Stroma provides an intestinal stem cell niche in the absence of epithelial Wnts

The epithelial lining of the intestine continuously renews itself every 3 to 5 days. In the crypts of the murine small intestine, long-lived, label-retaining stem cells in the +4 position marked by Bmi1, Tert and Hopx interact with and interconvert with more rapidly proliferating, radiation resistant, crypt base columnar cells that express markers including Lgr5, Olfm4, Lrig1 and Ascl2 (Barker et al., 2008; Buczacki et al., 2013; Powell and Saada, 2012; Takeda et al., 2011; van der Flier et al., 2009; Wong et al., 2012; Yan et al., 2012). This stem cell compartment gives rise to committed progenitor cells that proliferate rapidly and produce the diverse differentiated progeny that migrate up the villi before being shed into the lumen.

Wnt/β-catenin signaling plays a crucial role in maintaining normal proliferation in the intestinal crypt of the adult mouse. Secreted Wnts bind to LRP5/6 and Frizzled co-receptors present on epithelial crypt cells, leading to an increase in β-catenin protein (Clevers and Nusse, 2012). Activated β-catenin binds to the nuclear transcription factor TCF4 to drive a gene expression program that supports stem cell maintenance, proliferation and differentiation. Disruption of the Wnt/β-catenin pathway blocks intestinal proliferation. Embryonic knockout of TCF4 in intestinal epithelial cells leads to lack of proliferation in the inter-villus region of the neonatal small intestine, whereas inducible knockout of TCF4 and β-catenin in adults blocks proliferation in the crypt compartment (Fevr et al., 2007; Korinek et al., 1998; van Es et al., 2012). Conversely, stabilization of β-catenin by expression of constitutive active β-catenin or mutation of APC stimulates proliferation. Surprisingly, the evidence that secreted Wnt ligands regulate intestinal homeostasis in adult mice remains indirect. The strongest evidence comes from studies inhibiting or knocking out the Wnt ligand co-receptors Lrp5 and Lrp6, leading to a near total loss of epithelial proliferation (Kuhnert et al., 2004; Zhong et al., 2012).

The identity and cellular source of the Wnts that regulate the intestinal stem cell niche is not clear. Many of the 19 different Wnt genes are expressed in the small intestine, each with a distinct pattern of expression in diverse cell types of the epithelium and stroma (Gregorieff et al., 2005). These multiple Wnts may regulate diverse processes beyond the stem cell niche, including innate and adaptive immunity, injury repair, and intermediary metabolism (Cervantes et al., 2009; Davies et al., 2008; Zeve et al., 2012). The limited number of studies knocking out Wnts in mouse intestine have not identified defects in in vivo crypt stem cell proliferation (Cervantes et al., 2009; Farin et al., 2012). Paneth cells in the crypt base are a potential source of the Wnts that regulate stem cell proliferation. They express several Wnts, including Wnt3. In purified epithelial cell preparations, WNT3 from Paneth cells is required for the growth in culture of organoids derived from Lgr5-expressing intestine stem cells, although supplementation with R-Spondin 1 (RSPO1) is also required (Farin et al., 2012; Ootani et al., 2009; Sato et al., 2011, 2009). Based on these data, it has been proposed that Paneth cells form the niche for the isolated intestinal stem cells (Sato et al., 2011). However, depletion of Paneth cells or knockout of Wnt3 in the epithelial cells of the intestine did not show an obvious in vivo phenotype (Durand et al., 2012; Farin et al., 2012), indicating that other functionally important Wnts are made by epithelial or stromal niche cells. Several groups have demonstrated that stromal cells can support the growth of intestinal epithelium in culture (Farin et al., 2012; Lahar et al., 2011). Intestinal stromal cells express multiple Wnts (Farin et al., 2012; Gregorieff et al., 2005). Farin et al. demonstrated that purified stromal cells could support organoid formation from Wnt3 knockout epithelial cells, but the question of whether other epithelial-produced Wnts are essential ex vivo and more importantly, in vivo, remains unanswered.

