Maladaptive regeneration — the reawakening of developmental pathways in NASH and fibrosis

Abstract

With the rapid expansion of the obesity epidemic, nonalcoholic fatty liver disease is now the most common chronic liver disease, with almost 25% global prevalence. Nonalcoholic fatty liver disease ranges in severity from simple steatosis, a benign ‘pre-disease’ state, to the liver injury and inflammation that characterize nonalcoholic steatohepatitis (NASH), which in turn predisposes individuals to liver fibrosis. Fibrosis is the major determinant of clinical outcomes in patients with NASH and is associated with increased risks of cirrhosis and hepatocellular carcinoma. NASH has no approved therapies, and liver fibrosis shows poor response to existing pharmacotherapy, in part due to an incomplete understanding of the underlying pathophysiology. Patient and mouse data have shown that NASH is associated with the activation of developmental pathways: Notch, Hedgehog and Hippo–YAP–TAZ. Although these evolutionarily conserved fundamental signals are known to determine liver morphogenesis during development, new data have shown a coordinated and causal role for these pathways in the liver injury response, which becomes maladaptive during obesity-associated chronic liver disease. In this Review, we discuss the aetiology of this reactivation of developmental pathways and review the cell-autonomous and cell-non-autonomous mechanisms by which developmental pathways influence disease progression. Finally, we discuss the potential prognostic and therapeutic implications of these data for NASH and liver fibrosis.

Obesity and its metabolic consequences are among the most pressing health problems to date, with alarming projections for the future^1,2. Of these metabolic consequences, nonalcoholic fatty liver disease (NAFLD), defined by excess liver fat, has quickly become the most common chronic liver disease, with an estimated global prevalence of 25%^3–5. NAFLD ranges widely in severity, from simple steatosis, a prevalent and reversible state, to the necroinflammatory changes that characterize nonalcoholic steatohepatitis (NASH)⁶. NASH has no approved pharmacotherapy. As such, it is currently the second most common, yet the fastest growing, indication for liver transplantation in the United States⁷. However, livers available for transplantation are already limited and will not keep pace with the expected growth in the prevalence of NASH. Thus, new pharmaceutical targets are a great unmet need for an increasingly obese population and demand a comprehensive understanding of the pathophysiology of the disease.

Unlike other chronic liver diseases, the pathogenesis of NASH can be viewed as substrate-overload liver injury induced by an excessive lipid flux and increased de novo lipogenesis in hepatocytes as a result of obesity. Hepatocyte injury in turn provokes a multicellular repair response in which immune cells are activated locally or recruited from the bloodstream, the extracellular matrix (ECM) is remodelled to provide support, and hepatocytes proliferate to compensate for the lost functional hepatic mass⁸. Although this response is fundamentally similar to the response to viral infections or toxin exposure, chronic overnutrition, unlike such acute insults, can induce a persistent regenerative response, causing extensive accumulation of ECM and disrupting normal hepatic structure and function⁹. The resultant liver fibrosis is the strongest predictor of long-term clinical outcomes in patients with NAFLD–NASH^10–14 and an important end point in NASH clinical trials¹⁵. Fibrosis also increases the risk of development of end-stage liver disease (cirrhosis) as well as of hepatocellular carcinoma (HCC)¹⁶, the third-leading cause of cancer-related deaths worldwide¹⁷.

Developmental pathways, including Notch¹⁸, Hedgehog¹⁹ and Hippo–YAP–TAZ^20–22, are evolutionarily programmed to coordinate morphogenesis, lineage specification of hepatic progenitors and zonation establishment during development²³ but are thought to be quiescent in healthy liver unless ‘reactivated’ in acute liver injury to organize regeneration. However, studies over the past decade suggest that these pathways are persistently activated in obesity to cope with the chronic insult and that they contribute to the development of fibrosis. In this Review, we summarize the interwoven role of these developmental pathways in normal liver morphogenesis as well as in the wound-healing response. We also discuss how these protective mechanisms go awry in obesity-associated liver pathology and evaluate the potential therapeutic opportunities to target their reawakening.

Developmental signalling and morphogenesis

Notch

Signal transduction via Notch is dependent on juxtacrine interactions between cells expressing Notch receptors (NOTCH1–4) and cell surface-tethered Notch ligands (Jagged1 and Jagged2 or Delta-like ligands 1, 3 and 4)²⁴. Productive ligand–receptor interactions result in receptor proteolytic cleavage and release of the Notch intracellular domain, which cooperates with recombination signal binding protein for immunoglobulin-κJ region (RBPJ) and its co-activator Mastermind (MAM) to activate canonical Notch targets²⁵ such as Hairy and enhancer of split (HES) as well as HES-related with YRPW motif (HEY) genes, which are transcriptional repressors that regulate stem cell maintenance, boundary formation and cell-fate decisions²⁶ (FIG. 1a).

