ABSTRACT
Liver development is controlled by key signals and transcription factors that drive cell proliferation, migration, differentiation and functional maturation. In the adult liver, cell maturity can be perturbed by genetic and environmental factors that disrupt hepatic identity and function. Developmental signals and fetal genetic programmes are often dysregulated or reactivated, leading to dedifferentiation and disease. Here, we highlight signalling pathways and transcriptional regulators that drive liver cell development and primary liver cancers. We also discuss emerging models derived from pluripotent stem cells, 3D organoids and bioengineering for improved studies of signalling pathways in liver cancer and regenerative medicine.
Introduction
The vertebrate liver performs metabolic, endocrine and exocrine functions that are essential for maintaining physiological homeostasis, including bile secretion, cholesterol synthesis, drug detoxification, glucose storage, lipid turnover, metabolism and secretion of hormones and plasma proteins (Heslop and Duncan, 2019; Huppert and Iwafuchi-Doi, 2019; Si-Tayeb et al., 2010a). These operations are performed by hepatocytes, the principal functional cells of the liver, which constitute 80% of total parenchymal liver cell volume (Blouin et al., 1977; Gordillo et al., 2015; Si-Tayeb et al., 2010a). Among their functions, mature hepatocytes produce and secrete bile into the hepatic biliary ducts, where it is either released into the duodenum for digestion or transported to the gall bladder for storage (Huppert and Iwafuchi-Doi, 2019). By comparison, cholangiocytes, the biliary epithelial cells that line the liver bile ducts, only constitute 3% of total parenchymal liver cell volume (Si-Tayeb et al., 2010a). Non-parenchymal liver cells, such as hepatic sinusoidal endothelial cells, hepatic stellate cells and resident liver macrophages known as Kupffer cells, together comprise the remaining 17% of total liver cell volume (Gordillo et al., 2015; Si-Tayeb et al., 2010a).
There are four major lobes in the mature mouse liver that consist of smaller units termed liver lobules (Gordillo et al., 2015). Lobules are composed of sinusoids (liver capillaries), with hepatocytes arranged into cords that spiral out from a central vein and terminate at the portal triads, which includes the portal vein, hepatic artery and bile ducts (Fig. 1). The microscopic organization of the liver facilitates its functions, primarily through cellular zonation and the structure of the sinusoidal capillary. The liver lobule is divided into three major zones based on the spatial distribution of hepatocytes and their relationship to the portal or central vein, as well as their distinct patterns of gene expression. Hepatocyte zonation allows for division of labour among hepatocytes based on their location along the liver lobule and exposure to oxygen, nutrients and hormones (Halpern et al., 2017). This patterning is thought to be established in part by WNT (Planas-Paz et al., 2016) and glucagon signalling (Cheng et al., 2018), and has been recently reviewed in detail (Ben-Moshe and Itzkovitz, 2019). The sinusoidal capillary is a fenestrated capillary lined by specialized hepatic sinusoidal endothelial cells (HSECs), which form a permeable barrier. This physically separates blood from hepatocytes and hepatic stellate cells, but allows for the exchange of nutrients between the blood and these cell types. HSECs further regulate vascular tone, control sinusoidal capillary pressure and maintain hepatic stellate cell quiescence. Stellate cells function as sinusoidal pericytes, although their role in liver homeostasis remains elusive. These non-parenchymal liver cells have also been recently reviewed in detail (Shetty et al., 2018; Tsuchida and Friedman, 2017; Yin et al., 2013).
Cellular organization of the liver lobule. The liver is organized into microscopic subunits termed liver lobules that are composed of sinusoids with hepatocytes arranged into cords. Hepatocytes spiral out from a central vein and terminate at the portal triads, which include the portal vein, hepatic artery and bile ducts. Bile secreted from hepatocytes is drained into the bile canaliculi and flows to the intestine via the hepatic bile ducts. The sinusoidal capillary is lined by specialized hepatic sinusoidal endothelial cells that form a fenestrated and permeable barrier. This physically separates blood from hepatocytes and hepatic stellate cells but allows for exchange of nutrients. Stellate cells function as sinusoidal pericytes in the perivascular zone, known as the space of Disse. Kupffer cells are resident liver macrophages located in the sinusoids and respond to pathogens or inflammatory signals.
Cellular organization of the liver lobule. The liver is organized into microscopic subunits termed liver lobules that are composed of sinusoids with hepatocytes arranged into cords. Hepatocytes spiral out from a central vein and terminate at the portal triads, which include the portal vein, hepatic artery and bile ducts. Bile secreted from hepatocytes is drained into the bile canaliculi and flows to the intestine via the hepatic bile ducts. The sinusoidal capillary is lined by specialized hepatic sinusoidal endothelial cells that form a fenestrated and permeable barrier. This physically separates blood from hepatocytes and hepatic stellate cells but allows for exchange of nutrients. Stellate cells function as sinusoidal pericytes in the perivascular zone, known as the space of Disse. Kupffer cells are resident liver macrophages located in the sinusoids and respond to pathogens or inflammatory signals.
Liver organogenesis is tightly controlled by extrinsic signalling pathways and cell-autonomous transcription factors that converge to regulate the specification of the primitive liver bud, followed by cell proliferation, differentiation and maturation of hepatocytes and cholangiocytes. Conversely, dysregulation of the mature hepatic cell signature by genetic, epigenetic or environmental factors can initiate dedifferentiation (i.e. reactivation of developmental programmes), aberrant cell proliferation and carcinogenesis through genetic mutations and chromatin remodelling (Aiello and Stanger, 2016). In this Review, we examine the role of signalling pathways both in liver development and in primary adult liver carcinomas, including hepatocellular carcinoma (HCC) and cholangiocarcinoma (CCA). We highlight several prominent signalling pathways in both liver development and cancer, including fibroblast growth factor (FGF), bone morphogenetic protein (BMP), transforming growth factor β (TGFβ), WNT/β-catenin, Hippo and Notch signalling pathways. Finally, we identify in vitro differentiation models of pluripotent stem cells, 3D organoids and biofabricated livers as attractive preclinical tools for improving our understanding of hepatic development and disease.
