Liver damage leads to an inflammatory response and to the activation and proliferation of mesenchymal cell populations within the liver which remodel the extracellular matrix as part of an orchestrated wound-healing response. Chronic damage results in a progressive accumulation of scarring proteins (fibrosis) that, with increasing severity, alters tissue structure and function, leading to cirrhosis and liver failure. Efforts to modulate the fibrogenesis process have focused on understanding the biology of the heterogeneous liver fibroblast populations. The fibroblasts are derived from sources within and outwith the liver. Fibroblasts expressing α-smooth muscle actin (myofibroblasts) may be derived from the transdifferentiation of quiescent hepatic stellate cells. Other fibroblasts emerge from the portal tracts within the liver. At least a proportion of these cells in diseased liver originate from the bone marrow. In addition, fibrogenic fibroblasts may also be generated through liver epithelial (hepatocyte and biliary epithelial cell)–mesenchymal transition. Whatever their origin, it is clear that fibrogenic fibroblast activity is sensitive to (and may be active in) the cytokine and chemokine profiles of liver-resident leucocytes such as macrophages. They may also be a component driving the regeneration of tissue. Understanding the complex intercellular interactions regulating liver fibrogenesis is of increasing importance in view of predicted increases in chronic liver disease and the current paucity of effective therapies.
- Kupffer cell
- stellate cell
The adult human liver typically weighs approx. 1.5 kg. It is the largest internal organ and plays many pivotal roles in intermediary metabolism, and in the metabolism and clearance of xenobiotics. The liver is responsible for the disposal of bile pigments and for the generation of bile acids that are central to the maintenance of cholesterol homoeostasis and the absorption of dietary lipid from the intestine. The liver also plays an important role in lipoprotein metabolism and cholesterol homoeostasis. It is the site of synthesis for major serum proteins, including albumin, complement and clotting factors, and of catabolism of amino acids and the generation of urea. In the normal state, the liver is maintained at a size which provides substantial overcapacity. It also has a remarkable ability to regenerate in response to functional parenchymal loss and can return to normal size and functional state, even after 70% of the parenchyma is lost.
Chronic liver injury, irrespective of cause, is generally associated with the accumulation of matrix proteins, a process referred to as fibrosis. In parallel with this, there is a continued stimulus for regeneration, leading to further distortion of the hepatic architecture and vascular structures (portal veins, hepatic veins). This results in a transformation to a nodular architecture, so-called cirrhosis (Figure 1). Given the normal functional overcapacity of the liver, patients with cirrhosis can have apparently normal (compensated) liver function for long periods of time, but, with many, there is ultimately decompensation with catastrophic effects on the various strands of intermediary metabolism referred to above. In addition, the altered vasculature leads to the development of portal hypertension. To date, most therapies for chronic liver disease have targeted the aetiological agent: for example, there are now several effective antiviral agents for the treatment of hepatitis B and C. Similarly, there are a number of immunosuppressive agents that can be given for immune-driven processes such as autoimmune hepatitis. Although these agents often have a beneficial effect on the degree of fibrosis, there are currently few agents that are specifically developed to interfere with fibrosis. For some patients with end-stage liver disease, the only available option for treatment is orthotopic transplantation.
The present review will examine our current understanding of fibrosis, the experimental models available for its study and the potential therapeutic options currently being tested.
Accidental (e.g. adverse reaction to a normally therapeutic drug dose ) or deliberate (e.g. paracetamol overdose ) poisoning can lead to massive liver necrosis, acute liver failure and death. However, chronic liver damage and its associated progressive fibrosis is, quantitatively, a more significant clinical problem . Table 1 outlines the major causes of chronic liver disease in the human population. Viral infections are currently the major global cause of liver fibrosis. However, alcohol abuse is another important cause of chronic liver disease. Furthermore, over the last 25 years, it has become increasingly apparent that a similar spectrum to that observed with alcohol can be seen in individuals who are obese and/or diabetic. This so-called NAFLD (non-alcoholic fatty liver disease) can lead to end-stage liver injury and cirrhosis. The long-term socioeconomic impact of the ‘epidemic’ of NAFLD and obesity is yet to be established.
