^a Color codes: green, upregulation; dark green: higher upregulation; blue, downregulation; dark blue: higher downregulation; grey, no change; magenta, no correlation with liver fibrosis stage. ^† MFAP-4, microfibrillar-associated protein-4 (associated with elastin fibers). ^§ Endoglin mRNA was found upregulated in chronically HCV-infected patients compared to noninfected patients but not correlating with liver fibrosis stage.

3.2.1. Collagens and derived fragments.

By specifically analyzing the ECM components of biopsies from HCV-infected patients with liver fibrosis, levels of collagens type I, III and V were found increased from stages F0-F1 to F4 [59][79,80]. However, this is a hallmark of all fibrotic diseases [102]. These fibrillar collagens form arrays incorporating fibril-associated collagens with interrupted triple helices (FACIT), such as collagens XII and XIV [114]. Hepatic collagen XII amounts decreased with increasing stages of HCV-related fibrosis, whereas collagen XIV levels increased [59,84]. Collagen XVI, another FACIT not associated to collagens I, III, or V fibrils [115], was found underexpressed, specially at the cirrhotic stage of HCV-related fibrosis [59]. Collagen XVII is a transmembrane protein and a main component of hemidesmosomes, which are cell-ECM junctions anchoring epithelial cells to the basement membrane by interacting with integrins outside and with intermediate filaments of the cytoskeleton inside the cells [116]. Patients with advanced fibrosis/cirrhosis linked to chronic hepatitis C revealed focal positivity for collagen XVII in sinusoidal lining cells and few cholangiocytes, a pattern not observed in patients with cholangitis-linked cirrhosis [47]. Collagen XVII was also observed in the cytoplasm of HPC, and in the surrounding basal membrane in end-stage HCV-linked fibrosis. Collagen XVIII is a basement membrane collagen almost exclusively expressed in the liver, and hepatocytes were identified as its main source [28,117]. It can be cleaved at its C-terminus to generate endostatin, a powerful anti-angiogenic agent [117][118]. Collagen XVIII was found underexpressed in HCV-diseased liver [59], and through the subsequent down-expression of endostatin, hepatocytes could play an unforeseen role in neo-angiogenesis during hepatic neoplasia.

Studies attempting to define etiology-dependent molecular signatures of liver fibrosis identified serum collagen-derived biomarkers that variably evolve during chronic hepatitis B and C. Pro-C3 (N-terminal type III collagen propeptide), a fragment of collagen III formation, has been proposed as a serum predictive biomarker of fibrosis progression in patients with chronic hepatitis C [85]. These patients displayed higher serum levels of this biomarker than HBV-chronically infected patients [50]. By contrast, serum of HBV-infected patients had higher levels of the protease-cleaved 7S domain of the amino-terminal propeptide of type IV procollagen P4NP7S, a biomarker of type IV collagen formation. Chronic HBV-infected patients also had higher serum levels of markers of collagens III, IV and VI degraded by matrix metalloproteases than chronic HCV patients, as a reflection of a greater basement remodeling induced by chronic hepatitis B [50].

3.2.2. Enzymes of the ECM (Table 3).

3.2.2.1. Lysyl oxidases.

The lysyl oxidase family of enzymes comprises 5 members: LOX and LOX-like 1 to 4 (LOXL). They are copper-dependent secreted amine oxidases that cross-link monomers of collagen or elastin, to form insoluble fibrils. Hepatocytes of healthy livers do not express, or express only low amounts of lysyl oxidases [119]. In viral hepatitis C, LOX and LOXL1 expression is strongly enhanced in activated HSCs but not in hepatocytes [119]. Likewise, in viral hepatitis, collagen deposits are only observed in fibrotic zones, not around hepatocytes, in contrast to what is observed in fibrotic liver diseases of other etiologies. LOXL2 is strongly induced in fibrotic liver, where it localizes to regions of collagenous matrix and to a-SMA-positive fibroblasts-like cells [120]. More specifically, in fibrotic liver diseases related to active hepatitis C infection, LOX and LOXL2 were    detected at the fibrotic disease interface composed of fibroblasts, hepatocytes and neo-vasculature [120]. LOX was found to contribute to collagen stabilization in liver     fibrosis, promote fibrogenic activation of HSCs, and limit fibrosis reversal [121]. LOXL2 also mediates fibrotic matrix stabilization, and stimulates differentiation of HPC toward fibrogenic cholangiocytes [122]. The fibrosis-promoting activities of LOX and LOXL2 are susceptible to occur even after cessation of the chronic liver injury. Together with the global morphological reorganization of the liver due to ECM accumulation that alters liver metabolism, these impaired enzymatic activities might contribute to the persistence of clinical signs of liver disease after SVR in the case of chronic hepatitis C, in the absence of any therapy targeting the enzymes [120][123]. This feature may be even more pronounced when alcohol intake and/or metabolic syndrome complicate chronic infection. Thrombospondin-1 belongs to the family of matricellular proteins, which mediate cell-cell and cell-ECM interactions but are not primary structural ECM elements. It plays a role in collagen homeostasis, through its binding to fibrillar collagens, pro-LOX, matrix metalloproteases (MMPs) and TGF-b1 [71,72]. The core protein of HCV was found to downregulate LOX, while up-regulating the genes of collagen I (COL1A1) and thrombospondin-1 (THBS1) [73]. Similar gene expression profiles were obtained from mice xenograft tumors derived from HCV-infected human hepatocytes [74]. Mechanistically, HCV, and core in particular, activate the production of latent TGF-b1 by infected hepatocytes [53][73,74]. After its secretion in the microenvironment, latent TGF-b1 is then cleaved into active TGF-b1 by thrombospondin-1, the expression of which is upregulated by HCV [75], and the core protein in particular [73]. Active TGF-b1 finally activates  hepatic stellate cells, which contribute to fibrogenesis by (over) producing ECM components [73]. Interestingly, THBS1 is a TGF-b1 target gene, which creates a vicious loop where activated HSCs are committed to ECM overproduction.

3.2.2.2. Matrix Metalloproteases (MMPs) and their inhibitors.

Under physiological conditions of liver regeneration during wound healing, the normal equilibrium between MMPs and their inhibitors (tissue inhibitors of MMPs, TIMPs) is restored, whereas the MMP/TIMP ratio remains imbalanced during fibrosis (Figure 2). The mRNA and protein expression of several MMPs is modified during chronic hepatitis C: the gelatinases MMP-2 and -9 are enhanced [57,58,76], while the collagenases MMP-1 and -13 are down-regulated [57]. A recent study reported increased serum levels of MMP-2, -7 and -9 in patients with chronic hepatitis C, which correlated to fibrosis stage for MMP-7 [90]. Gelatinases contribute to the degradation of the   basement membrane collagen IV, while collagenases such as MMP-1 and -13 break down interstitial collagens I and III. During chronic hepatitis C, the upregulation of gelatinases will lead to a major reshuffling of the basement membrane, with the destruction of the structural support of cells, and the down-regulation of collagenases will amplify the deposition of insoluble collagen fibers and aggravate the fibrotic phenotype. As    mentioned earlier, hepatic collagen XIV is increased during HCV-related fibrosis [59,84]; since this collagen is collagenase-sensitive, this observation is in line with the down-regulation of collagenases [57].

