A classification of ductal plate malformations based on distinct pathogenic mechanisms of biliary dysmorphogenesis§

Authors


  • Potential conflict of interest: Nothing to report.

  • This work was supported by the Interuniversity Attraction Pole (IAP) Programme (Belgian Science Policy, to O.D., P.C., and F.L.), the D.G. Higher Education and Scientific Research of the French Community of Belgium (to F.L.), the Alphonse and Jean Forton Fund (to O.D. and F.L.), the Fonds de la Recherche Scientifique Médicale (Belgium, to F.L.), the European Community (EUNEFRON, GA#201590, to O.D.), and the National Institutes of Health (DK55534; to L.G.W.). M.P. was a recipient of “Bettencourt Schueller prix d'élan pour la recherche” and “Equipe Fondation pour la Recherche Médicale”. C.E.P. is Senior Research Associate of the FRS-FNRS (Fonds National de la Recherche Scientifique).

  • §

    See Editorial on Page 1795

Abstract

Ductal plate malformations (DPMs) are developmental anomalies considered to result from lack of ductal plate remodeling during bile duct morphogenesis. In mice, bile duct development is initiated by the formation of primitive ductal structures lined by two cell types, namely ductal plate cells and hepatoblasts. During ductal plate remodeling, the primitive ductal structures mature to ducts as a result from differentiation of the ductal plate cells and hepatoblasts to cholangiocytes. Here, we report this process is conserved in human fetal liver. These findings prompted us to evaluate how DPMs develop in three mouse models, namely mice with livers deficient in hepatocyte nuclear factor 6 (HNF6), HNF1β, or cystin-1 (cpk [congenital polycystic kidney] mice). Human liver from a patient with a HNF1B/TCF2 mutation, and from fetuses affected with autosomal recessive polycystic kidney disease (ARPKD) were also analyzed. Despite the epistatic relationship between HNF6, HNF1β, and cystin-1, the three mouse models displayed distinct morphogenic mechanisms of DPM. They all developed biliary cysts lined by cells with abnormal apicobasal polarity. However, the absence of HNF6 led to an early defect in ductal plate cell differentiation. In HNF1β-deficient liver, maturation of the primitive ductal structures was impaired. Normal differentiation and maturation but abnormal duct expansion was apparent in cpk mouse livers and in human fetal ARPKD. Conclusion: DPM is the common endpoint of distinct defects initiated at distinct stages of bile duct morphogenesis. Our observations provide a new pathogenic classification of DPM. (HEPATOLOGY 2011;)

Ductal plate malformations (DPMs) are characterized by the persistence of embryonic biliary structures after birth.1 They consist of biliary cell clusters or duct-like structures with elongated lumina and variable shape, and are found in several congenital diseases.2 DPMs are considered to result from lack of remodeling of the ductal plate during the fetal period. However, recent insight into the mode of biliary tubulogenesis identified a new step in biliary tubulogenesis,3 prompting the need to reassess how DPMs develop.

Bile duct development in the mouse starts around embryonic day (E)12.5 with the differentiation of hepatoblasts into biliary precursor cells. The latter form the ductal plate, a single-layered sleeve of cells located around the portal mesenchyme. Around E15.5, tubulogenesis is initiated by the formation of primitive ductal structures (PDS), which are developing ducts asymmetrically lined on the portal side by ductal plate cells, and on the parenchymal side by hepatoblast-like cells. When the PDS mature to ducts (E17.5 to birth), the cells on the portal and parenchymal sides differentiate to mature cholangiocytes, which then symmetrically line the duct lumina; simultaneously, the ducts become integrated in the portal mesenchyme, and the ductal plate cells not involved in tubulogenesis involute.4, 5

