ADAM10 and ADAM17 in Liver Regeneration and Fibrosis: Regulation of EGFR, c-Met and TNF RI Signalling

ADAM10 and ADAM17 are proteases that affect multiple signalling pathways by releasing molecules from the cell surface. As their substrate specicities partially overlaps, we investigated their concurrent role in liver regeneration and brosis, using three liver-specic decient mouse lines: ADAM10- and ADAM17-decient lines, and a line decient for both proteases. In the model of partial hepatectomy, double decient mice exhibited decreased AKT phosphorylation, decreased release of EGFR activating factors and lower shedding of HGF receptor c-Met. Thus, simultaneous ablation of ADAM10 and ADAM17 resulted in impaired EGFR signalling, while HGF/c-Met signalling pathway was enhanced. In contrast, antagonistic effects of ADAM10 and ADAM17 were observed in the model of chronic CCl 4 intoxication. While ADAM10-decient mice develop more severe brosis manifested by high ALT, AST, ALP and higher collagen deposition, combined deciency of ADAM10 and ADAM17 surprisingly results in comparable degree of liver damage as in control littermates. Therefore, ADAM17 deciency is not protective in brosis development per se, but can ameliorate the damaging effect of ADAM10 deciency on liver brosis development and results in decreased shedding of TNF RI, while ADAM10 deciency leads to increased levels of soluble TNF RI in serum. In conclusion, hepatocyte-derived ADAM10 and ADAM17 are important regulators of growth receptor signalling and TNF RI release, and pathological roles of these proteases are dependent on the cellular context.


Introduction
The liver is an organ with remarkable regenerative capacity. Yet, once this ability is compromised, liver transplantation often remains the only solution. Therefore, studying mechanisms contributing to the maintenance of its homeostasis is very important.
A well tolerated and widely used rodent model of liver regeneration is 2/3 partial hepatectomy [1]. In this model, two thirds of the liver are excised which induces hepatocyte proliferation in the remaining tissue to compensate for the original liver mass [2]. In the priming phase, the initial phase of regeneration after partial hepatectomy, differentiated G 0 hepatocytes are stimulated by cytokines and re-enter cell cycle.
Previous studies have shown that this inceptive part of healing process is governed to a large extent by two cytokines, tumor necrosis factor alpha (TNF-α) and interleukin 6 (IL-6) [3,4]. In the following phase, in which hepatocytes proceed through cell cycle, two growth receptor pathways were shown to play a crucial part: epidermal growth factor receptor (EGFR) [5] [6] and hepatocyte growth factor (HGF) receptor (c-Met) pathway [7][8][9]. Combined elimination of both EGFR and HGF/c-Met pathways fatally disrupted homeostasis in an unchallenged liver [10] and completely hindered liver regeneration after partial hepatectomy [11]. EGFR can be activated by different ligands: epidermal growth factor (EGF), transforming growth factor α (TGFα), amphiregulin (AREG), betacellulin, heparin-binding EGF-like growth factor (HB-EGF), epiregulin and epigen [12]. In the liver, a major source of AREG and TGFα are the hepatocytes themselves [13], HB-EGF is predominantly produced by a non-parenchymal liver cells [14], and EGF is supplied to the liver from intestinal cells [15]. Especially, HB-EGFR [16] and AREG [17] were reported as a potent stimulators of hepatocyte proliferation. In contrast to the EGFR pathway, the signalling pathway downstream of the cMet receptor is activated exclusively by HGF [18,19], which is mainly produced by a non-parenchymal liver cells [19]. Ligand binding leads to phosphorylation of growth receptors and activation of downstream mediators (such as ERK1/2, STAT3 and AKT) [2].
Several of the mentioned protein factors critical for liver regeneration are released from the cell surface by proteases. The most important of these are the membrane-bound zinc-binding proteases ADAM17 and ADAM10, whose activation leads to substrate cleavage proximal to the cell surface [20], in the process called ectodomain 'shedding'. The importance of ADAMs is highlighted by the fact that both, ADAM10- [21] and ADAM17- [22] de cient mice are not viable and die either prenatally or shortly (ADAM10) after birth (ADAM17). Observed lethal developmental defects were attributed either to insu cient Notch signalling (ADAM10-de cient mice) [21] or to aberrant EGFR signalling (ADAM17 de ciency) [22].
Although liver development was not compromised in ADAM10-or ADAM17-de cient animals, ADAM10 seems to be important regulator for liver homeostasis maintenance as liver-speci c ablation of ADAM10 resulted in spontaneous brogenesis, characterized by unbalanced differentiation of hepatocytes and deregulated expression of the bile acid transporters [23].