One approach to address the functionally relevant source of Wnts in the small intestine is to globally target their secretion. This can be achieved by knockout or inhibition of either of two genes, Porcn and Wls, that are indispensable parts of the core Wnt secretion machinery (Najdi et al., 2012; Proffitt and Virshup, 2012). The PORCN protein is a membrane bound O-acyl transferase that resides in the endoplasmic reticulum. PORCN palmitoleates all Wnts as they are synthesized (Biechele et al., 2011; Kadowaki et al., 1996; Tanaka et al., 2000). Palmitoleation is required for Wnts to bind to WLS, an integral membrane carrier protein that is essential for the secretion of all known vertebrate Wnts (Coombs et al., 2012; Najdi et al., 2012). Wnt palmitoleation is also required for secreted Wnt ligands to interact with Frizzled receptors at the cell membrane (Janda et al., 2012). Porcn is encoded by a single copy gene on the X chromosome and has no closely related homologs in the genome. Zygotic Porcn mutants exhibit gastrulation failure leading to early embryonic lethality (Barrott et al., 2011; Biechele et al., 2011, 2013). Several potent small molecule inhibitors of PORCN have been developed that phenocopy the biochemical effects of Porcn knockout, and may be of value in the treatment of Wnt-high diseases, including cancer and fibrotic disorders (Chen et al., 2009; Proffitt et al., 2013). Indeed, one PORCN inhibitor, LGK974, is currently in phase 1 trials in humans (Liu et al., 2013). However, due to potential detrimental effects on adult stem cell self-renewal, the role of pharmacologic inhibition of PORCN remains to be determined.

To dissect the role of epithelial and stromal Wnts in intestinal homeostasis and assess the potential toxicity of PORCN inhibition, we used genetic and pharmacological approaches to block Wnt production in the gut. We find that epithelial Wnts are dispensable for normal proliferation in the mouse intestine. By contrast, pharmacologic inhibition of Wnt production produces a graded decrease in proliferation. At intermediate doses, Wnt-responsive Lgr5 expression is markedly reduced, accompanied by impaired recovery from radiation injury, but without a decrease in the proliferation of transit amplifying cells. At higher doses of the PORCN inhibitor, global intestinal proliferation ceases. A stromal fraction enriched for myofibroblasts that endogenously expresses both Wnts and Rspo3 supports Porcn^Del organoid proliferation in the absence of supplemental RSPO1. These studies suggest that a stromal Wnt/RSPO3-producing niche is sufficient for normal and stressed intestinal homeostasis and support the role of the Lgr5 cell population in recovery from radiation injury.

Multiple studies have demonstrated a role for Wnt/β-catenin signaling in intestinal homeostasis, but whether intestinal epithelial cells provide any Wnts important in the niche remains an open question. To assess the role of the epithelial Wnts in intestinal homeostasis, we used a floxed allele of Porcn that is null for Wnt secretion after Cre-mediated excision (Biechele et al., 2013; Proffitt and Virshup, 2012). Porcn^flox and Porcn^WT mice were crossed with Villin-Cre mice to generate Porcn^Del/Villin-Cre or Porcn^WT/Villin-Cre mice, respectively. Villin expression begins in late embryogenesis in all epithelial cells of the intestine. Our hypothesis was that embryonic inactivation of Porcn in the intestinal epithelium would be lethal in the neonatal period, similar to what was observed in the Tcf4 or Lrp5/Lrp6 double knockout (Korinek et al., 1998; Zhong et al., 2012). Unexpectedly, Porcn^Del/Villin-Cre pups were viable, appeared phenotypically normal, and suckled and weaned without difficulty. We confirmed by genomic PCR and RT-qPCR that purified intestinal epithelial cells from Porcn^Del/Villin-Cre mice had complete excision of the Porcn gene (Fig. 1A,B). To confirm functional excision of Porcn, we took advantage of the observation that Wnt secretion from Paneth cells or other epithelial cells is required for purified crypts to form organoids in culture (Ootani et al., 2009; Sato et al., 2009). Indeed, purified isolated crypts from Porcn^Del/Villin-Cre mice did not form organoids in vitro (Fig. 1C,E). This result phenocopies both knockout of Wnt3 (Farin et al., 2012) and the effect of small molecule PORCN inhibition (Sato et al., 2011), confirming the functional inactivation of epithelial Porcn in Porcn^Del/Villin-Cre mice and supporting the central role of epithelial Wnts in ex vivo culture of purified intestinal stem cells.