Liver specification and hepatogenesis
Development of the liver involves the coordinated assembly of cells derived primarily from the mesoderm and endoderm (Fig. 2). Liver parenchymal cells are derived from posterior foregut endoderm (Tremblay and Zaret, 2005), from which pancreas progenitors are also derived (Deutsch et al., 2001) (Fig. 3). Starting at mouse embryonic day (E) 8.0 (Zhao and Duncan, 2005), inductive paracrine signals from the cardiac mesoderm and the septum transversum mesenchyme (STM) discriminate liver and pancreas domains. BMP, FGF and WNT signalling pathways play a large role in this process (Ang et al., 2018; Macchi and Sadler, 2020; Palaria et al., 2018; Wandzioch and Zaret, 2009). For example, treatment of in vitro foregut endoderm explant cultures with exogenous FGF1 and FGF2 is sufficient to induce hepatic cell fate (marked by albumin or Alb expression) in the absence of cardiac mesoderm (Deutsch et al., 2001), though FGF inhibition of whole-embryo culture only affects specification of the anterior liver bud (Wang et al., 2015). BMP signalling from the STM is also required for the induction of hepatic fate. In explant in vitro cultures containing foregut endoderm, cardiac mesoderm and STM, treatment with a BMP antagonist prevents Alb induction (Rossi et al., 2001). Additionally, treatment of embryo explant cultures with exogenous BMP4 enhances hepatic Alb expression (Wandzioch and Zaret, 2009), whereas BMP inhibition specifically attenuates posterior liver bud specification and results in an anterior shift of the hepato-pancreatic boundary (Palaria et al., 2018). In contrast to FGF and BMP signalling, early inhibition of WNT/β-catenin signalling promotes liver specification (Fig. 3). This is supported by evidence that β-catenin degradation via GSK3β-mediated phosphorylation in the posterior foregut endoderm is sufficient for induction of early liver markers through activation of the transcription factor Hhex (McLin et al., 2007). Similarly, WNT activation in Xenopus endoderm drives pancreas gene expression but represses alb and hepatic cell fate (Rodriguez-Seguel et al., 2013).
The liver is formed by the coordinated assembly of cells derived from the endoderm and mesoderm. Schematic of the relationship between cell types in the developing liver. Hepatoblasts are the parenchymal progenitor cells of the liver that derive from the posterior foregut endoderm. Paracrine signalling from the neighbouring septum transversum mesenchyme (STM) and endothelium instructs hepatoblasts to proliferate, migrate and differentiate into hepatocytes and cholangiocytes. At E8.5, the ventral region of the posterior foregut thickens and forms the liver bud, containing a single epithelial layer of hepatoblasts surrounded by the STM, endothelial cells and the basal lamina. Around E9.0, columnar hepatoblasts become pseudostratified and degradation of the basal lamina is initiated. Starting at E9.5, hepatoblasts begin proliferation, delamination and migration into the STM. At E10.5, significant hepatoblast expansion occurs, sinusoids are formed and the liver becomes a major site of haematopoiesis.
The liver is formed by the coordinated assembly of cells derived from the endoderm and mesoderm. Schematic of the relationship between cell types in the developing liver. Hepatoblasts are the parenchymal progenitor cells of the liver that derive from the posterior foregut endoderm. Paracrine signalling from the neighbouring septum transversum mesenchyme (STM) and endothelium instructs hepatoblasts to proliferate, migrate and differentiate into hepatocytes and cholangiocytes. At E8.5, the ventral region of the posterior foregut thickens and forms the liver bud, containing a single epithelial layer of hepatoblasts surrounded by the STM, endothelial cells and the basal lamina. Around E9.0, columnar hepatoblasts become pseudostratified and degradation of the basal lamina is initiated. Starting at E9.5, hepatoblasts begin proliferation, delamination and migration into the STM. At E10.5, significant hepatoblast expansion occurs, sinusoids are formed and the liver becomes a major site of haematopoiesis.
Key signalling pathways and transcription factors directing liver development. The liver parenchymal progenitor cells, known as hepatoblasts, are specified from the posterior foregut endoderm from which pancreas progenitors are also derived. Around mouse E8.0, inhibition of WNT/β-catenin and induction of BMP and FGF signals from the cardiac mesoderm and the septum transversum mesenchyme discriminate AFP+ liver and PDX1+ pancreas domains. The liver bud forms at E8.5 and BMP, FGF, TGFβ and WNT signalling from the surrounding mesenchyme and endothelium drive hepatoblast proliferation and migration. Around E11.5-E13.5, mesenchymal Notch and TGFβ signals promote periportal hepatoblast differentiation to KRT19+ cholangiocytes and inhibit ALB+ hepatocyte differentiation. WNT/β-catenin signalling plays a key role in the differentiation and expansion of hepatocytes, as well as controlling mature hepatocyte zonation. In contrast, cholangiocyte maturity is actively maintained through Notch, TGFβ and Hippo signalling. Lineage-specific transcription factors (e.g. Hnf4a, Pdx1) and key markers (e.g. Alb, Afp) associated with each cell type are indicated at each stage of development.
Key signalling pathways and transcription factors directing liver development. The liver parenchymal progenitor cells, known as hepatoblasts, are specified from the posterior foregut endoderm from which pancreas progenitors are also derived. Around mouse E8.0, inhibition of WNT/β-catenin and induction of BMP and FGF signals from the cardiac mesoderm and the septum transversum mesenchyme discriminate AFP+ liver and PDX1+ pancreas domains. The liver bud forms at E8.5 and BMP, FGF, TGFβ and WNT signalling from the surrounding mesenchyme and endothelium drive hepatoblast proliferation and migration. Around E11.5-E13.5, mesenchymal Notch and TGFβ signals promote periportal hepatoblast differentiation to KRT19+ cholangiocytes and inhibit ALB+ hepatocyte differentiation. WNT/β-catenin signalling plays a key role in the differentiation and expansion of hepatocytes, as well as controlling mature hepatocyte zonation. In contrast, cholangiocyte maturity is actively maintained through Notch, TGFβ and Hippo signalling. Lineage-specific transcription factors (e.g. Hnf4a, Pdx1) and key markers (e.g. Alb, Afp) associated with each cell type are indicated at each stage of development.
