Mechanisms of hepatic fibrosis.

Abstract

Liver fibrosis is a common sequel to a variety of liver diseases of varying causes. These include autoimmune disorders such as primary sclerosing cholangitis and primary biliary cirrhosis, gene defects such asα1-antitrypsin deficiency, cystic fibrosis and haemochromatosis, alcoholic liver disease, and viral hepatitis. The most common diseases that lead to the development of liver fibrosis in children are biliary atresia, cystic fibrosis, α1-antitrypsin deficiency and viral hepatitis. Liver fibrosis is a major health problem worldwide and in the United Kingdom alone accounts for approximately 6000 deaths annually. However, to date, there are few therapeutic strategies for treating liver fibrosis and liver cirrhosis, which represents the end stage of disease, frequently requires liver transplantation. In liver fibrosis, there is increased hepatic accumulation of extracellular matrix (ECM), most notably type I collagen, together with a wide variety of noncollagenous components such as proteoglycans and glycoproteins. Accumulation of ECM distorts the architecture of the normal liver, compromising its function, whereas qualitative changes in the ECM alter the normal function and phenotype of liver cells. Liver fibrosis may be viewed as a repair response to hepatocellular injury, and its underlying mechanisms are likely to share many features in common with scarring in other organs. There is now little doubt that the key event in liver fibrosis is the injury-induced activation of hepatic stellate cells (HSC, also known as lipocytes, fat-storing cells, or Ito cells) with their subsequent transformation to myofibroblastic cells that overproduce the complex array of ECM proteins ultimately deposited in fibrotic liver. Research in liver fibrosis has at its focal point the question of how these cells become activated after liver injury and the part played by these cells in shifting the homeostasis of ECM in the liver to the net deposition of ECM. In this article, the important aspects of the cell biology of liver fibrosis are reviewed, with particular concentration on recent advances in our understanding of the process of HSC activation. CELLS OF THE HEPATIC SINUSOID The hepatic sinusoid is of central importance to liver function. These small blood vessels permeate the liver and are fed at one end with blood from terminal branches of the portal and arterial blood supplies that flow through the sinusoids and feed into central venules. The lumen of the sinusoid contains Kupffer cells, liver-specific macrophages, that are important scavengers of microorganisms entering the liver in the portal circulation. Kupffer cells are key to the liver's response to injury and have a central role in the development of liver fibrosis. Hepatocytes, the cells of the liver that are responsible for the unique synthetic and metabolic capabilities of this organ, are arranged in palisades along sinusoids. This arrangement allows intimate contact between the brush borders of hepatocytes and the perisinusoidal space of Disse, which underlies the endothelium of the sinusoid. This endothelium differs importantly from that in normal capillaries, in that it is specialised to maximize exchange between hepatocytes and sinusoidal circulation. To this end, sinusoidal endothelial cells have multiple fenestrae through which macromolecules may freely pass between sinusoids and the space of Disse (1). Also, the basement membrane-like ECM underlying these endothelial cells, although having many components in common with typical basement membranes around capillaries (2), has a low-density structure that minimally impedes sinusoid-hepatocyte exchange. Hepatic stellate cells lie within the space of Disse and may be considered as liver-specific pericytes because they have cytoplasmic extensions encircling the adjacent sinusoid. These cells represent less than 5% of the total liver cells but are readily detectable by their cytoplasmic lipid granules, which contain more than 75% of the total body store of vitamin A sequestered as retinyl palmitate (3). During liver injury, HSCs proliferate and undergo activation with transformation to fibrogenic myofibroblasts, which deposit numerous ECM proteins into the space of Disse. As well as compromising macromolecular transfer between sinusoids and hepatocytes, fibrosis within the space of Disse has important consequences for cells surrounding this structure, as detailed later. CELLULAR ORIGINS OF LIVER MATRIX In keeping with their role as important structural components, the major types of ECM proteins (collagens, proteoglycans, glycoproteins) are widely distributed throughout the liver. Collagens type I, III, IV, V, and VI have all been found in liver. The interstitial or fibrillar collagens types I, III, and V are the major components that are localised to the liver capsule, the portal tracts, and surrounding