Liver tissue is composed of functional units that contain all the cells of the liver arranged around point(s) of entry (portal tract) and exit (central veins) of blood. There remains a debate about the structure of the functional hepatic unit with several models proposed. In part, this is because there is no impermeable barrier between the units in most species, including humans and rodents. A review of the various proposed functional units is beyond the scope of this review (for more information, see ); however, the most favoured in their simplest terms are referred to here as the liver lobule (first proposed by Kiernan) and the liver acinus of Rappaport (Figures 2A and 2B). Zonal expression of many genes fits with a lobular functional structure for the liver. However, the acinus is the unit of choice for histopathologists because it aids in the explanation of many pathological lesions (Figure 2C).
The liver receives a dual blood supply: there is a portal circulation which brings partially oxygenated blood from the gut, rich in compounds absorbed following digestion, and, in addition, some 30% of the hepatic blood comes via the hepatic artery. The two blood inputs become mixed at the edge of the portal tracts where the liver sinusoids are found. These, in essence, are the liver's capillaries and are lined by specialized fenestrated endothelial cells, controlling the flow of materials to and from the space between hepatocytes and endothelial cells (space of Dissé) . The space of Dissé contains a low density of extracellular matrix proteins and the hepatic stellate cell. Hepatocytes constitute the major epithelial cell type of the liver and extend along the sinusoids to the centrilobular region (lobule) or zone 3 (acini) as polarized epithelial cells. Liver macrophages (Kupffer cells) reside within the sinusoids. The major cell types of the liver and their functions are summarized in Table 2.
A striking feature of the liver is the heterogeneous expression of many genes along the sinusoid. A classic example is the expression of CYP (cytochrome P450) genes which are expressed at their highest levels in hepatocytes surrounding central veins  (see also Figure 2B). Xenobiotics requiring metabolism by CYP cause predominantly centrilobular/zone 3 damage, despite the fact that periportal/zone 1 hepatocytes are exposed first to ingested xenobiotics. The majority of exposure to toxins is through ingestion of pro-toxins that require enzyme-mediated conversion into a cytotoxic product (many directly cytotoxic substances are obviously avoided). Thus pro-toxin damage is located in the regions where the activating enzyme is expressed. Additionally, oxygen levels are also at their lowest in the centrilobular/zone 3 regions, which may have an impact on the ability of cells to respond to and survive injury. Hepatocytes are therefore more often the cell type that is damaged by hepatotoxins (although they are not intrinsically sensitive and have a number of active defence mechanisms, such as a capacity to maintain high levels of reduced glutathione to defend against oxidative stress ).
Enzyme-mediated toxicity has particular relevance to acute mechanisms of xenobiotic hepatotoxicity, but is the cause of only a minor proportion of chronic liver disease cases. Indeed, the major chemically induced cause of liver disease, alcohol, induces changes in gut permeability to bacterial membrane components, leading to a pro-inflammatory hepatic environment as a major part of its mechanisms of action [8,9]. Most experimental models of fibrosis employ pro-toxin/enzyme-mediated mechanisms to cause liver damage. An understanding of how toxins work in experimental models of fibrosis is therefore essential to avoid misinterpretation of data, particularly when the efficacy of a potential anti-fibrogenic agent is being examined.
CURRENT VIEWS OF LIVER FIBROGENESIS
Fibrogenesis is predominantly viewed as a repercussion of the wound-healing response that occurs after persistent liver tissue damage. These events are therefore discussed (Figure 3 and Animation 1 at http://www.BiochemJ.org/bj/411/0001/bj4110001add.htm).
