From a common progenitor to distinct liver epithelial phenotypes

Abstract

The vertebrate liver presents a fascinating case study for how cell form is optimized for function. To execute its duties the liver assembles two distinct lumen-forming epithelial phenotypes: 1) cords with a branched, capillary-like luminal network formed between hepatocytes (bile canaliculi); and 2) tubular ducts formed by biliary epithelial cells arranged around a central cavity and connected to the bile canaliculi. How these remarkably different epithelial polarity phenotypes are generated and joined into a contiguous luminal network are major unresolved questions. Recent studies have characterized the divergence of the two epithelial lineages from common progenitors, described the coordination of bile canaliculi formation with bile duct branching during biliary tree morphogenesis and implicated RhoA-dependent E-cadherin adhesion in the decision to polarize with hepatocytic or biliary phenotype.

Main Text

The liver’s two epithelial cell types - hepatocytes and biliary cells (also called cholangiocytes) - share all epithelial hallmarks, which include the establishment of distinct apical and basolateral surface domains that are separated by tight junctions. Yet both cell types differ drastically in how they organize these domains to suit their different functions: Hepatocytes arrange in cords that weave through a system of fenestrated blood vessels (the sinusoids). This allows hepatocytes to maximize their basolateral surface area used for molecular exchange with the portal blood supply and hence freely secrete serum proteins while absorbing solutes and xenobiotics. Bile duct cells, by contrast, organize into a closed monolayered tubule that transports bile. On the cellular level, bile duct cells organize in monolayers where each epithelial cell contributes its single apical domain to a central lumen, i.e. they are monopolar (Fig. 1, “Ductal”). Hepatocytes, by contrast, are multipolar because they simultaneously form luminal surfaces with multiple neighbors thereby generating bile canaliculi, a branched capillary luminal network (Fig. 1, “Hepatocytic”). While bile duct cell polarity is similar to that of most epithelial tissues, hepatocyte polarity is unique. In the following chapters I will discuss evolving concepts that explain how the different epithelial phenotypes might be generated and how they join to form contiguous luminal tracts.

Fig. 1. The organization of polarized surface domains in hepatocytes and bile duct cells.

Hepatocytes are multipolar; they form multiple luminal surfaces that interrupt their cell-cell contacting domains, and they have two basal domains. Bile duct cells are monopolar; they establish one luminal and one basal surface opposite from each other and perpendicular to their cell-cell contacting domains.

The polarization sequences in the embryonic and adult liver

The hepatic progeny originates from a monolayered epithelial tube, the foregut (Fig. 2A, E8.5). Hepatic specified foregut cells proliferate and invade the surrounding mesenchyme as nonpolar liver epithelial progenitors called hepatoblasts (HBs, Fig. 2A, E9.5), which give rise to both hepatocytes and biliary cells [1,2]. The first polarized structure to emerge from the HB mass in mammals is a monolayer of adherent, strongly Ecadherin-positive cells around the portal veins, called the ductal plate at around E15.5 in the mouse. Induced by the periportal mesenchyme, these cells acquire biliary characteristics [3]. Parts of the ductal plate develop into bile ducts, while others have been proposed to become hepatic stem cells [4,5]. HBs outside the vicinity of portal veins develop into hepatocytes. Bile canaluculi become apparent only in late gestation (E18), and continue to elongate postnatally.

Fig. 2. Models for the development (A) and maintenance (B) of the ductal and hepatocytic epithelial phenotypes.

A) hepatic-specified foregut epithelial cells lose E-cadherin expression and their basement membrane (BM) as they delaminate and proliferate as non-polar hepatoblasts (HBs), which, by default, develop into hepatocyte cords. Hepatocytes maintain low E-cadherin levels and establish no BM. Between E11.5-11.4 bile duct cells sharply diverge from this default differentiation program, re-acquiring high levels of E-cadherin and a BM as they first form a ductal plate around the portal vein (pV) and subsequently bile ducts. B) The intrahepatic bile duct branches into smaller ductules, which form a continuous lumen with the hepatocytic bile canaliculi. The biliary cells at the junction (i.e the Canal of Hering) are thought to represent the adult stem cell niche.