In parallel, expression and secretion of TIMPs, in particular TIMP-1, are enhanced to compensate for an increased activity of MMPs [57,124]. Elevated levels of MMP-2 and -9 were reported in the serum and/or liver specimens of chronic hepatitis C patients,   correlating with the fibrosis stage but not with the viral load [63,76][88]. The serum level of TIMPs was suggested as an indicator of hepatic fibrosis: increased serum levels of TIMP-1 in chronic hepatitis C patients correlated with the stage of liver fibrosis [86], better than those of TIMP-2 [91]. Interestingly, serum levels of MMP-9 and TIMP-1   decreased after SVR in fibrotic HCV patients treated with DAA, in parallel with the   regression of fibrosis [125]. When fibrosis etiologies are compared, it appears that MMP-10 and -11 are upregulated at late fibrosis stages of chronic hepatitis B, whereas at similar stages of chronic hepatitis C, MMPs-2 and -9 are upregulated [89]. In NAFLD and non-alcoholic steatohepatitis, MMP-9 and -10 were found at significantly higher levels than in chronic viral hepatitis B and C [89].

To study the underlying mechanisms of such enzyme involvement in HCV-related fibrogenesis, hepatoma cell lines stably expressing the HCV protein(s) of interest were generated. The highest levels of MMP-2 were secreted when primary HSCs were grown with hepatoma cells expressing HCV core, compared with HSCs grown alone or in the presence of “regular” hepatoma cells [58]. Hepatocytes expressing HCV core secrete high amounts of TGF-b1, and exhibit a transcriptional upregulation of the connective tissue growth factor (CTGF), a mitogen stimulating the production of ECM components and enzymes by HSCs. This suggests that HCV core is a key regulator of MMP(-2) expression, possibly through a TGF-b1-mediated upregulation of CTGF [58]. Along similar lines, Chang liver cells expressing HCV core displayed elevated levels of MMP-9 transcript and protein, not observed in NS5A-expressing hepatocytes [63]. This suggests a direct effect of the core protein on the regulation of MMP-9 protein synthesis, and could explain the elevated levels of hepatic MMP-9 in biopsies from patients with HCV, not observed in patients with non-alcoholic steatohepatitis. Interestingly, both core- and NS5A-expressing cells displayed elevated levels of cyclooxygenase-2 (COX-2), a key  enzyme of the biosynthesis of prostaglandins, mediators of inflammation. High hepatic levels of COX-2 were also reported in HCV-patients biopsies, which places core and NS5A at a regulatory hub between inflammation and fibrogenesis [63]. This is reminiscent of a strong correlation between the inflammatory activity of the liver of fibrotic HCV patients and the transcriptional levels of TIMP-1 [88]. Similar higher levels of MMP-9 were reported in the serum of chronically HCV-infected patients, and in HCV-infected hepatoma cells than in uninfected patients or cells, and MMP-9 displayed a greater  enzymatic activity [71]. COX-2 was also found overexpressed in HCV-infected hepatocytes. Mechanistic studies revealed a regulatory role of the HCV serine protease NS3, together with its cofactor NS4A, underlying such over-expression and -activity: HCV NS3 activated the ERK1/2 – p38 pathway, leading to the phosphorylation of ERK1/2 and p38. This led to nuclear factor-kB (NF-kB) activation, and translocation to the nucleus where it promoted MMP9 transcription, leading to MMP-9 (over)expression [83]. In  parallel, NF-kB promoted COX2 transcription, and once biosynthesized, COX-2 translocated to the nucleus where it acted as a transcriptional promoter of MMP9. COX-2 and MMP-9 contributed to inflammation and fibrosis through the synthesis of prostaglandins and ECM degradation, respectively [71]. As noticed earlier, HCV-mediated fibrosis  appears tightly linked to liver inflammation, with on one hand viral proteins (core, NS3, NS5A) acting as modulators of inflammation pathways, and on the other hand inflammation mediators such as COX-2 acting as regulators of liver fibrosis. This is also in line with the anatomical observations of gathering of mononuclear cells at the hepatic lobules in HCV-related fibrosis, a hallmark of inflammatory activity not observed in          alcohol-related fibrosis [46], and the presence of prominent aggregates of lymphocytes in peri-portal zones [46][47]. The non structural protein NS4B was also identified as an  activator of MMP-2 expression, both at the mRNA and protein levels [76]. Mechanistic studies revealed that NS4B activated the ERK/JNK pathway, which engaged the   transcription factor STAT3, thus contributing to MMP2 transcription.

3.2.2.3. A disintegrin and metalloprotease with thrombospondin motifs (ADAM and ADAM-TS).

ADAMs, comprising ADAM-TS, are membrane-bound or secreted zinc-proteases of the ECM, with a broad tissue distribution. Through their disintegrin and metalloprotease domains, they are therefore able to carry out cell adhesion and protease activities,    respectively. In the liver, they are involved in the regulation of epithelial cell regeneration after liver injury, and able to directly degrade ECM components, thereby promoting ECM rearrangement during wound healing and fibrosis. A strong correlation between ADAM-9, -28, -TS1 vs MMP-2 and a-SMA was identified in biopsies from patients with diseased liver of any etiologies [126]. Also no differences in ADAM expression were   detected in biopsies of different etiologies. The expression of ADAM-10 and -17 correlated with the severity of the fibrosis in patients with chronic liver diseases [127]. Higher expression of ADAM-TS2 was detected in cirrhotic than in normal liver, and correlated with TGF-b1 expression [94], independently of the etiology of the liver disease. However, ADAM-TS2 expression is indirectly linked to HCV-induced fibrogenesis. In an attempt to identify non invasive markers of liver fibrosis during chronic hepatitis C, the ELF   (Enhanced Liver Fibrosis) test was put forward; it evaluates the serum levels of TIMP1, HA, N-terminal peptide of procollagen type III (PIIINP), and includes age for statistical robustness [128]. This test has been shown to properly identify moderate and severe    fibrosis in the context of chronic hepatitis C [128], and in particular the serum levels of PIIINP could differentiate between mild and moderate/high HCV-related liver fibrosis [85]. Since ADAM-TS2 excises the N-propeptide of the fibrillar procollagens types I, II, III and V, one can infer that an implicit evaluation of ADAM-TS2 expression and activity is contained in the ELF test, in link with HCV pathogenesis and fibrogenesis.