Liver injury and the inflammatory response
In most cases of acute liver injury, hepatocytes are damaged and undergo both necrosis and apoptosis. Figure 2(C) demonstrates the typical effects observed shortly after treatment with the hepatotoxin CCl4. Hepatocytes located mainly around the central veins are damaged because CCl4 is a pro-toxin that requires metabolism by CYP for toxicity (mainly by CYP2E , see Table 4). Cell death is primarily via necrosis, although a proportion of hepatocytes may undergo an apoptotic mechanism of cell death, particularly following chronic exposure . Kupffer cells in damaged areas detect the release of intracellular contents and/or apoptotic cells and release a range of cytokines and chemokines that activate local resident leucocytes and promote the recruitment of circulating leucocytes to the region of damage [12,13]. Sinusoidal endothelial cells play a complementary role in this process through changes in the expression of adhesion proteins that promote the tethering of passing leucocytes, leading to their trans-endothelial migration to the site of injury . Major factors contributing to fibrogenesis include cytokines, of which many are released from Kupffer cells, but some are derived from myofibroblasts themselves (Table 3).
The response to tissue damage and infection is therefore analogous with activation of the innate immune system and the potential activation of the adaptive immune system. In some cases, the latter is followed by a breakdown in tolerance to self-antigens and therefore to the generation of T-cells and antibodies to self-antigens [15,16]. In cases where liver damage is drug-induced, it is probable that the drug and/or a metabolite(s) reacts with a protein(s), which is then seen by the immune system as foreign. Once the drug and neo-antigen are cleared, immune tolerance to unmodified protein is lost, resulting in continued attack of the self-antigen and a chronic autoimmune hepatitis. Typically, antibodies against smooth muscle proteins and/or a nuclear antigen are detected in affected individuals (in type 1 autoimmune hepatitis) or to microsomal xenobiotic-metabolizing enzymes such as CYPs (in type 2 autoimmune hepatitis) . Ticrynafen (tienillic acid) is converted into a metabolite which reacts with the generating enzyme (CYP2C9), against which antibodies are produced [17,18].
The inflammation associated with hepatic injury forms part of a regulated response to cell damage resulting in the removal of necrotic material. The liver, presumably because of its precarious function as a protector from endobiotics and ingested xenobiotics, has evolved the ability to regenerate . The cytokines involved in marshalling the inflammatory response to tissue damage also play an important role in the replacement of hepatocytes. Cytokines, such as TNFα (tumour necrosis factor α) and IL (interleukin)-6, prime remaining hepatocytes to be responsive to growth factors such as HGF (hepatocyte growth factor) and EGF (epidermal growth factor) . The stimulus for liver growth appears to be through a sense of its own functional output. The surgical removal of part of the liver (partial hepatectomy) in which there may be relatively minimal cell death, results in regeneration of the remaining liver to the required size, as if responding to the rise in the levels of a factor normally metabolized by the liver. This factor may be serotonin [20,21], which is primarily synthesized by the gut and would therefore pass directly to the liver via the hepatic portal vein.
Hepatocyte regeneration in most diseased states therefore comes from existing viable hepatocyte mitosis . Previous work suggesting that hepatocytes can be derived from the bone marrow [22,23] has now been shown, at best, to be a rare event (the presence of donor genetic material in recipient hepatocytes is now thought to occur via donor bone-marrow-derived macrophage fusion with recipient hepatocytes [24,25]). Liver-resident leucocytes (e.g. Kupffer cells) are derived from the bone marrow in addition to a proportion of other non-parenchymal liver cells [26,27]. A hepatocyte/bile duct epithelial cell progenitor cell, termed ‘oval cells’ in some species, is present in portal tract regions (at the level of the so-called ‘canal of Hering’), but only gives rise to hepatocytes when existing hepatocytes are prevented from replicating [28,29]. In this respect, the ‘streaming’ hypothesis in which hepatocytes are normally replenished (potentially from an oval cell progenitor) from replication in the periportal region , which could replace hepatocytes after acute centrilobular injury, is not now widely accepted [31,32]. Many of the factors which drive the innate immune system to clear away the damage also act in an apparent co-ordinate fashion to prime hepatocyte renewal. A less well understood component of this process is the remodelling of the extracellular matrix that occurs in damaged parenchyma. An additional factor, however, is that myofibroblasts may be essential for effective liver regeneration. Plasminogen (Plg)-deficient mice spontaneously develop liver fibrosis . The neurotrophin receptor p75NTR is expressed in myofibroblasts and appears to function in the transdifferentiation from hepatic stellate cells . Plg−/− p75NTR−/− mice have exacerbated pathology compared with Plg−/− p75NTR+/+ mice due to inhibited hepatocyte proliferation .