For this article, we took advantage of three mouse models with DPMs to investigate the dysmorphogenesis leading to DPM and to study the epistatic relationship between the transcription factors hepatocyte nuclear factor 6 (HNF6) and HNF1β and their downstream targets potentially involved in DPM. Mouse knockouts for HNF6 or with liver-specific inactivation of HNF1β show DPM associated with cholestasis.6, 7 HNF6 directly stimulates expression of HNF1β,6 but the effectors of HNF6 and HNF1β are not known. In pancreatic ducts, HNF6 controls primary cilia formation and stimulates expression of cystin1 (Cys1) and polycystic kidney and hepatic disease-1 (Pkhd1),8 two genes that control ciliogenesis and whose mutations are associated with cyst formation9-14; HNF1β stimulates expression of Pkhd1 in kidneys.15, 16 Moreover, mice deficient in cystin-1 (congenital polycystic kidney [cpk]) display DPMs.17, 18

Thus, we analyze here the dysmorphogenesis causing DPMs in HNF6- and HNF1β-deficient mice, as well as in livers deficient in cystin-1, which is identified as a common target of HNF6 and HNF1β. We focused on differentiation, apicobasal polarity, and ciliogenesis, and found that distinct defects initiated at distinct stages of bile duct morphogenesis may lead to DPMs.

Abbreviations

ARPKD, autosomal recessive polycystic kidney disease; cpk, congenital polycystic kidney; DPM, ductal plate malformation; E, embryonic day; HNF, hepatocyte nuclear factor; OPN, osteopontin; PDS, primitive ductal structures; PKHD1, polycystic kidney and hepatic disease-1; SOX9, sex-determining region Y–related HMG box transcription factor 9; TβRII, transforming growth factor receptor type II; W, week of gestation; ZO-1, zonula occludens-1.

Materials and Methods

Animals.

Wild-type, Hnf6, Hnf1bloxP/loxP-Alfp-Cre, and cpk mice6, 7, 19 were treated according to the principles of laboratory animal care of the National Institutes of Health and with approval from institutional animal welfare committees.

Human Fetal Liver.

Tissue samples were obtained in compliance with the French and the Belgian legislations, the 1975 Declaration of Helsinki, and the European Guidelines for the use of human tissues. Autosomal recessive polycystic kidney disease (ARPKD) fetuses had kidneys with radially oriented dilations of the medullary collecting ducts20 and had mutations in the PKHD1 gene.21

Immunofluorescence.

Human samples (five normal at 11 weeks of gestation [W], two ARPKD at 13W, one ARPKD at 22W, and one patient with HNF1B mutation at 4 days after birth) were formalin-fixed and paraffin-embedded. Mouse liver preparation and immunofluorescence analysis of 5-μm-thick (human) or 9-μm-thick (mouse) sections were as described22 (Supporting Table 1).

Quantitative Reverse-Transcriptase Polymerase Chain Reaction.

RNA from mouse liver was extracted with Tripure RNA Isolation reagent (Roche, Vilvoorde, Belgium), and quantitative reverse-transcriptase polymerase chain reaction (Supporting Table 2) was performed with SYBR Green Master Mix Reagent (Invitrogen, Merelbeke, Belgium). For quantification of Sec63, Pkhd1, and Cys1, copy number for each messenger RNA (mRNA) was normalized to β-actin mRNA copy number using standard calibration curves, and reported by reference to control values set at 100%.

Results

Transient Asymmetry During Bile Duct Development Is Differentially Controlled by HNF6 and HNF1β.

To investigate how DPMs develop in the absence of HNF6 or HNF1β, we first determined the differentiation status of the ductal cells lining DPMs in livers of Hnf6−/− mice and of mice with liver-specific inactivation of Hnf1b (Hnf1bloxP/loxP-Alfp-Cre). Because differentiation progresses from the hilum to the periphery, all livers were analyzed near the hilum. At E17.5, the cells lining biliary structures in Hnf6−/− mice did not express the biliary marker sex-determining region Y–related HMG box transcription factor 9 (SOX9), but most cells expressed the hepatocyte/hepatoblast marker HNF4. Higher E-cadherin (Ecad) expression in biliary cells as compared to parenchymal cells is typical for mouse fetal liver (Fig. 1A). The ectopic expression of HNF4 in biliary structures (arrowheads) was in line with our earlier observation that the lack of HNF6 generates hybrid hepatobiliary cells.23 Such cells were observed on the portal and parenchymal sides of the biliary structures, which suggests that HNF6 is required for differentiation of the two biliary layers. In control mice, expression of the transforming growth factor receptor type II (TβRII) is absent from the portal side of PDS at E15.5 but is detected on their parenchymal side.3 At E17.5, the expression on the parenchymal side progressively waned, leading to ducts with a limited number TβRII-positive cells (open arrowheads, Supporting Fig. 1) and devoid of TβRII at E18.5.3 In the absence of HNF6, expression of TβRII persisted on both sides of the biliary structures at E17.5 (arrowheads, Supporting Fig. 1), and this was again in line with our earlier data at E15.5 which showed elevated expression of TβRII in the liver.23

