Since its discovery in the late 1970s, hepatocyte heterogeneity and the resulting metabolic zonation of liver parenchyma has remained a fascinating and challenging issue of liver research (for review, see Gebhardt,1992a; Jungermann and Kietzmann,2000). According to the concept of metabolic zonation the various functions of the adult liver are spatially separated in different regions of the smallest structural units of the liver, the lobules. This spatial separation occurs in such a way that interdependent metabolic pathways (e.g., fatty acid synthesis and major NAPDH-producing reactions) are colocalized in overlapping zones to allow synergistic function, while opposing pathways (e.g., gluconeogenesis and glycolysis) are spatially separated in different zones to avoid interference and energy wastage. To achieve this complex topology of metabolic pathways, the expression of key enzymes of these pathways needs to be regulated by sophisticated control mechanisms.
For a long time, the basic structure of this complex regulatory network remained obscure, but recent research has broken the seal of the treasury. We are now beginning to understand the contours of a regulatory universe in the nutshell of the tiny liver lobuli, where topology meets combinatorics and dynamics. It emerges that regulatory mechanisms designed to govern embryonic development and tissue patterning also control the hepatocyte functional divergence that underlies the topological distribution of metabolism in adult quiescent liver. At the same time, these regulatory mechanisms hold keys for unlocking hepatocyte proliferation during liver regeneration, and for preventing the development of hepatocellular carcinoma. In particular, Wnt/β-catenin signaling has emerged as a central component of this network contributing in a variety of ways to the control of all of these processes (Burke and Tosh,2006; Spear et al.,2006; Gebhardt et al.,2007).
In this review, we aim to provide a brief overview of the considerable advances in this field. Although Wnt/β-catenin signaling also contributes in many ways to embryonic liver development, this topic will not be covered, because comprehensive reviews have appeared recently (Burke et al.,2006; Shackel,2007; Nejak-Bowen and Monga,2008). We will focus instead on the adult liver featuring fully developed metabolic zonation, while still maintaining the capacity for proliferation and differentiation. In the central part of this review, we will consider how low-level, sometimes even concealed activities of this pathway show gate-keeper functions with regard to metabolic zonation. By highlighting some examples of the combinatorial cross-talk of Wnt/β-catenin signaling with other signaling pathways and transcription factors, molecular mechanisms potentially contributing to this gate-keeper function will be emphasized. Last, but not least, important regulatory issues requiring future attention will be discussed.
ARCHITECTURE OF ADULT LIVER PARENCHYMA AND METABOLIC ZONATION
Adult liver parenchyma is composed of hundreds of thousands of lobules that can be considered as the smallest structural and functional units (reviewed in Lamers et al.,1989; Sasse et al.,1992; Ekataksin and Kaneda,1999). They appear as berry-shaped cylinders with a near hexagonal cross-section. At the periphery, the smallest branches of the portal vein and the arteria hepatica supply the blood to the sinusoids. The sinusoids are not just an anastomosing capillary network, but are subject to preferential grouping into repeated units, the hepatic microcirculatory subunits which are radially oriented toward the central veins of the lobules. There, the blood is collected and drains by means of the sublobular veins into the hepatic vein. The hepatocytes are arranged in slender cords along the sinusoids spanning the distance from periportal to pericentral. When viewed in three dimensions, they form a sponge-like cellular continuum, the liver parenchyma (Gebhardt,1992a). Thus, although each hepatocyte is part of this continuum, it occupies a unique position along the periportal to pericentral axis. This distance of the sinusoids is only a few cells long; approximately 9 to 10 hepatocytes in mice and almost twice as large in human liver (Ruijter et al.,2004).
Immunohistochemical studies during the last three decades have revealed that different metabolic pathways are heterogeneously distributed within liver parenchyma (reviewed in Jungermann,1987; Gebhardt,1992a). Usually, enzymes or other functional proteins (transporters, transcription factors, etc.) show decreasing or increasing gradients of different steepness over the distance from the portal triads (periportal) to the central vein (pericentral). Most interestingly, metabolic pathways performing opposing functions follow inverse gradients, i.e., are distributed in a more or less complementary manner throughout the parenchyma (Fig. 1). Examples comprise the opposite functions gluconeogenesis and glycolysis, ureogenesis and glutamine synthesis, or cholesterol biosynthesis and bile acid biosynthesis which, for each pair, predominate in periportal and pericentral zones, respectively. However, while some of the distribution gradients may partially overlap and can shift in response to nutritional needs (e.g., carbohydrate metabolism), others may be almost completely separated and seem less dynamic (e.g., ammonia metabolism; Gebhardt,1992a). In consequence, each hepatocyte performs a different spectrum of a continuum of biochemical functions according to its position along the periportal to pericentral axis.
