Embryonic cholecystitis and defective gallbladder contraction in the Sox17-haploinsufficient mouse model of biliary atresia

The gallbladder is a flexible organ of the biliary tract that regulates bile storage and discharge by the contractile movement of its smooth muscles. Malformation of the gallbladder and bile ducts can cause disease, including cholesterol gallstones, chronic inflammation and biliary atresia (reviewed by Portincasa et al., 2004, 2008; Asai et al., 2015). Congenital biliary atresia is a rare condition in newborn infants (Kohsaka et al., 2002; Mieli-Vergani and Vergani, 2009) that causes inflammation in the bile ducts and liver due to the blockage of bile flow (cholestasis). It is usually characterized by an aberrant gallbladder of reduced length, with irregular walls and indistinct mucosal lining, that appears to be closely associated with neonatal cholestasis associated with severe inflammation of the intrahepatic ducts (Desmet, 1992; Mack and Sokol, 2005; Mieli-Vergani and Vergani, 2009).

The etiology of human biliary atresia remains unclear. It is speculated to be caused by either environmental factors, such as viral infections or toxin exposure in genetically susceptible individuals, or by developmental errors during the specification and morphogenesis of bile duct epithelia (reviewed by Mack and Sokol, 2005; Mieli-Vergani and Vergani, 2009; Nakamura and Tanoue, 2013; Davenport, 2016). Whether a result of internal or external factors, epithelial defects and/or injury of the extrahepatic bile ducts may be associated with the onset of human biliary atresia by the neonatal stage (Mack and Sokol, 2005; Davenport, 2016).

The extrahepatic biliary structures (gallbladder, cystic duct, hepatic ducts and common bile duct) originate from the biliary primordium (Spence et al., 2009; Uemura et al., 2010, 2013), which expresses SRY-box 17 (Sox17), a core regulator of endoderm determination in mice and humans (Tam et al., 2003). Formation of the intrahepatic duct is regulated cooperatively by Sox9 and Sox4 (Poncy et al., 2015), albeit that their roles in the extrahepatic duct remain unclear. In a previous study (Uemura et al., 2013), perinatal lethality was observed in ∼90% of Sox17 heterozygous (Sox17^+/−) mice and the embryos displayed defective development of the gallbladder including defective bile duct epithelial wall and a biliary atresia-like phenotype (an abnormal accumulation of luminal decidual cells in the bile duct), together with severe embryonic hepatitis after the first biliary excretion into the fetal duodenum at ∼16.5 days post coitum (dpc). This timing of biliary atresia-like symptoms in mouse embryos is consistent with the hypothesis of early onset of biliary atresia in human fetuses (Tan et al., 1994; reviewed by Davenport, 2016). In human fetuses, bile acid synthesis occurs during the early organogenesis stages, during which bile starts to be excreted into the intestine near the end of the first trimester (Nakagawa and Setchell, 1990 and references therein). Furthermore, recent studies of naturally occurring outbreaks of sheep biliary atresia revealed that biliatresone, a causative toxin for biliary atresia, reduces Sox17 expression levels in the bile duct epithelia, and that silencing Sox17 mimics the effects of biliatresone on the bile duct epithelia (Lorent et al., 2015; Waisbourd-Zinman et al., 2016). Together with the extensive anatomical similarity of the extrahepatic biliary tracts and the associated blood vessels, nerves and smooth muscles between mice and humans (Higashiyama et al., 2016), the Sox17^+/− mouse embryo provides a useful experimental model with which to study the initial pathogenesis of biliary atresia.

In this study, we demonstrate the early onset of cholecystitis in the cystic duct-like gallbladder of Sox17^+/− mouse embryos, in which the reduction in sonic hedgehog (Shh) expression leads to defective contraction of the smooth muscles.

In wild-type embryos, the fetal gallbladder forms a pseudostratified columnar epithelium with epithelial folds by 17.5 dpc, while the cystic duct consists of a single-layered cuboidal epithelium with few epithelial folds throughout the fetal stages (Fig. 1A). The Sox17^+/− gallbladder shows severe hypotrophy and a single-layered cuboidal epithelium with few epithelial folds (Fig. 1A) (see also Uemura et al., 2013), similar in appearance to the cystic duct epithelia of wild-type embryos at the same stage (17.5 dpc). This is consistent with morphometric data showing a significant reduction in epithelial height, cavity and/or epithelial area of the Sox17^+/− gallbladder, such that these values are similar to those of wild-type cystic ducts (Fig. 1B).