Fibroblasts were hitherto considered to be a relatively inert ‘space-filling’ cell type secreting extracellular matrix proteins. However, there is increasing evidence in the liver to suggest that they play a role in tissue regeneration; they are now known to express a range of cytokines and chemokines [35–38]. In generic terms, a population of fibroblast-like cells is activated by (or activation is significantly enhanced by) the release of cytokines and ROS (reactive oxygen species) from Kupffer cells and other leucocytes [39–41] to proliferate and secrete proteases, extracellular matrix proteins and other factors. In the acute setting and under some chronic conditions in which fibrosis may be reversible, the fibroblast-like cells undergo apoptosis and Kupffer cells release factors which remodel the extracellular matrix through secretion of matrix metalloproteinases [42,43]. The end result is that the liver sinusoidal structure returns to normal, and the current theory is that the process of fibroblast-like cell activation is an essential component of this process, although this remains to be experimentally tested.
If damage to the liver persists, either continuously or through repeated acute episodes within a timeframe such that resolution has not had a chance to occur, fibroblast numbers increase and fibrosis develops (see Animation 2 at http://www.BiochemJ.org/bj/411/0001/bj4110001add.htm). As the liver disease progresses, the scarring becomes more extensive, impeding hepatocyte regeneration, and the function of the organ is compromised. For a period of time, fibrosis remains reversible if the primary cause of liver damage is removed or suppressed [44,45].
Early work on liver fibrosis originally proposed that hepatocytes were responsible for the production of scarring extracellular matrix protein in liver fibrosis . It was subsequently shown that hepatic stellate cells were a major source [47,48]. In a normal liver, the cells are referred to as ‘quiescent’ and function to store much of the body's vitamin A . Hepatocytes esterify vitamin A with fatty acids which are then stored by quiescent hepatic stellate cells [49,50]. In response to liver damage, quiescent hepatic stellate cells lose their vitamin A and transdifferentiate into a myofibroblast phenotype expressing α-smooth muscle actin [51–53], an actin isoform restricted to smooth muscle cells.
Heterogeneity of hepatic stellate cell-derived myofibroblasts
Under the appropriate in vitro culture conditions (see below) quiescent hepatic stellate cells undergo a phenotypically similar process of transdifferentiation to myofibroblasts as that which occurs in vivo in response to liver damage . It has recently emerged that quiescent hepatic stellate cells are not the only fibrogenic cell in the liver. Quiescent hepatic stellate cells can be found in the centrilobular and perisinusoidal regions of the liver lobule (Figure 4). Injury to the centrilobular region (e.g. in alcoholic liver disease) results in their activation. However, Ramadori and co-workers first suggested that portal tract fibroblasts rather than stellate cells were predominantly responsible for fibrogenesis in cases where damage was located in the periportal regions of the liver lobule (e.g. cholestasis) , a view that is now broadly accepted [56–58]. Figure 5 compares the fibrosis that occurs in rats treated with CCl4 (a centrilobular hepatotoxin) and bile duct ligation, which preferentially damages hepatocytes in the periportal region of the lobule. Both injuries result in fibrosis, but extensive α-smooth muscle actin myofibroblast immunostaining is only seen in the former, with cells positive for vimentin in the latter. Thus it is probable that a different population of vimentin-positive hepatic fibroblasts located in the periportal region is responsible for fibrogenesis in cases where periportal damage occurs. However, with more extensive injury, the disease progresses, and populations of cells from other regions probably contribute to fibrogenesis.
The complexity of establishing liver myofibroblast identity is compounded by the fact that each population of hepatic fibroblasts may be heterogeneous in its expression of fibrotic marker genes (and potentially other genes). In an elegant study from Brenner and co-workers, transgenic mice were generated in which red and green fluorescent protein reporter gene expression were driven by the α-smooth muscle actin and collagen IA1 promoters respectively . Both reporter genes were expressed in culture-activated hepatic stellate cells, but they were not co-expressed in all cells. In bile duct-ligated mice, pericentral and perisinusoidal myofibroblasts were both α-smooth muscle actin- and collagen IA1 marker gene-positive, whereas periportal fibroblasts only expressed the collagen IA1 reporter gene.