Recent genome-wide association studies, aimed at identifying genetic polymorphism behind the risk of developing HCV-related HCC, also unveiled a mechanistic scheme linking ADAMs to HCV pathogenesis. A single nucleotide polymorphism (SNP) was identified in the 5’-flanking region of MICA (Major Histocompatibility Complex class I-related chain A, or MHC class I polypeptide-related sequence A) as a susceptibility gene for HCV-induced HCC. In these patients, this SNP was associated with progression from chronic hepatitis C to cirrhosis and HCC [129]. MICA is a cell surface glycoprotein of epithelial and endothelial cells, monocytes and fibroblasts, involved in particular in  antiviral defense responses [130]. It functions as an indicator of cellular stress and activates circulating cytotoxic natural killer (NK) cells, key actors of immune surveillance,     deploying in particular anticancer activity. MICA expression is induced in cells undergoing oncogenic or viral stress [131]. In the context of HCV infection, membrane-bound MICA (mMICA) expression in HCV-infected hepatocytes was downregulated by the NS3/4A serine protease [74]. Hepatocytes expressing the cysteine protease NS2 or the polymerase NS5B exhibited decreased MICA expression when co-cultured with NK cells [70], supporting the notion that HCV NS2 and NS5B disable MICA in infected hepatocytes, thereby inhibiting the ability of these cells to respond to stimuli from NK cells. MICA is shed from the membrane by ADAM-9, -10 and -17, which generates soluble MICA (sMICA) [132][133,134]. ADAM-mediated shedding of MICA therefore leads to higher levels of sMICA and lower levels of mMICA. Since ADAM-9, -10 and -17 expression increases with the severity of fibrosis in patients with chronic liver diseases [127][131], more MICA shedding occurs in these pathologies, translated into higher levels of sMICA. Interestingly, sMICA is an inhibitor of the activity of NK cells, in particular their activity against HCC [135]. Elevated levels of sMICA were observed in patients with advanced HCC, associated with impaired activation of NK cells [131]. In the peculiar context of HCV-induced HCC, patients bearing the allele at higher risk of HCC occurrence     displayed higher sMICA levels [129]. This correlated with a poorer prognosis, and a higher risk of developing the liver disease in a context of escape from NK-mediated immune surveillance [129,131]. Finally, a correlation was found between high sMICA   levels, the higher risk allele and the development of HCC in HCV-infected cirrhotic patients who failed to develop a SVR [136]. Of note, therapies combining DAA and inhibitors of ADAMs are currently debated for this type of patients [131,135].

3.2.3. Proteoglycans (Table 3).

Proteoglycans (PGs) are membrane-associated and soluble proteins of the ECM. They are composed of long chains of sugar molecules, anchored on a short polypeptide chain: sugars can represent up to 90% of their weight. This is in contrast to matrix glycoproteins, comprising small sugar chains anchored to a long polypeptide chain, and where sugars do not represent more than 60% of their weight (see below ¶ 3.2.5.). Through their GAG chains, PGs interact with numerous regulatory molecules, such as cytokines, growth factors, hormones. This has repercussions on a myriad of normal and pathological cellular processes. PGs are classified as SLIPs (Syndecan-like integral membrane PGs), GRIPs (Glypican-related integral membrane heparan sulfate PGs), membrane-associated b-glycan and CD44, extracellular SLRPs (Small leucine-rich PGs) and hyalectans [137].

3.2.3.1. Membrane-associated PGs.

The SLIP family of PGs comprises four molecules, syndecans-1 to 4. Syndecan-1 harbors heparan and chondroitin sulfate GAGs (HS and CS, respectively), while other syndecans only bear HS. Interestingly, syndecan-1 and -4 were identified as co-factors of HCV   attachment and entry into hepatocytes [31,138,139]. Mechanistically, syndecan-1 forms with the tetraspanin CD81 a complex internalized with virions during viral entry. HCV core and syndecan-1 colocalized during the intracellular trafficking of virions [31]. When chronic infection of hepatoctytes was established, syndecan-1 expression was down-regulated [31]. This is in line with observations on biopsies collected from patients with HCC, including HCV-infected patients [140]: the down-regulation was more pronounced in the tumoral tissue than in peripheral non-tumoral zones, and correlated with the aggressiveness of the tumor. However, these data remain controversial, since other studies reported an up-regulation of syndecan-1 during cirrhosis, and in association with HCV-related HCC [141]. Syndecan-1 down-regulation correlated with the up-regulation of xylosyltransferase-2 [31], a key enzyme located in the ER/Golgi compartments,    involved in the biosynthesis of the tetrasaccharide linkage region of CS and HS PGs [142]. Xylosyltransferase-2 activity was found increased in the serum of patients with HCV-related liver fibrosis at early stages [95,96], evocative of a dramatic remodeling of PGs within the fibrotic ECM. Of note, this correlation became negative in the cirrhotic range of liver fibrosis [95]. Thus, higher enzymatic activity at the fibrotic stage could be seen as a compensatory mechanism against HCV-mediated liver injury; at the cirrhotic stage, as more damages have occurred, this compensation might become inefficient or disappear.

The GRIP family of PGs comprises the glypicans, anchored to the membrane through a glycosylphosphatidylinositol tail [143]. Among the 6 glypicans, the most studied is glypican-3 (GPC3), an oncofetal protein widely expressed during development, but   silenced in adult tissues. It is seen as a negative regulator of cell growth [144], and it is overexpressed in cancers such as embryonic tumors [145], and HCC [97]. In a cohort of HCV-infected patients with liver disease, high levels of GPC3 transcript and protein  appeared as good markers of early stages of HCC, discriminating from dysplastic   nodules [98]. Moreover, GPC3 was also strongly upregulated in early and advanced HCC, compared with normal tissue. As described for syndecan-1, GPC3 forms a complex with CD81. Mechanistic studies revealed the following loops: in resting liver, GPC3 is associated to CD81 at the plasma membrane, thereby preventing CD81 to bind the   cytosolic transcriptional repressor hematopoietically-expressed Homeobox (Hhex); this repressor therefore migrates to the nucleus, where it exerts its growth inhibitory role. In regenerating liver (notably after an injury), GPC3 binding to CD81 decreases, which leads to an increased binding of CD81 with Hhex in the cytosol; less Hhex therefore migrates to the nucleus, and proliferation is boosted [146]. During HCV-mediated liver disease and HCC, the virus, and in particular its membrane glycoprotein E2 which binds CD81, was shown to mimic GPC3 in infected hepatocytes, thereby interfering with the GPC3/CD81 binding [69]. The decrease of GPC3/CD81 entities leads to more CD81/Hhex cytosolic binding, so less nuclear Hhex. This fuels neoplastic proliferation. HCV is thus likely to enhance liver neoplasia by acting as a growth promoter of neoplastic hepatocytes, through its binding to CD81 to the detriment of GPC3 [69].