SOX17 is highly expressed in the developing gallbladder and cystic duct epithelia. It is enriched in the distal region of the developing gallbladder epithelium. This is in contrast to the restricted distribution of SOX9-positive epithelial cells to the proximal region of the cystic duct (Fig. 1C; Fig. S1), thus showing a mostly non-overlapping pattern between distal SOX17-positive and proximal SOX9-positive epithelial cells in the gallbladder of wild-type embryos. In the Sox17^+/− gallbladder, the SOX9-positive domain expands into the distal region in the gallbladder (Fig. 1D), despite the reduced proliferation in the proximal cystic duct domain of the Sox17^+/− versus wild-type gallbladder (Fig. S2). qPCR analysis confirms that Sox9 expression is significantly increased in Sox17^+/− gallbladder compared with wild type at 15.5 dpc (Fig. 1E). Considering our previous observation of hypoplasia and deciduation in the Sox17^+/− gallbladder epithelia (Uemura et al., 2013), these data suggest that the gallbladder epithelial cells might be replaced with the ectopic SOX9-positive cystic duct-like epithelial cells. In addition, such elevated expression of Sox9 in the Sox17^+/− gallbladder correlates with the subsequent expression of Sox4, which can act redundantly with Sox9 in intrahepatic duct formation (Poncy et al., 2015) (Fig. S3).

To examine the transcriptomic changes in the Sox17^+/− gallbladder, microarray analyses were conducted using gallbladder (distal) and cystic duct (proximal) segments of the Sox17^+/− and wild-type gallbladder primordia at 15.5 dpc, before the first secretion of bile fluid from the fetal liver (Uemura et al., 2013) (Fig. 2A,B). The transcriptomic analysis identified 279 upregulated and 501 downregulated genes in Sox17^+/− gallbladder compared with wild-type littermates at 15.5 dpc (n=4; Table S1). Gene set enrichment analysis (GSEA) confirmed Sox9 among the top 50 enriched/upregulated genes in the Sox17^+/− gallbladder (Fig. S4). Moreover, these data were compared with those of gallbladder-specific or cystic duct-specific genes from the same littermates [5361 or 4886 genes (n=2) expressed at higher or lower levels, respectively, in the gallbladder segment compared with the cystic duct segment of wild-type embryos]. Of the downregulated genes in the Sox17^+/− gallbladder, 207/279 (74.4%) overlap with the gallbladder-specific genes that we identified (Fig. 2A, Table 1). Furthermore, of the upregulated genes, 221/501 (44.1%) overlap with the cystic duct-specific gene list (Fig. 2B, Table 1). This is consistent with data showing the appearance of SOX9-positive cystic duct-like epithelia in the Sox17^+/− gallbladder during the late fetal stages (Fig. 1).

GSEA showed an enrichment of genes in the category ‘definitive hemopoiesis’ (Fig. S4). Gene ontology (GO) analysis confirmed that the 280 upregulated genes, other than the 221 cystic duct-specific genes, are involved in ‘myeloid cell differentiation’ (Klf1, Tal1, Epb42, Ahsp, Prdx3, Psen2 and Trim10) (Fig. 2B), suggesting the enrichment of immature hematopoietic cells (e.g. inflammation) in the fetal gallbladder region even at 15.5 dpc. Key marker genes of several hepatobiliary and digestive tract disorders were also upregulated in this dataset, including olfactomedin 4 (Olfm4) (Liu et al., 2010; Gersemann et al., 2012), ATP-binding cassette, subfamily B4 (Abcb4) (Fickert et al., 2004; Esten Nakken et al., 2007; Baghdasaryan et al., 2011; Gordo-Gilart et al., 2015) and polycystic kidney and hepatic disease 1 (Pkhd1) (Nakamura et al., 2010; Hartley et al., 2011). We therefore examined the expression levels of these three disease markers, in addition to two inflammatory markers, namely Cxcl10 (Leonhardt et al., 2006) and Serpine1 (Bessho et al., 2014), in both gallbladder and liver samples isolated from Sox17^+/− embryos with or without hepatic lesions (severe or mild phenotype group) at 17.5 dpc (Fig. 2C,D; Fig. S5A).

qPCR analysis showed a significant increase in the expression of Olfm4, Abcb4 and Cxcl10 in the Sox17^+/− gallbladder compared with the wild-type gallbladder (Fig. 2C,D, left two bars in gb), in addition to high expression levels of both Olfm4 and Cxcl10 in the severely affected livers (Fig. 2C,D, solid circles in ʻliver'). Serpine1 expression appeared to be higher, albeit not significantly, in both the gallbladder and liver of some severe phenotype Sox17^+/− embryos. Among these four genes, we found significant positive correlations in the increased expression levels of Olfm4 (only severe samples), Cxcl10 and Serpine1 (both severe and mild samples) between gallbladder and liver samples (Spearman test; Fig. 2E; see also Fig. S5B), suggesting a positive correlation in phenotypic severity between the gallbladder (i.e. cholecystitis) and intrahepatic regions (i.e. hepatitis) in Sox17^+/− embryos. Interestingly, Abcb4 expression was increased only in gallbladder and not liver tissues (Fig. 2C; Fig. S5A), suggesting a potentially useful marker specific for extrahepatic cholestasis (Schaap et al., 2009).