When considering metabolic zonation, this complex pattern is usually neglected and reduced to the simple dualism of periportal and pericentral functions. With respect to possible regulatory mechanisms, however, one should be aware that this view is a vast oversimplification.
Although it has been clear for more than two decades that cell–cell interactions as well as gradients of hormones, cytokines, and nutrients across the portocentral axis of the liver lobules play a role in metabolic zonation (for review, see Gebhardt,1992a,b; Gebhardt and Gaunitz,1997; Jungermann and Kietzmann,2000), the discovery that morphogen signaling by the Wnt/β-catenin pathway acts as a master regulator dominating most if not all other control mechanisms came as a real surprise in 2002. Independently, three groups of researchers demonstrated that glutamine synthetase (GS), a well-known marker of the most distal pericentral hepatocytes (Gebhardt and Mecke,1983), is strongly up-regulated in response to activation of β-catenin (Loeppen et al.,2002; Cadoret et al.,2002; Nicholes et al.,2002). Further proof that β-catenin is required for zonal GS expression came from studies with β-catenin knockout mice. Liver-specific loss of β-catenin in the late embryonic phase resulted in normal mouse development with slightly reduced liver mass (Sekine et al.,2006). The most prominent feature was the absence of GS from liver parenchyma. Likewise, other pericentral functions such as biotransformation by means of the cytochrome P450 enzymes CYP1A2 and CYP2E1 were also lost (Sekine et al.,2006). A role for β-catenin in the expression of cytochrome P450 isozymes has also been suggested by studies of mouse liver tumors (Loeppen et al.,2005). A pioneering study by Benhamouche et al. (2006) using an inducible liver-specific adenomatous polyposis coli (APC) gene knockout mouse model showed that APC knockout led to the activation of β-catenin followed by activation of proliferation, and hepatomegaly as well as a progressive loss of zonation. Specifically, the GS-positive area was extended toward the portal tract while periportal functions such as urea cycle enzymes and phosphoenolpyruvate carboxy kinase 1 (PEPCK1) were reduced or ablated (cf. Fig. 1). This apparent pericentralization of liver parenchyma could be blocked or even reversed by ectopic expressing of the Wnt inhibitor Dickkopf (DKK)1 in the liver (Benhamouche et al.,2006; cf. Fig. 1). Reed and coworkers extended these studies to show that both proliferation and altered zonation are dependent upon β-catenin but on neither APC (apart from interacting with β-catenin) nor cMyc, a downstream target of β-catenin signaling (Reed et al.,2008). Using transgenic mice with a scattered expression of an activated Ha-ras, Braeuning et al. (2007b) provided convincing results that β-catenin signaling is opposed by ERK1/2 activation, consistent with the suggestion that ras signaling in liver might also show a gradient, but inverse to that of β-catenin (Hailfinger et al.,2006). In toto, these results clearly demonstrate that activation of β-catenin is necessary and sufficient for establishing and maintaining liver metabolic zonation in adult mice. They further suggest that changes in zonation are due to altered gradients of upstream regulators of β-catenin and an altered balance with other (periportally dominating) signaling pathways yet to be identified (Gebhardt and Ueberham,2006).
This picture has been challenged strongly by results from MnSOD knockout mice, demonstrating that enhanced oxidative stress leads to the loss of zonal expression of periportal and pericentral genes such as PEPCK1 and GS, which are then expressed in a scattered distribution in single cells or small cell groups (Lenart et al.,2007). This indicates that cell-autonomous mechanisms can drive the expression of these genes in the absence of an organizing gradient. In this context, it should be mentioned that isolated single GS-positive cells persist (a) in primary hepatocyte cultures where cells from all zonal regions are randomly mixed (Gebhardt et al.,1994) and (b) in livers of APC knockout mice after ectopic expression of DKK1 (Benhamouche et al.,2006).
