synthesis by affecting the incorporation of myo-inositol into PI (Hoffmann et al., 1980; Thakur et al., 2011). Wild-type, Tg(gut:gfp), Tg(lyzc:egfp) or Tg(lc3:gfp) larvae were treated with 5 μM δ-HCH (Sigma) from 3.5 to 6 dpf and analyzed at 6 dpf.

For antibiotic treatment, zebrafish larvae were exposed to a cocktail of antibiotics from 3.5 to 6 dpf. Antibiotics consisted of ampicillin (Sigma, 100 μg/ml final concentration), kanamycin (Sigma, 5 μg/ml final concentration), and penicillin-streptomycin pre-mix (Invitrogen, 100 units/ml penicillin and 100 μg/ml streptomycin) in E3 media. Anti-inflammatory drugs consisted of 5-aminosalysilic acid (ASA; Sigma, 50 μg/ml) and 6-α-methylprednisolone (Sigma, 25 μg/ml) dissolved in 0.05% DMSO v/v in E3 media.

For chemical chaperone treatment, sodium 4-PBA (Sigma) was dissolved in E3 water and the fish exposed to various dosages of 4-PBA to optimize a nontoxic dosage that did not cause any developmental defects. A final concentration of 50 μM 4-PBA (from 3.5 dpf until 9 dpf) was used for the rescue experiments in this study. For control groups, larvae were treated with equivalent concentration of DMSO alone. The drug treatments were performed in 12-well plates containing 15 larvae each from three different biological clutches.

Data are representative of larvae from at least three different biological clutches. Statistical significance was calculated using a two-tailed Student’s t-test; P values of less than 0.05 were considered significant. The results are expressed as means, and standard deviations are indicted by error bars in the figures.

Gene expression and pathway analyses were performed as described previously using our microarray data deposited with GEO (GSE17711) (Stuckenholz et al., 2009; Thakur et al., 2011). We used Ingenuity’s pathway analysis tool (IPA; http://www.ingenuity.com) and the Gene Set Enrichment Analysis tool (GSEA; http://www.broad.mit.edu/gsea/) (Subramanian et al., 2005) to decipher the dysregulated pathways in the mutants. A P value of less than 0.05 (n=3) after adjusting for false discovery rate was considered significant.

The authors thank Drs Jeffrey Brodsky and Meir Aridor for helpful discussions; Dr Donna Beer Stolz and Mara Sullivan, Center for Biological Imaging, University of Pittsburgh for the TEM facility; Dr David Hyde, University of Notre Dame for cdipt^lop; and Dr Daniel Klionsky, University of Michigan, Ann Arbor for Tg(lc3:gfp) zebrafish.