Fig. 1 |. Overview of developmental pathways in mammalian cells.

a | Notch signalling. Canonical Notch signalling is activated by the ligand-to-receptor interaction between two neighbouring cells. Upon ligand binding, Notch receptor undergoes sequential cleavages by a disintegrin and metalloproteinase (ADAM) protease and γ-secretase, leading to the release and nuclear translocation of the Notch intracellular domain (NICD). NICD interacts with immunoglobulin-κJ region (RBPJ) and Mastermind (MAM) to initiate the transcription of downstream targets such as the HES and HEY family of genes. Notch receptor can also induce non-canonical signalling by regulating β-catenin protein degradation. b | Hippo–YAP–TAZ signalling. In the Hippo-off state, the inactive STK3/4–LATS1/2 cascade is unable to phosphorylate YAP and TAZ. Unphosphorylated stable YAP and TAZ translocate into the nucleus and bind to the transcription factor TEAD family to regulate transcription. c | Hedgehog signalling. Hedgehog (HH) ligand binds to the cell-surface receptor Patched (PTCH), releasing its inhibition of Smoothened (SMO). Active SMO prevents the phosphorylation and cytoplasmic sequestration of GLI protein by kinases including protein kinase A (PKA), glycogen synthase kinase 3β (GSK3β) and casein kinase 1 (CK1). Stabilized GLI enters the nucleus and promotes Hedgehog target gene expression. d | WNT–β-catenin signalling. Binding of the WNT ligand to its receptor Frizzled and co-receptor low-density lipoprotein receptor-related protein 5 or 6 (LPR5/6) recruits the destruction complex, consisting of axin, adenomatous polyposis coli (APC), and the kinases GSK3β and CK1α, to the cell membrane via Dishevelled (DVL). This coordinated effort prevents the destruction complex from phosphorylating β-catenin, which facilitates its degradation. Stabilized β-catenin translocates to the nucleus, binds to the T cell factor (TCF) transcription factor family and promotes the expression of WNT target genes.

Notch is a well-established regulator of intrahepatic bile duct development in vertebrates. It is required for the commitment to the biliary lineage by the embryonic liver bipotential progenitor (hepatoblast) during the initial single-layer ductal plate formation at embryonic day 14.5 (E14.5) as well as for ductal morphogenesis from E16.5 until postnatal day 2 during mouse liver development²³. These requirements are underscored by the finding of Notch signalling defects in patients with Alagille syndrome, a congenital disease characterized by bile duct paucity that results in cholestasis, with additional heart and kidney pathology²⁷. Consistently, studies in rodent models revealed that hepatoblast NOTCH2, the most abundant Notch receptor during liver development²⁸, and Jagged1 from portal fibroblasts²⁹ form the essential receptor–ligand pair for biliary lineage specification and morphogenesis^30,31. Bypassing this ligand requirement by expression of a constitutively active form of Notch in hepatoblasts directs cholangiocyte differentiation at the expense of hepatocytes³². Notch activation represses the expression of hepatocyte-enriched transcription factors, including hepatocyte nuclear factor 4α (HNF4α) and HNF1α, with a reciprocal upregulation of cholangiocyte-rich transcription factors such as HNF1β and sex-determining region Y-box 9 (SOX9), which in turn interact with HNF6 (REF.³³) and SOX4 (REF.³⁴). Conversely, RBPJ deletion in embryonic hepatoblasts prior to E14.5 prevents biliary fate commitment and even later deletion (after E16.5) disrupts tubulogenesis and the 3D structures of intrahepatic bile ducts^32,35. The sum total of these actions is to favour a biliary cell fate³⁶.

Hippo–YAP–TAZ

Hippo is considered a tumour suppressor pathway that responds to cell density and regulates organ size³⁷. Upon activation, the mammalian counterparts of the Hippo kinases found in Drosophila, serine/threonine-protein kinase 3 (STK3) and STK4, together with the scaffold protein salvador homologue 1 (SAV1), phosphorylate the kinases large tumour suppressor homologue 1 (LATS1) and LATS2. Activated LATS1 or LATS2, together with their scaffold protein (MOB kinase activator 1A (MOB1A) or MOB1B), phosphorylate the transcriptional co-activators YAP and TAZ³⁸. Phosphorylated YAP and TAZ are targeted for cytoplasmic retention and proteasomal degradation. The inhibition of Hippo signalling results in stabilization and nuclear translocation of YAP and TAZ, which bind to the TEAD family of transcription factors and regulate a myriad of cellular processes³⁹ (FIG. 1b).

In liver development, hepatoblast deletion of Yap1 impairs bile duct development and the survival of hepatocytes⁴⁰. Reciprocally, constitutive YAP–TAZ signalling in hepatoblasts promotes a biliary lineage commitment and ductal plate expansion in mice⁴¹.YAP also maintains the progenitor states of hepatoblasts by promoting HNF4α and forkhead box protein A2 (FOXA2) interactions with enhancers of embryonic genes before hepatocyte differentiation during liver development⁴². In adult mouse liver, YAP protein expression and activity is highly enriched in cholangiocytes⁴³ and display a gradient in the hepatic parenchyma, with the highest activity near the portal triads and the lowest near the central vein⁴⁴.