ADAM17 is an important regulator of TNF-α pathway. Besides cleavage of TNF-α itself, which is necessary in order to bind and activate its receptor, ADAM17 also sheds TNF receptor 1 (TNF RI) and TNF receptor 2 (TNF RII) [24], which leads to pathway inactivation. ADAM17 activity thus regulates TNF-α signalling in two directions and the outcome of ADAM17 activity on TNFα signalling therefore depends on the cellular context. Both liver macrophages and hepatocytes are a prominent source of TNF-α during liver injury [25] [26]. EGFR pathway is regulated by ADAMs [27] [28] through the release of EGFR ligands.
Surprisingly, liver-speci c ADAM17-de cient mice did not exhibit delayed hepatocyte proliferation after partial hepatectomy, even though levels of AREG were signi cantly reduced in these mice [25]. Similar to the EGFR pathway, HGF signalling is also regulated via the shedding process. Several studies have shown that the HGF receptor c-Met is cleaved by both ADAM10 and ADAM17 [29] [30].
In conclusion, ADAM10 and ADAM17 are implicated in several pathways crucial for liver regeneration. In this work, we aimed to examine the combined role of ADAM10 and ADAM17 in liver regeneration and liver brosis development, using mouse models de cient in ADAM10, ADAM17 or simultaneously de cient for both proteases.

Results
Generation of liver-speci c ADAM10-and ADAM17de cient mouse models To study the role of ADAM10 and ADAM17 in hepatic injury, we generated three liver-speci c de cient lines, using a mouse line expressing Cre under the human albumin promoter. Besides single de cient ADAM10 ΔAlb and ADAM17 ΔAlb lines, we also generated a double de cient ADAM10 ΔAlb ADAM17 ΔAlb line as both proteases share several substrates, which has been previously described as important in liver pathology. The ablation of Adam10 and Adam17 genes was veri ed by semiquantitative PCR using genomic DNA isolated from primary hepatocytes (Fig. 1a). Furthermore, lack of Adam10 and Adam17 expression was con rmed by quantitative PCR using RNA isolated from whole liver tissue (Fig. 1b).
Previously Muller et al. [23] reported that ADAM10-de cient mice, which had stronger expression of Cre driven by α-fetoprotein promotor, but same tissue speci city as our ADAM10 ΔAlb , develop spontaneous brosis. Interestingly, in our experimental setup low serum activity of alanine aminotransferase (ALT) and aspartate aminotransferase (AST), which are the markers of hepatocellular damage, indicated that neither ADAM10 ΔAlb , nor ADAM17 ΔAlb , ADAM10 ΔAlb ADAM17 ΔAlb mice developed spontaneous liver damage (Fig. 1c). Furthermore, the gross morphology of liver did not display any abnormalities in any of the mutant lines. ADAM10 ΔAlb ADAM17 ΔAlb mice have reduced AKT activation after partial hepatectomy As unchallenged animals (ADAM10 ΔAlb , ADAM17 ΔAlb , ADAM10 ΔAlb ADAM17 ΔAlb ) did not display any signi cant alteration in markers of liver damage when compared to their control littermates (Ctrl) littermates, we tested whether ADAM10/17 de ciency had an effect on model liver pathologies. First, we examined the regenerative ability of the liver by applying the model of 2/3 partial hepatectomy. The typical pattern of elevated levels of cytokine IL-6, corresponding to activation of resident liver macrophages, was observed in the serum of all treated animals three hours post hepatectomy with no difference between ADAM10 ΔAlb , ADAM17 ΔAlb , ADAM10 ΔAlb ADAM17 ΔAlb mice and Ctrl (Fig. 2a). As ADAM10/17-dependent release of c-Met into serum has been previously described to be associated with liver damage [30], we monitored levels of soluble c-Met in our model. We found that c-Met serum levels are elevated at 3 hours post hepatectomy in all lines irrespective of genotype. These levels remained signi cantly above unchallenged levels at all studied time points (3, 6, 40 and 72 hours post hepatectomy, Fig. 3d). ADAM10 ΔAlb and ADAM17 ΔAlb mice exhibited c-Met serum levels comparable to Ctrl littermates 3 hours post hepatectomy. However, shedding of c-Met was signi cantly inhibited in ADAM10 ΔAlb ADAM17 ΔAlb mice in comparison with Ctrl or single de cient lines (Fig. 2a).
To study the activation of pro-survival signals following the initial priming, we analysed the activation of AKT kinase, which plays a crucial role in proliferation of hepatocytes [31]. Immunoblot analysis of whole liver lysates of hepatectomized mice revealed comparable levels of AKT phosphorylation in ADAM10 ΔAlb and ADAM17 ΔAlb mice compared to Ctrl 6 hours post hepatectomy (Fig. 2b-c). In contrast, double de cient ADAM10 ΔAlb ADAM17 ΔAlb mice exhibited signi cantly reduced levels of phosphorylated AKT at this time point. This suggests that both ADAM10 and ADAM17 contribute to the release of factors which trigger proliferative signals after partial hepatectomy and that they are interchangeable in activation of AKT pathway.
