Regulation of hepatocyte identity and quiescence

Abstract

The liver is a highly differentiated organ with a central role in metabolism, detoxification and systemic homeostasis. To perform its multiple tasks, liver parenchymal cells, the hepatocytes, express a large complement of enabling genes defining their complex phenotype. This phenotype is progressively acquired during fetal development and needs to be maintained in adulthood to guarantee the individual’s survival. Upon injury or loss of functional mass, the liver displays an extraordinary regenerative response, mainly based on the proliferation of hepatocytes which otherwise are long-lived quiescent cells. Increasing observations suggest that loss of hepatocellular differentiation and quiescence underlie liver malfunction in chronic liver disease and pave the way for hepatocellular carcinoma development. Here, we briefly review the essential mechanisms leading to the acquisition of liver maturity. We also identify the key molecular factors involved in the preservation of hepatocellular homeostasis and finally discuss potential strategies to preserve liver identity and function.

Introduction

“While the heart resounds and attracts the music of the mandolin, there, inside, you filter and apportion, you separate and divide, you multiply and lubricate, you raise and gather the threads and the grams of life, the final distillate, the intimate essences”. With these words the Nobel literature laureate Pablo Neruda paid tribute to the liver, “the underground worker” with a “hidden alchemical chamber”, as he wrote in his Ode to the Liver. Indeed, the liver performs a wide variety of functions that are essential for the preservation of homeostasis in the organism. These functions include the synthesis of serum proteins and hormones, the extraction and metabolism of nutrients, xenobiotics and systemic waste products, the metabolism of lipids, bile acids and lipoproteins, the storage and usage of glucose, bile formation, and the neutralization of foreign antigens and microbes from the gut. All these tasks are conjointly performed by at least seven different types of cells: hepatocytes, cholangiocytes, sinusoidal endothelial cells, macrophages, different types of lymphocytes, dendritic cells and stellate cells, which are arranged in a parenchymal structure that facilitates their cooperative interaction. The liver parenchyma is perfused by two sources of blood: portal venous blood which is oxygen poor but rich in hormones, nutrients and toxins coming from the gut, pancreas and spleen; and oxygenated blood from the hepatic artery (Fig. 1). These afferent blood vessels serially branch to form the smallest capillary-size vessels known as the sinusoids. Sinusoids connect with the efferent hepatic veins (“central veins”) that collect sinusoidal blood carrying metabolized products and merge into large hepatic veins. Another network of conduits within the liver consists of the bile ducts that deliver bile to the gall bladder and the intestine, and together with the portal vein and the hepatic artery form the portal triad. Importantly, the liver parenchyma is not organized into connective tissue capsule-delimited repetitive units as occurs in other glands. Its anatomical organization is determined by these hemodynamic patterns of blood flow. The basic architectural unit of the liver is the liver lobule, formed by plates of 15–25 hepatocytes, the most abundant cell type in the liver, lined by sinusoidal capillaries that extend from the portal triad to the central vein. Blood flows from the portal vein and hepatic artery to the centrilobular vein, while bile moves in the opposite direction. Hepatocytes are normally quiescent polarized epithelial cells arranged in single cell thick cords [1]. Hepatocytes are held together by tight junctions, with their basolateral membranes facing the sinusoidal capillaries lined by endothelial cells, from where endocrine secretions to the blood are released, and their apical surfaces oriented toward the bile canaliculi that merge into the portal bile duct where bile is collected. Albeit liver function is the product of the concerted action of all cell types mentioned above, hepatocytes certainly play a central role. Interestingly, some important metabolic and secretory functions are unevenly distributed along the porto-central axis in the liver cell plate, a phenomenon known as “liver zonation” [2] (Fig. 1).

Fig. 1.

Schematic diagram of the liver lobule and zonated liver functions. Representation of the bile canaliculus and hepatic sinusoid, indicating the different cell types, blood vessels and bile duct. Arrows indicate the blood and bile flow. Metabolic functions distributed along the portal-central axis in periportal and pericentral hepatocytes are indicated. Mature liver functions and quiescence are maintained by transcription factors and mRNA splicing factors

To carry out their complex and varied functions, for which the preservation of their tridimensional organization within the lobule is essential, hepatocytes must express a large complement of enabling genes [3]. Understanding how the pattern of gene expression that defines hepatocellular identity is established has been the subject of intense study for more than two decades, and the principles guiding early hepatic development and differentiation are relatively well known [4–7]. These mechanisms are not only important from a developmental point of view, as hepatocellular identity clearly needs to be well preserved in the adult to maintain systemic homeostasis. Accumulating evidences indicate that the expression of liver-specific genes is reduced in chronic liver injury, a situation that would partly account for the progressive loss of function that occurs in chronic liver disease [8, 9]. Moreover, clinical and experimental observations indicate that hepatocellular dedifferentiation, which includes not only the loss of adult liver gene expression, but also the reactivation of genes expressed in the fetal hepatocytes, may be functionally related to the development of hepatocellular carcinoma (HCC) [10–14], the most serious complication of chronic liver disease and a deadly tumor [15].

However astounding the metabolic and secretory capacity of the liver may be, it is perhaps not the most impressive characteristic of this organ. The liver’s central position in systemic metabolism involves a prominent exposure to endogenous and exogenous damaging agents, including alcohol, viral infections and dietary components, which cause hepatocellular death and constitute the principal causes of liver disease [16]. To face these challenges, the liver has evolved an outstanding regenerative capacity mainly based on the proliferation of hepatocytes when a loss in functional parenchyma is sensed and most importantly liver function needs to be maintained while this growth response occurs [17]. Indeed, hepatocytes enclose a tremendous proliferative potential, perhaps only matched by hematopoietic cells [18]. Nevertheless, this potential is tightly controlled, as hepatocytes are mostly quiescent cells with an estimated lifespan of about 400 days in adult rat liver, where only 0.025 % of hepatocytes are typically engaged in DNA synthesis at a given time [19]. It is also worth mentioning that a great proportion of hepatocytes, up to 85 % in rodents and about 35 % in humans, are polyploid cells, i.e., contain four (tetraploid) or eight (octoploid) sets of chromosomes in one or two nuclei [20]. Interestingly liver ploidy is known to be modified during injury and regeneration, including chronic human liver disease [21]. However, the physiological significance of liver ploidy and of its pathological alterations are still not completely understood, and this is an area of active research recently reviewed elsewhere [20, 22].

A great deal of information has been produced over the past decades on the signals and mechanisms that upon parenchymal injury trigger hepatocyte proliferation and contribute to the cessation of the replicative response [17, 23–25]. All this knowledge is highly valuable, as it may lead to the design of protective and pro-regenerative therapies for the liver, but it is also relevant considering that unrestrained liver cell proliferation favors the development of HCC, and liver tumor cells co-opt many of these pro-mitogenic and survival pathways [26–28]. Moreover, recent studies have exposed the molecular mechanisms that tightly control the proliferative capacity of hepatocytes and operate under healthy normal conditions to preserve hepatocellular quiescence. Most interestingly, these current developments reveal that some of these genes involved in the establishment and/or preservation of hepatocellular identity are also able to keep hepatocytes in check, preventing untimely cell cycle entry.

In this review, we will summarize the essential aspects of the genetic control of liver development, the establishment of hepatocellular identity and the regulation of hepatocyte proliferation. These notions will help to understand how liver homeostasis is achieved. In light of current findings we will then discuss potential mechanisms underlying the loss of hepatocellular function and quiescence upon chronic liver injury, as well as prospective therapeutic strategies.