Pkhd1, which encodes the ciliary protein fibrocystin, was upregulated in several mild phenotype samples (without any gross anatomical hepatic lesions) (Fig. 2C; Fig. S5A). In Sox17^+/− gallbladders, we found a positive correlation between Pkhd1 and Olfm4 expression levels independent of gross anatomical hepatic lesions, in addition to the negative correlation between Pkhd1 and Abcb4 expression levels (Fig. S5C).

In addition, qPCR analysis of several fetal myeloid cell markers, namely Tal1 (Kelliher et al., 1996), Klf1 (McConnell and Yang, 2010) and Prdx3 (Park et al., 2016), revealed a tendency for an increase in their expression levels in the Sox17^+/− gallbladder region, together with a positive correlation among these three genes (Fig. S6A,B). Moreover, immunohistochemical analysis revealed that Gr1 (LY6G)-positive or F4/80 (ADGRE1)-positive spherical myeloid-like cells appeared to be enriched in the surrounding mesenchymal region of the Sox17^+/− gallbladder, albeit with no direct infiltration into the epithelial layers of the Sox17^+/− gallbladder (Fig. S6C).

Sox17 is expressed not only in the primordial gallbladder, but also in a population of hepatoblast progenitor cells at early somite stages (Okada et al., 2012). To specify the contribution of reduced Sox17 activity in the liver or gallbladder in the onset of embryonic hepatitis, we examined the liver phenotypes of Sox17^flox embryos with a conditional deletion of Sox17 using albumin (Alb)-cre and Pdx1-cre mice. In contrast to Alb-cre-dependent recombination of the ROSA^mTmG allele in liver hepatoblasts (Fig. 3A), Pdx1-cre was able to induce recombination in approximately half of the fetal gallbladder and cystic duct epithelia by 17.5 dpc (Fig. 3A), in addition to its efficient deletion of Sox17 in the pancreas and duodenum, with both tissues showing no detectable Sox17 expression at these stages (Spence et al., 2009; Uemura et al., 2010).

First we examined the liver and gallbladder phenotypes of Alb-cre/Sox17^flox/+ embryos, and these showed no developmental defects in liver and gallbladder development (data not shown). qPCR analysis confirmed no elevated expression of the inflammatory marker genes in gallbladder or liver of Alb-cre/Sox17^flox/+ embryos at 17.5 dpc (Fig. 2C,D, two righthand bars in each group), suggesting no appreciable link to hepatitis of the reduced Sox17 expression in the liver hepatoblasts. The Alb-cre/Sox17^flox/flox embryos also showed normal liver and gallbladder phenotypes (Fig. 3B).

By contrast, Pdx1-cre/Sox17^flox/flox embryos showed severe gross anatomical hepatic lesions in some embryos (4/55 embryos; Fig. 3B), together with elevated Cxcl10 and Serpine1 expression in the affected livers (Fig. 3C). Moreover, the Pdx1-cre/Sox17^flox/flox embryos showed varying degrees of shortening of the gallbladder and cystic duct (Fig. 3D), with hepatitis observed in fetuses with normal (type I, 3/34 embryos in Fig. 3D) and mild (type II, 1/10 embryos) phenotypes, but no hepatitis in individuals with a complete lack of the gallbladder (type III, 0/11 embryos). Moreover, histopathological and qPCR analyses confirmed that Pdx1-cre/Sox17^flox/flox embryos show biliary atresia-like phenotypes similar to those of Sox17^+/− embryos, such as epithelial deciduation, bile duct stenosis/atresia and reduced Shh expression (Fig. S7) (Uemura et al., 2013; this study). These data suggest that the hepatitis in Sox17^+/− embryos is caused by reduced Sox17 activity in the primordial gallbladder, and is not associated with the liver hepatoblasts.