Embryological origin of hepatic stellate cells and liver myofibroblasts
Establishing the embryological origin of hepatic stellate cells (i.e. the germ layer from which the cells are derived) is complicated by their apparent heterogeneity as a myofibroblast. To date, it has not been formally determined. Given that adipose, smooth muscle cells and fibroblasts are derived from the mesoderm and that stellate cells are lipid-storing cells with an ability to transdifferentiate into myofibroblasts, it is likely that at least a proportion of both hepatic stellate cells (and myofibroblasts) and liver fibroblasts are mesodermal in embryological origin. However, hepatic stellate cells and liver myofibroblasts express several genes normally associated with neural tissue, such as synaptophysin, nestin and glial fibrillary acidic protein . This supports the possibility that hepatic stellate cells (or a proportion of them) have an ectodermal origin, perhaps from the neural crest given its ability to generate a range of mesenchymal-like cells . However, this has recently been tested in transgenic mice [61a]. A transgenic mouse expressing the Cre recombinase transgene under control of a WNT-1 promoter /enhancer sequence (specific to neural crest cells) was crossed with a transgenic for a fluorescent protein whose expression was dependent on removal of a lox P-flanked stop cassette by Cre recombinase [61a]. Although fluorescent protein was found in all tissues known to be derived from the neural crest, there was no expression in desmin-positive perisinusoidal cells of the liver [61a]. This suggests that hepatic stellate cells are not derived from neural crest cells.
In the adult mouse, recent work suggests that some hepatic stellate cells are derived from the bone marrow . When bone marrow cells expressing GFP (green fluorescent protein) were transplanted into normal mice, hepatic stellate cells in the recipient mice were found to express GFP. In culture, the GFP-positive stellate cells transdifferentiated into α-smooth muscle actin-positive myofibroblasts . In CCl4-treated fibrotic animals, GFP also co-localized with α-smooth muscle actin . This observation has been confirmed in irradiated female mice receiving bone marrow from male donor mice. Myofibroblasts containing a Y chromosome were detected in approx. two-thirds of recipient fibrotic liver myofibroblasts, with little evidence of cell fusion . More recently, the scar-producing cells associated with bile duct ligation (periportal myofibroblasts) have also been shown to originate from the bone marrow in an animal model, but yet remain distinct from hepatic stellate cells in that they are CD45-positive . Examination of human liver sections from non sex-matched recipients of either bone marrow or liver transplants confirmed that bone marrow contributed to myofibroblast populations in patients who subsequently developed liver disease .
These data suggest that hepatic stellate cells and liver fibroblasts at least in part originate from a mesenchymal stem cell within the bone marrow in the liver disease state. However, there remain a number of observations which suggest that there are other sources from which fibrogenic cells may be generated. For example, Zeisberg et al.  reported that liver fibroblasts are derived from hepatocytes in vivo during fibrosis through an “epithelial-to-mesenchymal transition”, although most of these fibroblasts were not α-smooth muscle actin-positive. Similarly, Robertson et al.  have shown biliary epithelial cell “epithelial-to-mesenchymal transition” in a patient with primary biliary cirrhosis, which may account for the bile ductopenia and portal tract fibrosis that occurs with this disease.
There is an increasing body of literature concerning the plasticity of cells from several tissues which have normally been considered mature and terminally differentiated. In this respect, quiescent hepatic stellate cells (specifically the proportion that are CD133-positive) have been proposed as a progenitor cell not only for liver myofibroblasts, but also for hepatocytes and endothelial cells . Mesenchymal stem cells from the adult bone marrow have also been stimulated to form multiple cell types (including hepatocytes) in vitro [68,69]. However, it is possible that the in vitro environment imparts an unusually potent plasticity to many cells that enables them to differentiate in a way that would not occur in vivo (since bone marrow cells do not differentiate into hepatocytes appreciably in vivo [23,24]).