Overview of liver development and transcriptional control of hepatocellular differentiation

Liver parenchymal cells originate in the anterior portion of the definitive endoderm, one of the three germ layers that are formed during gastrulation. Coculture experiments in mouse embryos (E8.25) demonstrated that the foregut endoderm only undergoes hepatic specification if it is in contact with the developing cardiac mesoderm and the septum transversum mesenchyme [5, 7]. Hepatic specification refers to the first molecular evidence of liver development, which is characterized by the expression of albumin, transthyretin and α-fetoprotein (AFP) genes, indicating the cellular commitment to the hepatic differentiation fate even before organogenesis is detected [29]. The signals emanating from the cardiac mesoderm and the septum trasversum include fibroblast growth factors (FGFs) and bone morphogenic proteins-2 and -4 (BMP-2, BMP-4), respectively [5, 7]. Wnt signaling has been also involved in these early stages; however its role seems more complex (Fig. 2). While initially Wnt activity must be repressed to allow liver specification, later on Wnt signaling from mesoderm promotes liver development [7, 30]. These studies on intercellular signaling during liver specification were accompanied by significant efforts to elucidate the transcriptional events controlling hepatic development. The albumin gene is one of the best-characterized early markers of hepatic cells, and DNA footprinting studies of its regulatory regions in endoderm cells reveal the binding of forkhead box protein A1 (FOXA1, also called hepatocyte nuclear factors 3A, HNF3A) and GATA-4 transcription factors to the albumin enhancer even before its expression is activated [29]. These interesting observations indicate that FOXA1 and GATA-4 would interact with their respective binding sites even in the context of compacted chromatin. It was found that FOXA1 and GATA-4 behave as epigenetic modifiers, repositioning nucleosomes and opening the chromatin to facilitate the binding of other transcription factors that would respond to subsequent inductive signals. This ability to modify the chromatin landscape into a competent mode able to respond to developmental cues has led to consider FOXA and GATA family members as “pioneer factors” [4, 31]. The epigenetic mechanisms underlying this so-called chromatin prepatterning are currently being elucidated [32]. Consistently, deletion of FoxA1 and FoxA2 (or Hnf3B) genes, which have redundant roles in hepatogenesis, results in the impaired development of the hepatic lineage in vivo and in lack of response to FGF in cultured endoderm [33]. It is worth noting that while hepatocyte differentiation after specification of liver progenitors seemed to be independent of FOXA1/A2 [34], these factors play fundamental roles in the orchestration of metabolic functions in the adult liver [35]. A central role has been also identified for the homeodomain transcription factor HNF1β. Hepatic specification was not observed in Hnf1β null embryos and cultured ventral endoderm failed to express albumin upon FGF stimulation. Moreover, the expression of FOXA factors was also severely compromised in the prehepatic endoderm of HNF1β-deficient mice [36]. Taken together, these findings highlight the key function played by FOXA and GATA family members, along with HNF1β, in the regulation of early liver development.

Fig. 2.

Overview of signals (boxes with arrow) and regulatory genes (gray boxes) involved in the different stages of liver development and hepatocellular differentiation. Examples of relevant target genes expressed in each stage are indicated in the white boxes. MAT1A: methionine adenosyltransferase 1, GK glucokinase, LPK l-type pyruvate kinase, FASN fatty acid synthase, HK2 hexokinase 2, PKM2 pyruvate kinase M2. See text for details

Once the endoderm cells have been specified (E9 in mouse and day 22 in humans), committed cells from this region, now called hepatoblasts, adopt a columnar shape and generate a diverticulum in the transition to a pseudostratified epithelium. Hepatoblasts express a series of characteristic genes of the hepatocytic and biliary lineages such as albumin, AFP and cytokeratins 17, 18 and 19 (Fig. 2). These hepatoblasts proliferate and form a tissue bud limited by a basement membrane that is subsequently degraded by metalloproteases, allowing the hepatoblasts to expand into the septum transversum [5, 7]. The network of transcription factors involved in these steps of liver development has been established through studies in a series of knockout mice (Fig. 2). The hematopoietically expressed homeobox factor (HEX), which is induced in the ventral endoderm by FGF and BMP signaling [37], was shown to be essential for pseudostratification and hepatoblast proliferation [5]. Moreover, knockout of Hex in the hepatic bud or diverticulum resulted in the loss of the expression of key transcription factors including HNF6 (also known as Onecut-1) and HNF4α [38]. Interestingly, Hex expression was shown to depend in part on the zinc finger transcription factors GATA4 and/or GATA6, the latter factor being required to preserve the differentiation state of hepatoblasts [31]. Other important transcription factors involved at this stage are PROX-1 (prospero-related homeobox-1) and TBX3 (T-box transcription factor 3). PROX-1 promotes hepatoblast proliferation and migration [39], and TBX3 in turn controls Prox-1 gene expression [40]. The migratory defects of hepatoblasts from Prox-1 and Tbx3 null mice can be attributed in part to the high levels of E-cadherin detected in the liver buds of these embryos, which would maintain strong cell–cell contacts and thus hamper hepatoblast migration [40]. Another important observation in Tbx3 null embryos was the loss of expression of HNF4α and CCAAT-enhancer binding protein α (c/EBPα), two key regulators of hepatocyte differentiation that are otherwise highly expressed in hepatoblasts [5]. Other transcriptional regulators involved in hepatoblast migration are Onecut-1 and Onecut-2, which have been shown to promote degradation of the basal lamina surrounding the liver bud and the migration of hepatoblasts into the septum transversum. Consistently, Onecut-1- and Onecut-2-deficient embryos display a similar phenotype to Prox-1- and Tbx3-deficient mouse embryos [41].

After invading the septum transversum, between E9.5 and E15, hepatoblasts undergo intensive proliferation. WNT signaling is among the most prominent regulators of this proliferative response. Expression of β-catenin is maximal around E10 in mice, and WNT ligands are derived from endothelial and stellate cells [42]. Deletion of Ctnnb1, the gene coding for β-catenin, results in reduced proliferation, liver hypoplasia and decreased expression of liver-enriched transcription factors like c/EBPa and HNF4α [43]. Further important proliferative signals produced by the mesenchyma include hepatocyte growth factor (HGF), FGF-10, BMPs, transforming growth factor β (TGFβ) and retinoic acid [7].

Hepatoblasts are bipotential cells, capable of differentiating into either hepatocytes or cholangiocytes (biliary epithelial cells) (Fig. 2). Between E9.5 and E13.5 under the influence of signals generated in the periportal region, hepatoblasts adopt a cholangiocyte fate. These signals include TGFβ, NOTCH, WNT and FGFs such as FGF-2 and FGF-7 [5, 7]. It is believed that hepatoblasts away from the periportal region would be less exposed to these inductive signals and thus will adopt a hepatocytic fate. Nonetheless, in vitro experiments identified HGF as a growth factor favoring hepatocyte lineage differentiation [44]. Recent in vitro evidence also suggest that the differential expression of TGFβ receptor 2, which is induced by c/EBPb and repressed by c/EBPa in hepatoblasts, could determine their differentiation into cholangiocytes or hepatocytes, respectively [45]. Interestingly, hematopoietic cells may also be a source of signals for hepatocyte maturation. These cells migrate to the liver around E10, and by E12 they are closely associated with differentiating hepatocytes. Secretion of interleukin-6-related cytokines like oncostatin M (OSM) induces the expression of differentiation markers, such as carbohydrate metabolism enzymes and serum proteins through STAT3-mediated signaling [46–48].

The transcriptional repressor TBX3 appears to be essential for hepatocyte differentiation, as its elimination promotes biliary differentiation at the expense of hepatocyte specification [49]. As previously mentioned, TBX3 preserves the expression of key genes for hepatocyte maturation like HNF4α and c/EBPa, which are in turn silenced in periportal cells undergoing biliary differentiation [50, 51]. Concomitantly, TBX3 represses the expression of the transcription factor HNF6 that is required for biliary differentiation [52]. It has become apparent that not only the presence, but also the absolute expression levels of these transcriptional regulators are essential to define hepatic cell fate [31].