Hedgehog

Signal transduction via Hedgehog is initiated by the binding of Hedgehog ligand to the receptor Patched to release Smoothened, which stabilizes the Hedgehog-responsive transcription factor GLI⁴⁵ (FIG. 1c). During mammalian embryonic development, Sonic hedgehog (SHH) is highly expressed in the ventral foregut endoderm, the origin of liver and pancreas⁴⁶. Although the role of Hedgehog in liver development is not clearly defined, Shh expression disappears at the onset of liver bud formation, transiently increases in fetal hepatoblasts and then rapidly decreases again as progenitors differentiate into hepatocytes⁴⁷, suggesting that Hedgehog signal activity is temporally regulated and restricted to ensure progenitor expansion and differentiation.

WNT–β-catenin

When biologically active WNT ligands bind to the cell-surface receptor Frizzled, β-catenin translocates to the nucleus, where it binds to T cell factor (TCF) family transcription factors. In the absence of WNT ligands, β-catenin is phosphorylated by its destruction complex, including scaffold proteins AXIN1 and AXIN2, leading to its degradation⁴⁸ (FIG. 1d). During liver development, WNT–β-catenin signalling is crucial to the survival, proliferation and maturation of hepatoblasts. That is, early deletion of Ctnnb1 (the gene encoding β-catenin) in embryonic mouse hepatoblasts at E9.5 produces multiple defects, including intrahepatic bile duct paucity and disrupted hepatocyte maturity⁴⁹. Deletion of Ctnnb1 after E16.5 in embryogenesis (by use of albumin-Cre rather than Foxa3-Cre) results in a milder phenotype, including impaired hepatocyte proliferation and postnatal liver mass⁵⁰. WNT signalling also establishes liver zonation as β-catenin competes with HNF4α for binding to TCF to direct pericentral versus periportal gene expression in mouse hepatocytes⁵¹. Thus, persistent stabilization of β-catenin results in a pericentral phenotype across the liver lobule⁵², whereas hepatocyte-specific loss of WNT–β-catenin signalling in mice induces a wide-spread periportal signature^53,54. Pericentral hepatocytes have high expression of WNT–β-catenin targets such as glutamine synthetase and cytochrome P450 enzymes⁵⁵, which perform essential zonal metabolic functions.

Developmental signalling in liver maturation

It has become evident that a highly spatiotemporally regulated and interactive network of developmental pathways exists to precisely guide normal hepatocyte and biliary maturation. For example, β-catenin gain-of-function impairs cholangiocyte differentiation and duct morphogenesis in mice⁵⁶, a phenotype that mirrors Notch loss-of-function mice³². These findings are consistent with data showing Notch-induced β-catenin degradation⁵⁷ as well as the general concept that Notch and β-catenin favour opposing cell-fate decisions from the same progenitor. Notch and WNT–β-catenin also cooperate in a cell-non-autonomous way to regulate biliary maturation — in zebrafish, β-catenin induces hepatocyte Notch ligand expression to activate cholangiocyte Notch receptors and induce bile duct maturation⁵⁸.

The literature is replete with other, similar examples of pathway crosstalk in liver development. Portal mesenchyme-derived transforming growth factor-β (TGFβ) establishes a gradient that induces the differentiation of portal hepatoblasts into cholangiocytes⁵⁹ as well as the transdifferentiation of portal mesenchymal cells into portal myofibroblasts, the latter of which results in increased Jagged1, which activates hepatoblast Notch receptors adjacent to the portal vein⁶⁰. Active YAP–TAZ signalling favours biliary cell fate by repression of HNF4α and upregulation of TGFβ⁴¹, which could interact with Notch in bile duct formation, as discussed earlier. YAP–TAZ and WNT–β-catenin signalling activities display opposite zonal distribution in the liver lobule. Liver-specific Yap1 deletion in mice results in WNT-active, glutamine synthetase-positive hepatocyte expansion into the periportal region, whereas constitutive YAP activation causes disruption of metabolic zonation and downregulation of pericentral genes⁴⁴. The spatiotemporal regulation of developmental signalling and the coordination between these pathways underscore the precise nature of normal liver development (FIG. 2).

Fig. 2 |. A network of developmental signalling controls lineage commitment during mouse liver development.

Liver development is regulated by a highly controlled network of pathways in a spatiotemporal manner. Near the portal vein, portal fibroblast-derived Jagged1 activates NOTCH2 on hepatoblasts to induce biliary differentiation, in conjunction with transforming growth factor-β (TGFβ) and YAP–TAZ, by transactivation of biliary lineage-defining factors, including SOX9, hepatocyte nuclear factor 1β (HNF1β) and HNF6. In the absence of these signals, increased HNF1α and HNF4α promote a hepatocyte lineage. In parallel, responding to WNT signals near the central vein, β-catenin activates pericentral genes to establish normal liver zonation.

Liver injury response and regeneration

Hepatocyte replenishment after liver injury

The liver can fully restore its mass and synthetic function after toxin-induced acute injury or surgical resection. One of the most frequently applied rodent models is partial hepatectomy, in which the massive hepatocyte loss is recovered primarily by proliferation and hypertrophy of the remaining hepatocytes⁶¹. Similar lineage tracing and pulse-chase methods have shown that hepatocytes also display a proliferative capability to repopulate themselves in mouse models of chronic liver disease^62,63. Although several studies suggest that all hepatocytes proliferate modestly to participate in regeneration without zonal preference^64,65, other studies suggest that specialized subsets (that is, AXIN2⁺ pericentral⁶⁶, SOX9⁺ periportal⁶⁷ and zone-non-specific hepatocytes with high telomerase reverse transcriptase (TERT) activity⁶⁸) or leucine-rich repeat-containing G protein-coupled receptor 5-positive (LGR5⁺) cells⁶⁹ might preferentially be involved.