In addition to providing a general explanation for zonation of liver parenchyma, these fascinating discoveries raise several important questions: (1) Which Wnt factors are involved and where do they come from? (2) How do these Wnt factors determine a stable patterning of liver parenchyma? (3) Given the fact that the activity of β-catenin is highest in the pericentral zone and lowest in the periportal zone, how can this activity gradient be demonstrated? (4) How is the β-catenin activity gradient interpreted and translated into zonal gene expression?
Although we are far from definitive answers, certain findings that shed some light on these issues will be discussed in more detail below.
SOURCES AND ACTIONS OF WNT
Screening of different liver cell types for the expression of Wnt factors and Frizzled (Wnt receptor) genes revealed that each cell type seems to express a different set (Zeng et al.,2007). In addition, Wnt species from extrahepatic sites may reach the liver by means of lipoproteins (Neumann et al.,2009). Thus, the liver in general seems to provide a rich “Wnt-space” to which each cell is exposed and which may vary in different nutritional and/or diseased states. Because it is not yet known which Wnt factors are involved in the activation of β-catenin in hepatocytes in normal liver, it is hard to decide whether hepatocytes respond mainly to their own Wnt factors in an autocrine manner or are influenced also by neighboring cells, e.g. endothelial cells. An argument in favor of the latter assumption may be that Wnts produced by endothelial cells and hepatic stellate cells were found to have considerable influence on hepatocyte proliferation and differentiation (Matsumoto et al.,2008). It is not known, however, whether the contribution of Wnt factors of endothelial origin is needed in the adult quiescent liver or predominates only in conditions favoring proliferation.
Concerning the establishment of a Wnt/β-catenin gradient, we propose a model based on the simple reaction-diffusion mechanism of patterning originally developed by Turing (1952) and Gierer and Meinhardt (1972) which has been successfully adapted to describe animal pigmentation patterns and distribution of epidermal appendages (Sick et al.,2006). An adaptation of this model involving Wnts as activators and DKK1 as inhibitor (both of which are produced by hepatocytes) leads to a stable Wnt gradient with the highest activity around the central vein and the lowest activity in the periportal area (Krinner, Drasdo, Gebhardt, unpublished observations). Endothelial Wnt is superfluous in this model and may have some triggering influence only under conditions where the normal pattern is destroyed and must be re-established (e.g., after pericentral damage by carbon tetrachloride).
THE β-CATENIN ACTIVITY GRADIENT: FACT OR FICTION?
According to canonical Wnt signaling, the Wnt gradient should result in a similar gradient of β-catenin activation. However, no such gradient has ever been reported for normal adult liver. At most, some investigators described cytoplasmic and nuclear staining for nonphosphorylated β-catenin in the narrow pericentral zone occupied by GS-positive hepatocytes (Benhamouche et al.,2006; Moriyama et al.,2007), while other laboratories (including our own) have failed to reproduce this finding. Moreover, a reporter mouse driving enhanced green fluorescent protein under T-cell factor (TCF)/β-catenin detected activated β-catenin and TCF-dependent expression only in a few cells close to the central veins, while not even all GS-positive hepatocytes expressed the reporter (Moriyama et al.,2007). Similar observations have been made using the TOPGAL transgenic mice (Hu et al.,2007). Thus, while some evidence exists for activated β-catenin in hepatocytes adjacent to the central veins within part of the narrow zone of ammonia detoxification, no decreasing gradient toward the periportal zone is apparent from current literature.
On the other hand, studies of β-catenin knockout mice and liver tumors have revealed that the expression of many cytochrome P450 isozymes is also dependent upon β-catenin (Sekine et al.,2006; Loeppen et al.,2005). The zonal distribution of most of these isozymes, however, extends from the central veins to the midzonal area (Oinonen and Lindros,1998; Gebhardt et al.,2007) suggesting that β-catenin must show some effect in this region. Furthermore, the decreasing gradient of APC from periportal to pericentral in normal mice (Benhamouche et al.,2006) suggests a reciprocal gradient of active of β-catenin even, if it is not detectable by antibodies. When APC is knocked out, a zone of obviously activated β-catenin progresses from pericentral to periportal (Benhamouche et al.,2006), indicating a shift of this gradient toward the portal tract. In this overactive state, cytoplasmic and nuclear β-catenin is detectable by immunohistochemistry.