During the remainder of embryonic development, and early postnatal life, hepatoblasts committed to a hepatocyte cell fate progressively acquire the expression of genes and physiological functions characteristic of mature parenchymal cell. This maturation process is controlled by a dynamic and intertwined network of transcription factors. A study of the reciprocal interactions and cross-regulation, including the occupancy of regulatory regions within their respective promoters, identified a core of six liver-enriched transcription factors including: HNF1α, HNF1β, FoxA2, HNF4α, HNF6 and liver receptor homolog 1 (LRH1, also known as NR5A2) [53–55] (Fig. 2). The extensive cross regulation among these factors likely reflects the robustness of a network that is crucial for the terminal differentiation of the hepatocytes along the developmental process. A model in which hepatocyte maturation depends on the increase in the concentrations of these mutually regulated transcription factors has been proposed. In this model, threshold levels of each given factor would be reached at specific developmental stages, thereby allowing the simultaneous and timely recruitment of coactivators and the induction of target genes [5]. Nevertheless, as we will discuss in the next section, many if not all of these master regulators of embryonic liver development are also essential for gene regulation in the adult liver. It is not clearly understood how these transcription factors that are stably expressed throughout development and in adulthood fulfill distinct roles (i.e., transactivate different sets of genes) during organ maturation. One relevant example is the direct regulation of the transcription factor pregnane X receptor (PXR) by HNF4α which is essential in fetal hepatocytes, but lost in mature parenchymal cells [56]. In this regard, a very recent study demonstrated that the binding of HNF4α and FOXA2 to enhancer regulatory regions in target genes changed during hepatocyte differentiation in a very dynamic manner, and that binding site choice was modulated by the HIPPO–Yap signaling system [57].

Intriguingly, the general transcription factor TFIID has also been involved in the regulation of hepatic genes during embryonic liver development. Among the 14 TATA-binding protein (TBP)-associated factors encompassed in TFIID, TAF10 was shown to be critical for the expression of most hepatocyte-specific genes in embryonic liver, but not in the adult liver [58]. However, it was another TFIID component, TAF4, which proved to be necessary for activation of the postnatal hepatocyte gene expression program. Most interestingly, TAF4 directed the binding of HNF4α to its functional cognate sites in target promoters, and in turn HNF4α was required for the formation of the pre-initiation complex (basal transcriptional machinery including TFIID components) on target transcription start sites. This mutual interaction between HNF4α and TAF4 seems to be critical for the transcription of liver-specific genes during the very early neonatal maturation [59].

Establishment of adult liver features: metabolic competence, hepatocyte polarity and parenchymal functional zonation

The neonatal hepatocytes need to rapidly complete their maturation process to cope with the demands of systemic homeostasis. This involves the progressive acquisition of quiescence, their full metabolic activation, the development of cellular polarity and the porto-central distribution of certain key metabolic pathways (functional zonation). This complex phenotype is achieved to a great extent throughout the developmental process by the core set of liver-enriched transcription factors described above, which reach their peak expression levels at late stages of fetal development. However, multiple evidences obtained in genetic mouse models have shown that many of these core factors, together with a growing network of additional transcription factors, nuclear receptors and more recently mRNA splicing factors, also participate in the preservation of hepatocellular identity during adult life. For instance, it was shown already 20 years ago that C/Ebpa null mice died shortly after birth from hypoglycemia due to impaired expression of glycogen synthase (GS), glucose-6-phosphatase (G6PC) and phosphoenolpyruvate carboxykinase (PEPCK) [60]. These genes, together with other metabolic enzymes such as glucokinase (GK), l-type pyruvate kinase (LPK) and fatty acid synthase (FASN), are characteristic of the mature liver (Fig. 2). Interestingly, C/EBPa can also restrain hepatocellular proliferation [61], which may be attributed in part to its binding and inhibition of CDK2 and CDK4 [62]. HNF1α and HNF1β transcription factors share a high degree of homology and bind DNA as homo- or heterodimers in the promoter regions of many genes encoding serum proteins and metabolic enzymes [63, 64]. Hnf1β null mice fail to form visceral endoderm [36], and targeted deletion of Hnf1β gene in hepatoblasts resulted in severe impairment in bile acid sensing and hepatocyte fatty acid metabolism [65]. Hnf1α knockouts have a milder phenotype, but less than 15 % survive to postnatal day 42, expressing reduced levels of key metabolic enzymes and albumin, among other important serum proteins [66]. In a second model of Hnf1α knockout mice using the Cre-LoxP technique, mice survived longer and developed Laron dwarfism with reduced insulin-like growth factor 1 (IGF1) expression in liver, as well as non-insulin-dependent diabetes mellitus [67]. Early observations in mice lacking FOXA2 and FOXA3 showed an impaired glucose metabolism [35], while subsequent findings in mice with hepatocyte-specific FoxA2 deletion uncovered a fundamental role for this gene in bile acid homeostasis [68].

As previously mentioned, HNF4α is an important factor in early embryonic development. Hnf4α null embryos do not survive past 12 days post coitum (d.p.c), and their livers showed diminished expression of key transcription factors like HNF1α and other metabolic genes characteristic of the mature liver [50]. To evaluate the role of HNF4α at later stages in hepatogenesis, a conditional knockout mouse in which Hnf4α was deleted in the fetal liver by 15 d.p.c was developed [69]. By 18.5 d.p.c, these mice displayed strong impairment in the expression of glucose metabolism genes and lacked glycogen storage. Perhaps, one of the most remarkable observations in the livers of these embryos was the profound derangement in epithelial morphogenesis. The expression of key genes involved in the formation of cell junctions was dramatically reduced, and consistently normal cell contacts were absent in HNF4α-deficient livers [69]. These findings illustrate a fundamental role for HNF4α in the establishment of a sinusoidal architecture essential for adult liver function. These observations were also in accord with previous in vitro studies showing that re-expression of HNF4α in a dedifferentiated hepatoma cell line induced the formation of cell junctions and the expression of hepatocyte marker genes [70]. Interestingly, β-catenin deletion in the developing mouse hepatoblasts also resulted in the downregulation of HNF4α and c/EBPa expression, decreased numbers of tight junctions and the loss of cell polarity [43]. These observations attest to the complex nature of the mechanisms underlying the establishment and maintenance of epithelial organization of the liver parenchyma. A comprehensive review of this topic beyond the involvement of β-catenin and HNF4α has been recently published [1].

The central role for HNF4α in the maintenance of postnatal hepatocyte differentiation became evident upon conditional adult liver-specific deletion of the gene. In these mice, floxed Hnf4α was deleted with an albumin-Cre transgene, and by 45 days of age profound hepatic morphological and functional alterations were observed. These included centrilobular hypertrophic hepatocytes, intrahepatic lipid accumulation, reduced serum cholesterol and triglyceride levels, and markedly elevated circulating concentrations of bile acids [71]. Accordingly, the expression of multiple genes involved in lipid and lipoprotein metabolism, and in lipid and bile acids transport, as well as that of the transcription factors peroxisome proliferator-activated receptor α and γ (Pparα and Pparγ), was markedly deranged [71, 72]. Further in vitro experiments, also validated in these Hnf4α null mice, evidenced the pivotal role of this transcription factor in the preservation of epithelial and polarized phenotype of adult hepatocytes. HNF4α was found to actively repress the expression of master regulatory genes of the epithelial-to-mesenchymal transition (EMT) program in hepatocytes [73]. A complex mutual interaction between HNF4α and SNAIL, a key regulator of the EMT program, influencing the levels of specific stemness regulatory miRNAs was subsequently characterized [74]. These observations, together with the previously discussed involvement of HNF4α in fetal hepatic development, unambiguously demonstrate a central role for this factor in liver differentiation and homeostasis. However, it is clear that HNF4α regulatory functions change throughout development and maturation up to the establishment of the adult hepatocellular phenotype. Previously we alluded to the dynamic interaction of HNF4α with different enhancer regulatory regions in target genes during development [57]. Another important, but less characterized aspect that can significantly influence HNF4α function is the relative expression levels of its multiple isoforms. At least nine HNF4α isoforms can be generated through the usage of two alternative promoters (P1 and P2) and the alternative splicing of transcripts derived from these promoters [75, 76]. Hnf4α1 and Hnf4α7 are two prototypical isoforms derived from the P1 and P2 promoters, respectively, and diverge only in their N-terminal domains, with Hnf4α7 lacking one key activation function motif [77]. Through alternative splicing, Hnf4α1 and Hnf4α7 can give rise in turn to the HNF4α2 and HNF4α8 isoforms, respectively, that differ in ten amino acids in their C-terminal domains [78]. The structural differences between HNF4α isoforms are known to affect their interaction with transcriptional coregulators in a qualitative and quantitative manner, and thus strongly influence their gene-regulatory properties both in vitro and in vivo [78–82]. Most important from our current perspective is the fact that transcripts originating from the two different promoters are sequentially expressed during development. Indeed, P2 promoter-derived Hnf4α7/α8 isoforms are present throughout development, while P1 promoter-driven Hnf4α1/α2 isoforms are increased at birth [76]. Identifying the isoform-specific gene-regulatory activities of HNF4α variants is therefore essential to fully understand the role of this master regulator of liver development and function [82]. Furthermore, a switch in Hnf4α promoter usage from P1 to P2 has been reported in HCC cells, and this may have implications in cancer progression [83, 84].