Pivotal role for Kupffer cells in fibrogenesis
Although the status of hepatic myofibroblasts (their number and degree of activation) may directly control fibrosis severity, increasing attention is now been given to the role of other cells in the process. Macrophages within the liver or monocytes/macrophages recruited to the tissue are now known to exert a considerable influence. Their number increases in damaged liver and they are principally located around the regions of damage and fibrosis (Figure 6).
Thurman's group showed that inhibition of Kupffer cell function (using gadolinium chloride), reduced fibrosis in a CCl4 model of liver fibrosis . More recently, Duffield et al. , using a transgenic mouse model in which CD11b-positive cells can be conditionally stimulated to undergo cell death, have shown that liver-resident monocytes and macrophages not only promote fibrogenesis, but also are instrumental in the removal of fibrosis during any recovery phase.
The promotion of fibrogenesis by Kupffer cells is associated with the release of a range of pro-fibrogenic cytokines and ROS as part of their inflammatory response to liver damage. It is notable that the presence of Kupffer cells (normally present in low numbers in most isolations) in hepatic stellate cell cultures promotes a myofibroblast phenotype closely similar to the cells in vivo , but macrophages, in response to the milieu of cytokines and other factors, respond by releasing different mixtures of cytokines. It may be appropriate to suggest that macrophages differentiate into a variety of states. With regard to recovery from fibrosis, macrophages may differentiate into an anti-inflammatory phenotype and/or secrete proteinases that promote the degradation of scarring extracellular matrix proteins.
RESEARCH MODELS OF LIVER FIBROSIS
In vivo models of liver fibrosis
In view of the complex intercellular communication between cell types within the liver, and that extrahepatic cells may contribute to fibrogenesis, animal models remain an important experimental tool. In studies where the aim is to test the efficacy of a potential anti-fibrogenic agent, it is generally considered that drugs should be tested in at least two animal models (where the region of lobular damage is different so that effects on hepatic myofibroblasts derived from different regions of the lobule can be determined). Table 4 summarizes animal models of liver fibrosis most commonly employed. A vital issue with regard to the use of these models is that their use should be carefully matched to the mode of action of potential anti-fibrogenic agents. It is clear that treating the primary cause of liver damage in the human population would be the mode of therapeutic choice when available. The use of a potential anti-fibrogenic agent would be indicated when the primary cause of liver damage cannot be treated and/or fibrosis has progressed to cirrhosis. Accordingly, it is essential to ensure that the potential anti-fibrogenic agent does not modulate the hepatotoxicity of the experimental agent used to generate liver damage (i.e. that a potential anti-fibrogenic agent was not a CYP2E inhibitor if using the CCl4 model of liver fibrosis ) when using an animal model. Reductions in fibrosis could therefore be associated with reduced levels of liver damage, and not due to action of an anti-fibrogenic agent. Commonly used markers of liver fibrosis are given in Table 5.
Although the various animal models developed previously have served an important function in identifying basic mechanisms of liver fibrogenesis and testing new anti-fibrotic agents, it must be appreciated that all show some significant morphological differences with their human ‘counterparts’. This is perhaps least so with cholestatic models, particularly common bile duct ligation where the pattern of fibrosis is very similar to that seen when there is interruption of bile flow in humans. On the other hand, the commonly used CCl4 model is probably the most frequently used tool for investigating regression of fibrosis and cirrhosis, but the fibrous septa observed in animals, even on long-term CCl4, are generally much less developed compared with most human cirrhoses. Probably one of the most important differences between human and animal model diseases is the lack of significant alteration of vascular relationships in the latter; human cirrhosis is often confounded by liver cell loss through secondary hypoxic events, leading to parenchymal extinction. This does not occur in the commonly used animal models.
In vitro models of liver fibrosis
The most commonly used in vitro model of liver fibrosis is the isolation of quiescent hepatic stellate cells and their culture on plastic culture dishes in serum-containing culture medium. Under these conditions, cells commence a phenotypically similar process of transdifferentiation to the activated myofibroblast phenotype apparent in the liver in response to chronic liver damage (Figure 7), particularly when Kupffer cells are present (as a minor impurity) within the culture . The cells are amenable to trypsin subculture, although, once the cells have activated to a myofibroblast phenotype, the cells remain in the myofibroblast phenotype in subsequent passages. Where possible, cells should be used in primary culture since this period is the only time to examine effects on transdifferentiation.