To confirm that the defect in organoid formation was due to loss of secreted Wnts and not to a non-Wnt consequence of Porcn knockout (Covey et al., 2012), we employed two approaches. First organoid formation from Porcn^Del/Villin-Cre crypts was rescued by co-culture with WNT3A-secreting mouse L (L3A) or human HEK293 (STF3A) cells (Fig. 1D,E). Second, organoid formation was also partially rescued by co-culture with mouse embryo fibroblasts (MEFs) that we found express endogenous WNT3 (Fig. 1C,E; supplementary material Fig. S1A). However, Porcn^Del MEFs failed to rescue (Proffitt and Virshup, 2012). These findings are consistent with existing data that intestinal epithelial stem cells require a source of palmitoleated Wnts to proliferate and form organoids, and that ex vivo the Wnts can be supplied by co-culture with WNT3A-secreting cells.

We assessed the phenotype of the intestine in the Porcn^Del/Villin-Cre mice. The small and large intestines were grossly and microscopically normal (Fig. 2A). In addition, epithelial lineages including enteroendocrine, Paneth and goblet cells were not detectably altered in the absence of epithelial Wnt secretion, and there was no increase in apoptotic cells (Fig. 2B). Loss of epithelial Porcn did not alter the committed progenitor (also called transit amplifying) compartment, as assessed by short term EdU incorporation (Fig. 2C; supplementary material Fig. S1B). Importantly, β-catenin was observed in the nuclei of crypt cells of Porcn^Del/Villin-Cre mice, indistinguishable from controls (Fig. 2C). Consistent with intact β-catenin signaling, the expression of Axin2, Lgr5 and Olfm4 did not differ significantly between jejunum samples of control and Porcn-inactivated mice (Fig. 2D). These data demonstrate that Wnt/β-catenin/LGR5 signaling is ongoing in intestinal crypt cells in the absence of epithelial Wnt secretion. This strongly suggests that in Porcn^Del/Villin-Cre mice the Wnts important in intestinal proliferation are adequately supplied from the stroma.

We tested whether the unaltered proliferation and differentiation after epithelial loss of Wnt secretion resulted from compensatory changes in other signaling pathways regulating intestinal stem cell renewal. To assess this, gene expression profiling was performed on isolated epithelium and stroma from wild-type and Porcn^Del mouse intestine. As supplementary material Fig. S1C shows, there were only minimal changes in global gene expression detected, and there were no statistically significant alterations in known Wnt/β-catenin pathway regulators. There were no significant changes in expression of genes associated with the Notch and EGF pathways (supplementary material Fig. S1D,E). We conclude that epithelial Wnt secretion is not essential for normal murine small intestinal homeostasis in vivo, and its loss causes no readily detectable compensatory alterations in gene expression.

A critical role for intestinal stroma in the self-renewal of epithelial stem cells has previously been suggested by Farin et al. (2012). To test this, we isolated Porcn^Del crypts and Porcn^flox unfractionated stromal cells (‘fresh stroma’) from Porcn^Del/Villin-Cre mice and cultured them separately or together. Stroma added to Porcn^Del crypts supported cystic organoid formation (Fig. 3A). Crypts depleted of stroma require supplementation with RSPO1 to form organoids in culture. We reasoned that the added RSPO1 might be required to amplify the WNT3 signal from the Paneth cells. Remarkably, we found that Porcn^Del crypts supplied with stroma did not require exogenous RSPO1 (Fig. 3A).

Wnts signal at short range, and myofibroblasts are anatomically adjacent to the epithelial crypt, making them a candidate source of stem-cell supporting factors. We therefore prepared cultured adherent stromal cells highly enriched for myofibroblasts [vimentin positive (∼100%), alpha-smooth muscle actin (α-SMA) positive (>50%), desmin positive (4%)] (Fig. 3C). Notably, these cells supported organoid formation better than unfractionated stromal cells even in the absence of supplemental RSPO1 (Fig. 3A). The myofibroblast-enriched population expresses multiple Wnts (supplementary material Fig. S2C). In addition, these cells express abundant Rspo3 (Fig. 3B; supplementary material Fig. S2D). By contrast, the epithelium was a poor source of RSPOs, as assessed by both microarray and qPCR (Fig. 3B; supplementary material Fig. S2D). Thus, the myofibroblast-enriched stromal fraction can provide both Wnts and R-Spondin and support intestinal stem cell proliferation.