By contrast, cholangiocytes and cholangiocytederived oval cells, a biliary hyperplastic response to chronic injury, rarely give rise to hepatocytes^70–72. However, when hepatocyte proliferation is severely compromised, such as in mice with hepatocyte-specific ablation of genes essential for this proliferative response^73–75, cholangiocytes can differentiate into hepatocytes as a fail-safe mechanism.

Post-partial hepatectomy

By promoting hepatocyte survival and proliferation, developmental pathways are crucial to liver regeneration after partial hepatectomy. Consistent with its role in maintaining hepatoblast proliferation during development, β-catenin rapidly accumulates in the nucleus within minutes after partial hepatectomy⁷⁶ and drives the expression of cell cycle-related genes during rat liver regeneration⁷⁷. Thus, the deletion of Ctnnb1 in hepatocytes renders a delay in regeneration⁵⁰. Both Hedgehog and YAP–TAZ signalling also respond to partial hepatectomy. Hedgehog ligand expression increases transiently after partial hepatectomy, leading to peak activity at the time of maximal hepatocyte proliferation⁷⁸. Similarly, YAP activity increases within 24 hours in response to partial hepatectomy⁷⁹, with both YAP and TAZ being required for the coordination of normal regeneration^80,81. Hedgehog and YAP also cooperate in cell-autonomous and cell-non-autonomous ways to facilitate post-partial hepatectomy regeneration: Hedgehog pathway induces YAP signal activity and the activation of myofibroblasts, which is required for the proliferative response in hepatocytes with nuclear YAP accumulation, while inhibition of Hedgehog pathway in myofibroblasts impairs hepatocyte proliferation following partial hepatectomy⁸². By contrast, administration of Indian Hedgehog (IHH) protein after parenchymal resection accelerates liver regeneration in mice⁸³. Finally, liver Notch signalling is also increased after partial hepatectomy and small interfering RNA (siRNA) directed to Jag1 or Notch1 impairs hepatocyte proliferation during rat liver regeneration⁸⁴. However, these siRNA findings are due to impaired Notch signalling in liver sinusoidal endothelial cells rather than in hepatocytes, in which fine-tuned Notch signalling is required for the regulation of endothelial identity and homeostasis⁸⁵ in both normal liver and during regeneration in mice^86,87.

Hepatocyte reprogramming in liver diseases

Unlike post-partial hepatectomy regeneration, emerging evidence suggests that mature hepatocytes maintain plasticity and utilize Notch-mediated reprogramming to adopt features of cholangiocytes and/or progenitors in chronic liver injury. This observation is exemplified by mouse models of forced Notch activation in mature hepatocytes, which results in ‘biphenotypic’ hepatocytes with the expression of biliary markers, such as SOX9 and osteopontin (OPN), that are normally absent in hepatocytes^32,88. Consistently, endogenous Notch activity in mature hepatocytes is necessary for transdifferentiation into cytokeratin 19-positive (CK19⁺) cells that morphologically resemble cholangiocytes in response to cholestasis induced by bile duct ligation or by exposure to the biliary toxin 3,5-diethoxycarbonyl-1,4-dihydrocollidine⁸⁸, likely mediated by myofibroblast-borne Jagged1 and opposed by macrophage-derived WNT signals⁸⁹. Similarly, methionine choline-deficient diet-feeding in mice, which causes obesity-independent steatohepatitis, induces hepatocyte Sox9 expression and biliary transdifferentiation in a subset of hepatocytes⁹⁰, suggesting that Notch-mediated cell fate reprogramming is prevalent in different models of liver injury.

The biological role of cell fate conversion is intriguing. One possibility is that hepatocytes transdifferentiate to counteract the loss of cholangiocyte functions in cholestatic injury. In line with this idea, mice with hepatoblast-specific simultaneous deletion of Rbpj and Hnf6 lack peripheral bile ducts⁹¹ but are able to recover from postnatal cholestasis due to a functional biliary system formed by transdifferentiated hepatocytes in a Notch-independent process driven by TGFβ signalling⁹². This finding redefines the plasticity of hepatocytes as a possible path to biliary regeneration in extreme cases of intrahepatic bile duct deficiency such as in human Alagille syndrome. Similarly, in response to bile acids and cholestatic insults, signalling mediated by YAP is required for the transformation of hepatocytes into biliary progenitors in mouse models^93,94. However, ductular reaction — that is, the oval cell response, characterized by the expansion of biliary cells in the parenchyma frequently seen in NASH and other chronic liver diseases⁹⁵ — derives primarily from cholangiocytes, with only a minor contribution from hepatocytes^70–72. This observation suggests an alternative purpose for injury-induced cellular plasticity: an evasion mechanism by which cells retain their regenerative ability after insults. Evidence for this hypothesis arises from chimeric lineage-tracing models that reveal hepatocyte-derived ductal cells that resemble bipotential progenitors and are capable of re-differentiating back into hepatocytes upon the withdrawal of insults⁹⁶.