Recently, the morphological description of ductal and hepatocyte specification has been underpinned by RNA sequencing based transcriptome analysis of individual HBs taken at different points between HB emergence and late gestation [6**]. Guided by established markers for bipotential and lineage committed HBs, the authors performed RNA sequencing on sorted cells yielding two principle components that corresponded to the biliary and hepatocytic lineages. Their “pseudotime” transcriptome analysis of these components led them to several profound conclusions: 1) Hepatocyte specification begins shortly upon emergence of the liver bud from the foregut and HBs gradually move toward the hepatocytic fate in a synchronous manner. 2) HB-to-biliary specification represents a sharp branching from the default hepatocytic differentiation program during a limited developmental time window (between E11.5 and E14.5). This branching is associated with the silencing of HB and hepatocyte markers and the induction of biliary-specific markers. Chief among induced cellular processes/pathways were those related to cell adhesion, ERK1/2 signaling and tube morphogenesis.

Turnover of mature liver cells is slow, in the order of several months; however, upon hepatectomy differentiated hepatocytes re-enter the cell cycle to replace the lost liver mass. These observations prompted the view that liver homeostasis is maintained by proliferation of mature epithelial cells. Whether the adult liver, like other organs, also possesses bona fide stem cells that contribute to normal cell turnover or injury response is a still ongoing debate. While genetic lineage tracing has yielded conflicting conclusions (see [7]), isolated putative adult liver stem cell populations from normal liver are biliary in origin and can give rise to bipotential adult HBs [4] [8] [9*].

Taken together, current evidence thus indicates that a common “polarization sequence” might operate in both the embryonic and in the adult liver (Fig. 2): it starts with columnar epithelial cells (hepatic specified foregut/biliary hepatic stem cells) that become nonpolar cells (embryonic and adult HBs), and that repolarize with hepatocytic polarity by default unless induced to adopt ductal polarity. Remarkably, the hepatocytic model cell line WIF-B, one of the few polarized hepatocytic cell lines, spontaneously recapitulates this polarization sequence [10].

What causes branching into hepatocytic versus ductal polarity phenotypes?

No precise mechanisms have been elucidated to date, but current evidence points to the importance of two cellular processes in the polarity decision: cell-matrix- and cell-cell adhesion.

Bile ducts, like all other monolayered epithelial tissues, are surrounded by a basal lamina composed of collagen IV and laminin. By contrast, the space of Disse that separates hepatocytes from endothelial cells is devoid of a basement membrane [11] as it lacks laminin and the laminin-collagen IV crosslinker nidogen [12]. During ductal plate formation laminin (specifically, a laminin with an a1 chain) is initially provided by the portal mesenchyme, which also triggers laminin expression in the future biliary cells (specifically a5 chain laminin), most likely via TGFβ signaling [13]. Tanimizu et al. showed that the ability of HBs to form monolayered cysts in vitro depended on the presence of laminin in the 3D culture matrix [14] and on the activity of β1-integrin, a constituent of laminin and collagen receptors [13]. Biliary atresia, an early childhood liver disease characterized by bile duct malformation, has recently been associated with decreased levels of β1-integrin, laminin b1 and nidogen around diseased bile ducts [15]. This is consistent with an essential role of β1-integrin/laminin signaling in bile duct morphogenesis, which might be a general requirement for the establishment of monopolar tubular epithelial structures [16]. Our group established an experimental system that mimics the bipotency of HB polarization to address whether extracellular matrix (ECM)-signaling differences regulate the polarity phenotypes. In our model inducible overexpression of the AMPK-related kinase Par1b causes a switch from monopolar to hepatocyte-like polarization in kidney-derived MDCK cells. The polarity change is accompanied by reduced basement membrane deposition and focal adhesion formation and can be reversed by plating the cells on collagen IV substrates [17,18]. ECM-mediated integrin activation frequently leads to RhoA activation [19,20]. Indeed we measured lower RhoA activity in hepatocytic compared to monopolar MDCK-Par1b cells. Remarkably, RhoA depletion was sufficient to induce heptocytic polarity in MDCK cells. Conversely, pharmacological Rho activation in the hepatocytic cell line WIF-B promoted their monopolar organization [18,21*]. These findings point to RhoA signaling downstream of cell adhesion signaling as a putative key regulator of the polarity phenotypes.