Among the 207 downregulated genes in the Sox17^+/− gallbladder at 15.5 dpc (Fig. 2A, Table 1; Fig. S4), we focused on the altered expression of Shh and of the Shh-responsive gene Hhip. Whole-mount in situ hybridization analysis of wild-type embryos at 11.5-13.5 dpc revealed that Shh, as well as Sox17, was expressed in the epithelium of the distal portion of gallbladder (Fig. 4A). This is in contrast to the lack of positive signals for Ihh, the paralog of Shh, in the gallbladder (Fig. 4A). Ptch1, Gli1 and Hhip were highly expressed in mesenchymal tissues adjacent to the gallbladder epithelium. Analysis using the Shh^+/−(GFP) line confirmed expression of Shh in the bile duct epithelium of both the gallbladder and cystic duct, but not in the hepatic ducts, at 16.5 dpc (Fig. 4B). These expression data suggest a potential epithelial-mesenchymal interaction for SHH signaling in developing gallbladders of mid-to-late organogenesis stage embryos.

Next, we examined the expression levels of Shh and related genes using qPCR. Both Shh and Hhip were significantly downregulated in the Sox17^+/− gallbladder (P<0.05), as compared with the wild-type gallbladder at 15.5 dpc (Fig. 4C). The expression levels of Gli1, Gli3 and Ptch1 were not affected, whereas Gli2 expression tended to be reduced, albeit not significantly, in the Sox17^+/− gallbladders (Fig. 4C).

The liver and gallbladder phenotypes were examined in wild-type, Shh^+/− and Shh^−/− embryos at 14.5-17.5 dpc. The gross anatomical and histopathological analyses appeared normal in the Shh^+/− livers, which were similar to those of wild-type littermates. Shh^−/− embryos, by contrast, were headless and exhibited severe growth retardation in the liver, albeit without any sign of hepatic inflammation (Fig. S8). Whole-mount DBA (Dolichos biflorus agglutinin) staining for the bile duct revealed a shortened gallbladder and cystic duct, accompanied by a reduction in size of the whole liver in Shh^−/− embryos (14.5 dpc, Fig. 5A). However, in contrast to the shortened gallbladder length, transverse sectioning analysis revealed that the Shh^−/− gallbladder epithelium was histologically normal, and SOX17-positive cells were present and indistinguishable from those in wild type (Fig. 5B). The proportion of PCNA-positive or Ki67-positive cells among total cells showed no appreciable differences in the gallbladder epithelia among wild-type, Shh^+/− and Shh^−/− embryos at 14.5 dpc (Fig. 5B,C; Fig. S9). Cumulatively, these data suggest that there are no appreciable defects in epithelial proliferation of Shh^−/− gallbladder primordia, which is in sharp contrast to the hypoplasia of the Sox17^+/− gallbladder epithelium at the same stages (Uemura et al., 2013).

Next, we examined the gallbladder phenotypes of Sox17^+/−;Shh^+/− double-heterozygous embryos at 14.5-17.5 dpc (Fig. 5D-H). Histopathological analyses revealed no appreciable changes in epithelial morphology or anti-SOX9 or anti-SOX17 immunostaining intensity between Sox17^+/− and Sox17^+/−;Shh^+/− gallbladders, except for one severe sample showing luminal cell debris in the presumptive gallbladder region (1/4 Sox17^+/−;Shh^+/− embryos; Fig. 5D, right; Fig. S10). Quantitative data also confirmed that there were no significant alterations in epithelial height between Shh^+/− and Shh^+/+ gallbladders in either the Sox17^+/− or wild-type background (Fig. 5E).

In contrast to the lack of appreciable defects in the Shh mutant gallbladder epithelium, the formation of smooth muscle layers was considerably affected by the reduced dosage of Sox17 and Shh (Fig. 5F,G). Anti-α smooth muscle actin (SMA) staining of wild-type embryos revealed that SMA-positive cells were detectable within the distal mesenchymal region of the gallbladder at ∼11.5 dpc, when they expand in a distal-to-proximal manner, leading to the formation of smooth muscle layers surrounding the gallbladder region, but not the cystic duct region, by 15.5 dpc (Fig. S11). In the Sox17^+/−, Shh^+/− and Shh^−/− embryos, anti-SMA staining showed reduced signal intensities in the gallbladder mesenchyme at 14.5 dpc (Fig. 5F), in which SMA-positive cells appear to be located randomly and discontinuously in the mesenchymal region around the epithelium (Fig. 5F, insets). At later stages, the Shh^+/− gallbladders showed proper formation of the SMA-positive smooth muscle layer, similar to that of the wild-type gallbladder, in which SMA-positive cells completely surround the gallbladder epithelia (Fig. 5F, left two images). Both Sox17^+/− and Sox17^+/−;Shh^+/− embryos showed severely affected smooth muscle layers, in which SMA-positive cells appeared reduced and patchy in the surrounding mesenchyme (Fig. 5G, right two images). Morphometric analyses using anti-SMA staining revealed that the signal density was significantly reduced in the gallbladder mesenchyme of Shh^+/−, Sox17^+/− or Sox17^+/−;Shh^+/− embryos at 14.5 dpc, compared with that of wild-type embryos (Fig. 5H). Moreover, anti-SMA signals were more severely affected in Sox17^+/−;Shh^+/− gallbladders than in Shh^+/− gallbladders, albeit with no significant difference. These data suggest that the formation of smooth muscle layers in Shh^+/− or Sox17^+/− gallbladders is both delayed and defective, and this phenotype appears to be more severe in Sox17^+/−;Shh^+/− gallbladders.