Figure 1.

Abnormal biliary differentiation in embryonic Hnf6−/− livers and perturbed duct maturation in livers deficient in HNF1β. (A) Wild-type biliary cells (E17.5) are all SOX9+/HNF4. In Hnf6−/− livers (E17.5), biliary cells fail to express SOX9 whereas most cells express HNF4. In HNF1β-deficient mouse livers (E17.5), cells on the portal side are normal, whereas cells on the parenchymal side are still SOX9/HNF4+. (B) Ductal plate malformation in a 4-day-old patient with a mutation in HNF1B. The portal mesenchyme contains short cords of SOX9+ cells (arrows) and HNF4 cells (arrowheads). (C) Primitive ductal structures (PDS) with asymmetrical expression of SOX9 are found in a normal human fetal liver (W11), indicating that duct morphogenesis proceeds in humans as in mice. PDS and mature ducts lined by SOX9+ cells are found within the same liver. Dotted lines, developing ducts; pv, portal vein; *, duct lumen; scale bar, 40 μm.

In the absence of HNF1β at E17.5, differentiation of cells that lined the portal side of the DPM was normal throughout the liver: these cells were SOX9+/HNF4/Ecadhigh. In contrast, cells lining the parenchymal side were SOX9/HNF4+/Ecadlow, indicating that they had not matured to biliary cells but that HNF1β-deficient biliary structures were still asymmetrical (Fig. 1A). This was supported by the observation at E17.5 that all cells on the parenchymal side of the biliary structures expressed TβRII (arrowheads), whereas cells on the portal side no longer expressed TβRII (open arrowheads, Supporting Fig. 1).

At postnatal day 7, biliary cells had differentiated in Hnf6−/− and in Hnf1bloxP/loxP-Alfp-Cre livers because they were SOX9+/HNF4 (arrows, Supporting Fig. 2A). Therefore, embryonic biliary differentiation defects seemed to resolve, but this was not sufficient to allow normal tubulogenesis: Hnf6−/− livers showed DPM, and HNF1β-deficient livers showed heterogeneity, combining DPM and dysplastic ducts (Supporting Fig. 2A,B) within the same liver. Therefore, in HNF1β-deficient mice, a homogeneous embryonic phenotype gives rise to a heterogeneous postnatal phenotype. We concluded that the absence of HNF6 induces an early defect in biliary cell differentiation, whereas the lack of HNF1β leads to deficient maturation of PDS; both defects ultimately give rise to DPMs.

Transient Asymmetry in Human Bile Duct Development and Ductal Plate Malformation in a Patient with a HNF1B Mutation.

There are no HNF6 mutations reported in humans. In contrast, patients with HNF1B (TCF2) mutations present with renal cysts and diabetes syndrome (Mendelian Inheritance in Man #137920). There is phenotypic variability and in rare cases this syndrome is associated with bile duct paucity.24, 25 We therefore looked for the presence of DPMs in a patient with HNF1B mutation. This patient had multicystic kidneys and died at 4 days from pulmonary insufficiency; analysis of the HNF1B gene revealed heterozygous deletion of exon 6. Immunostaining of sections showed DPMs constituted of clusters of SOX9+/Ecad+ cells (arrows) and short cords of HNF4/Ecad+ cells (arrowheads) in the portal mesenchyme (Fig. 1B). Dysplastic ducts were also found (Supporting Fig. 3). Therefore, HNF1B mutation in humans can be associated not only with bile duct paucity but also with DPMs and duct dysplasia.