These studies demonstrate the plasticity of hepatocytes and their regulation by reactivated developmental signalling pathways (FIG. 3). These findings are likely to translate to patients as biphenotypic cells are observed in human chronic liver diseases^97,98, suggestive of cell identity conversion between hepatocytes and cholangiocytes or bipotential liver progenitors. Importantly, developmental pathways require precise control — for example, Notch can induce hepatocyte fate change in response to insults. Hepatocyte YAP can cooperate in dedifferentiation via the upregulation of Notch2 and Jag1; reprogrammed cells clonally expand and regenerate the hepatocyte compartment upon removal of activated YAP after in vivo engraftment into fumarylacetoacetate hydrolase-deficient (Fah^−/−) mice⁴³, a model of liver failure that can be rescued by successful repopulation of functional hepatocytes⁹⁹. In addition, YAP can activate cell death in injured mouse hepatocytes and proliferation in undamaged ones, a selective mechanism required for homeostasis and appropriate regeneration in mice¹⁰⁰. However, hepatocyte Notch and YAP must be turned off to enable the re-differentiation and repopulation upon injury cessation. It is possible that dysregulation of these signalling pathways occurs in chronic liver diseases such as NASH to transform a regenerative process into one leading to fibrosis and cancer.

Fig. 3 |. Cellular reprogramming during mouse liver regeneration.

In response to acute injury, hepatocytes proliferate to maintain liver functional capacity. However, when the insult persists, Notch and YAP signalling pathways are activated in a subset of hepatocytes, resulting in ‘biphenotypic’ cells with the expression of biliary and progenitor markers such as SOX9 and osteopontin (OPN). Biphenotypic cells can dedifferentiate into progenitor-like cells as a strategy to evade the insult, with the potential to re-differentiate back into hepatocytes upon injury cessation. Alternatively, biphenotypic cells can also transdifferentiate into cholangiocytes to cope with injury-induced cholestasis. When injury persists, ductular reaction or oval cell response is induced, originated mainly from cholangiocytes, which can contribute to liver regeneration when hepatocyte proliferative capacity is severely compromised. In the absence of hepatoblast Notch and hepatocyte nuclear factor 6 (HNF6), transforming growth factor-β (TGFβ) can functionally substitute Notch signalling to form a functional biliary system.

NASH and fibrosis: maladaptive regeneration?

Multiple hits in NASH pathogenesis

Unlike other chronic liver diseases, NASH is primarily the result of ‘multiple hits’ induced by obesity, wherein parallel pathogenic events interact with host genetics, environmental factors and crosstalk between tissues^8,101,102. Insulin resistance leads to liver steatosis — in a susceptible individual, the resultant lipotoxicity induces mitochondrial dysfunction, endoplasmic reticulum stress, and hepatocellular injury and death^103,104. Damaged and necrotic hepatocytes in turn leak cytokines and damage-associated molecular patterns that drive immune cell infiltration and hepatic stellate cell (HSC) activation to coordinate the wound-healing effort that replaces dead hepatocytes^105,106. Transient and reversible deposition of ECM during acute injury provides mechanical stability and serves as a scaffold to guide cells for the restoration of normal liver architecture¹⁰⁷. In fact, the inhibition of HSC activity after partial hepatectomy blunts ECM production and paracrine signals that support hepatocyte proliferation⁸². NASH, however, is associated with chronic HSC activation, resulting in a continuous accumulation of ECM¹⁰⁸ and progressive substitution of the liver parenchyma by fibrous tissue.

Developmental pathways in NASH

Several mouse models that recapitulate the full spectrum of NASH pathology have been described^21,109–113, enabling investigation of the potential role of developmental pathways in disease pathogenesis (FIG. 4).

Fig. 4 |. Pathogenesis of nasH and fibrosis in mouse models.

As a result of chronic substrate overload, some steatotic hepatocytes undergo cell death due to sustained lipotoxicity and endoplasmic reticulum (ER) stress, which releases factors such as damage-associated molecular patterns (DAMPs) and Hedgehog (HH) ligands, which activate resident Kupffer cells and hepatic stellate cells (HSCs), resulting in inflammatory and fibrogenic responses. Other hepatocytes become reprogrammed by Notch, HH and YAP–TAZ to initiate a maladaptive response that induces hepatocyte secretion of fibrogenic factors such as osteopontin (OPN) and HH ligands to activate HSCs. This process creates a regenerative niche that aggravates fibrosis as well as compensatory proliferation to restore loss of mass. A by-product of this proliferative response is genomic instability, which might contribute to tumorigenesis. HCC, hepatocellular carcinoma; IHH, Indian Hedgehog; NASH, nonalcoholic steatohepatitis; NICD, Notch intracellular domain.

Hedgehog.