To assess the contribution of immune and/or hematopoietic cells as a Wnt source for intestinal homeostasis, we crossed Porcn^flox/Villin-Cre mice to Vav-Cre mice to generate Porcn knockout in intestinal epithelial cells as well as hematopoietic and immune cells. These mice, which will be described in more detail elsewhere, were viable and their cultured adherent stroma also fully supported organoid formation from Porcn^Del epithelial cells (supplementary material Fig. S2A). Taken together, the data are most consistent with stromal myofibroblasts as a significant source of Wnts and Rspo3 both in vivo and in culture, although we cannot exclude the possibility that rare cell types form the niche.

As loss of Wnt secretion in the intestinal epithelium was so strongly predicted to produce a phenotype, we asked if knockout of Wls phenocopied the Porcn knockout. Wls is a dedicated Wnt carrier that is required to transport palmitoleated Wnts to the cell membrane. Wls, like Porcn, is required for the activity of all human Wnts (Najdi et al., 2012). Similar to the Porcn deletion, Wls^Del/Villin-Cre pups were viable and had normal development and growth. Histological appearance, proliferation, and β-catenin nuclear localization were identical in the intestines of one-year-old control (Wls^WT/Villin-Cre) and Wls-deleted (Wls^Del/Villin-Cre) mice (supplementary material Fig. S2B). These data confirm that inhibition of epithelial Wnt secretion by targeting either Porcn or Wls does not significantly impact intestinal development and homeostasis.

Substantial data indicate that Wnt signaling is essential for intestinal stem cell regulation, but the targeted inhibition of Wnt secretion from the intestinal epithelium produced no phenotype. Either, against all expectations, Wnts are not required, or the intestinal stroma is sufficient as the niche providing the Wnts regulating stem cell proliferation. To differentiate between these possibilities, we took advantage of a recently validated pharmacological inhibitor of Porcn, C59. This drug is a readily absorbed, bioavailable inhibitor of Porcn with a sub-nanomolar IC50 that inhibits Wnt/β-catenin signaling and blocks the growth of WNT1-dependent mammary tumors in mice (Proffitt et al., 2013). An added benefit of using a drug is the ability to titrate the dose to dissect intermediate phenotypes that are difficult to detect in gene deletion models. We reasoned that C59 should inhibit stromal as well as epithelial PORCN activity and, taken together with the results of epithelial Porcn knockout, be a good test of the role of stromal Wnt secretion.

We confirmed that C59 prevented the formation of intestinal organoids in culture, similar to the activity of IWP (supplementary material Fig. S3A) (Sato et al., 2011). We previously reported that orally administered C59 at 5-10 mg/kg/day inhibited the growth of MMTV-Wnt1 mammary tumors without intestinal toxicity (Proffitt et al., 2013). To identify a C59 dose that inhibited Wnt/β-catenin activity in the intestine, mice were administered various amounts of C59 daily for 2 days by gavage and their small intestines were harvested 20-24 h after the second dose. There was a dose-dependent reduction in the expression of the Wnt/β-catenin target gene Lgr5 and the crypt base columnar (CBC) cell marker Olfm4 (supplementary material Fig. S3B).

Mice receiving C59 50 mg/kg/day for 6 days maintained body weight and had normal intestinal structure and proliferation (Fig. 4A,C). This was despite the systemic C59 treatment causing a rapid, significant and persistent reduction in expression of the β-catenin target genes Lgr5, Axin2 and Ascl2, as well as the CBC cell marker Olfm4, but not Bmi1 in the small intestine (Fig. 4B; supplementary material Fig. S3C). As an independent approach to confirm that Lgr5 expression was reduced by C59 treatment, we treated Lgr5-IRES-EGFP-CreERT2 mice that express EGFP from the endogenous Lgr5 promoter with 50 mg/kg/day C59 for 6 days (Barker et al., 2007). We observed a marked decrease in EGFP expression in the C59-treated intestine (Fig. 4C). We confirmed that the C59-dependent decrease in Wnt/β-catenin target gene expression was also seen in isolated intestinal crypts (supplementary material Fig. S3D). Wnt signaling can regulate Paneth cell differentiation, and we observed that the Paneth markers lysozyme and MMP7, but not cryptdin nor the goblet cell marker MUC2, were also decreased after 1 week of treatment (supplementary material Fig. S4A,B). Thus, systemic, but not isolated epithelial inhibition of PORCN-dependent Wnt secretion suppresses Wnt/β-catenin target gene expression in the intestine, consistent with an important role for stroma. Unexpectedly, this suppression of Wnt/β-catenin signaling was not essential for short-term intestinal homeostasis.