In addition to the core set of transcription factors that determine hepatocyte maturation during development and preserve the hepatic phenotype in the postnatal liver, there is a large group of transcription factors that coordinate adult liver metabolism and function in a dynamic fashion. These factors include cAMP response element-binding factors (CREB), Kruppel-like factor 15 (KLF15), sterol regulatory element-binding protein 1 (SREBP1) and additional nuclear receptors like PPARα, PPARγ, PXR, LRH1, constitutive androstane receptor (CAR), liver X receptor (LXR), small heterodimer partner (SHP) and farnesoid X receptor (FXR) among others. These factors are involved in carbohydrate, lipid and bile acid metabolism, and also in the response and metabolism of xenobiotic compounds (reviewed in [85]). Their dysregulation, as that of other core factors such as HNF4α, may participate in the impairment of hepatic function in chronic liver injury as will be discussed later.

One less characterized aspect of hepatocellular maturation is the repression of a number of genes expressed in the fetal period. This process is complementary to the developmental acquisition of transcriptional competences described above. Understanding the mechanisms underlying this response may be of pathological significance, since the reactivation of fetal gene expression is increasingly recognized as a hallmark of the tumoral liver, the so-called oncofetal program. Prototypical oncofetal genes include those coding for the serum protein AFP, glypican-3, the miRNA precursor H19 and others such as methionine-metabolizing gene MAT2A, hexokinase 2 (HK2), pyruvate kinase M2 (PKM2) and Wilms’ tumor 1 (WT1) [5, 8, 85, 86] (Fig. 2). The detection of these genes may have potential value as biomarkers and predictors of the clinical course of the neoplastic disease [87]. However, in certain cases, their reactivation already occurs in pre-neoplastic liver lesions [10, 88], and some of them may have profound functional implications in the carcinogenic process [86, 89, 90] (Fig. 2). The mechanisms involved in the postnatal repression of fetal genes are not completely known. The zinc finger and homeoboxes factor 2 (ZHX2), whose expression is higher in the adult liver than in fetal liver, has been shown to inhibit the postnatal expression of AFP, glypican-3 and H19 in mice [91, 92]. Interestingly, expression of ZHX2 is reduced in HCC cells and tissues, and low ZHX2 levels were mechanistically linked to the upregulation of glypican-3 expression and tumor growth [93, 94].

As introduced above, a fundamental trait of the adult liver parenchyma is the heterogeneous distribution of metabolic activities along the porto-central axis. This separation of opposing or complementary functions is called liver zonation, and is essential to preserve metabolic homeostasis [95]. Serum protein synthesis occurs throughout the parenchyma, with albumin production being increased in the periportal zone [95]. However, glucose metabolism shows a clear zonal pattern, with gluconeogenesis taking place in the periportal area and glycolysis in the perivenous space [96]. Other zonated metabolic processes include lipogenesis, bile acid synthesis, xenobiotic metabolism and glutamine synthesis, which occur mostly or completely in the perivenous zone, while ammonia synthesis, fatty acid degradation and cholesterol synthesis take place in the periportal area [97, 98] (Fig. 1). Zonation patterns can be stable or dynamic. Those that do not change under variable hormonal or nutritional conditions are considered stable, like glutamine synthesis. However, most activities display a gradient-like distribution throughout the porto-central axis and a dynamic fluctuation adjusting to the physiological state. A number of signals and mediators including oxygen, metabolites, hormones such as insulin and glucagon, nervous input, circadian signals and morphogens like WNT factors have been proposed [97, 99, 100]. These patterns are defined by the differential expression of the genes coding for the various zonated enzymes and are established postnatally, as zonation in mice is not observed earlier than in the first week after birth [101]. Transcriptional mechanisms seem to be essential for liver zonation [102], and among the different gene-regulatory systems the WNT/β-catenin pathway plays a central role [99, 103]. Active β-catenin is found in pericentral hepatocytes, while that of its negative regulator, adenomatous polyposis coli (APC), is detected in periportal parenchymal cells [104]. Moreover, mice with liver-targeted inactivation of Apc display pan-lobular β-catenin signaling and the activation of a pericentral gene expression program throughout the liver. Conversely, forced expression of the WNT inhibitor Dickkopf-1 prevented this effect [104]. Moreover, a recent study using mice with liver-specific deletion of the WNT coreceptor low-density lipoprotein-related proteins 5 and 6 (Lrp5/6 null mice) demonstrated that WNT signaling is the major upstream effector of β-catenin activity in pericentral hepatocytes [105]. Although β-catenin is indeed a master regulator of liver zonation, there are other transcription factors, such as HNF4α that also play an important modulatory role. Early observations indicated that HNF4α is almost homogenously expressed throughout the liver lobule [106], although perhaps this point should be reevaluated given the existence of a broad diversity of HNF4α isoforms as described before [76, 97]. Nevertheless, it was observed that in hepatocyte-specific Hnf4α null mice, periportal hepatocytes reexpressed a subset of perivenous genes [107]. Further in vitro experiments in fetal rodent hepatocytes differentiated toward an adult hepatocyte phenotype demonstrated that pharmacological activation of the WNT pathway triggered the expression of the perivenous program [108]. In this model, it was found that LEF1, a member of the TCF/LEF family of transcription factors that bind consensus WNT response elements in target genes promoters, interacted with HNF4α to induce glutamine synthase gene expression, while HNF4α alone repressed the expression of this gene. On the other hand, it was shown that HNF4α by itself could activate periportal promoters in the periportal area, while β-catenin in the perivenous zone promoted the replacement of HNF4α by LEF1 leading to the silencing of these genes [108]. The interaction between the β-catenin system and HNF4α was later extended to other members of the LEF family such as TCF4 [109]. TCF4 could bind HNF4α-responsive elements in genes positively regulated by this nuclear receptor, allowing gene expression in the periportal region. Conversely, in perivenous hepatocytes, β-catenin would prevent HNF4α from binding to its regulatory sites in target gene promoters [109]. These findings illustrate the complex and still not fully understood molecular mechanisms regulating liver metabolic mosaicism. Important unresolved issues include the identity and source of the WNT ligands that orchestrate liver zonation and the mechanisms regulating periportal APC expression [97, 100, 103].