How can these inconsistent results be explained? First, it has to be taken into account that transgenic reporter constructs containing multimerized TCF binding sites should not be assumed to give a complete or definitive readout of Wnt signaling in vivo (Barolo,2006). In one of three caveats that apply to the use of such reporters, Barolo states “A lack of reporter activity should not be interpreted as compelling evidence that Wnts, β-catenin and/or TCFs play no role in a given context.” Second, the fact that the TCF reporter mouse used by Moriyama et al. (2007) does not respond in all GS-positive hepatocytes may indicate that canonical β-catenin signaling is not continuously active, but may be oscillating with different frequency or different phases in different hepatocytes (Gebhardt et al.,2007). Such an explanation would also fit to the existence of “isolated” GS-positive hepatocytes observed in MnSOD-transgenic mice (Lenart et al.,2007) and other situations (Umemura et al.,1993; Williams et al.,1993).
Third, it cannot be excluded that Wnt/β-catenin signaling is acting mainly in specific periods dedicated to certain cell-fate decisions. Thereafter, the primed cells maintain their functions by default as long as they are alive or until they are subject to further priming. Although this view is supported by observations from proliferating hepatocytes in vitro (Gebhardt et al.,1986) and is compatible with the observed changes after APC knockout (Benhamouche et al.,2006), direct evidence is still lacking.
In this context, it needs to be emphasized that the genes for most of the markers for hepatocyte heterogeneity (e.g., GS) have not yet been unequivocally identified as real targets of canonical Wnt/β-catenin signaling. It is therefore important to look more closely into the possible mechanisms by which Wnt/β-catenin signaling may affect metabolic zonation. Because β-catenin has been found to interact with a plethora of other transcription factors and is itself affected by different signals (Jin et al.,2008), the possible influence of these interactions for zonal gene expression in liver will be briefly discussed.
It is well known that the LEF/TCF family exhibits extensive patterns of alternative splicing and alternative promoter usage associated with repression as well as activation of target genes (Arce et al.,2006). It is not well understood which facets of this complexity actually contribute to canonical Wnt/β-catenin signaling in normal liver. However, this issue is potentially complicated, because it would appear that not all splice variants of TCF-4 are homogeneously distributed in liver parenchyma (Gaunitz, Gebhardt, unpublished observations). With specific reference to GS gene expression, we do not believe Glul to be a direct target of TCF/β-catenin activation. Instead, we favor an indirect mode of action (Fig. 2) based on the repressor function of TCF in the absence of activated β-catenin and the release of this repression upon activation of β-catenin. This may result in binding of STAT5 to the far-upstream enhancer of the GS gene and regulation of expression by growth hormone signaling (Werth et al.,2006; Gebhardt et al.,2007). A similar mechanism might be involved in the regulation of cytochrome P450 isoforms, because many of these are also affected by growth hormone signaling (Waxman and O'Connor,2006). Because TCF proteins interact with many co-activators, co-repressors, and antagonists (Arce et al.,2006), a large variety of similarly indirect mechanisms in the regulation of metabolic pathways can be envisioned.
Wnt/β-Catenin and FoxO in liver
Forkhead box O (FoxO) transcription factors in liver are known to be involved in the regulation of a vast variety of different functions including glucose and lipid metabolism (Barthel et al.,2005; Gross et al.,2008). They execute these functions in part by binding directly to DNA motifs and in part by interacting with a plethora of unrelated transcription factors including CEBPs, peroxisome proliferator-activated receptors (PPARs), CAR, and PXR (reviewed in van der Vos and Coffer,2008). For instance, insulin-like growth factor binding protein (IGFBP)-1, PEPCK, and glucose-6-phosphatase genes were found to be controlled by FoxOs (O'Brien et al.,1995; Rausa et al.,2000; Schmoll et al.,2000).
Apart from the down-regulation by insulin-PI3K-PKB signaling (Greer and Brunet,2005), FoxO proteins in liver are regulated by nutritional and hormonal factors (Imae et al.,2003). FoxO protein functions were also found to be activated by low levels of oxidative stress (Essers et al.,2004). Furthermore, Essers et al. (2005) described an evolutionarily conserved interaction between β-catenin, FoxO1 and FoxO3 by which FoxO-dependent transcription is activated in certain mammalian cells. Again, oxidative stress enhances both interaction with β-catenin and transcriptional activity (Essers et al.,2005). Recently, evidence was provided for a cross-talk mechanism between FoxO and TCF signaling in which β-catenin plays a central regulatory role leading to an inverse relation between the two signaling mechanisms (Almeida et al.,2007; Hoogeboom et al.,2008).