A remarkable feature of SLIPs and GRIPs is the ability of their ectodomain to be cleaved by ECM proteases, in a process called shedding. The resulting soluble domain then acts as an autocrine and/or paracrine signal in the microenvironment, per se and via the factors attached to its GAG chains. Elevated serum levels of shed syndecan-1 have been proposed as biomarkers of prediction of liver fibrosis, but not of liver inflammation in chronic hepatitis C patients [147]. Although of note, these data should be cautiously taken, since high serum levels of syndecan-1 are also observed during e.g. inflammation, sepsis, and vascular dysfunctions. Concerning GPC3, plasma levels of the shed form were found higher in patients with HCC than in healthy individuals; they were also higher in patients with HCC induced by HCV than by HBV or other etiologies [148]. The full length form of GPC3 also assayed in blood samples correlated with tumoral GPC3 expression only in patients with HCV-induced liver carcinoma. GPC3 has now demonstrated its added value as a biomarker of HCC, and has become a target of cancer   immunotherapy [149]. Indeed T cells expressing GPC3-specific chimeric antigen receptor (CAR-T) were generated, to target and destroy GPC3-positive HCC. Several clinical trials are currently ongoing to evaluate the safety and tolerability of such an immune strategy (e.g. clinicaltrials.gov identifiers NCT02395250, NCT02905188). Elevated serum levels of these shed PGs are an indication of increased shedding, and therefore of increased  systemic exposure to the factors retained in the GAG moieties, such as TGF-bs. These cytokines could then exert their fibrogenic and oncogenic effect at a distance from their site of biosynthesis. GPC3 was found to suppress the expression and signalling of TGF-b2, which led to activation of cell growth and cell cycle progression [150]; strategies aimed to suppress GPC3 expression were thus proposed as valuable options for the clinical management of GPC3-positive HCC patients. Concerning syndecan-1, paradoxical data were reported: indeed, retention of TGF-b1 and thrombospondin-1 in its GAG chains was shown to inhibit the conversion of latent TGF-b1 into its active form, thereby decreasing the availability of active TGF-b1 [151]. Due to increased shedding, shed syndecan-1 at higher amounts would enter the circulation together with its latent TGF-b1 cargo. Less activation and availability of TGF-b1 would then translate into less activation of HSCs, and therefore less synthesis of fibrogenic molecules (see also above ¶ 3.2.2.1.). This is somewhat conflictual with the correlation found between high serum levels of shed syndecan-1 and the prediction of liver fibrosis in chronic hepatitis C patients [147]. However, the precise mechanistic link between HCV pathogenesis and increased PG shedding is still missing, although the HCV-induced enhancement of protease expression could play a role (see above ¶ 3.2.2.).

The cluster of differentiation molecule 44 (CD44) is a membrane-associated adhesion glycoprotein of the ECM, considered as a “part-time PG” since covalent attachment of a CS or HS to its protein chain depends on developmental or pathological cues [152]. It is produced as standard (CD44s) and variant (CD44v) isoforms, and it is the receptor of hyaluronan (HA), a soluble non-sulfated GAG (see above ¶ 2.). Specific interactions  between CD44 and HA are of major importance for the maintenance of the proper (stem cell) niche properties under physiological conditions, and of the biology of cancer stem cells under pathological circumstances [152]. Together with other cell surface molecules, CD44 was proposed as a marker of cancer stem cells in HCC, which are responsible for tumor progression and aggressiveness. In particular, CD44s expression in HCC was  related to the TGF-b-mediated regulation of the epithelial-to-mesenchymal transition [153], and its level of expression in tumoral tissue was associated with negative prognosis [154]. In HCV-induced HCC, the variant 9 of CD44 was identified as a biomarker of liver cancer stem cells, and its expression correlated with tumor invasiveness and poor patient survival [155]. Interestingly, CD44v9 expression also correlated with the expression of markers of liver fibrosis (fibronectin) and of HSC activation (myosin light chain),   suggesting that CD44v9 could play a role at early stages of the HCV-induced liver   disease. In patients with chronic hepatitis C, high plasma levels of HA were associated to the progression of liver fibrosis, whereas low serum levels of the interferon        gamma-inducible protein 10 (IP-10) correlated with a favorable outcome to anti-HCV therapy [156]. IP-10 is a biomarker of liver inflammatory activity, correlating with lobular fibrosis and necroinflammation, but not with portal inflammation or fibrosis. From these observations, Matsuura and coworkers identified a mechanistic loop involving CD44. HCV-infected hepatocytes were found to secrete high amounts of IP-10, in response to the engagement and activation of the innate immune Toll-like receptor 2 (TLR2). This secretion depended upon the stimulation of TLR2 by HA, and induced an enhancement of CD44 expression. HA was then also able to bind CD44, and to further stimulate the production of IP-10. Additionally, CD44 was found to directly interact with TLR2 through its extracellular domain. The mechanistic loop is therefore as follows: high amounts of endogenous HA generated during HCV infection stimulate TLR2 in HCV-infected hepatocytes, which induces the production of the pro-inflammatory factor IP-10; IP-10 in turn stimulates the expression of CD44 which, as a receptor of HA and a co-receptor of TLR2, will enhance the cellular response to HA stimulation in a vicious loop aggravating the inflammatory profile [156]. Moreover, CD44 acted in synergy with TGF-b1 at early stages of cell transformation, and to confer stem properties to     transformed cells [153].

Another mechanistic loop underlying HCV pathogenesis and involving CD44 was    recently described [64]. HCV replication in hepatocytes induced the endogenous    expression of osteopontin, a matricellular protein. Osteopontin then localized at the ER, where it contributed to HCV replication, assembly and infectivity by direct binding to the HCV proteins core, NS3, NS5A, and NS5B. A pool of osteopontin was also processed by proteolysis inside HCV-infected hepatocytes, and secreted in the microenvironment where it exerted autocrine and paracrine signalling, through integrin aVb3 and CD44 binding. This latter binding activated the focal adhesion kinase which contributed to HCV replication, and to HCV assembly. Further data indicate that HCV replication was particularly enhanced in liver cancer stem cells expressing CD44 and EpCAM (epithelial cell adhesion molecule), and that osteopontin also activated HCV replication in these cells [65]. It did so by regulating the stemness of the CD44⁺/EpCAM⁺ cells, thereby   inactivating interferon signalling and fuelling HCV replication. CD44 is therefore a major actor of HCV pathogenesis, and together with its ligand osteopontin, emerges as a key regulator of cancer stem cells and HCV replication [65]. Recent studies dealing with CD44 also revealed that HCV would replicate at higher rates in liver cancer stem cells than in differentiated hepatocytes, which is of uttermost importance for the comprehension of HCV-induced liver disease, notably HCC.