Regulation of hepatocellular quiescence and proliferation

As introduced before, under normal conditions liver parenchymal cells are mostly quiescent, with less than 1–2 % of hepatocytes in the cell cycle at a given time and the rest remaining in a resting state (G0) [19, 110, 111]. The mechanisms involved in the maintenance of cellular homeostasis in the liver have been the subject of intensive research for many years. One early model, called the “streaming hypothesis”, proposed that hepatocytes by the portal tract had increased replicative potential, and the progeny of these cells would stream in a porto-central fashion in such a way that eventually all the lobule would be derived from this periportal population [112]. This hypothesis was recently supported by a lineage-tracing study of SOX-9-positive biliary cells in intrahepatic bile ducts, showing that labeled cells in the normal adult mouse liver expanded rather quickly (weeks to months) across the lobule in a porto-central direction, suggesting that almost all lobular cells would be derived from a biliary compartment [113]. However, a number of different labeling and tracing approaches looking at both hepatocytes and biliary epithelial cells that could eventually give rise to hepatocytes (reviewed in [114] ) do not sustain this streaming liver model, suggesting that the above-mentioned SOX-9 labeling studies should be interpreted more carefully [115]. Therefore, the current belief is that under normal conditions (i.e., in the absence of noxious stimuli), the cellular population of the liver is maintained by the replication of existing cells and progenitor cells have little implication in this maintenance liver remodeling.

In contrast to the quiescence of the normal healthy liver, upon organ injury, the loss of functional parenchymal mass and the ensuing inflammation trigger an extraordinary regenerative response. As we mentioned, this response likely evolved as a protective mechanism given the first-line exposure of the liver to environmental toxins coming from the diet, and the fundamental role of this organ in the biotransformation of potentially harmful endo- and xenobiotics. Liver injury may be triggered acutely by drugs, alcohol, toxins and by the surgical removal of liver mass. When this occurs on a healthy organ, hepatocytes rapidly reenter the cell cycle and proliferate to restore the original liver mass [17, 24]. Hepatocellular damage and regeneration also occur under chronic disorders, such as chronic viral infection, persistent alcohol consumption and metabolic conditions such as obesity and impaired iron management (hemochromatosis) [16]. Upon persistent injury, hepatocytes also proliferate up to a point when their replicative capacity becomes exhausted. Under these conditions, a ductular reaction takes place in portal areas from which progenitor cells emerge and are believed to differentiate into hepatocytes allowing liver regeneration [114, 116–118]. Moreover, a recent study demonstrated that in response to injury, hepatocytes can undergo reversible ductal metaplasia, expand as ducts and subsequently contribute to parenchymal restoration [119]. Nevertheless, the involvement of liver stem cells in regeneration during chronic liver injury has been also recently challenged. Genetic fate tracing experiments in adult mice with chronic liver injury failed to detect hepatocytes derived from biliary epithelial cells or mesenchymal liver cells; in fact, no hepatocytes were detected that were not derived from pre-existing hepatocytes [120]. The extent to which these stem cells, or other progenitor cells contribute to liver repair and hepatocellular regeneration is still a matter of debate and active research [116, 121, 122].

The mechanisms involved in acute liver regeneration have been extensively studied for more than 80 years. Most information has been gathered in mammalian models of partial hepatectomy (PH, i.e., the surgical removal of 2/3 of the liver mass), in which a broad range of regulatory signals generated within and outside the liver have been identified [25, 123–125]. After liver resection, virtually all hepatocytes rapidly abandon the quiescent state and enter the cell cycle in a synchronized manner, dividing once or twice until the original cell number is recovered. The liver to body mass ratio is precisely regulated in all examined species and, when this parameter is restored by compensatory parenchymal growth after liver resection, the regenerative process stops [18]. This observation suggests the existence of a master regulator of liver to body mass ratio that has been referred to as the “hepatostat”, the nature of which has not been resolved yet [25, 126]. Almost 10 years ago, to better understand the complex regenerative process, Nelson Fausto and colleagues classified the genes and mediators involved into three major and interconnected networks: cytokine, growth factor and metabolic [126, 127]. Based on a wealth of studies, including genetically modified mouse models, it was postulated that the innate immune system would “prime” hepatocytes to respond to the mitogenic action of growth factors, while fluctuations in the levels of key metabolites such as glucose, fatty acids and triglycerides, amino acids and its derivatives such as S-adenosylmethionine, and particularly bile acids would contribute to the onset and termination of the regenerative response after partial liver resection [128–137]. Nuclear receptors such as PPARα, PPARγ, CAR, LXR and the bile acid receptor, FXR, are increasingly recognized as modulators of hepatocyte proliferation during liver regeneration. These receptors are transcription factors that bind small lipophilic molecules, such as bile acids, steroid hormones, dietary lipids and xenobiotics, and trigger gene-regulatory mechanisms mediating metabolic adaptations as well as mitogenic signals. They could be part of the metabolic network regulating liver regeneration acting as sensors and transducers of the fluctuating concentrations of metabolites in the regenerating liver [138]. Interestingly, ligand-mediated activation of some of these nuclear receptors, such as FXR and CAR, can induce a marked hepatoproliferative response in a quiescent liver, attesting to their high pro-mitogenic potential [135, 139, 140]. An extensive revision of the signals and mediators of hepatocellular proliferation after PH is outside the scope of this article, and readers are referred to recent or more focused reviews on this particular topic [17, 23, 126, 133, 141]. Nevertheless, it should be pointed out that not a single component/gene of each of the three networks mentioned above seems to be absolutely essential for the completion of liver regeneration after PH. Although delays in DNA synthesis can be observed, a complete regeneration blockade, or 100 % lethality, has not been found in any genetically modified mouse model [25, 127]. These observations attest to the high degree of redundancy among these networks, a situation that guarantees the efficacy of the liver reparative and regenerative response after PH and also under different injury conditions.

One fundamental aspect of the regenerative response is that liver function needs to be preserved while hepatocytes replicate to maintain systemic homeostasis and survival. Moreover, some essential activities like gluconeogenesis have to be even enhanced to compensate for the loss of hepatic mass [141]. The mechanisms that allow the maintenance of differentiated liver functions during a period of rapid cell proliferation are not well understood. In this regard, the expression of key liver-enriched transcription factors that participate in hepatocellular differentiation has been examined in regenerating rodent livers. Earlier studies showed that expression levels of HNF1, HNF4α, HNF6, FOXA1 and FOXA2 remained essentially unchanged after PH [142–145]. However, the expression of C/EBPa and C/EBPb is markedly affected upon liver resection. C/EBPa levels are significantly reduced, while those of C/EBPb are upregulated early after the intervention [143, 146]. These changes appear to be important for normal hepatic regeneration, as C/Ebpb null mice have defective glucose homeostasis and display impaired liver regeneration [147, 148], and C/EBPa may restrain hepatocellular proliferative activity [3]. Nevertheless, cooperative interactions between liver-enriched transcription factors need to occur during regeneration to allow cell proliferation while maintaining liver function. Therefore, if indeed C/EBPa-deficient hepatocytes show enhanced proliferation, it is also true that C/EBPa null mice fail to express hepatic gluconeogenic enzymes [60].