Liver bud formation in mice begins at E8.5 with the exponential expansion of the hepatoblast progenitor pool and invasion of these cells into the surrounding mesenchyme (Fig. 2). Hepatoblast proliferation is largely regulated by BMP, FGF, TGFβ and WNT signalling from the STM and surrounding endothelium (reviewed by Zorn, 2008) (Fig. 3). Genetic deletion of either the STM transcription factor Gata4 or its downstream target Bmp4 prevents liver bud expansion (Rossi et al., 2001; Watt et al., 2007). In addition, conditional deletion of the WNT signalling gene Ctnnb1 from hepatoblasts with Foxa3-Cre significantly decreases hepatoblast proliferation and results in overall liver hypoplasia (Tan et al., 2008). At around E9.5, the hepatic epithelium becomes pseudostratified and the extracellular matrix is degraded on the basal surface (Si-Tayeb et al., 2010a; Zhao and Duncan, 2005) (Fig. 2). Degradation facilitates delamination from the liver bud into the STM and hepatoblast migration as epithelial cords (Zhao and Duncan, 2005; Zorn, 2008). The transcription factors HHEX and PROX1 play key roles in orchestrating hepatoblast delamination and migration, as well as the transition from a columnar to pseudostratified epithelial cell morphology (Bort et al., 2006; Sosa-Pineda et al., 2000). The onset of liver vascular development and haematopoiesis occurs in parallel at this stage, with signals such as vascular endothelial growth factor (VEGF) also playing a key role (Matsumoto et al., 2001; Si-Tayeb et al., 2010a). For instance, Vegfr2 (Kdr) deletion in mouse results in the absence of liver endothelial cells and prevents liver bud proliferation and invasion into the STM (Matsumoto et al., 2001), demonstrating an essential role for the endothelium in hepatoblast proliferation and outgrowth. How the parenchymal and non-parenchymal progenitors organize to form the primitive sinusoidal capillaries is not well understood; however, a recent single-cell transcriptomic analysis suggests that juxtacrine signalling may play a role (Lotto et al., 2020). Specifically, Notch and Ephrin signalling may facilitate direct cell-cell interactions between sinusoidal populations. Ephrin signalling is of particular interest as it has been shown to direct collective migration of hepatoblasts in zebrafish (Zhang et al., 2016).
Liver specification remains a complex stage of development, as hepatic induction is not directed by a single master regulator but is instead controlled by the concerted action of multiple transcription factors. In the endoderm, FOXA1/2/3, GATA4 and GATA6 act as pioneer factors, creating accessible chromatin and coordinating sequential binding with HNF4α during hepatic induction (Heslop et al., 2021; Horisawa et al., 2020). In support of this, combined deletion of Foxa1 and Foxa2 from foregut endoderm prevents liver bud formation (Lee et al., 2005), suggesting that FOXA factors are essential for liver specification. Similarly, deletion of GATA6 from pluripotent stem cells perturbs binding of FOXA2 and hepatic specification from definitive endoderm (Heslop et al., 2021). Although it is clear that FOXA and GATA factors are essential regulators of hepatic lineage induction, neither are sufficient on their own and additional transcription factors are required. Downstream of FOXA and GATA, HNF4α is expressed early in the hepatic endoderm and is required for hepatoblast differentiation (DeLaForest et al., 2011; Hayhurst et al., 2001). However, the direct role of HNF4α in hepatic induction is unclear, as HNF4α is dispensable for mouse liver specification (Li et al., 2000) but has an earlier role in hepatic progenitor formation from human pluripotent stem cells (DeLaForest et al., 2011). The importance of HNF4α and FOXA transcription factors in liver specification is further highlighted in studies of inducible hepatocyte-like cells (iHeps) where combined overexpression of HNF4α and FOXA1/2/3 in fibroblasts is sufficient to induce the hepatocyte programme (Sekiya and Suzuki, 2011). Taken together, additional research is needed to elucidate the transcriptional network governing hepatic cell fate induction.
Hepatoblast differentiation
Mouse hepatoblasts differentiate into either hepatocytes or cholangiocytes by E13.5 (Gordillo et al., 2015), although recent single-cell RNA-sequencing of the primitive liver revealed that differentiation may start as early as E11.5 (Yang et al., 2017) (Fig. 3). It has also been proposed that the hepatocyte lineage progressively differentiates over time as a ‘default’ fate, whereas cholangiocyte differentiation requires active divergence (Wang et al., 2020). To date, it is unclear whether an intermediate duct or hepatocyte progenitor exists prior to complete lineage differentiation. Characterization of duct and hepatocyte progenitors requires further investigation, but has been limited so far by the absence of adequate markers to distinguish hepatoblasts from early duct cells or fetal hepatocytes. Compounding this, the master regulators that drive specification of hepatocytes or cholangiocytes are also unknown. Together, the transcription factors FOXA2, HNF4α and CEBPA initiate and maintain the hepatocyte differentiation programme, including expression of the hepatic plasma proteins Afp (α-fetoprotein) and Alb (Alder et al., 2014; Karagianni et al., 2020; Kyrmizi et al., 2006; Li et al., 2000; Parviz et al., 2003; Thakur et al., 2019). The transcription factor TBX3 is also required for differentiation of hepatoblasts to hepatocytes, while suppressing duct cell fate (Lüdtke et al., 2009). Cholangiocyte differentiation is marked by increased KRT19 (CK19) and is driven by the transcription factors HNF1β, SOX4 and SOX9 (Poncy et al., 2015). It is important to note that, although FOXA2, HNF4α and CEBPA are essential for hepatocyte differentiation, neither factor is sufficient on its own. Similarly, liver-specific deletion (Alfp-Cre) of the earliest duct cell marker, Sox9, has no effect on bile duct formation (Antoniou et al., 2009); however, combined deletion of Sox9 and Sox4 impairs downstream Hnf1b activation and prevents cholangiocyte differentiation (Poncy et al., 2015).