There is a concern that differences in laboratory protocols may result in differences in the nature of hepatic myofibroblasts, and that hepatic stellate cell-derived myofibroblasts may be overgrown by portal tract myofibroblasts [55–57]. Indeed, at present, there is no universally agreed diagnostic marker for quiescent hepatic stellate cell populations that could be used to establish cell identity and purity. Nevertheless, the in vitro culture system will remain a valuable tool for studying liver fibrosis and can readily be used in high-throughput screening assays for potential anti-fibrogenic agents.
A technically more challenging in vitro model is the culture of liver slices [73,74]. Liver slices from normal or diseased liver will retain their native extracellular matrix, cell–cell contacts and cell density and may be a valuable model for short-term studies. However, even the thinnest tissue slices (∼200 μm thick, ∼10 cells thick) prevent effective access of nutrients to cells, and a large proportion of the slice, particularly hepatocytes, die after 24 h [75,76].
Hepatic stellate cells from rodents are readily subcultured for many passages and therefore, for studies using myofibroblasts in their pro-fibrogenic state only, immortalized cell lines are often not required. To study the process of transdifferentiation from the quiescent state, cells must be isolated repeatedly from animals, as myofibroblasts do not readily revert to the quiescent state. The value of a rodent hepatic stellate cell line (see Table 6) lies with their propensity for being transfected with plasmid constructs, because cells isolated from rodents are difficult to transfect with high efficiency. Human hepatic stellate cell lines are more common because of difficulties in most laboratories of accessing human tissue. Furthermore, in our hands, most human myofibroblast cultures appear to senesce at passage 2–5 (from 58 human preparations, only two cultures have remained proliferative for more than five passages). This may be due to isolation of cells primarily from elderly patients (mainly from partial hepatectomy specimens from individuals with a secondary tumour in the liver). Human myofibroblasts are also resistant to transfection, and, in our hands, only nucleofection was sufficiently effective (with repeatable high efficiency) for the transfection of a range of reporter gene plasmid constructs . Viral vectors have therefore been investigated from time to time for their ability to infect cells with high efficiency, including adenoviral [77,78], retroviral  and baculoviral  vectors. Although reported to be effective, the resources required to generate recombinant vectors has probably contributed to their lack of widespread use.
The in vitro model systems outlined above remain an effective way to screen a large number of potential compounds at an early stage for their ability to modulate the fibrogenic process using a variety of endpoints. However, as with many in vitro systems, there are limitations that prevent the identification of anti-fibrogenic agents with 100% accuracy. In vitro systems are closed artificial units that cannot allow for the absorption, distribution, metabolism and excretion of drugs that you see in whole-body studies. Liver myofibroblasts also reside next to the major drug-metabolizing cells of the body in vivo, but these are arguably functionally absent from most in vitro systems. Thus the replacement of in vivo animal models with in vitro models has yet to be achieved, and the former will probably always have a role to play in studies of liver fibrosis.
CLINICAL TREATMENTS FOR LIVER FIBROSIS
A number of reviews have appeared on approaches to the treatment of liver fibrosis [81–83] and so only brief discussion is given here. Currently, there is no therapeutic agent indicated and licensed for the primary treatment of liver fibrosis, although a number of therapies are in clinical trials (Table 7). A variety of potential drug targets have been shown to exist within hepatic myofibroblasts in experimental systems, but many of these are associated with other cell types which presents problems of drug specificity. Indeed, the surprisingly varied number of targets available suggests that there may be redundancy for each pathway in myofibroblasts. Thus the drug-mediated block of a pathway promoting fibrogenesis may be overcome by the remaining unrelated fibrogenesis-promoting pathways. For these reasons and because myofibroblasts produce a number of factors which act in an autocrine fashion to promote fibrogenesis, the efficacy of drug treatments that lead to myofibroblast apoptosis have been examined [84,85].