As noted above, the expression of most key liver-enriched transcription factors is not affected during hepatocellular proliferation. Their sustained levels after PH certainly contribute to preserve liver function under highly demanding metabolic conditions. This may be particularly important for HNF4α, given its central role as a master regulator of hepatocyte differentiation. However, two recent studies in which HNF4α expression was acutely downregulated in the adult mouse liver indicate that this nuclear receptor exerts a potent repressive effect on hepatocellular proliferation in the differentiated liver [149, 150]. Knockdown of Hnf4α by conditional genetic deletion reproduced the impairment in lipid metabolism found in mice with postnatal ablation of Hnf4α (Alb-Hnf4α-/- mice) [71], but also resulted in a potent hepatocellular proliferative response [149, 150]. Hepatocyte DNA synthesis and proliferation in these short-term knockout models were accompanied by profound changes in gene expression, which consistently included the induction of cell cycle and proliferation-associated genes. Interestingly, some of these genes were shown to be negatively regulated by HNF4α [150], which was in agreement with previous in vitro reports that identified cell cycle-related genes as direct HNF4α targets [151]. These findings demonstrate a potent antiproliferative effect of HNF4α that may be part of a tumor suppressor role for this gene, whose expression is otherwise downregulated in hepatocarcinogenesis [10, 152]. However, as indicated, HNF4α expression is not downregulated during liver regeneration after PH, suggesting that additional mechanisms must exist to allow cell cycle progression in the presence of normal HNF4α levels. These mechanisms might include the activation of specific signals during liver regeneration that could dictate the differential binding of HNF4α to its cognate regulatory sites in alternative gene promoters. One of these signals modulating HNF4α activity could be the WNT/β-catenin pathway. β-Catenin is activated very early after PH, and liver regeneration is impaired in hepatocyte-specific β-catenin knockout mice and when WNT signaling is prevented [105, 153]. As previously mentioned, a functional cross talk between HNF4α and the β-catenin system participates in the regulation of parenchymal liver zonation [103]. This suggests that similar interactions could operate during liver regeneration to facilitate hepatocellular proliferation, an issue that deserves further examination. Another factor that could modulate HNF4α activity during liver regeneration is cyclin D1. In addition to its well-known role in driving cell cycle through its interaction with cyclin-dependent kinases such as CDK4, cyclin D1 has been reported to control gene transcription in a CDK-independent manner. A recent study demonstrated that cyclin D1 was able to inhibit lipogenic gene expression in hepatocytes, and that this effect was due at least in part to the repression of HNF4α activity, to which cyclin D1 can bind via a “repressor domain” involved in the repression of other nuclear receptors [154]. We could speculate that besides lipogenic genes, this interference of cyclin D1 with HNF4α activity could also operate on the promoters of pro-mitogenic genes that are transcriptionally repressed by the nuclear receptor [150].

An additional system that may influence HNF4α promoter binding site choice during liver regeneration is the Hippo signaling pathway. We previously mentioned that this signaling system modulates HNF4α interaction with gene enhancers during hepatocyte differentiation, influencing enhancer selection and gene expression [57]. It has been recently demonstrated that Hippo/YAP is activated early during liver regeneration, possibly contributing to hepatocellular proliferation, thus suggesting that HNF4α transcriptional activity could be also modulated by this signaling pathway [155]. In addition to this interaction with HNF4α, the Hippo signaling system plays a fundamental role in the preservation of hepatocellular quiescence and differentiation on its own. The Hippo pathway, originally described in Drosophila, is highly conserved between species and is involved in the regulation of organ size, with a critical role in the liver [156]. Activation of this pathway is initiated by cell–cell contact and physical interactions with the extracellular matrix, although ligands of G protein-coupled receptors (GPCRs) and the epidermal growth factor receptor have also been shown to regulate Hippo pathway activity [157, 158]. In mammals, this pathway includes the serine/threonine kinases MST1/2 (mammalian STE-2-like kinases) and their downstream target kinases LATS1/2 (large tumor suppressor 1/2), which mediate the direct phosphorylation of the transcriptional co-activator YAP. Phosphorylation of YAP leads to its cytosolic retention and ubiquitin–proteasome-dependent degradation [157]. However, when YAP accumulates in the nucleus, it interacts with transcription factors, such as TEAD factors, and activates the expression of a wide variety of growth-related target genes [158, 159]. Overexpression of YAP in the liver of adult mice leads to overproliferation of hepatocytes and a dramatic increase in liver size [160]. Moreover, when YAP overexpression was maintained for longer periods of time, mice developed liver tumors [156]. The manipulation of different components of the Hippo pathway, such as the inactivation of MST1/2 kinases or the deletion of neurofibromatosis-2 (Nf2) gene involved in LATS1/2 activation, promote YAP nuclear accumulation and also result in hepatomegaly and liver cancer development [161–163]. A recent study has unveiled a novel biological activity of the Hippo/YAP pathway in mouse hepatocytes. Inducible and hepatocyte-specific overexpression of YAP resulted in rapid hepatocellular growth as expected; however, YAP activation also led to hepatocellular dedifferentiation to a ductal/progenitor-like fate, and this occurred in a significant proportion of hepatocytes suggesting that most mature hepatocytes harbor this capacity. Activation of NOTCH signaling by Hippo/YAP was involved in this hepatocyte-to-ductal dedifferentiation response [164]. Interestingly, liver YAP/TEAD transcriptional activity has been already observed in the early stages of experimental hepatocarcinogenesis, as well as in human early dysplastic nodules [165]. Furthermore, a very recent report demonstrated that targeting Yap with siRNA-lipid nanoparticles in a genetic mouse HCC model restored hepatocyte differentiation and resulted in pronounced tumor regression [166]. Collectively, these reports indicate that the Hippo/YAP pathway plays a fundamental role in the preservation of hepatocellular quiescence and liver cell fate, and therefore its precise regulation is essential for the maintenance of hepatic homeostasis.

Loss of hepatocellular identity and quiescence during chronic liver injury and carcinogenesis

Chronic liver injury and inflammation lead to the development of liver cirrhosis, a condition affecting 1–2 % of the world’s population that causes more than 1 million deaths annually [167, 168]. Liver cirrhosis involves the progressive deterioration of hepatic architecture, with the substitution of the parenchymal component by abundant extracellular matrix (ECM) and the appearance of regenerative hepatocellular nodules [169]. Importantly, the metabolic zonation characteristic of the healthy liver parenchyma is also lost in cirrhotic nodules [170], an alteration that is related to another hallmark of liver cirrhosis, which is the development of hepatic insufficiency. Loss of hepatic functions is a strong indicator of the progression of cirrhosis and, as previously mentioned, cirrhosis is also the major driver of HCC development. Accordingly, in the clinical setting, the Child–Pugh classification, a score system which principally includes biochemical indicators of liver function, is the most consistent and robust predictor of patients’ death due to cirrhosis [171]. However, accumulating evidence suggest that hepatic insufficiency cannot be only ascribed to the loss of liver parenchyma and the circulatory alterations caused by fibrosis, but also to the gradual decrease in the expression of hepato-specific genes [8, 10, 86, 172]. These alterations in the cirrhotic liver are frequently accompanied by unrestrained cell proliferation and the upregulation of genes characteristic of fetal and transformed hepatocytes, such as WT1, gankyrin and as recently reported the HCC marker glypican-3 [10, 13, 14, 88, 173] (Fig. 2). Experimental and clinical observations indicate that loss of hepatocellular identity may indeed be related to the deterioration of liver function reflected in the Child–Pugh score and to the development of HCC. Expression profiling studies in peritumoral cirrhotic liver tissues identified a gene signature with strong potential for survival prognosis [12]. Furthermore, this same gene signature recently allowed the classification of early-stage cirrhosis patients into prognostic subgroups with increasing risk of disease evolution, including decompensation, Child–Pugh class progression, HCC development and death [174]. Importantly, in addition to genes related to inflammation and cell proliferation, this genetic signature included changes in genes indicative of fetal and dedifferentiated status [12].

In spite of all these evidences, the molecular mechanisms driving hepatocellular dedifferentiation during chronic liver disease and hepatocarcinogenesis are not completely known. Impaired activity and expression of some of the previously discussed liver-enriched transcription factors may certainly play a fundamental role in this pathological process. For instance, early reports showed marked downregulation of C/EBPb and FOXA2 (HNF3β) expression in the liver of cirrhotic rats [175]. FOXA2, which is required for normal bile acid homeostasis, was later found reduced in the liver of patients with cholestatic syndromes and shown to contribute to disease progression [68]. HNF1α expression is also reduced in experimental liver cirrhosis and hepatocarcinogenesis, as well as in human HCC tissues [152, 176–178]. Moreover, adenoviral delivery of HNF1α in human HCC cell lines resulted in partial restoration of the liver-specific gene expression profile, reduced cell proliferation and in vivo tumorigenicity [178]. Together, these findings suggest that the downregulation of this transcription factor may thus participate in hepatocellular dedifferentiation and unrestrained proliferation.