Lgr5-expressing CBC cells are relatively radiation resistant and proliferate after radiation injury (Hua et al., 2012). We reasoned that suppression of Wnt/β-catenin signaling by PORCN inhibition might impair the response to injury. C57BL/6 mice were therefore treated with 50 mg/kg/day C59 for 6 days. Twenty hours following the final dose of C59, mice were irradiated with a single dose of 12 Gy, and sacrificed when they became ill 5 days later. In the C59 treated mice, the duodenum was dilated and the small intestine was markedly shortened (supplementary material Fig. S4C). Histologic examination demonstrated a marked loss of crypts and villi throughout the small intestine in Wnt-suppressed compared with control mice (Fig. 5A). Proliferation of the epithelial layer was assessed by EdU incorporation 5 days after radiation. As expected, increased proliferation was observed in the epithelium of the control mice due to tissue regeneration. However, the C59-treated mice showed a marked reduction in EdU incorporation in the small intestine (Fig. 5B). Thus, intermediate inhibition of Wnt/β-catenin signaling by C59 significantly reduced Lgr5 expression in the CBC cells and impaired recovery from radiation injury. This is consistent with a recently proposed role for Lgr5-expressing cells in the response to radiation damage (Hua et al., 2012; Metcalfe et al., 2014).

We tested whether epithelial Wnts are required for recovery from radiation injury. Porcn^WT and Porcn^Del/Villin-Cre mice were treated with the same radiation protocol as above. Three out of five wild-type, and five out of five Porcn^Del/Villin-Cre mice survived 8 days after radiation. All the mice were ill and had significant weight loss. Histologic examination of the intestines of surviving mice showed successful ongoing epithelial regeneration regardless of genotype, with no detectable differences between wild-type and knockout mice (Fig. 5C). Hence, epithelial Wnt production is not required for intestinal regeneration after radiation injury. The radiation sensitivity seen in C59-treated, but not Porcn knockout, mice is consistent with functionally important Wnts coming from the stromal niche.

C59 given once daily at 50 mg/kg significantly reduced expression of the CBC cell markers Lgr5, Ascl2, and Olfm4, decreased expression of Paneth cell markers and prevented an effective response to radiation injury. However, the persistence of EdU incorporation and nuclear β-catenin (supplementary material Fig. S3E) in these mice suggested residual Wnt activity might be sufficient to maintain a stem cell population, as has been suggested for the related compound LGK974 (Liu et al., 2013). To test this, we increased the C59 dose to twice daily to increase trough drug levels. Mice treated with 50 mg/kg per dose, twice daily, were moribund within 6 days. At this dose intensity, intestinal crypts began to disappear in the duodenum and jejunum on the second and fourth day of treatment, respectively (Fig. 6A).

After 6 days of high dose treatment, there was a global reduction in proliferation, as indicated by loss of all crypts in the small intestine and absence of EdU incorporation (Fig. 6A,B). This toxicity is unlikely to be nonspecific, as there was a decrease in proliferation rather than an increase in apoptosis (Fig. 6B; supplementary material Fig. S5A). These dramatic findings resemble the proliferation defect observed after total loss of Wnt/β-catenin signaling in the epithelium in mice with conditionally deleted β-catenin or TCF4 alleles (Fevr et al., 2007; van Es et al., 2012). We conclude that full inhibition of PORCN activity in both epithelium and stroma of the intestine blocks proliferation and intestinal homeostasis.

Our study indicates that stromal Wnts are sufficient to maintain mouse small intestinal homeostasis. Tissue-specific knockout of Porcn and Wls shows that epithelial Wnt production is dispensable for normal intestinal stem cell development, self-renewal, proliferation, and the response to radiation-induced injury. Conversely, using a small molecule inhibitor of Wnt production we find that varying levels of systemic PORCN inhibition produce distinct phenotypes in the gut. Moderate global inhibition of Wnt secretion markedly reduced Lgr5 expression and impaired intestinal homeostasis after radiation injury, whereas more complete inhibition of Wnt secretion immediately affected stem cell function, similar to results seen after genetic knockout of key Wnt/β-catenin pathway components. The ability to moderately inhibit PORCN function and not impair short-term intestinal homeostasis suggests that drugs inhibiting PORCN will have a therapeutic index allowing clinical use.