Determining the specific signals that control hepatoblast differentiation into hepatocytes or cholangiocytes is an area of active investigation. Hepatoblast cell fate decisions are regulated by signalling gradients that are dependent on the location in the liver parenchyma, whereby only hepatoblasts in contact with the portal vein, in a region known as the ductal plate, differentiate into cholangiocytes (Gordillo et al., 2015; Zorn, 2008). The ductal plate consists of a single layer of SOX9+ cells and gives rise to cholangiocytes and periportal hepatocytes (Carpentier et al., 2011). With respect to lineage decisions in general, it is common for the same signal to both promote one cell fate and inhibit the alternative cell fate. Together, Notch and TGFβ signals from the periportal mesenchyme promote cholangiocyte differentiation and inhibit hepatocyte differentiation (Chen et al., 2018) (Fig. 3). For example, Notch signalling directly activates the duct transcription factor Sox9 (Zong et al., 2009), expression of which is described as the first evidence of cholangiocyte differentiation (Lemaigre, 2020). Several transgenic mouse studies have demonstrated that Notch signalling activation in hepatoblasts or hepatocytes results in ectopic SOX9+/HNF1β+ cells and ultimately drives biliary cell fate commitment (Tchorz et al., 2009; Zong et al., 2009). Additionally, deletion of either the Notch ligand Jag1 in periportal mesenchyme or the Notch transcription factor Rbpj in Foxa3-derived hepatoblasts reduces bile duct formation (Lemaigre, 2020; Zong et al., 2009). Early work from Frédéric Lemaigre's group demonstrated the essential role of TGFβ signalling in cholangiocyte differentiation. Treatment of ex vivo mouse E12.5 liver explants with TGFβ induced biliary marker expression, and in vivo inhibition of TGFβ signalling at E10.5 reduced liver duct cell differentiation (Clotman et al., 2005). Further analysis of single and double Onecut1 and Onecut2 mutant livers revealed that the transcription factors ONECUT1 and ONECUT2 together suppress TGFβ signalling in hepatocytes. Treatment of a mouse hepatoblast cell line with TGFβ ligands activates Sox9 but represses Alb and Hnf4a, suggesting that TGFβ signalling promotes cholangiocyte differentiation at the expense of hepatocyte differentiation (Antoniou et al., 2009). Furthermore, TGFβ signalling to periportal hepatoblasts represses the hepatocyte transcription factor Cebpa through SOX9 and in turn induces expression of the duct cell factor Cebpb (Lemaigre, 2020).
The Hippo signalling pathway has been proposed as a duct cell driver (Lemaigre, 2020), but direct evidence for its role in cholangiocyte specification, rather than maintenance, is completely absent from the literature. Homozygous deletion of the Hippo co-activator Yap1 in mice is embryonic lethal at E8.5 prior to hepatic specification (Morin-Kensicki et al., 2006). Studies of liver-specific Hippo inactivation have mostly used Alb-Cre driver models, in which efficient deletion occurs at E18.5, long after hepatoblast differentiation (Nguyen et al., 2015; Patel et al., 2017). Both in embryonic and in inducible models in which Yap1 is deleted from ALB+ hepatocytes results suggest that hepatocyte differentiation is relatively unaffected whereas bile duct formation is impaired (Zhang et al., 2010). These results are possibly confounded by compensation from the YAP1 paralogue WWTR1, better known as TAZ. Using an overexpression model, Lee et al. suggested that YAP promotes duct differentiation through TGFβ signalling and represses hepatocyte differentiation by inhibiting HNF4α (Lee et al., 2016). In this study, YAP was hyperactivated with Alb-Cre in hepatocytes late in embryonic development, and it is impossible to confirm whether the increased number of KRT19+ cells observed in the transgenic livers are KRT19+ duct or progenitor cells. In addition, there is confounding evidence that the Hippo pathway may also influence hepatocyte differentiation through an enhancer switching mechanism (Alder et al., 2014). Overlap of HNF4α-, FOXA2- and TEAD-binding sites in hepatoblasts suggested that TEAD may cooperate with HNF4α and FOXA2 to drive hepatocyte cell fate decisions (Alder et al., 2014), although this model has not been directly tested. Taken together, the significance of the Hippo signalling pathway in hepatoblast cell fate decisions requires further investigation.
Literature describing the role of WNT/β-catenin signalling in hepatoblast differentiation is relatively limited and also contradictory. The number of KRT19+ cholangiocytes is reduced in Foxa3-Cre; Ctnnb1flox/flox embryonic mouse livers (Tan et al., 2008), and several other reports suggest that β-catenin drives cholangiocyte specification through activation of TGFβ signalling and Sox4 expression (Decaens et al., 2008; Lemaigre, 2020). However, others have shown that Ctnnb1 deletion from E12.5 mouse hepatoblasts with Alfp-Cre has no effect on Sox9 expression or other biliary markers, suggesting that WNT/β-catenin signalling is dispensable for cholangiocyte specification (Cordi et al., 2016). Furthermore, a re-analysis of Foxa3-Cre; Ctnnb1flox/flox livers also demonstrated normal cholangiocyte differentiation (Cordi et al., 2016), which is inconsistent with the earlier report (Tan et al., 2008). Although the literature surrounding the role of WNT signalling in bile duct cell differentiation is unclear, these two separate studies have consistently determined that β-catenin is required for hepatocyte differentiation (Fig. 3), and that its absence results in parenchymal disorganization (Cordi et al., 2016; Tan et al., 2008).
Hepatocyte and cholangiocyte maturation
Although hepatocyte and cholangiocyte differentiation is mostly complete by E18.5 (Gordillo et al., 2015), liver cell maturation is an ongoing process that begins shortly after cell differentiation, terminates and continues postnatally (Zorn, 2008). Significant hepatocyte expansion occurs during this postnatal maturation phase and is driven by WNT/β-catenin signalling (Apte et al., 2007; Lade and Monga, 2011) (Fig. 3). Hepatocyte functional maturation is marked by albumin secretion, glycogen storage and metabolic activity of cytochrome P450 enzymes (Baxter et al., 2015; Hannan et al., 2013). The final maturation step in hepatocyte development is partially controlled by the oncostatin M (OSM) cytokine pathway, hepatocyte growth factor (HGF) signalling and the continued inhibition of Notch and TGFβ signalling (Ang et al., 2018; Chen et al., 2018; Hannan et al., 2013; Kamiya et al., 2001). Liver non-parenchymal cell types – endothelial, Kupffer and stellate cells – are responsible for paracrine OSM and HGF signalling to nearby maturing hepatocytes. To date, the exact signals that regulate hepatocyte maturation have not been fully elucidated. This is reflected in the fact that hepatocytes derived from human pluripotent stem cells in vitro are not functionally mature (Ang et al., 2018; Baxter et al., 2015; Camp et al., 2017; Hannan et al., 2013; Takebe et al., 2015). The hepatocyte microenvironment, which contains different extracellular matrix components, is another important feature that also needs to be considered for its role in cell maturation (Shiojiri and Sugiyama, 2004).