One fundamental transcription factor downregulated in liver cirrhosis and hepatocarcinogenesis is the nuclear receptor HNF4α. At variance with the sustained expression of HNF4α during liver regeneration after PH, HNF4α levels have been found consistently reduced in human liver cirrhosis and HCC. A first report identified lower levels of HNF4α transcripts in advanced human liver cirrhosis [10], and this finding was confirmed and extended to HCC tissues in subsequent human and experimental studies [152, 179, 180]. The mechanisms leading to the downregulation of HNF4α expression in liver injury and HCC are likely multifarious. It has been described that HNF4α expression in human liver cells is under the positive influence of HNF1α; therefore, the previously mentioned decay in HNF1α expression in chronic liver injury and HCC could be involved in HNF4α abatement [181]. Loss of HNF4α expression in the cirrhotic and transformed liver may be also mediated by active repressive mechanisms. One such mechanism could be triggered by the pro-fibrogenic cytokine TGFβ through the induction of the transcription factor WT1. WT1 is not expressed in the mature healthy liver, but becomes reactivated in the cirrhotic and tumoral hepatocytes, inversely correlating with HNF4α mRNA levels [10, 86]. Overexpression of WT1 in hepatocytes led to HNF4α downregulation, while its knockdown in human HCC cells resulted in the reactivation of liver-specific genes, including that of HNF4α [86]. Activation of mitogen-activated protein kinase signaling, which is frequently found in HCC [26], also reduces HNF4α gene transcription by preventing the recruitment of transcription factors and the basal transcriptional machinery to HNF4α-regulatory regions and proximal promoter [182]. Most interestingly, upon cellular stress, induction of P53 can impair P1 promoter-driven transcription of HNF4α in human HCC cells [183], while P53 is also able to repress the transactivation function of HNF4α1 protein on HNF4α target genes [184]. Therefore, the stress-associated upregulation of P53 expression in chronic liver injury may certainly contribute to the inhibition of HNF4α expression and activity [185]. A different mechanism leading to HNF4α downregulation that has been recently exposed involves the oncoprotein gankyrin. Gankyrin was found to bind HNF4α and promote its proteasome-dependent degradation in HCC cells without affecting HNF4α mRNA levels [186]. Accordingly, gankyrin and HNF4α expression levels showed an inverse correlation in experimental and human HCC tissue samples [187]. The recently appreciated induction of gankyrin expression in chronic hepatitis and liver cirrhosis suggests that gankyrin-mediated downregulation of HNF4α protein levels may contribute to the loss of hepatocellular identity from early stages of chronic liver disease up to HCC development [173]. Furthermore, as we previously mentioned, persistent activation of YAP also leads to the deregulation of hepatocellular differentiation and to the development of aggressive HCC characterized by a proliferative signature [166]. Several experimental studies have provided direct evidence on the role of HNF4α in the preservation of hepatocellular identity and quiescence in the context of chronic liver injury and carcinogenesis. For instance, deletion of HNF4α in the liver of adult mice significantly increased the progression of chemically induced tumors, supporting a tumor-suppressor role for this gene [187]. Complementarily, adenoviral delivery of HNF4α improved hepatocellular differentiation, liver function and reduced fibrosis and quelled the growth of xenografted human HCCs or endogenous tumors in rodents [179, 180, 188]. Reintroduction of HNF4α partially recovered hepato-specific gene expression and blocked hepatocyte EMT both in vitro and in vivo during hepatocarcinogenesis, effects that were attributed in part to its ability to inhibit β-catenin activation and signaling [180, 188]. Nevertheless, according to the varied nature of its direct transcriptional targets, the antitumoral effects of HNF4α are likely mediated through multiple mechanisms. In addition to metabolic genes, HNF4α also controls the expression of signal transduction and inflammatory regulators, the cell cycle inhibitor CDKN1A (P21), and as recently shown that of key antitumoral miRNAs [151, 189, 190].

All the evidence summarized above indicates that impaired expression and activity of liver-enriched transcription factors are important determinants in the progression of chronic liver disease. However, in addition to the unquestionable role of these factors, very recent reports have demonstrated that preservation of accurate pre-mRNA splicing is also essential to maintain hepatocellular homeostasis. Moreover, this additional mechanistic layer may be compromised as well in liver disease and thus contribute to the deterioration of liver function. Alterations in alternative splicing of different genes have been described in chronic hepatitis, cirrhosis and in HCC tissues [191]. Dysregulated splicing can be ascribed to mutations affecting splice sites or splicing enhancers or silencers, the activation of cell signaling pathways affecting the splicing machinery, or the abnormal expression of splicing factors [191–193]. One such splice regulator is SLU7, a splicing factor critical for the correct selection of the 3′ splice site during the second step of splicing [194]. The expression of SLU7 is downregulated in chronic liver disease and HCC, and loss of SLU7 in hepatocytes was initially found to promote the aberrant production of a protumorigenic variant of the P73 tumor suppressor [195]. A subsequent study demonstrated that adequate expression of SLU7 is essential to preserve hepatocellular differentiation and quiescence. Knockdown of Slu7 in adult mouse liver resulted in impaired glucose and lipid metabolism, refractoriness to key metabolic hormones, loss of hepatocyte-specific gene expression and the reactivation of a fetal gene expression pattern. Moreover, SLU7 downregulation was accompanied by the induction of pro-mitogenic genes and hepatocyte proliferation in the absence of injury, indicating that this splicing factor is indeed a gatekeeper of hepatocellular quiescence [196]. Importantly, loss of SLU7 expression impaired the gluconeogenic activity of hepatocytes while eliciting a switch to a glycolytic phenotype, reminiscent of that found in tumors [197, 198]. Interestingly, an increase in hepatic glycolytic activity has been recently described both in experimental cirrhosis and in cirrhotic patients in the early stages of the disease. This was regarded as an adaptive response to preserve energy homeostasis in the face of liver injury and failing oxidative phosphorylation [199], and could well be mediated by reduced SLU7 expression. Mechanistically, SLU7 was shown to influence the splicing of Srsf3, an RNA-binding protein and a splice regulator as well, leading to the formation of a degradation-prone splice variant of this factor and its subsequent downregulation. Interestingly, the expression levels of SLU7 and SRSF3 were directly correlated when analyzed in control, cirrhotic and HCC human liver tissues [196]. SRSF3 has been recently reported as a crucial factor in the maintenance of hepatocellular function, and hepatocyte-specific Srsf3 null mice present impaired hepatocyte maturation and profound alterations in glucose and lipid homeostasis by 1 month of age [200]. Aberrant splicing and expression of key genes such as HNF1α, the insulin receptor, albumin and enzymes involved in cholesterol metabolism were detected upon Srsf3 deletion. Marked signs of liver injury and upregulation of oncofetal markers were also found in the absence of SRSF3 [200]. Follow-up of liver-specific Srsf3 null mice revealed the spontaneous development of liver tumors in the context of chronic liver disease between 12 and 24 months of age [201]. Together, these findings suggest that the effect of SLU7 on Srsf3 expression may be functionally relevant, as the phenotype of the Slu7-depleted mice overlapped to a great extent with that of Srsf3 liver-specific null mice. Nevertheless, SLU7 also exerted additional functions important for hepatocyte homeostasis that exceeded its role as splicing factor, as it was identified as a coactivator in the cAMP pathway regulating cAMP-CREB-mediated gluconeogenic gene transcription [196]. At variance with SRSF3, SLU7 depletion did not result in liver injury, and therefore the phenotypic changes found upon its inhibition were not secondary to a situation of parenchymal damage. Interestingly, also different from SRSF3, changes in SLU7 levels also had a profound impact on HNF4α expression in hepatocytes. Transcription of HNF4α isoforms driven by the P1 promoter was reduced upon Slu7 knockdown, while P2 promoter activity was markedly enhanced. As previously discussed, P2-derived HNF4α isoforms are characteristic of the fetal and transformed liver and code for HNF4α variants with reduced transactivation potential [76, 77, 83]. The mechanisms underlying HNF4α P1/P2 promoter usage switch upon SLU7 reduction are not known. They may be indirectly mediated through the upregulation of WT1 gene expression, a strong repressor of HNF4α which is also found induced in SLU7-depleted livers [10, 196]. Nevertheless, further studies into this important response are warranted. Be that as it may, collectively, these findings identify a fundamental role for mRNA splicing regulators such as SRSF3 and SLU7 in the preservation of liver identity, adding another layer of complexity to the mechanisms governing hepatocellular homeostasis.