3.2.3.2. Soluble extracellular PGs.

These PGs are classified in extracellular SLRPs, and pericellular PGs and hyalectans. SLRPs are extracellular PGs, and represent the largest family of PGs, characterized by a small core protein comprised of a central region of leucine-rich repeats [143]. Most SLRP members carry chondroitin-, dermatan- or keratan-sulfate chains, but a few lack any GAG. Main SLRPs comprise biglycan, decorin, asporin, fibromodulin, and lumican.  Pericellular PGs such as the heparan sulfate PGs perlecan and agrin are main components of basement membranes; in the liver, they surround blood vessels and the biliary  compartment (mature bile ducts and the HPC niche) in portal zones [143]^,[157,158]. They bind collagens (specially type IV), fibronectin and laminins (see below ¶ 3.2.5.), thereby assisting collagen fibrillogenesis, and playing a key role in ECM stabilization and in the maintenance of the functional status of mature connective tissues. Hyalectans (a.k.a.  lecticans) are components of the pericellular or interstitial matrix, residing in close    vicinity of the cell membrane, important for the communication between cells and ECM as well as for the regulation of proliferation and migration. They form a ternary binding complex with HA and the matricellular protein tenascin-R, which creates large aggregates in the ECM. The four members of the hyalectan family are versican, aggrecan, brevican and neurocan [143][159].

The SLRP biglycan is involved in collagen fibril assembly, in an ECM-bound form. After proteolytic cleavage, it is released as a soluble form. Proteomic analyses of the ECM of liver biopsies of chronic hepatitis C patients with liver fibrosis revealed a downregulation of the ECM-bound form of biglycan [59]. However, serum levels were found not significantly altered in HCV patients compared to normal controls [160]. Biglycan is a CD44 ligand, and also binds the innate immune receptors TLR2 and TLR4 [161]. It is generally considered as a pro-inflammatory molecule, however its role and the explanation behind its downregulation during HCV-induced liver disease remain unclear. As biglycan, decorin plays a role in collagen fibrillogenesis (its name comes from the fact that it  “decorates” collagen I fibrils). It binds growth factors, notably TGF-b1. The Glisson’s capsule of the liver is strongly enriched in decorin. During HCV-induced liver disease, decorin deposition was identified as an early and sensitive indicator of active ECM  remodeling after liver injury [92]. This feature correlated with increased TGF-b1     expression, for low to moderate stages of fibrosis. However, in this study, a comparable pattern was observed for liver disease induced by other etiologies. As biglycan, decorin is able to modulate TLR2 and TLR4-mediated signalling, and play a role in inflammation; however, our knowledge about decorin and (HCV-induced) liver fibrosis remains piecemeal [162]. Fibromodulin and lumican are SLRPs decorated with keratan sulfate GAG chains, and participate in collagen fibril assembly. Both molecules are biosynthesized by HSCs, and are weakly expressed in normal liver [60]. Fibromodulin transcript and protein were found overexpressed in patients with HCV-induced cirrhosis, in parallel with collagen I expression; however, in this study, a similar overexpression was observed for other etiologies of liver fibrosis. Mechanistic analyses revealed that the overproduction of fibromodulin was under the control of oxidative stress, locally    induced by the liver injury [60] (and by HCV-induced liver injury in particular [163]). Lumican was also found overexpressed in liver fibrosis [102], and in HCV-induced fibrosis in particular, on liver biopsies [84] and specifically in the ECM [59]. Lumican expression correlated with the severity of the (HCV-related) fibrosis [102][84][81], and high levels of lumican were reported in the plasma of HCV-positive patients with liver disease [84][81]. However, the exact roles of fibromodulin and lumican in (HCV-related) liver disease remain to be determined.

Of the four hyalectans, only versican exhibits a wide distribution; brevican and neurocan are only expressed in neural tissue, and aggrecan in cartilage. Versican biosynthesis is  stimulated by TGF-β1 and PDGF [143]. In the ECM, it interacts with matricellular   proteins, fibronectin and chemokines via its core domain or GAG chains. It also binds CD44, integrin β1, and at the surface of macrophages, toll-like receptors; versican is therefore involved in key cellular functions such as adhesion, proliferation, migration and invasion. It has been reported as a good biomarker of advanced liver fibrosis in  patients with HCV infection or NAFLD, since its hepatic mRNA and serum levels were higher at advanced stages of the disease than at earlier ones [93]. Concomitantly, the levels of ADAM-TS1, involved in versican proteolysis, were enhanced, and higher in liver fibrosis with than without steatosis [134]. Extracellular vesicles are nano‐sized  particles shed into body fluids by many cell types that carry various bioactive molecules; among them, plasma microvesicles (100-1000 nm in size) are key messengers of cellular communication. Enhanced microvesicle levels were linked to disease activity and   progression, e.g. microvesicles isolated from patients with HCV-related decompensated cirrhosis induce vascular hypocontractility, contributing to portal hypertension and  circulatory dysfunctions [164]. Versican-positive microvesicles were found elevated in patients with HCV-induced cirrhosis, and further increased with HCC [103]. In patients who had received DAA, the levels of versican-positive microvesicles dropped at the end of treatment and remained low throughout the 48-week follow‐up [164]. This supports the view that DAA‐induced eradication of HCV could promote a reversal of fibrosis, and versican-positive microvesicles could be a potential early biomarker of liver fibrosis.

3.2.4. Matricellular proteins (Table 3).

These proteins are non-structural components of the ECM, in contrast to structural   elements such as collagens and fibronectin. As their name indicate, they form a bridge between the matrix (matri-) and the cells (-cellular). They bind to other ECM proteins and cell surface receptors, growth factors, cytokines and MMPs, thereby modulating the   activity and accessibility of these factors, and mediating enzymatic activities to regulate ECM homeostasis [165]. Main members of this family are thrombospondins, tenascins, osteopontin (a.k.a. SPP1) and CCNs [named from the first 3 proteins identified in this family: Cysteine-rich angiogenic inducer 61 (CYR61), CTGF, and Nephroblastoma overexpressed protein (NOV)]. Notably, CCN2 is also known as the connective tissue growth factor (CTGF) [166]. Increased serum levels of CYR61 or CCN1 were proposed as biomarkers of HCV-induced HCC post-SVR [14] (see ¶ 1). The link between HCV pathogenesis and thrombospondin-1 has already been described in ¶ 3.2.2.1.

3.2.4.1 Tenascins.

There are four tenascins, C, R, W and X, but only tenascins-C and -X are ubiquitously expressed. Tenascin-C is present in all organs during fetal development, but weakly  expressed in normal adult tissue. During mechanical stress or pathological tissue remodeling, such as wound healing, fibrogenesis or tumorigenesis, its expression is enhanced. A chronic liver injury leads to the activation of HSCs, with concomitant enhanced secretion of tenascin-C in chronic hepatitis C [104]. Serum tenascin-C has therefore been suggested as a biomarker to discriminate fibrotic/cirrhotic patients with active hepatitis C from healthy controls, and from those with HCV eradication after antiviral therapy [105]. Tenascin-C serum levels were found good indicators of ongoing hepatic injury and inflammation in fibrotic/cirrhotic patients. After SVR, serum levels of this protein returned to the baseline observed in healthy individuals, suggesting the reversion of HSCs to their quiescent state [105]. More specifically, serum levels of large splice variants of tenascin-C were proposed as useful markers of the inflammatory activity during chronic hepatitis C, and in particular of the degree of piecemeal necrosis [167]. However, somewhat discordant results were reported, since tenascin was found under-expressed in the liver ECM of chronically HCV-infected patients, in correlation with the fibrosis stage [59]. In fine, no mechanistic hypothesis was put forward concerning the link   between HCV infection, liver disease and tenascin-C expression.