To examine the contractile ability of the smooth muscle layers in the fetal gallbladder, we analyzed the network of smooth muscle cells in wild-type and Sox17^+/− embryos at 17.5 dpc by whole-mount anti-SMA staining (Fig. 6). In the wild-type gallbladder, most of the smooth muscle fibers run in a circular direction (Fig. 6A), whereas in Sox17^+/− gallbladders the smooth muscle cells appear to be distributed randomly and irregularly (Fig. 6B), with a swirling pattern distinct from that seen in the wild-type circular smooth muscle layer. This is consistent with the fragmented pattern of the smooth muscle layer observed in transverse sections (Fig. 5G).

Next we examined KCl-induced contraction (the relative changes in the maximum luminal diameter) of the fetal gallbladder in wild-type and Sox17^+/− embryos with or without gross anatomical hepatic lesions (severe or mild phenotype group) at 17.5 dpc (Fig. 6C,D). In wild-type gallbladders, the KCl-induced contraction level was 87.4±1.9% (n=33; Fig. 6D). However, in the Sox17^+/− gallbladder, KCl treatment caused no appreciable contraction: 94.2±2.9% (n=11) and 99.4±4.4% (n=12) in the mild and severe phenotype groups, respectively (Fig. 6D). In particular, the contraction level of the severe phenotype group was significantly reduced in the Sox17^+/− gallbladders compared with wild type.

To examine non-induced muscle contraction in the fetal gallbladder at the perinatal stage in vivo, we isolated the liver and gallbladder from wild-type and Sox17^+/− embryos at 17.5 dpc and measured the rate of autonomous gallbladder contraction under a dissection microscope for 10 min at room temperature. In wild-type gallbladders, the average rate of circular contractile movement was 0.48±0.28 times/min (n=5; Movie 1). This is in contrast to the rare and longitudinal muscle contraction in Sox17^+/− gallbladders at 0.09±0.06 times/min (n=5; Movie 2). These data show that the defective formation of the circular smooth muscle layer affects muscle contractility in the Sox17^+/− gallbladder. This might contribute to the onset of cholestasis and biliary atresia in the Sox17^+/− mouse model.

Using an in vitro culture system in which each gallbladder is held within a single cylindrical groove of an agarose gel (Fig. 7A,B), we carried out a rescue experiment using recombinant SHH-soaked beads, which can weakly to moderately induce upregulation and downregulation of endogenous Gli2 and Shh expression, respectively, in the 13.5 dpc gallbladder explants (Fig. S12). Wild-type gallbladder explants elongated in a distal-to-proximal manner to form a tubular structure along the cylindrical space within the agarose gel. Furthermore, these gallbladders displayed a well-developed pseudostratified columnar epithelium with epithelial folds (n=14/17; Fig. 7A,C, left), in addition to well-developed SMA-positive smooth muscle layers surrounding a SOX9-negative gallbladder epithelium (n=11/14; Fig. 7D,F, left). By contrast, the control (PBS) explants of Sox17^+/− gallbladders displayed a single-layered cuboidal epithelium without any epithelial fold formation (n=12/12; Fig. 7C). The SMA-positive smooth muscle layer developed poorly around the SOX9-positive gallbladder epithelium (n=10/11; Fig. 7D), recapitulating the defective phenotypes of the Sox17^+/− gallbladder in vivo. Interestingly, the addition of SHH-soaked beads rescued several of the Sox17^+/− gallbladder phenotypes, resulting in a well developed smooth muscle layer organized around a pseudostratified epithelium (n=10/13; Fig. 7C,D, right two images). Morphometric analysis confirmed that the epithelial height and area of the luminal cavity in the Sox17^+/− gallbladder were comparable to those of wild-type explants following SHH treatment (n=11; Fig. 7G). These findings suggest that reduced SHH signaling results in defective smooth muscle formation and epithelial fold formation in the Sox17^+/− gallbladder.