The limited availability of samples from patients with HNF1B mutations precludes analysis of the morphogenesis of DPMs. Therefore, we speculated that DPMs develop similarly in patients and in Hnf1bloxP/loxP-Alfp-Cre mice. This was supported by our observations that bile duct development in humans proceeds by transient asymmetry, like in mice (Fig. 1C). Indeed, the liver of a normal fetus at the 11th week of gestation (W) showed PDS with asymmetrical expression of SOX9. Because maturation of ducts is not equal throughout the liver, ducts entirely lined by SOX9+ cells were also found within the same liver.

Ciliogenesis and Polarization of Cholangiocytes Is Impaired in the Absence of HNF6 and HNF1β.

HNF6 controls the formation of primary cilia in the pancreas and HNF1β regulates genes involved in cilium function in mouse kidneys.8, 15 Therefore, we investigated how ciliogenesis proceeds in the biliary tract of Hnf6−/− and Hnf1bloxP/loxP-Alfp-Cre mice. Primary cilia were identified as acetylated tubulin-stained dots in wild-type livers at E17.5 (Fig. 2). In contrast, little or no cilia were detected on HNF6- and HNF1β-deficient biliary cells. In wild-type cholangiocytes, the basal body, which is detectable by γ-tubulin staining and constituted by two centrioles, is assembled near the apical pole of the cells. In HNF6- and HNF1β-deficient biliary cells, the centrioles were randomly distributed (Fig. 2).

Figure 2.

Deficient ciliogenesis and apicobasal polarity in biliary cells in the absence of HNF6 or HNF1β at E17.5. Biliary cells do not show primary cilia (acetylated tubulin [AcTub]) in Hnf6−/− and Hnf1bloxP/loxP-Alfp-Cre livers. Mucin-1 is absent and centrioles (γ-tubulin, arrows; yellow dotted lines; biliary epithelium) are randomly located instead of being apical as in controls. White dotted lines; developing ducts or cysts; pv, portal vein; *, duct lumen; white and yellow scale bars, 50 and 10 μm, respectively. KO, knockout; TOPRO labels nucleic acids.

The lack of cilia at the apical pole was associated with polarity defects. The apical markers mucin-1 and osteopontin (OPN) were not expressed in the absence of HNF6 or HNF1β (Fig. 2 and Supporting Fig. 4). The Golgi marker GM130 was randomly distributed instead of being located along the apicobasal axis between the nucleus and apical pole (Supporting Fig. 4). The basement membrane component laminin, normally connected to the basal pole, formed a continuous layer encircling the ducts in wild-type livers. In the absence of HNF6 or HNF1β the laminin layer was continuous along the basal pole of cells located at the portal side of biliary structures, but was irregular and discontinuous along the cells lining the parenchymal side (arrowheads in Supporting Fig. 4). Phenotypic variability was observed in Hnf6−/− livers, in which some cystic structures were entirely lined by a near continuous layer of laminin (Fig. 2); this variability was unrelated to the hilar-periphery axis. Tight junctions separate the apical pole from the basolateral pole of cells. In the absence of HNF6, zonula occludens-1 (ZO-1) was localized at the apical/lateral boundary on cells of the parenchymal and portal side (Supporting Fig. 4). When Hnf1b was deleted, ZO-1 was barely detected in cells of the parenchymal side but was clearly detectable at the portal side. On the portal side, ZO-1 staining did not show the expected punctuate pattern of tight junctions but showed coverage of the apical surface on serial confocal sections, suggesting lack of apical pole (Supporting Fig. 4).