3.2.4.2. Osteopontin.

Osteopontin or SPP1 (secreted phosphoprotein 1) is an adhesive phosphorylated acidic glycoprotein, existing as a full length form and cleaved fragments. It is produced by HSCs, cholangiocytes, hepatocytes, HPC [168], macrophages and T helper lymphocytes, and its production is enhanced by PDGF and inflammation. Its serum levels are viewed as markers of liver fibrosis from several etiologies (including chronic hepatitis C), significantly correlating with the fibrosis stage, liver insufficiency, portal hypertension, and the presence of HCC [160,169,170][171]. In patients with chronic hepatitis C, osteopontin gene expression enhancement was part of a gene expression signature of moderate (F2) compared to mild (F1) fibrosis [82]. Mechanistic studies revealed a vicious loop linking osteopontin to HCV replication and pathogenesis: endogenous or exogenous osteopontin favors viral replication in hepatocytes [106], as well as assembly and infectivity of viral particles [64]. It likely does so through direct interactions with viral structural and non-structural proteins (core, NS3, NS5A, NS5B), in the ER and at the membrane of lipid droplets, both sites of HCV replication and assembly. Osteopontin is cleaved intracellularly, and released fragments are secreted in the extracellular space, where they bind CD44 and the aVb3 integrin in an autocrine and paracrine manner [64]. These cues stimulate HCV replication in the ER through the activation of the focal adhesion kinase (FAK), and promote HCV assembly at lipid droplets. In turn, HCV indirectly stimulates the endogenous production of osteopontin, notably through the activation of oxidative stress [172] and of cellular kinases (MAPK, JNK, PI3-K, MEK1/2). Therefore a vicious circle is generated. Further investigations revealed that osteopontin enhanced HCV  replication in a peculiar population of hepatic cells, i.e. CD44⁺ cancer stem cells [65]. Interestingly, osteopontin was described as a regulator of stemness of these cells, which are among the targets of HCV infection in the liver, and the cells of origin of liver cancers (HCC or intra-hepatic cholangiocarcinoma) [173]. This can be linked to the observation that increased serum levels of osteopontin unfavorably correlated with the early     recurrence of HCV-related HCC [174]. Osteopontin was also described as an inducer of ductular reactions [172]. In the recent study of ductular reactions in chronic liver disease of various etiologies [47], HPC (stem cells) from ductular reactions of patients with chronic and active hepatitis C displayed gene signatures related to metabolism and hepatocytes, with gene networks enriched for cell movement and receptor activity, not observed in patients with cholangitis. Osteopontin is therefore at a pivotal position: as a regulator of stemness of cells targeted by HCV during infection - in link with its activity as an inducer of ductular reactions - and as an enhancer of infection. Along similar lines, HPC play a key role in HCV pathogenesis, since they are HCV target cells and cells of origin of liver cancers. A more specific link between osteopontin, stemness, ductular  reactions and fibrosis occurrence was established [172]: osteopontin^-/- transgenic mice less readily developed chemically-induced liver fibrosis than wild type animals, suggesting that osteopontin expression is increased after liver injury. Osteopontin of the diseased liver ECM then decreased hepatocyte proliferation, induced stem cell expansion and ductular reactions, and concomitantly, fueled the up-regulation of collagen I in HSCs via its binding to aVb3 and CD44, contributing to an enhanced fibrogenic response [107,172] together with the modulation of TGF-b signaling [168]. Another vicious circle is therefore unraveled: osteopontin is directly involved in the initial response to the hepatic insult, and promotes ductular reactions, which contribute to sustain injury and to liver fibrosis.

Lastly, a genetic link between osteopontin and chronic hepatitis C was discovered;   indeed, studies of genetic polymorphism revealed SNPs in the promoter region of    osteopontin in chronic hepatitis C patients. In particular, the SNP at nucleotide -443 (C or T) showed an association with the activity of hepatitis C. This activity was defined as the serum levels of alanine aminotransferase (ALT): low (ALT > 30 IU/l), medium (30 < ALT < 80 IU/l) or high (ALT > 80 IU/l). The frequency of T/T homozygotes prevailed in the  medium- and high-activity groups, whereas C/T heterozygotes prevailed in the low-activity group [175].

3.2.4.3. CCN2 or CTGF.

CCN2 or CTGF is a mitogen and a matricellular protein expressed by hepatocytes, HSCs, vascular endothelial cells and cholangiocytes [96][176], and overexpressed in chronic liver diseases [177]. CTGF binds to sulfated GAGs and HSPGs (Figure 1), thereby acting as a local reservoir that can attract competent cells to fibrotic sites. It also plays an    autocrine role on HSCs, leading to abundant production of ECM molecules. Elevated serum and liver levels of this cytokine were found in fibrotic/cirrhotic patients chronically infected with HCV [177,178]. However CTGF levels did not correlate with the stage of fibrosis, and were higher in patients with progressive fibrosis than in those with end-stage cirrhotic liver disease [176]. HCV induced CTGF overexpression, which in turn stimulated the production of procollagen-I by HSCs, in a TGF-β1-dependent manner [57,58][179]. The viral protein core was suggested as a main trigger of CTGF overproduction [58].

3.2.5. Adhesive glycoproteins.

The main components of this family of ECM proteins are fibronectin, laminins and nidogens. They are formed of small sugar chains anchored to long polypeptide chains, and organised as structural modules enabling a great diversity of protein-protein and ECM-cell interactions. This modular organisation contributes to ECM cohesion. Fibronectin is the main component of the space of Disse in normal liver; it is produced at high rates during liver fibrosis. Fibronectin is found in the plasma as its full length form  produced by hepatocytes, and as isoforms produced by activated HSCs, defined by the presence of alternatively spliced domains [fibronectin containing extra-domains A or B (EDA or EDB), variable domain]. This variable domain is also found glycosylated at a specific threonine residue, known as the oncofetal fibronectin (oFN) isoform [108].  Laminins and nidogens are main components of basement membranes, involved in the formation of a dense network of ECM proteins. Under physiological conditions, laminins are absent from the space of Disse, but present in the basement membrane that surrounds bile ducts and are in contact with HPC [180].

3.2.5.1. Fibronectin.

Circulating levels of the fibronectin isoforms were analyzed in patients with liver fibrosis related to chronic hepatitis C or to other etiologies [108]. Total fibronectin did not correlate with any parameter in either group. However, levels of fibronectin isoforms taken individually were significantly higher in patients with chronic hepatitis C compared to healthy controls, and high levels of EDA and oFN correlated with high scores of liver  fibrosis (F2 to F4). Additionally, low levels of both isoforms were associated with the absence of significant liver fibrosis. By contrast, in patients with liver disease unrelated to HCV, none of the isoforms correlated with any parameter. When examining specific  aspects of the fibrosis, it was found that EDA elevation significantly correlated with  necrosis, and oFN predicted inflammation [181]. A slightly better correlation was found when combining oFN plasma levels with those of some elements of the complement system. Similarly the combination of oFN with a fraction of the complement correlated better with the fibrosis score than oFN alone [181].