E-cadherin is present at the cell surface of both biliary cells and periportal hepatocytes (hepatocytes outside the periportal zone express N-cadherin instead). Nonetheless, hepatocytes initiate lumen formation at the very cell-cell contacting domains that in tubule-forming epithelia are reserved for the establishment of E-cadherin-based adherens junctions. This makes it likely that adherens junction formation and the associated signaling events differ between hepatocytes and biliary cells. Immuno labeling consistently shows a significantly stronger E-cadherin signal at biliary compared to hepatocyte cell-cell contacts (e.g. [3], human protein atlas https://www.proteinatlas.org/ENSG00000039068-CDH1/tissue/liver). This might be due to differential gene expression as suggested by recent RNA sequencing data [22] or it could result from differences in the extent of adherens junction formation; the latter promotes E-cadherin stability on the protein level [23]. E-cadherin knockout during HB differentiation [24] caused defects in bile duct but not hepatocyte morphogenesis, which might indicate that E-cadherin-mediated adhesion is more important for biliary than for hepatocyte tissue organization. Cell culture studies support this notion and suggest furthermore that the level of E-cadherin-adhesion defines the epithelial polarity phenotype: Thus, in the hepatocytic cell line HepG2, E-cadherin was dispensable for polarization [25]. Increased E-cadherin levels, on the other hand, led to the formation of a horizontal tight junction belt characteristic of monopolar cells [26]. In the MDCK model, substitution of E-cadherin for an adhesion-deficient mutant promoted lumen establishment at cell-cell contact sites as in hepatocytes, at least transiently [27]. To assess E-cadherin adhesion strength Borghi et al. developed a FRET-based force sensor, which measures the tension that arises in E-cadherin molecules that are linked to the circumferential actin cytoskeleton via their cytoplasmic domain and are also engaged in homotypic interactions with neighboring cells [28]. We utilized this sensor to compare E-cadherin adhesion in hepatocytic polarized WIF-B cells to WIF-B cells that adopt monopolar organization upon Rho-activation. We also compared hepatocyitc polarized MDCK-Par1b cells to their monopolar control counterparts. In both experimental systems hepatocytic polarity was associated with lower E-cadherin tension [21*].

Since it has been proposed that traction forces generated by cell-matrix adhesion are balanced by cell-cell adhesion forces [29] it is conceivable that the differences in cell-ECM interactions between monopolar cells and hepatocytes contribute to the observed differences in their E-cadherin tension. Biochemically, the coupling of integrin- to E-cadherin-adhesion in single cell assays has been linked to Src and RhoA/Rho-kinase signaling [30]; the latter fits well with the role of RhoA activity in the polarity decision.

Taken together, the evidence suggests that RhoA-dependent E-cadherin-mediated cell-cell adhesion promotes biliary over hepatocytic polarization. Given that cell-cell adhesion requires close cell-cell contact, this hypothesis seems at odds with the well-established observation that culture conditions, such as spheroid cultures, which maximize cell-cell contacts, improve hepatocytic polarization [31]. Wang et al. modulated the compaction of primary hepatocyte spheroid cultures by microfluid pressure and showed that an increase in cell-cell contact area accelerated their polarization [32]. Interestingly, E-cadherin immunofluorescence labeling did not differ between compact and noncompacted cells. This suggests the hypothesis that compaction facilitates E-cadherin independent cell-cell adhesion signaling mechanisms that promote luminal surface formation at cell-cell contact sites. Candidates for this role are nectins, ubiquitously expressed adhesion molecules of the IgG-super family that precede E-cadherin in junction formation at new cell-cell contact sites in MDCK cells [33] but have yet to be studied in the liver.