Hepatocyte maturation is maintained by the combined and continuous action of a few key transcription factors: FOXA1/2/3, HNF4α and CEBPA (Hayhurst et al., 2001; Karagianni et al., 2020; Reizel et al., 2020; Thakur et al., 2019). Together, these factors bind cis-regulatory elements that activate the hepatocyte gene programme and maintain hepatocyte identity. Loss of any of these factors results in hepatocyte dedifferentiation, reduced expression of hepatic genes and liver disease (Hayhurst et al., 2001; Lee et al., 1997; Reizel et al., 2020). Two important features of hepatocyte maturation are hepatocyte heterogeneity and metabolic zonation, whereby hepatocyte function is determined by location within the liver lobule (Gordillo et al., 2015). Periportal hepatocytes near the portal triad are responsible for albumin secretion and are also enriched for genes involved in oxidative phosphorylation, the urea cycle and glycogen synthesis, as well as PCK1, a major regulator of gluconeogenesis (Aizarani et al., 2019; Halpern et al., 2017). Hepatocytes in the central zone surrounding the central vein are responsible for bile acid biosynthesis and are enriched for GLUL, a glutamine synthetase, whereas hepatocytes in the remaining mid zone express cytochrome P450 enzymes and are responsible for metabolism (Aizarani et al., 2019; Halpern et al., 2017). This hepatocyte diversity and patterning that occurs postnatally is largely determined by WNT and Notch signalling (Gordillo et al., 2015; Lade and Monga, 2011). WNT signals from central vein endothelial cells establish the central hepatocyte zonation pattern (Halpern et al., 2017), whereas Notch signals establish periportal hepatocyte zonation (Brosch et al., 2018). A key step in hepatocyte functional maturation involves the switching of hepatocyte metabolism from glycolysis (glucose breakdown) to gluconeogenesis (glucose generation) (Gordillo et al., 2015). Finally, it should be noted that hepatocyte maturation status is not strictly unidirectional or static; for example, hepatocytes also contribute to the adult liver progenitor cell population that arises during regeneration in response to liver injury to replace both hepatocyte and cholangiocyte populations (Tarlow et al., 2014).
Although TGFβ signalling is required for cholangiocyte differentiation, this pathway is inhibited by SOX9 during remodelling and maturation of the ductal plate (Antoniou et al., 2009). Ductal plate remodelling in mice occurs between E14.5 and E17.5 when cholangiocytes form dilations known as primitive ductal structures, which eventually organize into a mature, branched tubule network (Gordillo et al., 2015; Lemaigre, 2020). Primitive ductal structures are bilayered but asymmetrical, with SOX9+ cholangiocytes lining the portal side and HNF4α+ hepatoblasts lining the parenchymal side (Antoniou et al., 2009). Starting around E18.5 and continuing postnatally, TGFβ signalling further stimulates biliary fate in periportal hepatoblasts, creating mature, symmetrical ducts lined with cholangiocytes on both sides. The process of liver bile duct maturation and tubulogenesis is driven by two transcription factors targeted by Notch signalling: SOX9 and HES1 (Antoniou et al., 2009; Lemaigre, 2020). In support of this, Sox9-deficient livers display an increase in asymmetrical primitive duct structures at E18.5 with an immature HES1 immunostaining pattern, suggesting that cholangiocyte maturation and morphogenesis is delayed (Antoniou et al., 2009). The extracellular matrix also plays a vital role in the maturation of ductal structures. Specifically, laminin α5 is required for the formation of mature ductal lumens of the proper size and KRT19+ cell symmetry (Tanimizu et al., 2012). The WNT/β-catenin pathway is dispensable for bile duct morphogenesis (Cordi et al., 2016), though appropriate levels of β-catenin are required for cholangiocyte maturation. Importantly, hyperactivation of β-catenin in SOX9+ cells at E14.5 results in bile duct dysplasia and dedifferentiation (Cordi et al., 2016), suggesting that aberrant activation of WNT signalling molecules during development can lead to a precancerous phenotype.
Alterations in signalling pathway activities contribute to liver cancer
Many of the developmental signals and fetal genetic programmes outlined above that regulate hepatogenesis also re-appear in the context of liver disease, including primary liver cancer. Primary liver cancer is the second leading cause of cancer-related deaths in the world (Villanueva, 2019). The major etiological factors for liver cancer are a history of liver cirrhosis, chronic viral hepatitis infections, lifestyle and environmental factors (e.g. smoking, excessive alcohol consumption, obesity and diabetes) and a variety of genetic mutations (Forner et al., 2018; Villanueva, 2019). According to The Cancer Genome Atlas (TCGA), mutations in liver cancers are most often seen in TP53 (p53), which is associated with approximately 25% of cases (Fig. 4), followed closely by mutations in CTNNB1 (β-catenin). Interestingly, many additional mutations identified in liver cancer are in genes associated with chromatin remodelling complexes (e.g. ARID1A, ARID2) (Fujimoto et al., 2012), supporting a role for epigenetic dysregulation in the pathogenesis of liver cancer (Macchi and Sadler, 2020). Together, genetic, epigenetic and environmental conditions initiate chronic inflammation, liver fibrosis, cirrhosis, and cell dysfunction, followed by progression to carcinoma (Forner et al., 2018; Villanueva, 2019) (Fig. 5).
Genetic mutations in hepatocellular carcinoma and cholangiocarcinoma. The top genetic mutations in reported cases of liver cancer, including hepatocellular carcinoma (HCC) and cholangiocarcinoma (CCA), are plotted as a percentage of the total cases and categorized by gene function. Data are curated from The Cancer Genome Atlas (TCGA) and filtered using a curated list from the Cancer Gene Census, including 415 cases from liver and intrahepatic bile duct samples (TCGA-LIHC and TCGA-CHOL projects). Whole-exome sequencing was performed on tumour and adjacent normal tissue specimens collected from patients diagnosed with HCC (The Cancer Genome Atlas Research Network, 2017) and CCA (Farshidfar et al., 2017) prior to chemotherapy.
Genetic mutations in hepatocellular carcinoma and cholangiocarcinoma. The top genetic mutations in reported cases of liver cancer, including hepatocellular carcinoma (HCC) and cholangiocarcinoma (CCA), are plotted as a percentage of the total cases and categorized by gene function. Data are curated from The Cancer Genome Atlas (TCGA) and filtered using a curated list from the Cancer Gene Census, including 415 cases from liver and intrahepatic bile duct samples (TCGA-LIHC and TCGA-CHOL projects). Whole-exome sequencing was performed on tumour and adjacent normal tissue specimens collected from patients diagnosed with HCC (The Cancer Genome Atlas Research Network, 2017) and CCA (Farshidfar et al., 2017) prior to chemotherapy.
