In this work we showed that biliary tubulogenesis starts with the formation of asymmetrical PDS lined by biliary cells on the portal side and by hepatoblasts on the parenchymal side. The PDS constitute a leading front of biliary tubes which develop along a longitudinal axis from the hilum towards the periphery and in which the hepatoblasts differentiate to biliary cells along the radial axis (). To our knowledge this mode of tubulogenesis differs from other known models which have been classified as wrapping, budding, cavitation, or cord or cell hollowing27–30
. Indeed, none of these models fit with bile duct development. We found no evidence for wrapping or budding, and unlike in the known models of cavitation and hollowing, there is no uniform circumferential signal at the onset of biliary tubulogenesis.
Figure 7 (A) Model for the formation of bile ducts in mouse liver. The transition from single layered ductal plate to formation and maturation of primitive ductal structure is controlled by a HNF6-SOX9-C/EBPα cascade and by Notch and TGFβ signaling. (more ...)
How ducts grow along the longitudinal axis remains unknown. However, our data allow us to propose a model for differentiation along the radial axis. Indeed, the expression of hepatoblast markers on the parenchymal side of the PDS suggests that the asymmetry of PDS results from the apposition of hepatoblasts to biliary cells of the ductal plate. The binding of TGFβ ligands on the parenchymal side indicates that hepatoblasts may differentiate to biliary cells under influence of TGFβ to generate radially symmetrical ducts lined exclusively by biliary cells. Our experiments with cultured hepatoblasts correlate well with this in vivo expression profile: TGFβ can stimulate biliary markers and repress hepatoblast markers.
TGFβ signaling is unlikely to be the sole determinant of initiation of biliary tubulogenesis. Indeed, TGFβ is produced by a large portion of the periportal mesenchyme, and the mesenchyme is nearly completely encircled by the single-layered ductal plate. Yet, only a limited number of ducts form at focal areas of the ductal plate. Our data show that HES1 is expressed in several cells of the ductal plate but its expression is highest on the portal side of the PDS, in areas participating to duct development. Given its known role in initiation of biliary tubulogenesis15
, high Notch signaling, via HES1, may act with TGFβ to promote differentiation along the radial axis.
SOX9 is shown here to be a specific and early marker of biliary cells. Therefore, SOX9 expression can be used for diagnostic purposes in biliary diseases. In addition, determining its expression will be useful to detect biliary contamination during programmed differentiation of stem cells to hepatocytes for cell therapy of liver diseases. The same holds true for OPN, a known marker for adult bile ducts31
which we now also found to be an embryonic biliary marker, but starting to be expressed later than SOX9 (Supplementary figure 1
In the absence of SOX9, biliary morphogenesis is initiated but delayed from e15.5 until birth. It eventually produces functional ducts. SOX9 controls differentiation along the radial axis, since an excessive proportion of asymmetrical PDS persist near the hilum until birth. Serial sections along the hilum-periphery axis did not show evidence for a role of SOX9 in determining the longitudinal axis of duct development. Indeed, PDS were found in SOX9-deficient livers all along the hilum-periphery axis (not shown). In the absence of SOX9, HES1 expression was low or absent on the parenchymal side of the PDS at e18.5. Stimulation of HES1 by SOX9 was found by others to occur in the pancreas20
. These data suggest that SOX9 may control HES1 in liver in a cell-specific way. Moreover, since expression of HES1 is perturbed on the parenchymal side of the PDS and not on their portal side in SOX9-deficient livers, the control exerted by SOX9 in liver is also cell type-specific. Our data also suggest that the low level or absence of HES1 on the parenchymal side of e18.5 PDS contributes to the delay in bile duct maturation of SOX9-deficient livers. However, we cannot eliminate the possibility that this expression of HES1 simply reflects immaturity, without contributing to it.
Our data establish a link between SOX9 and TGFβ signaling. Indeed, TβRII was repressed on the portal side of wild-type asymmetrical PDS, but not in SOX9-deficient PDS. How this impacts on the maturation of PDS cannot yet be determined with certainty. However the data suggest that during PDS formation, the cells on the portal side of the PDS instruct hepatoblasts to constitute the parenchymal side via cell-cell contacts. Such cell-cell contacts would together with TGFβ signaling promote biliary differentiation of hepatoblasts along the radial axis of the developing tubes. We speculate that persistent TGFβ signaling on the portal cells of the PDS in the absence of SOX9 perturbs the ability of these cells to induce biliary differentiation of cells on the parenchymal side.
Along the same lines, the persistent expression of C/EBPα in SOX9-deficient PDS may contribute to delayed maturation of the ducts. Indeed hepatoblasts devoid of C/EBPα preferentially differentiate to cholangiocytes rather than to hepatocytes11
, suggesting that the opposite, namely persistence of C/EBPα, may inhibit biliary development.
Other potential targets of SOX9 are genes coding for extracellular matrix (ECM) proteins. The delay in extracellular matrix (ECM) deposition around the PDS in SOX9 knockouts, the potential role of ECM in biliary morphogenesis32
, and SOX9’s ability to stimulate ECM gene transcription in liver33
, suggest that control of ECM production by SOX9 may be important for bile duct development. An involvement of SOX9 in repressing hepatoblast markers is unlikely since overexpression of SOX9 in cultured BMEL cells did not inhibit HNF4 expression (not shown). The transient nature of the defect in SOX9 knockouts may result from compensation by other SOX factors. Indeed, our preliminary data indicate that several SOX factors are expressed in developing liver (not shown).
Finally, humans with SOX9 deficiency suffer from campomelic dysplasia (OMIM #114290). Hepatic defects have not been reported, but considering the transient nature of the biliary anomalies, these might have been overlooked in humans. In any case, SOX9 dysfunction deserves to be investigated in cases of neonatal biliary anomalies of unknown origin.