How do biliary cells and hepatocytes arrange to form interconnected luminal tracts?

The hepatic ductal network is aptly called the biliary tree with the common hepatic bile duct as trunk and a hierarchical branched network arising from the two main hepatic ducts of the left and right liver lobes. Its smallest ductules connect to the hepatocyte bile canaliculi (Fig. 2B). How the two epithelial tissues link their respective luminal networks into a contiguous entity is only beginning to emerge. Expanding on earlier studies, which visualized the biliary tree by 3D reconstruction of serial liver sections [34], [35], Tanimizu et al combined thick section immunofluorescence on cleared tissue with carbon ink injections into the common bile duct to investigate how biliary cells organize into the biliary tree over the course of liver development [36**]. Antibodies to ductal markers identified cells as biliary; ink staining indicated their incorporation into bile ducts that were outlined by the dye. The authors found that most mature biliary cells were incorporated into bile ducts at all developmental stages; by contrast, the branching of smaller ductules from the larger ducts happened only late in gestation. Notably, bile duct branching coincided with a spurt in bile canaliculi formation and with the first appearance of bile in the intestine at E18 in the mouse. Pharmacological inhibition of the hepatocyte bile acid transporter Mrp2, which pumps bile acids into the bile canaliculi, prevented bile duct branching. Arias’ group had previously shown that the main bile acid taurocholate dramatically stimulates bile canaliculi formation in cultured primary hepatocytes [37]. The bile acids that reach bile ducts from these newly formed bile canaliculi might constitute signals for ductal branching into narrower ductules, which then connect to the bile canaliculi. Taurocholate signaling for bile canaliculi formation involves activation of the kinase LKB1 and its effector AMPK [37], a signaling cascade that has also been linked to branching morphogenesis in the lung [38]. It will be interesting to decipher whether bile acids elicit different morphogenetic effects on the two epithelial cell types via similar or distinct signaling pathways.

Summary and Outlook

(1) The isolation of embryonic and adult hepatic stem cells that are biliary in nature and give rise to bipotential hepatoblasts and (2) the discovery of a default hepatoblast differentiation program in which the ductal lineage escapes hepatocytic fate during a restricted time window, are recent developments towards the characterization of liver stem and progenitor cells. They will inform the development of physiologically relevant in vitro systems that recapitulate the branching of the liver polarity phenotypes to identify regulatory mechanisms involved. Although hepatic progenitors have been driven into one or the other lineage and polarity phenotype [14,39,40], an experimental system that steers a clonal hepatic progenitor into a ductal or hepatocytic polarization program with equal efficiently has not been reported to date. (3) Cell culture studies suggest that lack of a basal lamina, unique to hepatocytes, also contributes to their unique polarity. Whether weak E-cadherin-mediated cell-cell adhesion, which accompanies low RhoA activity in hepatocytic polarized cells, is due to crosstalk or mechanical coupling between cell-matrix and cell-cell adhesion signaling remains to be determined. How cell-cell contacts promote lumen formation at cell-cell contact sites in hepatocytes and whether/how E-cadherin-adhesion antagonizes hepatocytic polarization are key questions to be addressed. (4) Bile acids not only emerged as potent initiators of bile canaliculi formation but they also link bile canaliculi with bile duct development. Thick-section microscopy of cleared tissue, and intravital live cell imaging, which has recently been adapted to the liver [41*], are suitable techniques to characterize this process and its signaling mechanisms in detail.