Liver architecture is disorganized in hepatocellular carcinoma. Schematic of liver progression to HCC and the underlying disrupted cellular organization. Liver cancer pathogenesis results from an accumulation of genetic mutations (e.g. TP53, CTNNB1) and/or exposure to adverse environmental factors (e.g. viral hepatitis infections). Together, these conditions initiate chronic liver inflammation, fibrosis, cirrhosis and cell dysfunction. Progression to hepatocellular carcinoma (HCC) is observed when the spatial and cellular organization of the mature liver is lost. Upon chronic injury, 90% of livers accumulate fibrotic scars, cirrhosis and hepatocyte dysfunction. In the healthy liver (left), hepatocytes are spatially organized around portal triads and interspersed with bile canaliculi and hepatic sinusoids. Within an HCC tumour nodule (right), this cellular architecture is lost and the liver develops fibrotic scarring, necrosis, vascular invasion and immune cell infiltration. In addition, a higher nucleus to cytoplasmic ratio, trabecular architecture, increased cellularity and extracellular matrix deposition are also features present in HCC. This figure shows a stylized schematic of the common histological abnormalities; however, HCC tumours are often uniform in appearance and do not generally display all of the described features within one nodule.
Liver architecture is disorganized in hepatocellular carcinoma. Schematic of liver progression to HCC and the underlying disrupted cellular organization. Liver cancer pathogenesis results from an accumulation of genetic mutations (e.g. TP53, CTNNB1) and/or exposure to adverse environmental factors (e.g. viral hepatitis infections). Together, these conditions initiate chronic liver inflammation, fibrosis, cirrhosis and cell dysfunction. Progression to hepatocellular carcinoma (HCC) is observed when the spatial and cellular organization of the mature liver is lost. Upon chronic injury, 90% of livers accumulate fibrotic scars, cirrhosis and hepatocyte dysfunction. In the healthy liver (left), hepatocytes are spatially organized around portal triads and interspersed with bile canaliculi and hepatic sinusoids. Within an HCC tumour nodule (right), this cellular architecture is lost and the liver develops fibrotic scarring, necrosis, vascular invasion and immune cell infiltration. In addition, a higher nucleus to cytoplasmic ratio, trabecular architecture, increased cellularity and extracellular matrix deposition are also features present in HCC. This figure shows a stylized schematic of the common histological abnormalities; however, HCC tumours are often uniform in appearance and do not generally display all of the described features within one nodule.
Hepatic cancers can develop from any cell lineage of the liver, but, because of the relevance to development, this Review will focus on primary cancer of the hepatocytes and cholangiocytes: HCC and CCA, respectively. HCC is the most prevalent form of liver cancer (90%), whereas CCA is more rare (10%) (Ghurburrun et al., 2018; Whittaker et al., 2010). The majority of HCC cases (90%) arise in the context of chronic liver injury and cirrhosis; however, a small proportion of cases (10%) arise spontaneously or without pre-existing liver disease (Ghouri et al., 2017; Zhang and Friedman, 2012). Unfortunately, because the early stages of HCC are often asymptomatic and tumour growth escapes detection, effective treatment options are limited at diagnosis and the 5-year survival rate is only about 18% (Villanueva, 2019). The 5-year survival rate for CCA is even lower, approximately 10% (Ghurburrun et al., 2018). HCC and CCA survival rates are also low owing to extremely high reoccurrence rates even with early interventions such as surgical resection (Ding et al., 2019).
Subclasses of hepatocellular carcinomas are defined by distinct molecular and clinical features, including genetic mutations, levels of the oncofetal protein AFP, which is present in both the fetal liver and HCC, as well as differentiation status (Hoshida et al., 2009). An HCC proliferative subclass with TP53 mutations, high AFP levels and poor hepatocyte differentiation is associated with worse clinical outcomes (Villanueva, 2019). The spatial and metabolic zonation of hepatocytes observed in the mature human liver is lost in HCC, in addition to the lack of expression of cytochrome P450 enzymes (Aizarani et al., 2019). A common theme among the most aggressive HCC subtypes with the worst prognoses is the loss of maturation markers, such as FOXA2 and HNF4α, and reactivation of developmental programmes and embryonic genes (Kitisin et al., 2007; Yong et al., 2013). For example, genes with expression normally restricted to embryonic hepatoblasts, including AFP, KRT19, DLK1, EPCAM, GPC3, MYC and SALL4, and genes involved in cell proliferation, are often hyperactivated in HCC (Andrisani et al., 2011; Si-Tayeb et al., 2010a; Yong et al., 2013). It is not surprising then that the genes that normally support liver progenitor cell growth and differentiation also drive cancer cell growth and dedifferentiation when abnormally expressed in mature hepatocytes and cholangiocytes. Signalling pathways regulating the transcription of genes that tip the balance between cell proliferation and differentiation are obvious suspects underlying the molecular mechanisms driving liver cancer. This suggests a link between hepatic oncofetal genes and the signalling pathways that have been implicated in HCC and CCA pathogenesis.
Mutations in the WNT signalling pathway promote HCC
After birth, significant postnatal hepatocyte proliferation results from a dramatic increase in nuclear β-catenin levels (Apte et al., 2007; Lade and Monga, 2011). The link between the WNT/β-catenin signalling pathway and liver cell proliferation also reappears during carcinogenesis. The downstream effector of the WNT signalling pathway, β-catenin, has a strong association with HCC and as many as 70% of HCC cases have elevated β-catenin levels (Whittaker et al., 2010). In addition, nuclear localization of β-catenin is indicative of advanced stage HCC (Kim et al., 2019). The β-catenin gene, CTNNB1, is one of the most highly mutated HCC genes and accounts for over 20% of cases (Fig. 4). High levels of β-catenin promote excessive cell proliferation through cell cycle target genes such as MYC and CCND1 (Kitisin et al., 2007). However, excess β-catenin alone is not sufficient to drive HCC progression (Whittaker et al., 2010) and mutations in CTNNB1 are associated with a less aggressive, non-proliferative HCC subtype (Fitamant et al., 2015; Hoshida et al., 2009; Villanueva, 2019). In addition, other genes in the WNT signalling pathway (e.g. AXIN1) are often mutated in HCC (Fig. 4) (Ghurburrun et al., 2018). These results support the idea of a ‘two-hit hypothesis’ whereby a single genetic mutation is not sufficient for carcinogenesis, but requires an additional ‘hit’ – either a second mutation or an unfavourable environmental factor, such as inflammation or injury. Further support for this model comes from studies in CCA in which KRAS mutations are the most frequent. However, introduction of an oncogenic Kras mutation in mouse cholangiocytes does not initiate CCA, unless also combined with a mutation in Pten (Ikenoue et al., 2016). Whether the combination of a Kras mutation in an inflammatory environment also results in CCA has not been determined.