In addition, ectopic SHH signaling had no appreciable effect on the percentage of PCNA-positive or Ki67-positive cells in the Sox17^+/− gallbladder epithelium (Fig. 7E,H; Fig. S13), or on the number of ectopic SOX9-positive epithelial cells in the Sox17^+/− gallbladder (epithelial area in Fig 7G, Fig. 7F,I). The Sox17^+/− gallbladder explants showed shredding of some epithelial cells even in vitro (Fig. S14A; see also Fig. 7C, right, inset). However, the remaining Sox17^+/− epithelial layer exhibits proper subcellular localization of both E-cadherin and ZO-1 (TJP1) even in the region near the damaged site (Fig. S14A). Moreover, the expression levels of Cxcl10 and Olfm4, although not Serpine1, showed a tendency to increase in the Sox17^+/− explants as compared with the wild-type explants (Fig. S14B), but such increased expression, at least of Olfm4, could not be rescued by SHH-soaked beads (Fig. S14B). These data suggest that most defective phenotypes in the Sox17^+/− gallbladder epithelium are SHH independent. The present results are summarized in Fig. S15.

First, we demonstrated the presence of ectopic SOX9-positive cystic duct-like epithelia in the Sox17^+/− gallbladder, together with the cooperatively increased expression of Sox4, which functions redundantly with Sox9 in intrahepatic ducts (Poncy et al., 2015). Sox9 is highly expressed in the cystic duct epithelium, in addition to the intrahepatic, pancreatic and common bile ducts (Furuyama et al., 2011). One possible explanation for the ectopic Sox9/Sox4-positive cystic duct epithelia in the proximal gallbladder region is an expansion of the cystic duct region toward the gallbladder domain as a consequence of the reduced proliferation and luminal deciduation of the Sox17^+/− gallbladder epithelial cells (Uemura et al., 2013). However, the reduced epithelial proliferation is also observed in the cystic duct domain of the Sox17^+/− gallbladder primordium, in addition to there being lower proliferative activity in the proximal cystic duct region than in the SOX17-positive distal region (Uemura et al., 2013). Although SOX17 and SOX9/SOX4 have similar sequence specificity in their DNA-binding HMG box domains (Kanai et al., 1996; Mertin et al., 1999), they have distinct characteristics for β-catenin binding and dimerization (Wilson and Koopman, 2002; Sinner et al., 2007; Kamachi and Kondoh, 2013). Hence, the cystic duct-like phenotype in the proximal Sox17^+/− gallbladder might be caused partially by the appearance of ectopic SOX9/SOX4-positive cells in the bile duct epithelia, instead of by the loss of SOX17-positive cells. Further studies are required to more precisely define the hierarchy downstream of SOX17 or SOX9/SOX4 in the specification of the gallbladder and cystic duct epithelia of wild-type and Sox17^+/− embryos.

By the perinatal stage, cystic duct-like Sox17^+/− gallbladders showed increased expression of the inflammatory/cholestasis-associated markers Abcb4 and Olfm4, in addition to Cxcl10, a key chemokine gene characteristic of the early immune response (Fig. 2C,D). The expression levels of several fetal myeloid cell markers also showed a tendency to increase in the Sox17^+/− gallbladders, together with the enrichment of Gr1-positive or F4/80-positive myeloid-like cells in the gallbladder mesenchymal region (Fig. S6), suggesting fetal cholecystitis in Sox17^+/− embryos. Abcb4 encodes a lipid translocator for phosphatidylcholine, which transfers bile into bile salt micelles for the protection of the apical surface of the bile duct epithelia (Fickert et al., 2004; Esten Nakken et al., 2007; Baghdasaryan et al., 2011; Oude Elferink and Paulusma, 2007). Abcb4 null mice display severe sclerosing or chronic cholangitis (Fickert et al., 2004; Esten Nakken et al., 2007; Baghdasaryan et al., 2011). Since a significant elevation of ABCB4 was also reported in human patients with extrahepatic cholestasis (Schaap et al., 2009), Abcb4 upregulation in Sox17^+/− gallbladders in vivo might be caused partially by its adaptive response to minimize cellular damage by decreasing bile salt toxicity. Olfm4 encodes a secreted protein with one olfactomedin-like domain (also known as an intestinal stem cell marker; Gersemann et al., 2012) and is upregulated in inflammatory bowel diseases such as ulcerative colitis and Crohn's disease (Liu et al., 2010; Gersemann et al., 2012). It is also possible that the in vivo upregulation of these bile duct disease markers is an effort to compensate for the defective barrier of the bile duct epithelia in the Sox17^+/− gallbladder.