The first developmental pathway to be linked to the pathogenesis of NASH is Hedgehog signalling, which increases in hepatocytes, cholangiocytes and HSCs in response to injury to coordinate the wound-healing effort¹¹⁴, likely due to increased Hedgehog ligand expression from a near-zero basal level in normal livers¹¹⁵. Whereas Hedgehog signalling can be restorative after partial hepatectomy⁷⁸, Hedgehog signalling contributes to various NASH pathologies, including inflammation and fibrosis in mouse models^116–118. Dysregulation of Hedgehog might even exacerbate obesity-associated steatosis as in vivo mouse studies have revealed an intersection between Hedgehog–GLI protein signalling with the circadian clock to regulate hepatic lipid metabolism^119,120. Importantly, liver Hedgehog activity is evident in human NASH and positively correlates with disease severity and fibrosis stage¹²¹. SHH expression is increased in ballooned hepatocytes and drives HSC activation and ductular reaction¹²². Thus, in histological analyses of human liver biopsy samples from the PIVENS trial, which compared pioglitazone, vitamin E and placebo for the treatment of patients with NASH but without diabetes, it was observed that improved liver injury and fibrosis was associated with a reduction of SHH⁺ hepatocytes¹²³. Similarly, liver IHH is increased in patients with NASH compared with normal and steatotic livers²¹.

Hippo–YAP–TAZ.

Aberrant activity of YAP and TAZ is implicated in the pathogenesis of chronic liver diseases. For example, one study explored the role of autophagy in the regulation of YAP and found that hepatocyte-specific deletion of Atg7 resulted in impaired autophagy, liver fibrosis and cancer development, driven by increased levels of the autophagy substrate YAP in mice¹²⁴. Hepatocyte YAP and TAZ expression are also closely linked to NASH. TAZ protein is increased in hepatocytes derived from patients with NASH compared with simple steatosis and, in mouse models fed a NASH-promoting diet, hepatocyte TAZ action promotes the steatosisto-NASH transition by inducing injury, inflammation and fibrosis, largely mediated by increased secretion of IHH, a direct transcription target of TAZ²¹. Another study demonstrated that YAP and TAZ are activated in metabolically stressed hepatocytes in mouse models and in patients with NASH, inducing the expression of cysteine-rich angiogenic inducer 61 (CYR61), which is a chemokine that promotes inflammation and fibrosis²².

YAP is also an important regulator of non-parenchymal cell biology in liver diseases¹²⁵. YAP drives HSC activation and a fibrogenic programme^126,127, regulated by Hedgehog signalling^82,128. YAP is also essential to cholangiocyte homeostasis and contributes to ductular proliferation upon injury^93,94. Accordingly, YAP⁺ ductular reaction cells accumulate in human NASH, which is positively correlated with fibrosis¹²⁹.

WNT–β-catenin.

Although WNT–β-catenin regulates hepatocyte proliferation after partial hepatectomy¹³⁰, its role in NASH and fibrosis is less well defined. However, multiple indirect effects exist. Hepatocyte WNT–β-catenin signalling regulates lipogenesis¹³¹ and mitochondrial functions¹³² and interacts with FOXO1 to increase gluconeogenesis¹³³. The effects of WNT–β-catenin in non-parenchymal cells are similarly unclear, with most evidence coming from in vitro studies that have contradictory outcomes. For example, β-catenin activation is required for HSC quiescence in cultured primary rat HSCs¹³⁴ but knockdown of β-catenin can inhibit the proliferation and fibrogenic activation of a rat HSC line¹³⁵.

Notch.

Hepatocyte Notch activity is absent in healthy livers but upregulated in obesity, where it contributes to maladaptive outcomes¹³⁶. The post-developmental role of Notch in the liver was first discovered in genetic (leptin-deficient ob/ob mice) or diet-induced mouse models of obesity, in which Notch was found to increase hepatic glucose production by co-activating FOXO1 at gluconeogenic promoters¹¹². Notch was subsequently shown to activate de novo lipogenesis and increase hepatic steatosis by activation of the mechanistic target of rapamycin complex 1 (mTORC1) in mice fed a high-fat diet^137,138. These data suggest that Notch might uncouple these two insulin-regulated processes (de novo lipogenesis and hepatic glucose production) in liver, which is rare among mouse models of altered insulin action. For example, FOXO1-deficient mice show reduced glucose production but increased de novo lipogenesis¹³⁹.

Studies suggest an additional, obesity-independent and lipid-independent effect of hepatocyte Notch signalling on liver pathology. Forced hepatocyte Notch activation induces OPN expression and secretion, which activates HSCs and provokes marked fibrosis even in lean mice despite unchanged liver triglyceride content. Conversely, mice with hepatocyte-specific Notch loss of function are protected from NASH-induced pathology¹⁸. Importantly, liver Notch activity, as assessed by downstream transcriptional target expression of the HES and HEY families, is also positively correlated with NAFLD and NASH severity in patients¹⁴⁰. Hepatocyte HES1 staining is higher in livers with NASH and fibrosis than in healthy ones and, like SHH, is inversely correlated with treatment response in patients from the PIVENS trial¹⁸.