The Hippo pathway effectors YAP/TAZ are hyperactive in HCC and CCA
The Hippo signalling pathway acts as a tumour suppressor mechanism that modulates cell proliferation and organ homeostasis in many tissues (Manning et al., 2020). In the adult liver, Hippo signalling regulates hepatocyte proliferation, liver size and regeneration (Manmadhan and Ehmer, 2019; Wang et al., 2018). Although mutations in Hippo pathway genes are rare, the downstream Hippo effector proteins, YAP1 (or simply YAP) and TAZ, are consistently hyperactivated in HCC and CCA (Bai et al., 2012; Fitamant et al., 2015; Li et al., 2012; Patel et al., 2017; Wang et al., 2018). A progressive liver cancer phenotype is also observed in mouse models in which YAP and TAZ are elevated or constitutively active. Hyperactivation of YAP/TAZ in mouse hepatocytes through transgenic overexpression or genetic deletion of upstream regulatory kinases results in expansion of atypical ductal cells, hepatocyte overproliferation, liver enlargement and progression to HCC (Dong et al., 2007; Fitamant et al., 2015; Manmadhan and Ehmer, 2019; Nguyen et al., 2015; Patel et al., 2017; Wang et al., 2018; Yimlamai et al., 2014; Zhou et al., 2009).
The subcellular localization of YAP/TAZ is particularly important and is regulated by Hippo pathway activity through phosphorylation (Manning et al., 2020). When Hippo signalling is active, phosphorylated YAP/TAZ are retained in the cytoplasm. When Hippo signalling is inactive, the unphosphorylated forms of YAP/TAZ translocate to the nucleus where they act as co-activators of TEAD transcription factors (Manning et al., 2020). Together, TEAD and YAP (or TAZ) recruit chromatin remodellers to activate the expression of genes that control cell proliferation (Manning et al., 2020). Under normal conditions in the liver, YAP is weakly cytoplasmic in hepatocytes but strongly nuclear in cholangiocytes (Patel et al., 2017; Yimlamai et al., 2014). However, in HCC, hepatocytes gain nuclear YAP expression and it is estimated that nuclear YAP is detected in about 70% of cases (Tschaharganeh et al., 2013). This is consistent with the observation that YAP is particularly enriched at mitotic chromatin (Manning et al., 2018). There is also evidence to suggest that an activated YAP signature serves as a biomarker for a particularly aggressive subtype of HCC associated with a progenitor or duct-like phenotype (Fitamant et al., 2015). In addition, AFP and GPC3 are suspected targets of YAP and this may contribute to the elevated AFP and GPC3 levels observed in HCC (Li et al., 2012; Zhou et al., 2009). Further chromatin immunoprecipitation studies are needed to confirm the direct binding of YAP to AFP and GPC3 regulatory elements. Although hepatocyte-activated YAP is described as a metastatic driver of HCC, it is important to note that YAP activation on its own is insufficient for uncontrolled hepatocyte proliferation, but requires an additional signal – either inflammation or injury (Su et al., 2015).
Tissue engineering to model liver development and disease
In vitro liver model systems are powerful tools for studying liver development, regeneration and disease. In particular, pluripotent stem cell differentiation is an attractive approach to investigate the unique combination of signals and transcription factors required for progression from pluripotency through to functionally mature liver cells. Multiple groups have applied concepts from the developing mouse liver to direct embryonic and induced pluripotent stem cells towards hepatocyte-like cells in vitro. Early in liver differentiation protocols, the TGFβ family member activin A is used to guide pluripotent stem cells toward definitive endoderm, in the presence or absence of FGF2 or WNT3A (Baxter et al., 2015; Hannan et al., 2013; Ogawa et al., 2013; Si-Tayeb et al., 2010b; Takebe et al., 2013). Successful differentiation is marked by loss of pluripotency transcription factors [e.g. NANOG, OCT4 (POU5F1) and SOX2] and gain of endoderm factors (e.g. GATA4 and SOX17). Next, definitive endoderm cells are exposed to BMP, FGF and either activin A or TGFβ inhibition (Ang et al., 2018; Hannan et al., 2013; Raggi et al., 2020 preprint). This further restricts differentiation to foregut endoderm, a stage that is marked by FOXA2 and HHEX. Specification to the hepatic lineage (hepatic endoderm and hepatoblasts) follows with the addition of BMP4 and FGF (±TGFβ and WNT inhibition) and is defined by the presence of AFP, HNF4A, PROX1 and TBX3 (Ang et al., 2018; Baxter et al., 2015; Hannan et al., 2013; Ogawa et al., 2013; Raggi et al., 2020 preprint; Si-Tayeb et al., 2010b; Takebe et al., 2013). Differentiation efficiency is relatively high to this stage, with approximately 80-90% of cells positive for AFP or HNF4A (Hannan et al., 2013; Raggi et al., 2020 preprint; Si-Tayeb et al., 2010b). Further differentiation and maturation of hepatocytes is marked by loss of AFP, enhanced albumin levels and enhanced cytochrome P450 enzyme function; however, the efficiency is comparatively low and the growth factors (e.g. HGF, OSM, dexamethasone), small molecules (Siller et al., 2015) and culture techniques (e.g. 2D versus 3D) used to achieve mature hepatocytes vary considerably. For example, Ang et al. combine signals that drive 2D hepatocyte differentiation with NOTCH and TGFβ inhibitors to simultaneously prevent cholangiocyte differentiation (Ang et al., 2018). Gordon Keller's group uses 3D aggregation and cAMP treatment of 26-day differentiated hepatocyte-like cells to further improve maturation and metabolic enzyme activity (Ogawa et al., 2013). In the final stage of their 2D protocol, Takeishi et al. expose hepatocytes to bile acid and fatty acids to induce maturation (Takeishi et al., 2020), whereas Max Paganelli's group includes WNT activation and varies the duration of OSM and dexamethasone treatment (Raggi et al., 2020 preprint). Taken together, the field has made significant progress in stem cell-derived hepatocyte differentiation in the last decade and the focus now shifts toward generating chemically defined protocols and large-scale production for clinical translation (Harrison et al., 2020 preprint; Sekine et al., 2020).