In animal models recovering from chronic liver damage (i.e. the cause of damage is stopped), fibrosis reversal is associated with the apoptosis of liver myofibroblasts . NF-κB (nuclear factor κB) is a transcription factor that regulates the expression of a range of genes associated with inflammation . NF-κB becomes constitutively active in myofibroblasts in vitro and induces the expression of several pro-inflammatory genes [e.g. ICAM (intercellular adhesion molecule) and nitric oxide synthase 2] . NF-κB is also constitutively active in myofibroblasts in vivo, suggesting that it is not an artefact of the in vitro culture system (F. Oakley, personal communication). NF-κB functions to prevent the apoptosis of myofibroblasts [84,87] as it does in other cells (e.g. hepatocytes) when they are exposed to pro-inflammatory mediators such as TNFα [89,90]. Inhibition of NF-κB therefore leads to the apoptosis of myofibroblasts in vitro  and in vivo [84,85]. The stimulation of liver myofibroblast apoptosis using gliotoxin (a fungal toxin that inhibits NF-κB , but which also works directly on the mitochondria to promote apoptosis ), resulted in a rapid recovery from fibrosis, and using this compound provided the first evidence in vivo that myofibroblast apoptosis may be an effective therapeutic approach to liver fibrosis treatment . Using existing licensed NF-κB inhibitor drugs such as sulfasalazine to achieve the same ends may be a realistic treatment option . More recently, targeting gliotoxin to hepatic myofibroblasts using a recombinant single-chain antibody to a surface antigen on the cells (thereby retaining functional liver macrophages), reduced fibrosis in vivo in a sustained injury model in contrast with free gliotoxin . The specific removal of liver myofibroblasts in a ‘non-inflammatory’ manner while leaving macrophages in place therefore appears to be an effective anti-fibrogenic approach. By removing the liver myofibroblasts, the potential of specific drug targets within these cells becoming redundant with time is avoided. The converse, however, is that employing cell-death mechanisms as a therapeutic approach requires careful targeting to avoid adverse effects. Furthermore, it remains to be seen whether myofibroblast reductions as part of an anti-fibrogenic approach has any deleterious effects on liver regeneration.
Treatments in which drug targets have opposite effects in hepatocytes and liver myofibroblasts may turn out to be safer options, particularly when disease progression is to be slowed (in contrast with stimulating a reversal of cirrhosis). The PXR (pregnane X receptor) is a nuclear receptor transcription factor that regulates the expression of genes associated with endobiotic and xenobiotic clearance in hepatocytes . PXR function is normally activated through contact with a range of ligands, including existing licensed drugs (e.g. rifampicin) and endogenous compounds (steroids, bile acids) . Recent work from our laboratory has shown that the PXR is expressed in myofibroblasts, is capable of binding to its DNA enhancer response element in myofibroblast nuclear extracts and is transcriptionally functional on transfected reporter gene constructs . The PXR regulates a specific set of genes in myofibroblasts. Ligand activators of the PXR reduce the expression of pro-fibrogenic cytokines such as TGF (transforming growth factor) β and inhibit proliferation . PXR activators therefore act in an anti-fibrogenic manner in human liver myofibroblasts in vitro . Using mice with a disrupted PXR gene, the role of the PXR in mediating an anti-fibrogenic effect was unequivocally established .
The fibrotic response in the liver is dependent on multiple cell types, both resident and recruited to the liver, and is regulated by a vast array of factors in vivo. A full understanding of the response remains a goal at present, particularly the benefit of fibrosis to tissue function and viability. The ability to specifically ‘knock out’ cell types in a tissue should help to resolve many questions in the same way that transgenic technology has resolved questions surrounding a specific gene's function. Clinically employed anti-fibrogenic agents may then become available.
Abbreviations: CYP, cytochrome P450; EGF, epidermal growth factor; GFP, green fluorescent protein; IL, interleukin; NAFLD, non-alcoholic fatty liver disease; NF-κB, nuclear factor κB; Plg, plasminogen; PXR, pregnane X receptor; ROS, reactive oxygen species; TGF, transforming growth factor; TNF, tumour necrosis factor
- © The Authors Journal compilation © 2008 Biochemical Society