large blood vessels (2,4-9). Collagen types I and VI have a sparse distribution in the lobule, and collagen type III is in periportal and centrilobular zones (4,5). Of the noncollagenous proteins, undulin, fibronectin, and laminin have been localised to portal tracts (10,11) and the lobule (2,11). Within the lobule, matrix is found mainly in the perisinusoidal space, which has an atypical low density basement membrane-like structure, a complex structure containing minor amounts of interstitial collagens, microfilamentous collagen type VI, collagen type IV (in human, but not in rat perisinusoidal space [2,12]), glycoconjugates (laminin, fibronectin, vitronectin, entactin), heparan sulphate proteoglycan, and hyaluronate (2,12-16). In several studies, investigators have examined the cellular origins of liver matrix using techniques applied to liver sections (immunocytochemistry, in situ hybridisation) or techniques such as immunoassay and Northern blot analysis, applied to isolated hepatic cell types maintained in culture. Endothelial cells isolated from rat liver synthesize collagen types I, III, IV, laminin, and fibronectin (17-21). Matrix synthesis has also been attributed to hepatocytes(22-24), although these cells are considered to play a minor role in liver matrix synthesis overall, and collagen type I originally detected in hepatocyte cultures may derive from contaminant HSCs (25). There is abundant evidence from several experimental approaches that activated HSCs elaborate a wide variety of the collagenous and noncollagenous ECM proteins that are known to accumulate in fibrotic liver (17,20,21,23,24,26-38). Cirrhotic liver contains approximately 6 times more ECM than does normal liver, and the relative proportions of ECM components are altered so that, although collagen types I, III and IV are all increased, the change for type I is disproportionate and its ratio to types III and IV therefore increases (2,27,39,40). This switch to collagen type I's synthesizing in excess of collagen type III is also a feature of HSCs activated in vitro(34). Because HSCs are confined to the space of Disse, it is not surprising that profound increases in collagenous and noncollagenous proteins occur in this region in liver fibrosis (2,4,7,11,41). There is an established temporal pattern of ECM deposition in the space of Disse during development of liver fibrosis: In early injury collagen types III and V and fibronectin accumulate(39), but in chronic injury there is increasing deposition of collagen types I and IV, undulin, elastin, and laminin(9). Hyaluronan, a minor component in normal liver space of Disse, is increased more than eightfold(42), and dermatan and chondroitin sulfate increase in proportion to heparan sulfate proteoglycans. This marked accumulation of basement membrane components and collagen type I in the space of Disse is accompanied by loss of fenestrae from sinusoidal endothelial cells(1,2), in an overall process termed capillarisation. Capillarisation of the space of Disse inhibits the transfer of macromolecules between the sinusoidal circulation and hepatocytes, and constricts and distorts the structure of the sinusoid, causing portal hypertension. In cirrhotic liver, the vascular structures at each end of the sinusoid become linked by dense ECM, which extends into the parenchyma, and the whole liver anatomy becomes infiltrated with collagen type I-rich bands easily identifiable in liver sections. At a microanatomic level, the gross qualitative and quantitative changes in matrix may contribute to the disease-associated alterations in the phenotype of hepatocytes, stellate cells, and endothelial cells, as described later. STELLATE CELL ACTIVATION-A PROCESS CENTRAL TO LIVER FIBROSIS Activation of HSCs is an invariant feature of liver diseases that ultimately lead to fibrosis. The process of HSC transformation has been intensively investigated in progressive liver fibrosis in animals. Administration of a single intraperitoneal bolus of CCI4 to rats induces hepatocyte injury and zonal necrosis together with a rapid increase within 48 to 72 hours of the desmin-positive cell population (desmin is a marker for rat HSCs) selectively in areas of injury, demonstrated by their increased incorporation of bromodeoxyuridine (43,44). Increased bromodeoxyuridine labelling has also been demonstrated in desmin-positive cells in rats subjected to chronic bile duct ligation (45,46). In acute CCI4 treatment, HSC numbers and liver histology return to normal within 1 week, but continuous treatment with CCI4 twice weekly results in sustained HSC proliferation and in appearance of activated HSC-expressing α-smooth muscle actin and various ECM products (17,26,46-49). Results in more limited studies with primates(50) and in human liver fibrosis confirm this pattern of expansion of the HSC population, which is transient in such acute insults as those in patients