3.2.5.2. Laminin and nidogen.

As described above, HPC produce during fibrosis an excess of fibrogenic mediators (TGF-β1 and -β2, PDGF, CTGF), which fuel HSC proliferation and activation. In an  attempt to understand how the niche of HPC influences their behaviour, Lorenzini et al. analyzed the liver tissue from various liver injury models and compared it with healthy tissue, and addressed the functional significance of the laminin - HPC interaction. In liver tissue from patients with chronic hepatitis C, a close contact between HPC, myofibroblasts and laminin was observed [182]. Laminin, produced by myofibroblasts and HPC, deposited as a sheath surrounding HPC, which contributed to the cohesion of the niche around HPC. Laminin also deposited in the space of Disse, in a phenomenon called capillarization of sinusoids observed in liver disease related to chronic hepatitis C, but not to other etiologies [183]. In vitro  experiments revealed that HPC grown on     laminin-enriched ECM maintained their stemness properties, while HPC grown on   fibronectin differentiated into hepatocytes. Laminin strongly repressed the expression of C/EBPa, an early hepatocyte gene encoding a liver-selective transcription factor [182]. Laminin therefore contributed to maintain HPC in their undifferentiated status, which could fuel the deregulation of HPC proliferation underlying liver cancer.

Nidogen was found expressed on cell membranes of biliary epithelia, and not on hepatocytes in normal liver tissue. In HCV-infected diseased tissue, nidogen staining was seen on cell membranes of periportal hepatocytes [183]. However no functional analysis was performed, which could explain such location.

3.2.6. Elastin.

Elastin is a minor ECM component in normal liver, which nevertheless plays a key role to confer strength and elasticity to the organ. It is cross-linked from tropoelastin by LOX to give rise to insoluble elastin fibers, in a similar reaction as collagen fibrillogenesis [24,27]. Elastin fibers are stable components of the ECM, although proteolytically processed by MMP-2, -9 and the elastase MMP-12. TGF-β1-mediated activation of HSCs leads to increased expression of tropoelastin [184], and the deposition of insoluble elastic fibers in the diseased liver could lead to irreversibility of fibrosis [27]. Elastin was found overexpressed in fibrotic and cirrhotic livers, compared to normal organ. This overexpression correlated with the severity of the hepatic disease [102]. In liver biopsies from patients with chronic hepatitis C, increasing amounts of elastin correlated with the severity of liver fibrosis [84][83]. This was accompanied by an increase of expression of fibulin-5 and microfibrillar-associated protein-4, proteins associated to elastic fibers and involved in elastogenesis [59][185,186]. More selectively, microfibrillar-associated protein-4 was found overexpressed in liver biopsies [84], more specifically in the ECM fraction of liver biopsies [59], and also at high levels in the serum of HCV-positive patients [84], correlating with the severity of the disease [109,110]. Mechanistically, elastin processed into small peptide fragments acts as a chemo-attractant for cells of the immune system that will locally secrete inflammatory cytokines (Figure 1). Elastin peptides can also stimulate the proliferation of activated HSCs and their secretion of MMP-12 [27], thereby aggravating the imbalance between MMP and TIMP, and contributing to fibrogenesis (Figure 2).

3.2.7. TGF-b and HCV pathogenesis (Table 3).

In most organs, such as the liver, TGF-βs are secreted as a large latent TGF-β complex, formed by the mature TGF-β protein non-covalently attached to the precursor protein (latency-associated peptide), linked to large latent TGF-β-binding proteins (LTBPs) [187]. LTBPs play a key role in the activation of this latent complex, by targeting it to the cell surface followed by its secretion into extracellular areas, where activation likely occurs, notably via thrombospondin-1 (see above [61], ¶ 3.2.2.) and furin [72][188]. Of the three TGF-β isoforms (TGF-β1, -β2 and -β3) described in mammals, TGF-β1 is the most extensively studied in human liver diseases. Under normal conditions, TGF-β1 is mainly   secreted from Kupffer cells, while hepatocytes only secrete small amounts [72]. Upon any liver injury in general, and chronic hepatitis C in particular, high levels of TGF-β1 and -β2 are secreted in the space of Disse by hepatocytes and neighboring cells [57,66], fueling fibrogenesis [34,189] (Figure 2), whereas TGF‐β3 was reported as an anti-fibrotic cytokine in the liver. Indeed, TGF-β3 was shown to inhibit TGF-β1 expression at the transcriptional level, and suppress collagen synthesis [190]. TGF-β1 and -β2 secretion contributes to HSC activation and transformation into myofibroblast-like cells [66]. Indeed TGF-β1-mediated activation of HSCs [191] results in: (i) vitamin A loss accompanied by a decrease in serum retinoid levels [192]; (ii) α-SMA expression [42,58,61]; (iii) greater TGF-β1 secretion [72], in a vicious loop. TGF-βs are therefore key regulators of fibrosis, involved in chronic liver diseases, which contribute to all disease stages, from initial liver injury to fibrosis, cirrhosis, and HCC [193].