Purified epithelial stem cells can form organoids and expand ex vivo in the presence of Wnt3-producing Paneth cells and exogenous RSPO1. Here, we found epithelial Wnts supplemented with recombinant RSPO1 can be replaced by an intestinal myofibroblast-enriched stromal fraction that endogenously produces Wnts and RSPO3. Taken together, the data are consistent with the hypothesis that stromal cells can form a Wnt- and RSPO3-producing niche for intestinal epithelial stem cells in the absence of epithelial Wnt production.

Is there a single essential source of Wnts in the small intestine? We found that both epithelial and hematopoietic Porcn function are dispensable for intestinal homeostasis. Our data on stroma as a source of both Wnts and RSPO3 ex vivo are consistent with myofibroblasts in close proximity to the epithelial stem cells forming the Wnt-producing niche. However, this stands in contrast to a paper published while this work was in revision (San Roman et al., 2014). The authors of that paper reported that inducible short-term double deletion of Porcn in both intestinal epithelium (in Porcn^flox/villin-creER^T2 mice) and subepithelial myofibroblasts (in Porcn^flox/Myh11-creER^T2 mice) did not alter intestinal homeostasis. However, that study did not determine the efficiency of Porcn deletion in the myofibroblasts and whether the Myh11-creER^T2 driven excision in fact abrogated the organoid-supporting ability of the stroma.

Several publications have suggested that there are redundant sources of Wnts supporting the intestine (Farin et al., 2012; San Roman et al., 2014). Indeed, the intestinal stroma contains multiple cell types capable of making Wnts, including endothelial cells, macrophages, neurons, fibroblasts, and myofibroblasts. Wnts produced in combinations of these cells could produce a cocktail of redundant Wnt ligands maintaining intestinal homeostasis in vivo. Alternatively, our data are also consistent with the niche being a specific myofibroblast population adjacent to the crypts that produces both Wnts and RSPO3.

Our study complemented the genetic knockout of PORCN with pharmacologic inhibition. One unexpected finding was the broad therapeutic range for pharmacologic PORCN inhibition. We previously reported that as little as 5 mg/kg/day of C59 blocked proliferation of a Wnt1-dependent mammary tumor, yet here, also in C57BL/6 mice, a 20-fold higher dose was required for inhibition of intestinal stem cell proliferation. Our data suggest that even small amounts of Wnt secretion can maintain an intestinal stem cell niche in the absence of external stress. This robust network of stem cell regulators suggests that inhibition of Wnt production may be effective for diseases with pathological Wnt elevation at doses that do not perturb normal stem cell niches.

We noted that Lgr5/Ascl2/Olfm4 expression could be reduced by C59 without immediate effect on overall intestinal architecture, similar to genetic knockout of Lgr5 cells with diphtheria toxin (Metcalfe et al., 2014). However, several of the mice treated at the intermediate dose of C59 become ill after 18 days. Examination of the small intestine revealed patchy loss of proliferation and lack of crypts in the proximal small intestine (supplementary material Fig. S5B,C). This may be due to differential sensitivity of two distinct populations of stem cells in small intestine. Buczacki et al. recently demonstrated that label retaining cells (LRCs) in the +4 position of the crypt are secretory precursors of Lgr5 cells and serve as a reserve pool of stem cells after intestinal damage (Buczacki et al., 2013). We speculate that long-lived LRCs are relatively insensitive to C59-mediated Porcn inhibition and sustain intestinal homeostasis in the absence of Lgr5 stem cells. Impaired crypt homeostasis would slowly occur as the LRCs were depleted at the intermediate dose, or rapidly if they are completely deprived of Wnts at the high dose of C59.