recovering from paracetamol overdose(38) but proceeds to subsequent transformation of HSC and ECM deposition in chronic disease (39,51). Proliferation of HSCs and activation in liver disease can be mimicked accurately in culture, and this model has been used extensively to study the cell biology of the activation process. Hepatic stellate cells freshly isolated by proteolytic digestion of normal liver have the same features as HSCs in intact liver, having a relatively compact phenotype and containing cytoplasmic granules rich in vitamin A esterified to fatty acids (retinyl palmitate) (52). Freshly isolated HSCs show little ECM synthesis or α-smooth muscle actin expression. However, during the first 7 days of culture in serum-containing medium, HSCs proliferate, release most of their vitamin A and gradually spread out on the plastic substrate as they transform to larger myofibroblastic cells that have greatly increased synthetic capacity for a wide variety of collagenous and noncollagenous matrix components (33,35,36,53-55). Activated HSCs also show reorganisation of the cytoskeleton, becoming positive for α-smooth muscle actin(47) and having increased content of fibrillar actin. FACTORS INFLUENCING MATRIX SYNTHESIS IN THE LIVER-IMPORTANCE OF TRANSFORMING GROWTH FACTOR BETA Studies of HSC activation in vitro have shown that this process is regulated by a wide variety of growth factors and cytokines derived from other liver cells (paracrine factors) or from HSCs themselves (autocrine factors). Damaged hepatocytes produce HSC-activating factors(56,57) including insulin-like growth factor(IGF) and IGF-binding proteins (58). Although binding of IGF to its receptors on HSCs is reduced during their activation(59), receptors for the potent HSC mitogen platelet-derived growth factor (60) are increasingly expressed during HSC activation in vivo and in vitro(61). Sinusoidal endothelial cells and platelets activated after liver injury are potential sources of this important HSC mitogen (62), as are HSCs themselves after activation(63). Infiltrating inflammatory cells also produce HSC mitogens such as interleukin-1 and tumour necrosis factor-α (TNF-α). Mast cells, which normally represent a minority cell type in normal liver, accumulate in fibrotic human liver (64) and in rat bile duct ligation model of fibrosis (65). In human liver, mast cell accumulation correlates with the extent of collagen deposition. These cells are a rich source of a variety of mediators, including the serine protease tryptase, heparin, histamine, and basic fibroblast growth factor (bFGF), which promote mitosis and increase collagen production by fibroblasts(66). Kupffer cells produce the HSC mitogens TNF-α, IL-1 and a transforming growth factor-α-related protein(67,68). They are also an important source of transforming growth factor-β1 (TGF-β1)(69): This protein is considered to play such an important role in fibrosis that it deserves detailed discussion. TGF-β mRNA upregulation occurs in human injured liver(70,71) and in rat liver after CCl4 intoxication, bile duct ligation, iron and alcohol intoxication, and murine schistosomiasis (72-78). In rat CCl4, injury, TGF-β1 mRNA increases dramatically with a similar time course to the expansion of the Kupffer cell and HSC populations, and TGF-β1 mRNA is strongly localised to these cells in fibrotic liver (74,77,79). However, TGF-β isoforms are produced to some degree by a wide variety of liver cell types, including endothelial cells, bile duct epithelial cells, hepatocytes, and HSCs (76-78,80). TGF-β upregulates HSC expression of collagens types I, III, and IV, fibronectin and laminin (34,69,81-83) and accelerates transformation of quiescent HSCs to myofibroblasts(81). Because activated HSCs both produce and respond to TGF-β1, it may be an important autocrine effector on these cells. The profibrogenic effects of TGF-β are enhanced by its protection of ECM from protease-mediated remodelling. In fibroblasts, TGF-β1 inhibits production of the matrix-degrading enzymes, interstitial collagenase, stromelysin, and plasmin by inhibition of urokinase plasminogen activator(84,85), while simultaneously upregulating expression of protease inhibitors including tissue inhibitor of metalloproteinase-1 (TIMP-1) and plasminogen activator inhibitor(84). After activation, HSCs express signalling receptors(type I and II receptors) for active TGF-β, as well as the nonsignalling receptor type III, the proteoglycan β-glycan(86). β-Glycan and the proteoglycan decorin may sequester TGF-β and constrain its activity in tissues. Administration of decorin to animals with experimental kidney disease effectively prevents the development of fibrosis in this organ (87,88). Injection of an expression vector for decorin intramuscularly into rats reduces TGF-β activity in kidney and markedly reduced experimental kidney fibrosis (89). In contrast, TGF-β1 overexpressor transgenic