Genetic analysis demonstrated that embryonic hepatitis could be induced by a conditional Sox17 deletion in the primordial gallbladder, but not in the liver hepatoblasts (Fig. 3; Fig. S7). The expression analysis of inflammatory markers also showed a positive correlation between cholecystitis and the severity of embryonic hepatitis in Sox17^+/− embryos (Fig. 2D,E). These data suggest that the cholecystitis causes the biliary atresia and subsequent hepatic inflammation in Sox17^+/− embryos. It is also possible that, at the late organogenesis stages of the extrahepatic bile ducts, the inflammatory responses of the bile duct epithelia and its surrounding mesenchymal tissues cause severe defects in both bile duct morphogenesis and epithelial barrier function, leading to increased bile leakage and inflammatory responses in the fetal gallbladder by 17.5 dpc.

The present genetic and in vitro rescue experiments revealed that SHH signaling may be crucial for the proper formation of smooth muscles downstream of Sox17 in the normal development of the gallbladder during the late organogenesis stages. HH signaling in the endodermal epithelium plays a major role in the mesenchymal development of the digestive tract. Loss of HH signaling has also been shown to decrease cell proliferation in the underlying mesenchyme, resulting in thinner walls of the stomach and intestine (Mao et al., 2010; reviewed by van den Brink, 2007). Moreover, SHH signaling enhances smooth muscle formation in the intestine, lung, and urinary bladder via other molecules, including Ptch and Gli family members (Apelqvist et al., 1997; Li et al., 2004; Caubit et al., 2008; DeSouza et al., 2013), suggesting that smooth muscle cells might be a target of SHH signaling. These previous reports help to explain the delayed and aberrant formation of smooth muscle layers in the developing Sox17^+/− gallbladder, and they are also consistent with several previous reports showing deficient smooth muscle layers in the fetal gallbladder caused by Foxf1 haploinsufficiency, a downstream mediator of HH signaling, in mouse organogenesis stage embryos (Kalinichenko et al., 2002; Madison et al., 2009).

It was recently suggested that the deformation, inflammation and repair processes of the bile ducts are closely associated with SHH signaling. For example, Jung et al. (2015) reported that SHH pathway activation was observed, especially in the cholangiocytes of the peribiliary glands, in human patients with biliary atresia. In pathological studies, some have reported that Hh expression is considerably increased in fibrotic damaged biliary diseases in response to injury (Omenetti et al., 2008; Omenetti and Diehl, 2011; Cui et al., 2013; Hu et al., 2015). It has also been speculated that the SHH signaling pathway regulates the epithelial-mesenchymal transition of cholangiocytes (Omenetti et al., 2008; Jung et al., 2015). The effects of excess SHH signaling encompass defective hepatobiliary ducts in a zebrafish model (Cui et al., 2013; Tang et al., 2016), suggesting that excess SHH signaling can have detrimental effects on the proper development and maintenance of the biliary duct. The current phenotype of reduced SHH signaling in the defective gallbladder might be explained by the distinct roles of SHH signaling during developmental and early pathogenic stages (i.e. smooth muscle layer formation and tubulogenesis) versus late pathogenic and inflammatory processes (i.e. transdifferentiation of damaged mesenchymal cells into smooth muscle cells for tissue repair). It is also possible that SHH signaling must be maintained at an appropriate level, since both excess and reduced SHH signaling may possibly lead to defective formation and maintenance of the biliary duct system.

Finally, the present study has shown defective gallbladder contraction in Sox17^+/− perinatal embryos (Fig. 6; Movies 1 and 2). In particular, Sox17^+/− gallbladders with hepatic lesions have severe defects in contraction, suggesting a potential contribution of the defective muscle layers to the onset of cholestasis in the Sox17^+/− embryos soon after the first biliary excretion into the fetal duodenum. Since well-developed smooth muscle layers are restricted to the gallbladder wall, in contrast to poor formation of the smooth muscle layers around the cystic and common bile ducts (Higashiyama et al., 2016 and references therein), the defective gallbladder contraction directly affects the flow of bile from the fetal gallbladder into the duodenum in the Sox17^+/− embryos. This symptom is also consistent with the ultrasonographic diagnostic features for human biliary atresia, such as the ʻnon-contractile' gallbladder with its reduced length, irregular wall and indistinct mucosal lining (Kanegawa et al., 2003). Further studies are required to establish the contribution of fetal gallbladder contraction in the onset of biliary atresia-like syndrome in individual Sox17^+/− embryos.