Coffer and coworkers recently found that GS is a target of FoxO3 (Coffer, van der Vos, Gebhardt, unpublished observations). Apparently, FoxO3 is enriched in the pericentral zone of liver parenchyma and shows nuclear localization in GS-positive hepatocytes (Fig. 3). Based on these results, we suggest that GS in liver may be regulated alternatively in an indirect mode and a direct mode of β-catenin action by means of interaction with TCF and FoxO3, respectively (Fig. 2). This dual regulation may contribute to the particularly stable phenotype of GS-positive cells. Furthermore, we speculate that it may explain the peculiar influence of oxidative stress observed in MnSOD knockout-mice (Lenart et al.,2007) on the distribution of GS-positive cells.
Because other FoxO transcription factors also depend on β-catenin and oxidative stress (Manolagas and Almeida,2007; van der Vos and Coffer,2008), a considerable number of targets may be regulated in a comparable way. For instance, FoxO1 has been shown to stimulate the expression of PEPCK and G6Pase in isolated hepatocytes and to suppress sterol response element binding protein (SREBP)-1 as well as glycolytic and lipogenic enzymes (Zhang et al.,2006). In line with these results, a gain of function mutation of FoxO1 targeted to liver and pancreatic β-cells resulted in diabetes arising from a combination of increased glucose production and impaired β-cell compensation (Nakae et al.,2002). These findings may suggest a possible mechanism by which the predominant pericentral action of β-catenin may influence remote periportal functions such as gluconeogenesis.
Wnt/β-Catenin and PPARs in Liver
Wnt/β-catenin signaling interferes with various nuclear receptors (Shah et al.,2003; Mulholland et al.,2005). One particular example is PPARs, which play an important role in lipid and glucose homeostasis (Villacorta et al.,2007). Among the many binding partners of PPARs, β-catenin is of particular interest in the context of this review.
As demonstrated in colon cancer cells, PPARγ seems to be present in complexes with β-catenin and TCF (Jansson et al.,2005). There is a clear physical interaction between the Tcf-binding domain of β-catenin and a β-catenin binding site on PPARγ (Liu et al.,2006). This leads to the inhibition of PPARs under conditions of high β-catenin activity, while activation of PPARγ may in turn result in enforced degradation of β-catenin (Liu and Farmer,2004). Hypothetical effects on gene transcription resulting from these interactions are displayed in Figure 4. Furthermore, PPARγ has been found to be a positive target of Wnt/β-catenin signaling, at least in colon cancer cells (Jansson et al.,2005), providing a possible basis for a negative feedback effect of PPARγ on β-catenin signaling.
Because PPARγ also forms heterodimers with members of the retinoid X receptor family (Willson et al.,2000), there is a further level of competition between different binding partners for PPARγ. It is conceivable that competition of this nature contributes to the differential regulation of periportally and pericentrally expressed genes. Most interestingly, a further level of regulation is introduced by the influence of the various agonists and antagonists of PPARγ (Handeli and Simon,2008). However, the positive or negative effects of these modulators on β catenin do not necessarily correlate with their effects on PPAR. It therefore remains controversial whether or not PPARγ may be a valuable target for the treatment of (colorectal) cancer (Gupta and Dubois,2002).
In contrast to PPARγ, expression of PPARδ is repressed by Wnt/β-catenin signaling most probably through interaction with β-catenin/TCF-4–responsive elements in the PPAR promoter (He et al.,1999). Thus, different members of the PPAR family may well respond differently. Interactions with potential effectors are also of considerable importance here, as suggested by studies linking cPLA2α, PPARδ and Wnt7β-catenin signaling (Han et al.,2008). So far, it is not known to what extent these findings are cell type-specific or can be generalized.