At postnatal day 7, a few Hnf6−/− biliary cells showed normal location of cilia (acetylated tubulin, arrows), mucin-1, and laminin, indicating partial restoration of apicobasal polarity (Supporting Fig. 2A). A fraction of HNF1β-deficient cells expressed mucin-1; no cilia were detected on these cells, the basal bodies remained randomly distributed, and ZO-1 still showed apical coverage (Supporting Fig. 2A). In patients with HNF1B mutation, ZO-1 was irregularly expressed in dysplastic ducts and the DPM did not express ZO-1 (Supporting Fig. 3). We concluded that HNF6 and HNF1β are required for normal development of cilia and for polarization of the cells.

HNF6 and HNF1β Stimulate the Expression of Cystin-1 in Liver.

The lack of cilia in HNF6- and HNF1β-deficient livers prompted us to investigate whether the expression of genes controlling ciliogenesis or cilium function was affected in these mice. These genes included a set that is regulated by HNF1β in kidneys,15, 16, 26 as well as those associated with hepatorenal fibrocystic diseases27 (Supporting Table 2). In the absence of HNF6 at E17.5, Pkhd1 and Sec63 mRNA levels were up-regulated, but were unaffected in Hnf1b knockout (KO) livers (Fig. 3). In contrast, expression of Cys1 was reduced in both mutants. The other genes tested had normal expression (data not shown); this suggests that HNF1β targets in kidney, such as Pkhd1, Pkd2, Nephronophthisis 1 (Nphp1), Intraflagellar Transport 88 (Ift88), Kinesin family member 12 (Kif12) are not regulated by HNF1β in liver, and that the transcriptional network operating in bile duct morphogenesis is distinct from that in kidneys.

Figure 3.

Expression of genes required for cilium function at E17.5. The list of genes tested is shown in Supporting Table 2. Expression of Sec63 and Pkhd1 is up-regulated in the absence of HNF6. HNF6 and HNF1β are required to stimulate expression of Cys1. Means ± standard error of the mean, n = 5 (Hnf6−/−) or n = 6 (Hnf1bloxP/loxP-Alfp-Cre), *P < 0.05.

Cystin-1 Is Required Prenatally for Normal Polarization of Biliary Cells and for Duct Morphogenesis.

Cys1 was the only common target of HNF6 and HNF1β. Cpk mice have a mutation in the Cys1 gene and display renal disease resembling ARPKD, in association with DPMs.12, 17 Therefore, Cys1 is a candidate effector of HNF6 and HNF1β. We investigated how cyst morphogenesis is initiated in cpk mice and human ARPKD fetuses. Cysts were already prominent in cpk mice at E17.5, and were lined by SOX9+/HNF4 cells. Except for a few cells (arrowhead, Supporting Fig. 5), most biliary cells no longer expressed TβRII (open arrowhead, Supporting Fig. 5). Cysts in human ARPKD fetuses at 13W and 22W were lined on the parenchymal and portal sides by cells expressing SOX9 (Fig. 4A). Therefore, biliary cells in cpk embryos and human fetal ARPKD showed normal differentiation. This was also the case after birth (Supporting Fig. 2A).

Figure 4.

Abnormal biliary tubulogenesis in ARPKD fetuses and in cpk embryos. (A) In ARPKD fetuses at W13 and W22, developing ducts are expanded and lined by SOX9+ cells. (B) Biliary cell polarity is perturbed in cpk embryos at E17.5. The number of cells expressing mucin-1 (Muc1) and showing primary cilia (acetylated tubulin [AcTub]) is reduced; developing ducts give rise to cysts. In cpk livers, E-cadherin (Ecad) expression often extends toward the apical pole and more extensively covers the basal pole (arrows). Laminin expands along the lateral (arrows in inset) and apical poles of biliary cells. pv, portal vein; *, duct or cyst lumen; white and yellow scale bars, 50 and 20 μm, respectively.