Animal experiments were performed in strict accordance with the Guidelines for Animal Use and Experimentation established by the University of Tokyo (approval ID: P13-763, P14-877), the Tokyo Medical and Dental University (approval ID: 0140007A, 0150259C2, 0160024C2, 0170248C2) and the University of Utah (approval ID: 14-01003). The Sox17^+/− embryos at F9-10 generation were obtained from wild-type females [C57BL6 (B6) strain; Clea, Japan] mated with the Sox17^+/− male mice (Kanai-Azuma et al., 2002), which were intercrossed and maintained at F8-9 backcross generation to the B6 strain. In this mating system, the Sox17^+/− embryos (F9-10) show ∼70% neonatal lethality, while the remaining survivors can be grown until adulthood without any signs of hepatitis, albeit with a small gallbladder. Sox17^+/−(GFP), Sox17 ^flox/flox (B6×129sv background; 13.5-17.5 dpc) (Kim et al., 2007), Alb-cre (Postic et al., 1999), Pdx1-cre (Hingorani et al., 2003) and Shh^+/−(GFP) (Harfe et al., 2004) mice were also used in this study.

Tissues were fixed in 4% paraformaldehyde in PBS for 12 h at 4°C. For sections, the fixed samples were embedded in paraffin and serially sectioned (4 µm thick). For comparative analyses of transverse sections of the gallbladder, sections at the level of maximum diameter (at least four sections in each individual sample) were used. Immunohistochemistry was conducted by a standard protocol (see the supplementary Materials and Methods) and in situ hybridization was performed as described by Hiramatsu et al. (2009). Details of reagents and methods are provided in the supplementary Materials and Methods.

For length and area measurements, ImageJ 1.48V software (National Institutes of Health, USA) was used. For measurement of signal density of anti-SMA-stained sections, we conducted grid analyses of the HRP-positive reactions. A grid (each box=10×10 µm) was overlaid on the image of the stained section and the signal density (d) calculated by d=b/(a+b), where (a) the signal passed through more than two sides of the box and (b) the signal was fragmented or passed through one side of the box.

The Sox17^+/− and wild-type gallbladders at 15.5 dpc were used for microarray expression analysis according to the method developed by Huang et al. (2009a,b). GSEA was carried out according to Subramanian et al. (2005). qPCR analysis was used to determine marker gene expression levels relative to Gapdh. Details of these methods are provided in the supplementary Materials and Methods.

Whole livers, including the gallbladder and biliary tract (17.5 dpc), were maintained in 10% fetal calf serum-DMEM (Sigma) at 37°C, and prepared for time-lapse imaging using a dissection microscope (Olympus SZX16) equipped with a video recording system (Olympus DP71 camera; imaged every 2 s for a total of 10 min). Each gallbladder was separated from the liver and imaged before and after 45 mM KCl treatment for 10 min. Maximum luminal diameter of the gallbladder before and after KCl was measured using ImageJ 1.48V, and the relative change in maximum luminal diameter estimated as an indication of gallbladder contraction.

The gallbladder was isolated from the fetal liver using a dissection microscope at 13.5 dpc. To apply the pressure that would normally be exerted by the liver and to promote directional growth, the gallbladders were placed in a groove of a 1.5% agarose gel plate prepared using a stainless steel needle of 0.26 mm diameter, which is similar to the liver gap around the gallbladder at 16.5 dpc (see Fig. 7B). Beads soaked in mouse recombinant SHH (1 mg/ml; SRP6004, Sigma) or PBS were placed on the distal tip of the gallbladder and cultured in 10% fetal calf serum-DMEM at 37°C for 72 h.

All quantitative data are represented as mean±s.e.m. Student's t-test, Mann–Whitney U-test and ANOVA tests were used to determine overall differences between two groups or among more than two groups. Where differences existed, Tukey's test was also used to compare each value with every other value. The correlation between genes/groups in both gallbladder and liver samples was estimated using Spearman's rank correlation test. P<0.05 was considered statistically significant.

We thank Drs A. Asai, S. Elliott, K. Nakamura and A. Tanoue for critical reading of the manuscript; Prof. Dr S. J. Morrison for providing Sox17^+/−(GFP) mice; and Dr M. Kawasumi, Y. Kuroda, Ms Y. Uchiyama and I. Yagihashi for helpful support.