Although there is still a lack of detailed information concerning hepatocyte heterogeneity, we believe the various interactions with members of the PPAR family to play an important role in the regulation of zonal gene expression in adult liver. An example is given by the influence of the PPAR agonist dihydroepiandrosterone, which stimulates GS expression in the pericentral zone (Mayer et al.,1999). So far, however, it is not fully established whether this is due to interactions of dihydroepiandrosterone with PPAR or with the androgen receptor. It would be of considerable interest to study the influence of PPAR agonists and antagonists on the zonal localization of other PPAR target genes, particularly those involved in lipid metabolism.
β-Catenin and HIF in Liver
The transcription factor hypoxia inducible-factor 1 (HIF1) is induced by hypoxia and is involved in the regulation of a great number of physiologically important genes including nearly all glycolytic enzymes, glucose transporters and LDH-A (Kietzmann et al.,2006; DeBerardinis et al.,2008).
HIF1 is a heterodimer composed of a HIF1α and a HIF1β subunit and binds to hypoxia response elements. In rat liver, all HIF-α-subunits were found associated with the less aerobic pericentral zone (Kietzmann et al.,2006). Recent investigations have shown that β-catenin interacts directly with HIF-1 and that this complex binds to DNA specifically through HIF-1. β-Catenin enhances HIF-1 activity independently of TCF/Lef (Kaidi et al.,2007). It has also been found that HIF-1α prevents acetylation of β-catenin by removing human arrest-defective (hARD)1 protein from β-catenin and, in doing so, represses the transcriptional activity of the β-catenin/TCF4 complex (Lim et al.,2008).
An interaction between β-catenin and HIF1α in the pericentral zone may therefore enhance the expression of glycolytic enzymes and regulate liver zonation partially through HIF1α.
Of interest, HIF1α is not only responsive to hypoxia. Several studies indicated that HIF1α is responsive to many other factors including hormones such as platelet-derived growth factor, epidermal growth factor, fibroblast growth factor, transforming growth factor-beta (TGFβ), and insulin-like growth factor-1, even under normoxia (Kietzmann et al.,2006). Furthermore, β-catenin has been proposed to be a key switch between various reactive oxygen species-sensitive transcriptional programs (Hoogeboom and Burgering,2009).
As HIF1α transcription factors are stabilized by hypoxia and as hypoxia also occurs physiologically during development, where β-catenin and Wnt are known to have critical roles, it is possible that the cross-talk between hypoxia and β-catenin signaling has considerable implications during liver development (Kaidi et al.,2007).
β-Catenin, Further Transcription Factors and Combinatorics
Apart from the examples discussed above, β-catenin is known to interact with a variety of other transcription factors and hormone receptors. Although it is not within the scope of this review to provide a comprehensive overview of these possibilities, some examples are worth mentioning. First, β-catenin has been described to interact physically with the androgen receptor (Pawlowski et al.,2002; Yang et al.,2002), thereby enhancing the androgen-induced transcriptional response (Truica et al.,2000; Singh et al.,2006). With respect to zonal heterogeneity in liver, these findings may explain the enlargement of the GS-positive zone by androgens (Sirma et al.,1996; Mayer et al.,1999). Furthermore, binding of the co-activator glucocorticoid receptor interacting protein (GRIP)1 to β-catenin may synergistically enhance the activity of both the androgen receptor and Lef1 (Li et al.,2004).
Second, the signaling activity of β-catenin can be repressed by activation of the vitamin D receptor (Easwaran et al.,1999). Again, this interaction can be modulated by vitamin D receptor-antagonists (Shah et al.,2006) leaving much room for nutritional modulation of metabolism in liver.
Third, a recent example illustrating the importance of the interaction of β-catenin and hepatocyte nuclear factor (HNF)4α-driven transcription in controlling liver zonation was provided by Colletti et al. (2009).
Last, but not least, there is evidence for the interaction of the TCF4/β-catenin complex with the powerful transcription factor c-Jun (Nateri et al.,2005).