The apical pole marker OPN was equally expressed in wild-type and cpk biliary cells at E17.5, but a lower number of cells showed cilia and mucin-1 in cpk mice (Fig. 4B). E-cadherin did not show the expected basolateral location in cpk biliary cells, because it extended toward the apical pole and covered the basal pole more extensively (arrows, Fig. 4B). The laminin layer was thickened and irregular, and was fragmented along the parenchymal side of the cysts. Laminin also expanded along the lateral and apical membranes (arrow, Fig. 4B), suggesting that the basal and lateral poles were not correctly set. This phenotype persists after birth: some cells did not express mucin-1 and showed apical location of laminin (arrow, Supporting Fig. 2A). The localization of ZO-1 was not restricted to the apical/lateral boundaries, but often extended to cover the apical surface. In wild-type mice at E17.5, serial sections (1.5 μm) observed by confocal microscopy disclosed the expected belt of ZO-1 expression, i.e., two dots when the focal plane intersects the belt, and linear staining when the focal plane sections an extended belt-like structure (Supporting Fig. 6). In cpk cysts, the intensity of ZO-1 staining was stronger and a higher number of successive confocal sections showed a linear staining of ZO-1, revealing that ZO-1 extensively covered the apical pole (scheme in Supporting Fig. 6). Interestingly, fetuses with ARPKD resulting from PKHD1 mutation had a phenotype similar to that of cpk mice (Supporting Fig. 6). In addition, the same extension of ZO-1 staining was detected on the portal side of DPM in HNF1β-deficient livers (Supporting Fig. 4).

We conclude that DPM and cyst formation in cpk mice and human ARPKD results from dysmorphogenesis affecting the parenchymal and portal sides of developing ducts, with normal differentiation but perturbed apicobasal polarity. Expression of cpk is controlled by HNF6 and HNF1β, but cystin-1 is not their main effector: except for the fragmented laminin layer surrounding the cysts and perturbed apicobasal polarity, there is limited phenotypic overlap between HNF6- and HNF1β-deficient livers and cpk livers. Therefore, the analysis of the various mutants indicates that distinct mechanisms operate to generate DPMs and cysts.

Discussion

Recent insight into the mechanism of bile duct morphogenesis, and in particular the transient asymmetry,3 shown here to occur in humans as in mice, prompted us to evaluate the morphogenesis of DPM, by investigating differentiation, polarity, and ciliogenesis. We studied three mouse models with DPM and mainly focused on embryos at E17.5. This stage enabled us to investigate the ductal plate, the PDS, and their maturation. The morphogenic mechanisms leading to DPM differed in the three models (Fig. 5): differentiation of biliary precursors from hepatoblasts was perturbed in HNF6-deficient fetuses, maturation of PDS failed in HNF1β-deficient livers, and abnormal expansion of ducts occurred in cpk mice and human ARPKD. Considering that DPM is the common endpoint in these animal models and human cases, but that the pathogenic mechanism leading to DPM differs in all those instances, we propose to classify (Fig. 5) the DPM according to distinct defects in (I) differentiation of biliary precursor cells, (II) maturation of PDS, or (III) duct expansion during development.

Figure 5.

Classification of DPM based on distinct pathogenic mechanisms. Normal bile duct morphogenesis is illustrated at the top. DPMs can arise by three mechanisms: (I) differentiation of hepatoblasts to ductal plate cells is abnormal and associated with perturbed polarization and cyst formation; (II) PDS are formed but fail to mature to bile ducts; this is associated with formation of cysts and abnormal polarity; (III) differentiation of ductal plate cells and maturation of PDS proceeds but duct expansion is perturbed; this is associated with defects in cell polarity.