In patients with chronic hepatitis C, increased serum levels of TGF-β1 and -β2 were  reported [87][67], together with high amounts of TGF-β1 and -β2 mRNA in liver specimens [194], compared to healthy individuals. These correlated with the degree of liver fibrogenesis but not with HCV RNA levels [67][111,195]. Higher circulating levels were also found in patients with HCC, compared to patients with liver cirrhosis, and these levels were the highest when the liver disease was caused by HCV infection or HBV/HCV co-infection, compared to other etiologies [113]. TGF-β1 mRNA in liver   biopsies strongly correlated with procollagen type I mRNA, as a further indication of the link between TGF-β1 expression and liver fibrosis/cirrhosis [111]. LTBP-1 and TGF-β1 protein levels were found upregulated in liver tissue from patients with chronic hepatitis C [112,195]. Analyses selectively performed on the ECM fraction of liver biopsies     revealed large deposits of LTBP-1 and -4 in HCV patients at stage F3, suggestive of a pivotal role of the large latent TGF-β complex at the F2-F3 transition [59]. The mechanistic crosstalk between HCV and TGF-β synthesis and activity has been investigated in various in vitro or in vivo models of chronic HCV infection. HCV-infected hepatocytes implanted into mice generated nodular tumors enriched in cancer stem-like cells. These hepatocytes were shown to recruit activated murine fibroblasts into the xenograft stroma, by secretion of TGF-β1 [194]. This led to stromal changes, such as activation of HSCs (a-SMA expression and enhanced synthesis of MMP-2), and appearance of markers of tumor-associated fibroblasts: collagen I, CTGF, vimentin, fibroblast-specific protein 1. HCV infection also resulted in overproduction of TGF-β1 in liver cells isolated from  patients with chronic hepatitis C [66]. HCV core, expressed in human hepatoma cells, strongly enhanced TGF-β1 mRNA expression in these cells, and upregulated TGF-β1 promoter [196]. The MAPK pathway was found involved in the engagement of TGF-β1 by the HCV core protein. Other studies have similarly addressed the crosstalk between this viral protein and TGF-β1-mediated fibrogenesis, as described earlier in link with enzymes of the ECM (¶ 3.2.2.1. and 3.2.2.2.). The expression of HCV core, E1 and E2 proteins in human hepatocytes revealed that core and E2 were responsible for the    enhanced secretion of TGF-β1, through the overproduction of the glucose-regulated protein 94 (GRP94), a chaperone protein of the ER lumen and an ER stress marker [66]. This further emphasizes the link between TGF-β1 production, and oxidative and ER stress, all induced by HCV [72][77]. GRP94 upregulation would engage the NF-κB pathway, which would in turn trigger the enhancement of TGF-β1 production. The  exogenous addition of the HCV NS3 protease to cultured hepatocytes led to increased expression of collagen I, TGF-β1 and TGF-β type I receptor [73] ; additionally, NS3 directly interacted with the TGF-β type I receptor at the surface of HCV-infected cells, and in HCV-infected chimeric mice, an anti-NS3 antibody attenuated HCV-induced liver fibrosis. HCV NS3 present in extracellular areas, possibly arising from the leakage from injured hepatocytes, would therefore function via its direct binding to the TGF-β type I receptor and its activation, thereby enhancing liver fibrosis [73]. Moreover, the functionally-active protease domain of NS3 was found required for TGF-β1 activity, i.e. for the activation of the TGF-β1 promoter [72]. This implies the presence of the cofactor NS4A with NS3. NS3-4A could then enhance the proteolytic activity of thrombospondin-1 and furin for the cleavage of the latent TGF-β complex and subsequent activation of TGF-β. NS3-4A also interacted with SMURF2, a negative regulator of the TGF-β signaling. Through this interaction, NS3-4A blocked the negative regulation of TGF-β signaling, thereby enhancing the cellular response to TGF-β [75]. Accordingly, hepatocytes expressing either non-structural proteins (NS3 to NS5B) or NS3-4A of HCV, once stimulated by TGF-β, exhibited a pro-tumoral transcriptional program, with enhanced expression of genes involved in cell proliferation, negative regulation of cell differentiation,  epithelial-to-mesenchymal transition and vasculature development [75], all genes related to carcinogenesis. Expressing HCV NS5A into hepatocytes also induced the production of high levels of TGF-β1, and NS5A was found important for the activation of the TGF-β1 promoter [72]. Interestingly, NS5A associates with the membrane of the ER [197], where it contributes to ER stress and induces an ER-to-nucleus signal transduction pathway involving NF-kB [77][198]. This pathway could contribute to chronic inflammation and liver fibrosis. Again, this confirms the link between HCV infection, virus-induced ER stress, and TGF-β1 overexpression and activity. As already noted for NS3, NS5A directly interacted with the TGF-β type I receptor at the surface of cells expressing NS5A [78], and together with other viral proteins, might therefore contribute to liver fibrogenesis through the engagement of TGF-β signaling.

TGF‐β2 was found upregulated by the co-expression of two proteins of HCV, core and NS2, most markedly in hepatocytes compared to HSCs and Kupffer cells [67]. Mechanistically, HCV-induced ER stress following hepatocyte infection caused the translocation of the liver-enriched and hepatocyte-specific transcriptional regulator cAMP responsive element binding protein H (CREBH) from the ER to the Golgi apparatus, and its subsequent activation. The activated CREBH was then transported to the nucleus, where it targeted and bound the DNA CREBH response element and the CRE responsive element-like sequence. Both elements modulate the expression of their target gene, TGFβ2 [190]. Through this pathway, HCV therefore enhanced TGF-β2 expression in hepatocytes [67]. After its release in the extracellular milieu, TGF-β2 exerted effects on hepatocytes and HSCs in autocrine and paracrine manners, increasing the expression of profibrotic molecules (TGF-β1 and -β2, α-SMA, collagen I), thereby promoting liver fibrosis [67].

Endoglin is a transmembrane glycoprotein belonging to the TGF-β receptors family, where it constitutes, together with betaglycan, the TGF-β type Ⅲ receptor family.   Endoglin interacts with the TGF-β receptors types I and II, and acts as a key switch by producing different variant forms, adjusting ligand affinity, and creating links with a versatile receptor network, thereby modulating the specific outcome of TGF-β-dependent and -independent pathways [199]. It is expressed as a dimer at the surface of proliferating vascular endothelial cells, of quiescent and activated HSCs, but not of hepatocytes. It binds TGF-β1 and -β2, and in HSCs, it is associated to the TGF-β type Ⅱ receptor, and shows highest expression during maximal cell activation. It is activated by phosphorylation through the activity of the TGF-β type Ⅱ receptor part. Its expression is upregulated during liver damage and transiently induced in HSC by TGF-β1. Circulating levels of endoglin were higher in patients with chronic hepatitis C than in healthy individuals [100], and correlated with the severity of the liver disease, with highest values in HCV-positive cirrhotic patients. Intra-hepatic levels of endoglin were also higher in HCV-infected livers than in uninfected biopsies [68,100], as also observed for TGF-β1 levels. Endoglin expression in the liver correlated with the METAVIR score. This   suggests that increased levels of released TGF-β1, linked to HCV-induced liver        fibrosis/cirrhosis, may be responsible for the endoglin upregulation, in a positive feedback loop where the ligand (TGF-β1) stimulates the expression of (one of) its receptor (endoglin). This could possibly create a vicious circle, with a sustained or continuous   activation of the TGF-β receptors, leading to the production of overly abundant components of the ECM by activated HSCs, and therefore to fibrosis. This poses endoglin as a key element of HCV pathogenesis, playing a crucial role in liver fibrogenesis. Mechanistic studies revealed that endoglin upregulation in HCV-infected hepatocytes was mainly due to the HCV core protein [68]. In these cells, core-induced overexpression of     endoglin contributed to the activation of the TGF-β signaling pathway, with the    transcription of target genes involved in cell proliferation and acquisition of (cancer) stem cell properties. In an attempt to discover novel determinants of HCV-related liver fibrosis progression, a joint French/Swiss study group identified an endoglin variant, Thr5Met, the frequency of which was higher among HCV/fibrosis patients than among HCV/controls [200]. This variant was even depleted in patients without fibrosis. An  additional SNP was found in a regulatory region of the endoglin gene, in link with   increased risk of liver fibrosis in HCV patients, and predicted to decrease the DNA-binding affinity of HNF4a (involved in the expression of liver-specific genes) [200]. This would lead to a lower transcription of liver-specific genes, and, as mentioned earlier, to the maintenance of hepatic progenitors in their undifferentiated state, leading to the occurrence of (HCV-infected) cancer stem cells underlying HCC [47].