Notch activity in non-parenchymal cells can also contribute to NASH and other chronic liver diseases through a variety of proposed mechanisms. Notch promotes a pro-inflammatory (that is, ‘M1’) polarization of macrophages in CCl₄-induced injury¹⁴¹ and alcohol-induced injury mouse models¹⁴². Notch also seems to have a cell-autonomous effect in HSCs as γ-secretase inhibitor treatment of CCl₄-treated rats¹⁴³ or a siRNA against Notch3 in a rat HSC line¹⁴⁴ repressed fibrogenic gene expression. Similarly, HSC Notch inhibition dampens Hedgehog activity and the HSC–myofibroblast transition in cultured primary mouse HSCs and suppression of HSC Hedgehog signalling results in downregulation of Notch activity and in accumulation of myofibroblasts in a mouse model of bile duct ligation¹⁴⁵. In addition, a study using mice with HSC-specific overexpression and deletion of hyaluronan synthase 2 found that hyaluronan activates HSC Notch via CD44 signalling, whereas HSC-specific Notch1 deletion attenuated bile duct ligation-induced fibrosis, providing a causal link between hyaluronan, HSC Notch and fibrosis development¹⁴⁶. These data suggest that both hepatocyte and non-parenchymal cell Notch activity can activate HSCs to create a fibrotic niche for regeneration.

Future directions and therapeutics

Unanswered questions

The studies discussed enhance our understanding of the pathogenesis of NASH but there remain multiple unanswered questions. First, what is the mechanism by which developmental pathways are induced in response to injury or NASH? That is, what are the upstream signals that activate hepatocyte Notch and YAP–TAZ activity as well as WNT and Hedgehog ligand production in injured livers? Additionally, how are these signals turned off correctly in normal development and regeneration? Owing to the contextual importance of cell type and injury-induced activation, it is unlikely that these data can be derived using in vitro modelling, although data from stem cell-derived liver organoids are encouraging¹⁴⁷.

In addition, developmental pathways show a coordinated response in cellular reprogramming in the liver^43,82 but the mechanisms underlying this response are unclear. One intriguing idea is that of a ‘regulon’, in which multiple genes are controlled by the same regulatory molecule. For example, insulin-like growth factor 2 mRNA-binding protein 3 (IGF2BP3) regulates the expression of thousands of genes in HSCs via interactions with mRNAs involved in proliferation and transdifferentiation¹⁴⁸. Upon acute injury, hepatocyte IGF2BP3 modulates the expression of fetal markers required for regeneration, with checks and balances provided, in part, by YAP to achieve a balanced regenerative response in both hepatocytes and HSCs¹⁴⁹. The dysregulation of this pathway might theoretically result in acute liver failure and fibrosis but the specific role of this regulon and others in the wound-healing response in NASH requires further study.

Future directions — NASH-driven HCC?

Although a small percentage (2.4% to 12.8% depending on the cohort) of patients with NAFLD develop HCC^10–14, NASH is the most rapidly growing contributor to HCC in the United States^150,151. Pathogenic mechanisms underlying NASH-induced HCC have been comprehensively reviewed elsewhere¹⁵², with a focus on the role of DNA damage response, inflammation, autophagy and the gut microbiome. However, the chronic activation of developmental pathways in obesity suggests another mechanistic link between NASH and HCC. In NASH, these pathways could provoke uncontrolled hepatocyte proliferation and genomic instability, while simultaneously altering the surrounding microenvironment to one that favours malignant transformation and tumour growth. In fact, several groups have documented roles for these developmental pathways in liver tumorigenesis. Notch accelerates tumour proliferation via the direct transcriptional activation of cell cycle-related genes¹⁵³ and insulin-like growth factor 2 (IGF2)-mediated mitogenic signalling¹⁵⁴ to drive liver oncogenesis in mice. YAP–TAZ rewires glutamine metabolism to support proliferation¹⁵⁵, while inducing NUAK2 to increase tumour growth¹⁵⁶ and the chemokine monocyte chemoattractant protein 1 (MCP1; also known as CCL2) to promote an inflammatory tumour microenvironment^157,158. β-Catenin pathway mutations, frequently seen in human HCCs, induce a unique metabolism that depends on fatty acids and glutamine to sustain tumour proliferation^159,160 but also help tumour cells to evade T cell surveillance and resist checkpoint inhibitors^161–163. Intriguingly, the interactions between these pathways that are seen in development and regeneration are also preserved in HCC — for example, Notch is required for YAP–TAZ activity in hepatocarcinogenesis, while β-catenin signalling inhibits the tumorigenic functions of the YAP–TAZ–Notch signalling loop¹⁶⁴ (FIG. 5).

Fig. 5 |. The role of developmental pathways in liver cancer.

In the injured and fibrotic liver, hepatocyte developmental pathway activity results in proliferation and cellular transformation. Notch activation induces a range of cell-cycle genes, the autocrine mitogen Igf2, and biliary and progenitor-associated genes (that is, Sox9 and Spp1) that regulate the tumour microenvironment (TME) and cellular plasticity. Enhanced cell–cell contact and extracellular matrix (ECM) stiffness as well as a variety of nutrient factors, such as bile acids and cholesterol, increase YAP–TAZ stability, nuclear translocation, and transcription factor TEAD activity to augment proliferation and influence cellular metabolism and the TME to favour tumorigenesis, and might drive further Notch activity. In parallel, β-catenin activation, in part due to driver mutations in Ctnnb1, induces a T cell factor (TCF)-dependent proliferative response, alters hepatocyte metabolism via regulation of glutamine synthetase (encoded by Glul) and peroxisome proliferator-activated receptor-α (encoded by Ppara), and promotes tumoural T cell exclusion. HSC, hepatic stellate cell; NICD, Notch intracellular domain; RBPJ, immunoglobulin-κJ region.