Given the variety of interactions of β-catenin with transcription factors, the competition of these factors for limited concentrations of β-catenin and the fact that many of them can also interfere with each other or with other factors such as CCAAT enhancer binding protein (C/EBP) in a β-catenin–independent manner, there is clearly the potential for an immense variety of combinations participating in a large interdependent regulatory network. Changes in the concentrations of activated β-catenin would be expected to influence competitive binding reactions impacting on the whole network, thus providing a molecular framework for establishing and adjusting the β-catenin activity gradient along the portocentral axis. These features are fully compatible with, and support the concept of, postdifferentiation patterning (PDP) developed recently for explaining the role of Wnt/β-catenin in hepatocellular heterogeneity of gene expression (Gebhardt et al.,2007). Moreover, because many of the different transcription factors are also targets of a plethora of different hormonal, nutritional and metabolic signals, the integration of these signals into a regulatory response allowing the adaptation of liver metabolism to the complex and ever-changing needs of the body seems possible. It will be the challenge of future research, in particular of a systems biological approach (c.f. Kestler and Kühl,2008), to elucidate the structure of this regulatory network in detail and to define its role in the control of normal and pathological liver metabolism. In the latter respect, the recent identification of TCF7L2 as a diabetes susceptibility gene in genome-wide association studies (Grant et al.,2006) provides a clear indication for the role of deranged Wnt signaling in susceptibility and development of complex and possibly age-related diseases (Manolagas and Almeida,2007; Jin,2008).
Overexpression of β-Catenin in Adult Liver
As mentioned above, the normal activity of Wnt/β-catenin signaling mediating zonal expression of metabolic genes does not result in substantial hepatocyte proliferation. Indeed, while the stimuli governing the low frequency of hepatocyte mitosis in normal liver remain elusive, a role for Wnt/β-catenin signaling seems unlikely, because the activity of this signaling pathway appears to be highest in pericentral hepatocytes, whereas the mitotic index is highest in periportal hepatocytes (Grisham,1962; Gebhardt,1992a).
Partial hepatectomy, which is the most frequently used model for inducing proliferation of all hepatocytes, leads to the up-regulation of β catenin, together with a range of growth factors and cytokines (Michalopoulos,2007). Under these conditions, nuclear translocation of β-catenin is induced by hepatocyte growth factor in a Wnt-independent manner (Monga et al.,2002), resulting in nuclear localization within 15–30 min of partial hepatectomy (Monga et al.,2001). A similar activation of β-catenin is observed during hepatocyte growth factor-induced hepatomegaly in mice (Apte et al.,2006). However, regeneration after partial hepatectomy in β-catenin knockout mice is attenuated without being completely blocked (Tan et al.,2006) indicating that short-term activation of β-catenin seems supportive, but not sufficient to drive liver cell proliferation. Under these conditions, the balance needed to maintain normal liver mass is not disturbed.
In contrast, the long-term activation of Wnt/β-catenin signaling above normal levels results in considerably enhanced hepatocellular proliferation and may even lead to hepatocellular carcinomas. For instance, overexpression of mutated forms of β-catenin has been found to stimulate hepatocyte proliferation (Miyoshi et al.,1998; Cadoret et al.,2001). Likewise, the liver-specific knockout of APC in mice leads to the massive appearance of bromodeoxyuridine (BrdU) -labeled hepatocytes and causes hepatomegaly within a few days (Benhamouche et al.,2006; Reed et al.,2008). Similar results have been found in APCMin/+ mice (Goessling et al.,2008). Of interest, the hepatomegaly phenotype was rescued by loss of β-catenin, but not by loss of c-Myc, which is sufficient to rescue the phenotypes of APC-loss in the intestine (Reed et al.,2008). Thus, long-term activation appears to be a strong proliferation stimulus, leading to an increase in liver mass. The fact that a substantial proportion of HCC and an even greater proportion of hepatoblastomas carry mutations in β-catenin or in other members of this signaling pathway such as Axin (for review, see Gebhardt et al.,2007; Austinat et al.,2008; Cavard et al.,2008; Takigawa and Brown,2008), suggests that carcinogenesis is just a tiny step away from β-catenin overexpression. Indeed, slight differences in transgenic mice with overexpression of mutated forms of β-catenin may determine whether hepatocellular carcinomas are induced (Colnot et al.,2004) or not (Harada et al.,2002). When overexpression of Ras is simultaneously induced in the latter model, however, development of carcinomas is frequent (Harada et al.,2004).