mice show increased matrix deposition in liver and kidney (90). TGF-β1 neutralising antibodies may also be useful therapeutically, in that their topical administration inhibits development of fibrosis in skin wounds(91). Taken together, these observations underscore the central role of TGF-β1 in fibrosis. TGF-β1 is secreted as a latent form of 112 kDa, consisting of an inactive precursor consisting of a homodimer of active 25-kDa TGF-β noncovalently linked to a homodimer of latency-associated peptide(Fig. 1). A large pool of latent TGF-β exists in serum and is bound to proteoglycans within ECM. It is therefore likely that the supply of active TGF-β in tissues is regulated mainly by local activation mechanisms. The major pathway of latent TGF-β activation in liver is considered to involve its binding, through oligosaccharide residues on the latency associated peptide, to the IGF receptor-II/mannose-6-phosphate receptor (IGF-II/M6P receptor). Once bound, proteolytic cleavage within the latency-associated peptide releases the active 25-kDa homodimeric form of TGF-β (92) (Fig. 1). The proteases that are important physiologically in this cleavage are uncertain, but the serine protease plasmin may play a role. Plasmin itself derived from the cleavage of circulating plasminogen by such plasminogen activators as urokinase. As activated HSCs express IGF-II/M6P receptors(93) and produce urokinase (94), they apparently express a complete system for latent TGF-β activation. However, TGF-β activation by HSCs may be relatively inefficient in the absence of factors derived from other liver cells, such as sinusoidal endothelial cells (93).FIG. 1: Schematic showing the pathway of transforming growth factor-β (TGF-β) activation and its consequences for matrix homeostasis. IGF-II/M6P = insulin-like growth factor II/mannose-6-phosphate receptor; PAI-I = plasminogen activator inhibitor type I.Although blocking TGF-β synthesis or activation appears an attractive prospect for therapy of fibrosis, this protein has important physiological functions in dampening immune responses and in regulating the proliferation and differentiation of cells, particularly epithelial cells. An important recent finding is that the profibrogenic, rather than antiproliferative effects of TGF-β1 may be mediated by connective tissue growth factor (95). This protein is induced in cells in response to TGF-β1 and then acts in an autocrine fashion to upregulate transcription of ECM genes. Studies of connective tissue growth factor and its receptors are at an early stage, and nothing is known of their possible role in liver fibrosis. However, connective tissue growth factor may be an attractive therapeutic target in this disease. RETINOIDS AND HSC ACTIVATION Recent advances in understanding the importance of retinoids in HSC activation deserve detailed discussion. Hepatic stellate cells contain more than 75% of the total body store of vitamin A as retinyl palmitate. The loss of cytoplasmic vitamin A is a dramatic feature of HSC activation, and the liver content of vitamin A is depleted during HSC activation in advanced liver fibrosis and cirrhosis. Retinoids are known to be important in influencing phenotype transformations in several cell types and loss of vitamin A may play an active role in modulating HSC activation. The major product of retinyl palmitate hydrolysis by HSCs is retinol(96). When added exogenously to cultured HSCs in vitro, retinol inhibits HSC proliferation in response to serum or platelet-derived growth factor, and also their synthesis of ECM components (97-100). These findings suggest that loss of retinoids from HSCs may promote their activation. However, the finding that hypervitaminosis A, in which HSCs become loaded with retinoid granules, is sometimes associated with their fibrosis (101), argues against this hypothesis; and to date, the role of vitamin A in liver fibrogenesis remains unclear. One possibility is that completion between different pathways of retinol metabolism within HSCs dictates its ultimate effects. It is therefore relevant that in recent studies investigators have shown that, after activation, HSCs produce 14-hydroxy-4, 14-retro-retinol, which belongs to a novel class of retinoids termed retro-retinoids(102). 