Cilia were mostly absent in HNF6- and in HNF1β-deficient livers. The presence of cilia in cpk mice dismissed the possibility that reduced expression of Cys1 in HNF6- and HNF1β-deficient livers causes the near absence of cilia in the latter two mouse models. However, the lack of cilia in the absence of HNF6 or HNF1β at E17.5 may result from mispositioning of the basal body, and from the global perturbation of biliary cell polarity. Apicobasal polarity was strongly affected, as shown by the absence of OPN, abnormal location of centrioles and Golgi apparatus, and fragmented appearance of laminin. Whereas these polarity criteria were equally affected in the absence of HNF6 or HNF1β, the location of tight junctions differed in the two mouse models. Tight junctions were detected normally by ZO-1 immunostaining and electron microscopy (data not shown) in HNF6 knockout livers. In contrast, in the absence of HNF1β, ZO-1 was barely detected in cells of the parenchymal side of the biliary structures and was mislocated on the cells of the portal side. Cpk mice and human fetuses with ARPKD also showed mislocation of ZO-1. They displayed the same apical coverage of ZO-1 as on the portal side of developing ducts in HNF1β-deficient livers. Therefore, an HNF1β–cystin-1 cascade may control ZO-1 location. Earlier work has shown that HNF1β controls bile duct polarity in vitro and that this process requires laminin.28, 29 Our data now show that laminin expression in vivo is reduced in HNF1β-deficient livers, thereby uncovering a cross-talk between the biliary cells and extracellular matrix in the control of tubulogenesis.

In HNF6 knockout livers, biliary cell differentiation is abnormal. Perturbed transforming growth factor β signaling generates hybrid hepatobiliary cells,23 and this hybrid character persists at later stage of gestation as shown here at E17.5 by the coexpression of HNF4 and SOX9, and by the presence of microvilli, glycogen, well-developed endoplasmic reticulum, and large nuclei with large nucleoli (data not shown). Our work also reveals an unexpected regulation of SOX9. Indeed, SOX9 mRNA levels are reduced in Hnf6−/− livers at E15.5, whereas normal levels are restored at E17.5.3 SOX9 protein is undetectable at E15.5 (not shown); here, we found that it remains absent at E17.5, indicating that SOX9 expression is controlled by HNF6 at the transcriptional and posttranscriptional levels.

In HNF1β-deficient livers, biliary cells on the portal side appeared well-differentiated because they were SOX9+/HNF4/TβRII. They also expressed the Notch effector Hes1 (Homolog of Hairy/enhancer of Split-1) (data not shown). In contrast, biliary cells on the parenchymal side were SOX9/HNF4+/TβRII+ and expressed low levels of Hes1 (data not shown). Therefore, at E17.5, the biliary structures still displayed a PDS-like configuration. It cannot be determined if perturbed Notch or transforming growth factor β signaling, which is suggested by the PDS-like expression of Hes1 and TβRII, causes or results from the lack of PDS maturation. Still, our data identify HNF1β as a critical regulator of PDS maturation and show for the first time that deficient maturation is a cause of DPMs.

During normal biliary tubulogenesis, differentiation and polarity progress concomitantly.3 When HNF6 or HNF1β is inactivated, differentiation, polarity, and tubulogenesis are all affected. In contrast, in cpk mice and patients with ARPKD, differentiation does not seem affected whereas polarity and tubulogenesis are perturbed. Therefore, polarity and differentiation are associated or separated, depending on the DPM model studied. Interestingly, lumina still formed in all models. We also measured the expression of key planar polarity genes in the three mouse models, but found no strong evidence for a HNF6–HNF1β–cystin-1 cascade regulating planar polarity (Supporting Fig. 7).

Cyst expansion in cpk mouse kidneys depends on excess proliferation, but at E17.5 in the liver, no excess proliferation was seen: the percentage of proliferating biliary cells measured by phospho-histone H3 (PHH3) staining was 2.06% (58 PHH3+ biliary cells/2811 biliary cells; quantification on three livers) as compared to 1.8% (41 PHH3+ biliary cells/2254 biliary cells). Therefore, the mechanism of cyst formation in liver may differ from that in kidneys. This is in line with our observations that the transcriptional network regulating bile duct development is distinct from that in kidneys.

Previous work suggests that Hnf6 and Hnf1b are linked in a common gene network.6, 7 Importantly, the present work reveals that HNF1β cannot be considered as the sole effector of HNF6, and that cystin-1 is not the main effector of HNF6 or HNF1β in differentiation and morphogenesis. We speculate that the three genes may coordinately regulate biliary functions not uncovered in the present study.

Acknowledgements

We thank Dr. Laure Collard for providing information on the patient with HNF1B mutation, and Chaozhe Yang for help.

Ancillary