The role of Lgr5+ and Paneth cells in the response to radiation damage is of great recent interest (Buczacki et al., 2013; Hua et al., 2012; Metcalfe et al., 2014; Roth et al., 2012). We addressed the role of Wnt production in the radiation response and found that global pharmacologic, but not epithelial-specific, inhibition of PORCN caused markedly increased sensitivity to radiation stress in the intestine. Although PORCN inhibition both reduced Lgr5 expression and modified Paneth cell differentiation, Paneth cell depletion reportedly does not affect recovery from radiation (Metcalfe et al., 2014). Our data are therefore most consistent with the model that PORCN-dependent Lgr5+ cells are required for recovery from radiation, and suggest the possibility of synergistic toxicity in clinical settings.

In conclusion, taking advantage of the essential roles of Porcn and Wls in Wnt secretion, we have demonstrated that epithelial Wnts are not vital for intestinal homeostasis or recovery from radiation injury, while confirming that they are required for ex vivo cultures. In addition, we provide strong evidence that stromal Wnts play a crucial role in the maintenance of small intestine homeostasis in vivo.

Porcn^flox mice (Biechele et al., 2013) were backcrossed to C57BL/6 mice at least for six generations. Porcn^flox mice were crossed with BL/6 Villin-Cre mice. Age- and gender-matched mice were used as controls for all experiments. Lgr5-IRES-CreER^T2-EGFP mice were obtained from Jackson Laboratories (Barker et al., 2007). Wls^flox mice, studied at the Van Andel Research Institute, were from Richard Lang (Carpenter et al., 2010). All mouse procedures were approved by the respective Institutional Care and Use Committees (IACUC). Genotyping and PCR are described in Table S1 and the Materials and Methods in the supplementary material. PORCN inhibitor C59 was suspended in a mixture of 0.5% methylcellulose and 0.1% Tween 80 by sonication for 30 min and then administrated by gavage as described (Proffitt et al., 2013).

Intestine was harvested, cut longitudinally, washed, and macerated in pieces not exceeding 2 mm in size. Fragments were incubated in ice-cold PBS containing 2 mM EDTA for 40 min with gentle shaking every 10 min. The solution was pipetted up and down for 40 times followed by 3 min of gravity sedimentation. Supernatant fractions containing released cells were collected after each sedimentation step. This procedure was repeated three times, resulting in three collected fractions. After the first round, intestinal pellets were washed with PBS three times. Fraction 3 was used for subsequent crypt isolation. Crypts were enriched by centrifugation at 200 g for 2 min and counted using a phase contrast microscope. As judged by microscopy, this routinely yielded >90% pure crypts. All the procedures were performed at 4°C. Crypt culture was performed in 48-well plates using 6000 crypts per well closely following conditions described by Sato et al. (2009).

Tissues remaining after crypt isolation were subjected to an additional round of pipetting for 40 times to remove most of the remaining epithelial cells, and washed once with PBS and with serum-free DMEM containing 1% Glutamax and 1% Penicillin-Streptomycin (both reagents from Life Technologies). Thereafter, tissues were digested for 3 h in 6 ml of serum-free DMEM containing 1% Glutamax, 1% Penicillin-Streptomycin and 2 mg/ml of Collagenase/Dispase (Roche). The digestion solution was replaced with fresh digestion solution every 60 min. At the end of this digestion step, tissues were further dissociated by vigorous pipetting. To inhibit proteolytic activity and cellular aggregation, at this step suspensions were supplemented with 5% fetal calf serum. Thereafter, samples were passed through a 70 µm cell strainer, centrifuged at 400 g for 4 min, washed once in PBS and counted. At this point, stromal cells were either directly mixed with epithelial cells (‘fresh stroma’) or cultured for 5 days in RPMI1640 containing 10% FCS, 1% Penicillin-Streptomycin and 1% Glutamax (‘cultured stroma’).

To test the ability of stroma to support epithelial crypt proliferation, fresh or cultured stroma (50,000 or 25,000 cells, respectively) was mixed with epithelial crypts (generally 6000 crypts) in 15 ml tubes and pelleted via centrifugation at 400 g for 4 min. Following centrifugation, supernatant was carefully removed and then pellets were re-suspended in 50 µl of Matrigel per well equivalent and then directly distributed into 48-well plates. The Matrigel was allowed to gel for 30 min at 37°C and then each well was supplemented with complete crypt culture medium. Each experiment included wells containing stroma alone to estimate amounts of contaminating epithelial stem cells.

Intestine was harvested immediately after sacrifice and washed extensively with PBS. The small intestine was cut in two identical lengths, small fragments were collected for RNA isolation, and the remaining small intestine was flushed with 4% formalin and prepared for formalin fixation and paraffin embedding as a Swiss Roll.