14-Hydroxy-4, 14-retro-retinol is a potent proliferative factor for lymphocytes (103), but whether it has similar effects on HSCs has not been assessed. Retinoids exert their actions on gene transcription not directly but through binding to nuclear retinoic acid receptors. This growing family of receptors has been divided into two types, RARs and RXRs, which themselves are divided into subtypes of α, β, and γ receptors. During their activation, HSCs show diminished expression of RAR-β (104,105) which, in other cell types, is required for retinol inhibition of cell proliferation. Reduction in RAR-β may contribute to HSC proliferation by reducing their normal responsiveness to the antiproliferative effects of retinoids (105). This would be exacerbated by loss of the cellular retinoids themselves during activation. Strategies for increasing sensitivity of HSCs to retinoids or for targeting the enzymes involved in their metabolism may be fruitful for inhibiting liver fibrosis. EXTRACELLULAR MATRIX AS A DETERMINING FACTOR OF LIVER CELL PHENOTYPE There is a growing awareness that the ECM is not just a passive framework binding tissues together, but that it also regulates the phenotype and function of cells. During liver fibrosis, the major alterations in liver matrix are associated with radical changes in the phenotype and function of hepatocytes, endothelial cells, and HSCs. The specific role of matrix in regulating liver cell function has been explored in vitro. Culture of hepatocytes on a basement membrane-like matrix or in gelatin gels maintains their cuboidal shape and expression of albumin and cytochrome P450, but when cultured on collagen type I or plastic they become flattened, spread out on the substrate and show diminished expression of these proteins(106,107). Endothelial cells also require a basement membrane-like matrix for maintenance of their normal fenestrae, which are lost when cultured on collagen type I (108). The nature of the substrates on which HSCs are cultured have pronounced effects on their phenotype. Hepatic stellate cells, freshly isolated from normal liver and plated on a basement membrane-like matrix resembling their pericellular matrix in normal liver, conserve their compact phenotype, rich in retinoid granules, and show minimal production of collagen type I(54). In contrast, culture of HSCs on plastic, collagen type I or individual basement membrane components (heparan sulfate proteoglycan, laminin, collagen type IV), results in their proliferation and transformation within 7-14 days to the activated, retinoid-depleted myofibroblastic phenotype producing collagen type I (35,55,109). These observations suggest that damage to the normal HSC subendothelial matrix in early hepatic inflammation might disrupt the normal cell-matrix interactions crucial to maintaining HSC quiescence, thereby initiating HSC activation. Matrix damage in injured liver may be mediated by matrix metalloproteinases, such as gelatinase B, released from invading inflammatory cells or resident Kupffer cells(110), or by stromelysin and gelatinase A, produced by activated HSCs themselves (111-113). Hepatic stellate cell activation might also be promoted by matrix produced by other cell types, such as EIII-A type fibronectin, which is produced by hepatic endothelial cells after liver injury and has been shown to stimulate HSC activation in vitro (19). The importance of matrix in regulating HSC activation is underscored by findings that HSC activation induced by culture on plastic is reversed after plating of these cells onto basement membrane-like matrix (114). This appears to be an in vitro correlate of the rat recovery model of fibrosis discussed later, and studies of the cell biology of the phenotype reversal might give pointers to designing drugs that accelerate this process. Hepatocytes, HSCs, and endothelial cells recognise their pericellular matrix through adhesion molecules, with integrins probably playing the dominant role. Integrins are heterodimeric transmembrane proteins consisting of noncovalent associations of α and β subunits that are universally expressed by cells (115). The specificαβ subunit structure determines specificity for different matrix components. Common matrix recognition sites for many of the integrins have been defined, the first being the tripeptide motif RGD (Arg-Gly-Asp). Between them, the β1, β3 and β4 integrins in combination with α subunits recognise most of the constituents of the ECM. In the normal liver, each of the different cell types displays a distinct repertoire of integrins that is altered in liver diseases(116,117). The changes in expression of both the integrins and their respective ECM ligands in liver disease may influence liver cell function. Integrin-mediated mechanisms may be important in regulating ECM homeostasis in the liver, in that integrins regulate production of ECM and matrix metalloproteinases by fibroblasts and chondrocytes, and integrin-mediated signalling may integrate with those from soluble growth factors to regulate cell proliferation and phenotype (118). Results in recent studies show that activated human HSCs express α1β1,α2β1, αvβ1, andα6β4; and in fibrotic liver, integrinsα2, α4, and αv are increasingly expressed in HSCs relative to their expression in normal liver(119,120). Integrins may be important in communicating changes in ECM to HSCs in regulating HSC activation. REDUCED MATRIX DEGRADATION MAY CONTRIBUTE TO LIVER FIBROSIS-ROLE OF METALLOPROTEINASE INHIBITORS Limited damage to skin or internal organs frequently