So far, hepatocytes produced in the tissue culture dish have an immature phenotype, owing to our general lack of understanding of the final steps in hepatocyte functional maturation (Baxter et al., 2015). Furthermore, isolated primary adult hepatocytes rapidly dedifferentiate in 2D culture (Akbari et al., 2019a), suggesting that novel in vitro culture systems and/or elucidation of the signals required to maintain hepatic identity are needed for both stem cell-derived hepatocyte maturation and long-term primary hepatocyte culture. Several groups have identified a benefit of 3D models that include multiple cell types present in the developing liver. For example, 3D organoid culture systems and bioengineering of implantable livers have the unique advantage of controlling the ratio of liver cell types and replicating the hepatic microenvironment by addition of extracellular matrix components. This has recently been accomplished through the co-culture of stromal and/or endothelial cells with hepatocytes (Camp et al., 2017; Sekine et al., 2020; Stevens et al., 2017; Takebe et al., 2013, 2017). These models have already advanced our understanding of the relationship between cellular maturation and 3D interactions, as it is now understood that 3D liver cell cultures have improved cell structure and metabolic function (Akbari et al., 2019b; Camp et al., 2017; Hu et al., 2018; Ogawa et al., 2013; Prior et al., 2019b). Recent work demonstrated that maturation of in vitro-derived human hepatocytes was improved after the hepatocytes were organized into a decellularized liver scaffold, together with cholangiocytes, endothelial cells, fibroblasts and mesenchymal cells at appropriate cell ratios (Takeishi et al., 2020). Although this impressive study highlights the progress that the liver field is making towards understanding hepatocyte maturation and liver bioengineering, the cells are unfortunately still immature compared with isolated human hepatocytes. Researchers have now begun building novel tools that improve our understanding of hepatic stellate and sinusoidal endothelial cell biology (Gage et al., 2020; Vallverdú et al., 2021) to support the overall goal of generating and maintaining fully mature hepatocytes in vitro.
Liver organoids are a useful research tool for studying signalling interactions, performing genetic manipulations and understanding cell development (Akbari et al., 2019a; Prior et al., 2019b). For example, LGR5+ hepatoblasts can be isolated from an E10.5 mouse liver and cultured in vitro as organoids (Prior et al., 2019a). Differentiation is directed towards the hepatocyte or cholangiocyte fate using separate media culture conditions; however, formulation of media with the ability to differentiate multiple liver cell types simultaneously is under development (Harrison et al., 2020 preprint). The challenge of generating optimal culture media also applies to liver organoid co-culture systems, which consist of heterogeneous cell populations and also require different signals for specific cell types. Another important application of organoid culture is in vitro disease modelling, including for primary liver cancers. In an elegant study from Meritxell Huch's group, ‘tumouroids’ were derived from HCC or CCA samples and the phenotypic hallmarks of each cancer subtype were recapitulated in vitro (Broutier et al., 2017). Transplantation into immunocompromised mice demonstrated their potential for tumour growth and metastasis. Furthermore, the tumouroids were used to screen for anti-cancer drugs and several unique targets were developed (Broutier et al., 2017). A recent mouse tumouroid study demonstrated that FGF signalling is a driver of CCA by expressing FGF receptor 2 fusion proteins in Tp53 mutant liver organoids (Cristinziano et al., 2021). The organoids developed into CCA tumours when transplanted into immune-deficient mice, and treatment with an FGFR-specific inhibitor effectively reduced tumour volume. Overall, the ability to control organoid cell composition and environment are crucial biological advantages for studying hepatic cell differentiation and liver cancer mechanisms in an in vitro organoid system. For example, the inflammatory signals and fibrotic cells found in mouse liver cancer models are excluded from liver cancer organoids, which effectively removes secondary effects from the analysis. In addition, small molecule inhibitors or suppressive lentiviral delivery approaches can be employed to assess the role of signalling pathways in hepatoblast organoid differentiation in complete isolation.
These studies highlight the utility of 3D organoids and biofabrication techniques for understanding the complexity of signalling pathways and transcription factor networks in liver cell development and cancer research. Further optimization of protocols to include co-culture systems and appropriate vascularization should significantly improve their use as liver research tools. Another important consideration is scalability of bioengineered livers, where generation of a therapeutic hepatocyte dose remains a current limitation of this approach. Additionally, whether these ex vivo 3D model systems recapitulate the genetic and epigenetic modifications of native liver cells is only beginning to be explored. Although some important outstanding questions remain, it is clear that liver organoids and 3D-printed livers hold great research potential. The discovery of exogenous signals that drive and maintain hepatocyte maturity will be fundamental for the development of personalized regenerative therapies to treat primary liver cancers.
Conclusions and future outlook
The liver is a highly complex organ that relies on multiple signals from surrounding tissues to guide proper cell development and maintain hepatic identity. It is becoming increasingly apparent that the same developmental signals not only reappear in liver cancer, but may be underappreciated oncogenic drivers. Thus, studies of liver developmental biology are paramount for improving our understanding of liver cancer biology and for providing more targeted and personalized treatments. Although the major signalling pathways in liver development are known, some important outstanding questions still remain. For example, what are the initial developmental signals that activate Foxa2 and Hnf4a during hepatic specification? Which signalling pathways further promote differentiation and maturation of hepatocytes? Answers to these questions are particularly important for regenerative medicine approaches, where the aim is to use small molecules that mimic secreted signals during development to generate mature, functional hepatocytes in vitro from embryonic or adult liver stem cells.
Acknowledgements
The authors would like to thank Nicole A. J. Krentz (Stanford University) and Sigrid Alvarez (University of British Columbia) for their critical reading of the manuscript, as well as David Schaeffer (University of British Columbia) for his consultation on liver cancer pathology. The results published here are in whole or part based upon data generated by the TCGA Research Network (https://www.cancer.gov/tcga).
Footnotes
Funding
Funding was provided by the Canadian Institutes of Health Research (FRN-153006). J.L. and S.D. are recipients of studentships from the Natural Sciences and Engineering Research Council of Canada.
References
Competing interests
The authors declare no competing or financial interests.