For confocal imaging of EGFP, the small intestine was washed with PBS and then perfused with cold 4% paraformaldehyde, prepared as a Swiss Roll, and fixed for an additional 2 h. Samples were then incubated in 15% sucrose solution for 24 h, followed by 30% sucrose for another 24 h, all at 4°C. Samples were then embedded in OCT and stored at −80°C.

Synaptophysin (Lifespan Biosciences, Cat. #LS-C49473), β-catenin [Becton Dickinson, Cat. #(421)610154] and lysozyme (Abcam, Cat. #ab108508) antibodies were used at dilution of 1:50, 1:150 and 1:5000, respectively. Antigens in formaldehyde-fixed and paraffin embedded intestinal tissues were retrieved by boiling in citrate buffer, pH 6, for 10 min. Thereafter, samples were blocked in 1% BSA for 60 min. Tissues were incubated with primary antibodies for 60 min, washed and subsequently incubated for 60 min with secondary antibody diluted 1:200. Sections were mounted in DPX medium and analyzed using a Leica DM2000 microscope. Apoptotic cells were detected by TUNEL assay, using ApopTag Plus Peroxidase In Situ (Millipore, Cat. #S7101). Ki67 and β-catenin staining in Wls^flox samples were performed as described (Zhong et al., 2012). Stromal cells cultured for 6 days were then cultured on glass coverslips for 2 days, fixed in 2% PFA in PBS for 15 min and permeabilized with 0.2% Triton X-100 for 10 min. Thereafter, samples were washed once in PBS followed by the staining procedures described above using primary antibodies to Vimentin (Cell Signaling, Cat. #5741), Desmin (Cell Signaling, Cat. #5332) and Smooth muscle actin (α-SMA; Abcam, Cat. #ab-7817) diluted 1:100 in PBS containing 1% BSA. Secondary ant-rabbit and anti-mouse antibodies (Invitrogen Alexa Fluor 594 goat anti-rabbit and mouse (Cat. #A1102, #A11005) were diluted 1:500. After staining, samples were mounted in Vectashield medium containing DAPI and analyzed using a LSM710 Carl Zeiss confocal microscope.

Cell proliferation in vivo was assessed by EdU incorporation. EdU and Click-iT EdU Alexa Fluor 555 Imaging Kit, were purchased from Life Technologies (Cat. #A10044 and #C10338, respectively). Two hours before sacrifice mice were injected with 0.5 mg EdU in 150 μl PBS (∼16.66 mg/kg). Incorporated EdU was visualized following the manufacturer's instructions and mounted in fluorescent mounting media with DAPI (VectaShield, Cat. #H-1200). When EdU staining was performed on OCT embedded samples, incubation time for the Click iT reaction cocktail was reduced to 1 min.

Epithelial (crypt cells) and stromal cells from Porcn^Del/Villin-Cre and Porcn^WT/Villin-Cre mice were harvested as described in the Materials and Methods in the supplementary material. RNA was isolated using the RNeasy purification kit from Qiagen. Labeled cRNA was prepared and hybridized to MouseWG-6 v2.0 Expression BeadChip Kit (Illumina) according to the manufacturer's protocols. The gene expression data were extracted by GenomeStudio (v1.7.0) software. After normalization by median centering, significance analysis of microarrays (SAM) with less than 10% false discovery rate (FDR) was used to compare the samples. Signaling pathways are downloaded from Molecular Signatures database (MsigDB 2.5, http://www.broadinstitute.org/gsea/msigdb/). The genes in each pathway were centered using Cluster software and their heatmaps were generated with TreeView software. Microarray data are available at Gene Expression Omnibus with accession number GSE56911.

Data were analyzed using Prism 5 software and Excel. A two-tailed t-test was performed in Excel for Mac 2011 version 14.3.2.

We thank Kyle Proffitt, Tracy Covey, and other past and present members of the Virshup lab as well as Sven Pettersson, Yunhua Zhu, Dmitry Bulavin and Alexandra Pietersen for reagents and helpful discussion. We acknowledge the invaluable assistance of Chung Hock Lee and the Duke-NUS animal facility, and Patrick Tan and the Duke-NUS Genome Biology Facility.