completely resolves with regeneration of normal tissue architecture without fibrosis. This healing response requires the remodelling of the provisional matrix initially deposited as a rapid response to form a barrier to infection. Various enzymes are considered to mediate matrix remodelling, including plasmin, cathepsins, elastase, inflammatory cell-derived proteases and matrix metalloproteinases(MMPs). The MMP family of enzymes are zinc and calcium ion dependent proteinases which together are capable of degrading most ECM components(121). Within this family are: stromelysins which degrade a wide variety of collagenous and noncollagenous proteins;gelatinases A and B (also called 72-kDa and 92-kDa type IV collagenases, respectively) which degrade laminin, fibronectin, and collagen type IV, important components of basement membranes; andinterstitial collagenases with activity against fibrillar collagen types I, II, and III, the major components of fibrotic matrix. Because of their potential for tissue damage, MMPs are subject to complex regulation(Fig. 2). Firstly, they are subject to transcriptional regulation by specific stimuli, especially such inflammatory mediators as IL-1, TNF-α and growth factors. Secondly, they are released from cells as catalytically inactive proenzymes that require cleavage of an N-terminal propeptide to achieve activity. Thirdly, the active forms are controlled by endogenous proteinase inhibitors, also with a wide tissue distribution, called tissue inhibitors of metalloproteinases (TIMPs)(122). The TIMPs block the active site of the MMPs and also inhibit activation of the progelatinases.FIG. 2: Factors that influence the degradation of collagen by interstitial collagenase. PMA = phorbol myristate acetate.The balance between the activities of specific matrix-degrading enzymes and rate of matrix synthesis is maintained in normal tissues, so that quantity and quality of ECM remain in balance. Liver fibrosis results from matrix synthesis overwhelming the normal matrix-degrading capacity of the liver. There is accumulating evidence that matrix deposition in fibrosis is promoted by deficiencies in matrix degradation. Because the fibrotic neomatrix in liver is rich in collagen type I, the activity of interstitial collagenase may critically influence fibrosis. In support of this, several investigators have indicated that activity of this enzyme is decreased in liver homogenates as liver fibrosis progresses in humans and in animal models(123-127). Evidence from our group suggests that decreased interstitial collagenase activity in fibrotic liver arises from inhibition of this enzyme by TIMPs. We have shown that TIMP-1 mRNA and protein is increased three-to fourfold in human liver cirrhosis of varying origins (chronic active hepatitis, primary biliary cirrhosis, primary sclerosing cholangitis, biliary atresia) compared with its presence in normal liver (128,129). TIMP-1 mRNA expression significantly correlates with extent of collagen deposition in these livers (129), and increases in TIMP-2 mRNA in fibrosis have also been noted. Hepatic stellate cells may be an important source of TIMP-1 and -2 in fibrotic liver (129,130). In our more recent studies, we have further demonstrated that mRNA for TIMP-1 and -2 increases with similar time course to expression of procollagen type I mRNA during progressive liver fibrosis in rats treated with CCl4 or after bile duct ligation(131,132). In these studies of liver fibrosis in humans and rat models, the expression of interstitial collagenase mRNA remains remarkably stable. These findings suggest that, in injured and fibrosing liver, increased TIMP relative to interstitial collagenase may inhibit the natural remodelling of fibrotic matrix by this enzyme. Further evidence for this derives from our studies of natural recovery from liver fibrosis (133). Rats treated for 4 weeks with carbon tetrachloride gradually recover their normal liver histology when left untreated for a further 4 weeks. Within the first week of the recovery period, there is marked dissolution of the collagen fibrous septae and a decrease in liver hydroxyproline that correlate temporally with reductions in mRNA for TIMP-1 and -2. Because interstitial collagenase mRNA expression remains unchanged during recovery, we suggest that the degradation of fibrotic matrix in recovering livers, consisting mainly of collagen type I, is caused by recovery of the enzymatic activity of interstitial collagenase. If this mechanism accounts for the spontaneous recovery from liver fibrosis sometimes seen in humans, then therapies based on inhibiting TIMP expression or its binding to collagenase may be envisaged. In marked contrast to interstitial collagenase, mRNA, and protein for another MMP, gelatinase A (72-kDa type IV collagenase) is chronically increased in fibrotic compared with that in normal human liver(129,134). Results of in vitro