Translational perspectives

Although our knowledge of the mechanisms involved in the preservation of hepatocellular quiescence and differentiation is far from complete, the evidence presented above allows the identification of a number of key molecular players. From a translational point of view, this information could be applied to design new preventive and therapeutic strategies for chronic and neoplastic liver disease. We have discussed the important role of nuclear factors such as HNF1α and HNF4α in the establishment of the mature hepatic phenotype during development, the consequences of their downregulation in the mature liver and their impaired expression in cirrhosis and HCC. Ideally, restoration of HNF1α and HNF4α levels in the chronically injured liver could help to preserve liver function and even forestall tumor development or progression. Theoretically, this could be achieved by gene therapy strategies with viral vectors harboring the desired cDNA [202]. In support of this possibility are the previously alluded studies showing how adenoviral gene transfer of HNF1α and HNF4α in rodent models partially restored liver-specific gene expression, hepatocellular function and even attenuated fibrosis and HCC development [178–180, 188]. Moreover, delivery of HNF4α to rats with established cirrhosis also resulted in the reversal of fibrosis and the restoration of some functional parameters [203]. Nevertheless, this beneficial effect was only fully observed in animals at early stages of fibrosis and not in those with advanced disease [203]. Such difference may be attributed to the difficulty of resolving thick and cross-linked fibrous septa found in progressed cirrhosis [204]. However, efficient transduction of the cirrhotic liver with adenoviral viral vectors may also be hampered by collagen deposits [205]. This situation might be circumvented using other types of viral vectors such as adeno-associated viruses (AAV) [206], albeit delivery of the therapeutic factor to a significant proportion of the parenchyma would still be needed to influence overall liver function and even more to prevent HCC. In this regard, an interesting study recently showed that AAV vector-delivered HNF4α to rats with decompensated cirrhosis reversed terminal chronic hepatic failure. Apparently, this effect only required transduction of a modest fraction of hepatocytes, and improvement in hepatic function began within 48 h after infection and lasted at least for 100 days [207]. Nevertheless, administration of AAV vectors to patients with chronic liver disease is not exempt from risks [208], and also strategies to avoid the immune response to AAVs found in humans need to be optimized [209].

An alternative approach would be to reactivate the expression of endogenous liver differentiation factors, including not only HNF1α and HNF4α, but also the splicing regulators discussed above, SLU7 and SRSF3. Understanding the mechanisms leading to the inhibition of the expression of these important factors might provide novel targets for intervention. For instance, microRNAs, miR-24 and miR-629 [190] as well as miR-34a [210], have been found to potently repress HNF4α expression in HCC cells, suggesting that interference with the expression or activity of these miRNAs could somehow restore HNF4α levels and inhibit HCC growth. On the other hand, recent reports have identified mi-RNAs that are physiological targets of hepatocyte nuclear factors and whose expression is also reduced in chronic liver injury and hepatocarcinogenesis. Interestingly, in vitro and in vivo studies demonstrated that mi-RNAs such as miR-194, or miR-124 and miR-134, mediate to a significant extent the antitumorigenic effects of HNF1α and HNF4α, respectively [178, 190, 211]. Together, these studies support the potential of mi-RNA-based therapies in chronic liver disease and cancer, either by reactivating the expression of master regulators of mature liver phenotype or by mimicking their effects [212]. Nevertheless, many important issues need to be solved before a safe and efficacious application of these mi-RNA therapies can be implemented, including a comprehensive knowledge of all potential targets of the mi-RNA to be overexpressed or downregulated to avoid unwanted effects. Also, the mi-RNA delivery strategies need to be improved to optimize stability, biodistribution and interaction with the endogenous RNA machinery [213]. In this respect, mi-RNA delivery approaches based on AAV vectors with serotypes showing enhanced liver targeting are worth further explorations [214].

Chemopreventive strategies with agents that preserve hepatocellular differentiation and inhibit liver carcinogenesis are also being considered. Perhaps, the most studied molecule is s-adenosylmethionine, the universal methyl donor and an endogenous metabolite of methionine synthesized at high levels in hepatocytes [215]. Fifteen years ago, s-adenosylmethione was already described as a “hepatotrophic” factor based on observations performed in cultured hepatocytes. It is well known that primary hepatocytes rapidly lose liver-specific gene expression and reactivate a genetic program characteristic of fetal and even transformed hepatocytes. s-Adenosylmethione treatment preserved liver-specific gene expression, importantly including that of HNF4α, and maintained hepatocellular quiescence [10, 128, 216–218]. The pre-clinical and clinical evidences of the reduction of s-adenosylmethionine levels in chronic liver injury, as well as data on its hepatoprotective and chemopreventive effects, are abundant and readers are referred to more specific reviews [215]. Based on all this information, large clinical studies focussed on defined patient groups and with better-demarcated end points assessing the hepatoprotective effects of s-adenosylmethionine are warranted [219].

Interference with dysregulated pathways that have been related to hepatocellular dedifferentiation, uncontrolled proliferation and neoplastic conversion could be also an interesting approach to preserve liver homeostasis. One paradigmatic example would be the HIPPO/YAP signaling pathway. As previously summarized, under normal conditions YAP is phosphorylated by a tumor suppressor kinase cascade leading to YAP cytoplasmic retention and degradation. When YAP is overexpressed, or its phosphorylation impaired, it translocates to the nucleus and acts as a co-activator of TEAD transcription factors, leading to cellular dedifferentiation, growth and ultimately HCC development [160, 164]. A recent in vitro screening of US FDA-approved small molecules identified several compounds, all belonging to the porphyrin family, which disrupted the physical interaction between YAP and TEAD and blocked YAP/TEAD-mediated gene expression [220]. Most interestingly, these compounds inhibited YAP-induced hepatomegaly without affecting liver size in control mice and reduced preneoplastic liver foci formation in chemically induced rats [165, 220].

Finally, the cellular and molecular knowledge generated over the past decades on liver development may also have important implications for the generation of new cell-based therapies for liver failure and congenital liver diseases [118, 221]. Although the topic is beyond the scope of this article, among the different strategies the development of hepatocytes from mouse and human fibroblasts is worth mentioning. These strategies produced functional hepatocyte-like cells through the expression of defined transcription factors such as HNF1α, HNF4α, FOXA1, FOXA2 and GATA4 [222, 223], or through the generation of multipotent progenitor cells that were subsequently differentiated into hepatocytic cells upon stimulation with factors known to drive hepatocellular differentiation, such as OSM, FGF-2 and HGF, among others [224]. Importantly, when transplanted into mice, these fibroblast-derived hepatocyte-like cells were able to repopulate the liver and protect from acute and/or chronic liver injury [222–224].

Conclusion

Although far from comprehensive, the evidence summarized in this article highlights the extraordinary plasticity of the liver. This mostly quiescent organ is capable of undergoing growth when required, while maintaining its fully differentiated functions to preserve systemic homeostasis. The mechanisms underlying such “acrobatic” behavior are currently being elucidated and this knowledge will surely be translated in the near future to clinical applications. The fundamental genes involved in the control of liver quiescence and identity, most of them discussed here, not only constitute targets to prevent or treat liver disease. The sometimes underappreciated central position of the liver in systemic health makes these investigations even more relevant, and liver disease is increasingly being related to global health problems such as cardiovascular disease and diabetes [225, 226].