Most interestingly, many metabolic features associated with basal activity of this signaling pathway are maintained at higher levels of Wnt/β-catenin signaling, despite the increased cell proliferation. This is particularly obvious after partial hepatectomy when most metabolic functions of the liver are slightly suppressed but do not disappear. Furthermore, even hepatocellular carcinomas frequently maintain such metabolic features (Cadoret et al.,2002; Loeppen et al.,2002,2005), which can reflect pericentral or periportal functions depending on whether the tumors carry β-catenin mutations or not (Braeuning et al.,2007a). These observations are indicative of step-wise differential interpretation of relative activity gradients of Wnt/β-catenin signaling even when the absolute level of activity is higher than in normal liver. The preservation or even up-regulation of certain metabolic functions may have an important impact on the growth rate and aggressiveness of the respective tumor (Gebhardt et al.,1989; Gebhardt and Williams,1995; Cavard et al.,2008).
An Integrative Model of β-Catenin Function in Liver
In conclusion, we hypothesize that the different actions of Wnt/β-catenin signaling can be attributed to three levels of activity of this pathway (Fig. 5). At a low level of activity, differentiation phenomena occur leading to formation of hepatoblasts and subsequently of mature hepatocytes (cf. Hussain et al.,2004; Decaens et al.,2008; Nejak-Bowen and Monga,2008). The low level of β-catenin in the nucleus allows interactions only with binding partners having the highest affinity. At the intermediate level of activity characteristic of adult liver, postdifferentiation patterning of hepatocytes is apparent and forms the basis of metabolic zonation (Gebhardt et al.,2007). Under the assumption that most binding partners (transcription factors) only differ slightly in their affinities for β-catenin, competition between these binding partners for a limited amount of nuclear β-catenin creates different outputs for each concentration of β-catenin in this range. Thus, this level of activity features the richest combinatorial interplay of β-catenin with a plethora of transcription factors involved in the regulation of metabolic pathways, enabling fine-tuning of the metabolic activities (Fig. 5). Any derangement of this interplay, for example, by mutations of transcription factors interacting with β-catenin, may lead to a nonphysiological output contributing to the development of metabolic diseases as, for instance, highlighted by the association of mutations in TCF7L2 with diabetes type II (Grant et al., 2006). At a high level of activity, Wnt/β-catenin leads to proliferation of hepatoblasts or hepatocytes (Fig. 5). At these concentrations of nuclear β-catenin, most binding partners are (almost) saturated leading to a uniform (metabolic) response. In addition, low affinity binding partners may become affected. If this activation is extremely high and persists for extended periods of time, proliferation may become autonomous and may favor initiation and progression of hepatocellular tumors.
Wnt/β-Catenin and Beyond
Within less than a decade, Wnt/β-catenin signaling has emerged as an important player in liver development, maturation, normal function and neoplastic transformation (Gebhardt et al.,2007; Cavard et al.,2008; Takigawa and Brown,2008). However, there are already clear indications that other morphogens, for example, members of the Hedgehog and TGF protein families, also take part in this spectrum of events. The contribution of Hedgehog signaling to metabolic control has recently been hypothesized (Gebhardt et al.,2007) and subsequently demonstrated in cultured hepatocytes subjected to siRNA-mediated knock-down of APC, leading to enhanced expression of Indian hedgehog (Ihh; Ueberham and Gebhardt, unpublished observation). Similar findings have been reported from APC knockout mice (Reed et al.,2008), suggesting that activation of β-catenin signaling may result in an up-regulation of Hh signaling. Given the fact that Wnt/β-catenin signaling and Hedgehog signaling pathways show mutual interactions in several tissues (van den Brink et al.,2004; Yang et al.,2008) in ways which can result in either stimulation or inhibition on many different levels (Watt,2004; Alvarez-Medina et al.,2008; Roop and Toftgard,2008), it is conceivable that at least some of these interactions are relevant in adult liver. Future studies concerning liver function and hepatocellular heterogeneity should focus on the cross-talk of these two important morphogen signaling pathways and their influence on the respective downstream transcription factor network.
In conclusion, the combinatorial interplay of β-catenin with many different transcription factors is emerging as the molecular basis of metabolic zonation in the liver. Disturbance in this balance of activities may contribute to many complex metabolic diseases and may even underlie the currently unexplained phenomenon of specific metabolic preferences in related tumors.
The author thanks Dr. Michael Cross (University of Leipzig) for critically reading the manuscript. The excellent technical assistance of Mrs. Doris Mahn is gratefully acknowledged. This work was supported by the German Federal Ministry for Education and Research, BMBF within the program “HepatoSys” (FKZ 0313081F).