studies indicate that activated HSCs may be an important hepatic source of this enzyme (112,113). It seems paradoxical that gelatinase A expression increases in fibrotic liver where type IV collagen actually accumulates. However, the catalytic activity of this enzyme may be constrained by simultaneous increases in the TIMPs. In addition, there are mechanisms that may localise activation of gelatinase A to the immediate vicinity of HSCs such as binding to membrane-type MMP, the gelatinase A-activating protein present on the HSC cell surface (135,136), and binding to αvβ3 integrin present on these cells (119). In that gelatinase A degrades several basement membrane components, its activity may promote HSC activation by releasing growth factors such as TGF-β or bFGF from the matrix or by interfering with the ability of basement membrane to maintain HSC quiescence as described above (54). Our recent findings suggest that gelatinase A may be an important autocrine growth factor for HSCs(137) although its mechanism of action remains to be defined. Gelatinase A is also mitogenic for glomerular mesangial cells(138), which have many similarities to HSCs. CONCLUSIONS AND FUTURE DIRECTIONS This review has concentrated on the central role of activated HSCs in tipping the balance of matrix homeostasis in the liver toward fibrosis. The processes that initiate and perpetuate HSC activation in liver disease result from complex and interacting stimulatory mechanisms, involving soluble growth factors and cytokines released from injured hepatocytes, Kupffer cells, infiltrating inflammatory cells and HSCs themselves. Once HSCs become activated, their response to the repetitive generation of soluble growth factors in chronic liver disease or recurrent liver injury is likely to be enhanced by their increased expression of receptors for mitogens such as platelet-derived growth factor and fibrogenic factors such as TGF-β. Loss of retinoids from HSCs after liver injury and altered expression of nuclear retinoic acid receptors may chronically enhance their proliferative and fibrogenic responses, as described. Changes in the pericellular matrix of HSCs during activation may occur as a net result of degradation of the normal basement membrane-like matrix by gelatinase A and other HSC derived MMPs, and its replacement by a fibrotic neomatrix rich in collagen type I. These matrix changes may sustain HSC activation by integrin-dependent signalling mechanisms either directly or by changing HSC responsiveness to soluble growth factors. Once activated, HSCs not only produce the major components of the fibrotic neomatrix in liver but also produce TIMPs, which promote the stability of this matrix. As they become activated, HSCs switch from producing urokinase plasminogen activator to producing plasminogen activator inhibitor(94,139), which may further inhibit collagen breakdown by reducing production of plasmin, an enzyme that is considered important for activating interstitial collagenase. Our current understanding of the cell and molecular biology of HSC activation provides clues for developing antifibrotic therapies. Because liver fibrosis in humans is usually detected clinically when it is in its advanced stages, strategies should be directed at simultaneously preventing HSC activation and encouraging remodelling of the collagen type I-rich neomatrix of fibrosis. These might be directed at inhibiting specific fibrogenic factors, with antifibrotic effects already reported in studies blocking TGF-β1. Blocking fibrogenic messengers downstream of TGF-β1, such as connective tissue growth factor, may be a better approach in vivo with fewer side effects. Alternatively, reducing expression of TIMPs, and thereby unmasking the natural matrix-degrading capacity of liver, is a novel approach that might be effective in advanced fibrosis. Perhaps paradoxically, MMP inhibitors may be be effective therapeutically if selectively targeted at MMPs such as gelatinase A or stromelysin, which may damage the normal liver matrix and set the stage for its replacement by fibrotic neomatrix. This strategy may also block MMP-mediated processing or liberation of inflammatory cytokines such as TNF-α (140). Future studies of the detailed mechanism of HSC activation may also provide clues to therapies based on antagonising intracellular pathways or gene regulatory factors fundamental to activation, thereby simultaneously blocking matrix and TIMP production by these cells. The discovery of activation pathways specific to these cells, possibly based on their unique features of retinoid metabolism, would be particularly advantageous in targetting therapies avoiding systemic alterations in matrix homeostasis. Acknowledgment: Supported in part by Wellcome Trust, United Kingdom; The Medical Research Council, United Kingdom; the James Knott Trust, United Kingdom; and the Wessex Medical School Trust, United Kingdom.