An important caveat to these data is that most of the mouse studies described earlier were performed in toxin-induced, carcinogen-induced, or oncogene-induced tumours and might not necessarily be extrapolatable to NASH-induced HCC. Although mouse models of NASH–HCC have been described, including prolonged exposure to high-fat, high-carbohydrate diets or combinations of diet and toxins with genetic alterations¹⁶⁵, there is still no robust model of NASH-derived cirrhosis that precedes HCC nor a reliable and time-efficient NASH–HCC mouse model that strongly reproduces human disease progression and the complex features of HCC at histological, genomic and transcriptomic levels. New studies that combine genetic manipulations of these developmental pathways with mouse models of diet-induced NASH are necessary to determine their role in NASH-induced HCC.

Therapeutic possibilities

Dissecting the cell-autonomous and cell-non-autonomous effects will guide our understanding of the pathophysiology of liver diseases but might also enable the development of targeted therapies. Most developmental pathways depend on cell membrane ligand–receptor interactions, opening possibilities for therapeutic tractability. For example, hepatocyte Notch activity promotes HSC activation and fibrosis, though it is unclear by which ligand, borne by which cell or by which mechanism the ligand is regulated in NASH. Dissecting this regulation and blocking specific Notch ligand–receptor interactions might enable the inhibition of pathological Notch reactivation in hepatocytes, while leaving intact homeostatic Notch functions in other cell types such as liver sinusoidal endothelial cells.

Alternative therapeutic modalities might also be effective. Proof-of-principle studies with small-molecule γ-secretase inhibitors efficiently blocked Notch-dependent increases in hepatic glucose production^137,166 and liver fibrosis¹⁸ in mouse models, although regulatory approval is unlikely owing to substantial intestinal adverse effects167. However, Notch-induced goblet cell metaplasia can be avoided by using liver-directed antisense oligonucleotides to the γ-secretase-targeting subunit nicastrin¹⁶⁸. Mice treated with nicastrin antisense oligonucleotides show ameliorated fibrosis with established NASH without intestinal toxicity¹⁸. Similar benefits were seen with γ-secretase inhibitor-encapsulated liver-directed nanoparticles¹⁶⁹. Other therapeutic approaches, such as hepatocyte-specific N-acetylgalactosamine (GalNAc)-modified siRNA¹⁷⁰, might similarly maintain therapeutic efficacy with minimal off-target adverse effects. In fact, GalNAc-siRNA targeting hepatocyte TAZ reversed NASH and fibrosis in a mouse model of NASH¹⁷¹. Additionally, lipid nanoparticles containing siRNA against Ctnnb1 specifically silenced hepatic β-catenin¹⁷² and improved porphyria-induced liver injury¹⁷³ and HCC¹⁷⁴ in various mouse models.

Combination therapy with dual-pathway inhibitors might provide additional benefit. As alluded to earlier, Notch and YAP–TAZ share overlapping functions in cell fate specification and might form a feedforward activation loop in tumorigenesis. Similarly, Notch and TAZ activation in hepatocytes activate HSCs via different secreted factors — OPN and IHH, respectively. Thus, it might be beneficial to simultaneously inhibit two pathways to achieve maximal treatment efficacy for NASH and fibrosis.

Conclusions

Developmental pathways are crucial to the expansion of progenitors and lineage specification in the liver. In embryogenesis, these pathways are coordinated in a spatiotemporal and cell type-specific manner to ensure the precision and accuracy of development. In a healthy liver, they lie dormant or restrictedly active in a cell type-specific or liver zone-specific manner. These pathways are reawakened when responding to liver injury and assemble a coordinated wound-healing response by inducing proliferation and creating a regenerative niche. Upon cessation of the acute insult, developmental pathways return to their baseline quiescent state and the regenerative process tapers off. However, in NASH, lipotoxicity remains a persistent injury stimulus and this evolutionarily programmed regenerative response becomes maladaptive and ultimately a culprit in disease progression. Understanding the mechanisms of this response will refine our knowledge of NASH pathogenesis but might also lead to novel therapeutic opportunities for a disease that is highly prevalent and has high morbidity.

Key points.

• Nonalcoholic steatohepatitis(NASH), defined as liver lipid accumulation accompanied by injury and inflammation, is a prevalent condition without approved pharmacotherapy, leading to huge unmet needs for an increasingly obese population.

• Developmental pathways, including Notch, Hedgehog, Hippo–YAP–TAZ and WNT–β-catenin, are fundamental regulators of cell fate decisions and morphogenesis in liver development.

• Developmental pathways are also activated to promote regenerative proliferation and/or cellular reprogramming in response to liver injury.

• In the face of chronic liver disease such as NASH, persistent and maladaptive activation of developmental pathways leads to exacerbated NASH, fibrosis and liver cancer.

• Inhibitors targeting developmental pathways can ameliorate NASH and fibrosis in preclinical models, uncovering novel therapeutic possibilities for patients with NASH.