Reduced amounts of EGFR activating factors in serum of ADAM10 ΔAlb ADAM17 ΔAlb mice 6 hours after partial hepatectomy As the EGFR pathway is a well-established target of both ADAM10 and ADAM17, we next analysed whether reduced AKT activation might result from aberrant EGFR signalling. Therefore, we rst checked the extent of release of AREG, a well-known EGFR ligand, in primary ADAM10 ΔAlb ADAM17 ΔAlb hepatocyte cultures. AREG levels in culture media were measured by ELISA at 6 hours in cultures of untreated ADAM10 ΔAlb ADAM17 ΔAlb and Ctrl hepatocytes in the presence of 10% FBS. ADAM10 ΔAlb ADAM17 ΔAlb primary hepatocytes exhibited a dramatically reduced ability to effectively shed AREG (Fig. 2d).
Compromised AREG shedding in ADAM10 ΔAlb ADAM17 ΔAlb hepatocytes prompted us to check the release of EGFR ligands in vivo -in serum from peripheral blood of mice after partial hepatectomy. To evaluate the effectiveness of EGFR ligand shedding, we incubated Ctrl hepatocytes in the presence of serum obtained from ADAM10 ΔAlb ADAM17 ΔAlb mice or Ctrl 6 hours post hepatectomy for 30 minutes and determined its ability to trigger EGFR phosphorylation. Immunoblot analysis revealed that ADAM10 ΔAlb ADAM17 ΔAlb triggered signi cantly lower phosphorylation of EGFR than serum from Ctrl ( Fig. 2e-f). This result strongly suggests that the shedding of EGFR ligands is compromised in ADAM10 ΔAlb ADAM17 ΔAlb mice at the beginning of the regeneration process in the liver. ADAM10 ΔAlb ADAM17 ΔAlb hepatocytes show no defect in proliferation in response to partial hepatectomy Lower levels of AKT phosphorylation in ADAM10 ΔAlb ADAM17 ΔAlb at early time points after hepatectomy suggested a slower proliferative response; therefore we next examined hepatocyte proliferation in the following stages of liver regeneration.
To follow liver mass restoration, we monitored changes in liver weight to body weight ratio in ADAM10 ΔAlb ADAM17 ΔAlb mice consecutively three days after hepatectomy. In this time frame, ADAM10 ΔAlb ADAM17 ΔAlb mice gained liver mass to comparable extent as Ctrls, reaching increases of 50 % within 72 hours (Fig. 3a). To examine whether the described increase re ected the rate of hepatocyte division, we next analysed hepatocytes proliferation using immunohistological staining of proliferating cell nuclear antigen (PCNA) (Fig. 3b). For quanti cation, only nuclei corresponding to hepatocytes, identi ed by their typical round shape, were counted, while all non-parenchymal cells were excluded from the analysis. Surprisingly, no difference in number of PCNA positive nuclei was detected between ADAM10 ΔAlb ADAM17 ΔAlb and Ctrl mice at 40 hours post hepatectomy (Fig. 3b). This suggests that despite the initial decrease in the activation of EGFR pathway, ADAM10 ΔAlb ADAM17 ΔAlb mice are able to proceed with hepatocyte proliferation as effectively as Ctrls.
HGF triggers stronger signalling through cMet in ADAM10 ΔAlb ADAM17 ΔAlb hepatocytes To explore the ability of ADAM10 ΔAlb ADAM17 ΔAlb to compensate for lower EGFR activation, we examined other signalling pathways important for hepatocyte proliferation. The HGF/c-Met pathway was a prospective candidate, as it plays an important role in liver regeneration and c-Met is a substrate of both ADAM10 and ADAM17. To nd out whether HGF/c-Met pathway is dysregulated in ADAM10 ΔAlb ADAM17 ΔAlb mice in the model of partial hepatectomy, we next analysed serum levels of both HGF and the soluble form of the HGF receptor c-Met. ELISA analysis revealed a prominent increase of serum HGF levels mice 40 hours post hepatectomy, however no signi cant difference was observed in any of the studied mouse lines, ADAM10 ΔAlb , ADAM17 ΔAlb , ADAM10 ΔAlb ADAM17 ΔAlb compared to their Ctrl littermates (Fig. 3c, S1). Together with increased levels of HGF, levels of c-Met were signi cantly elevated in Ctrl animals as well as in ADAM10 ΔAlb and ADAM17 ΔAlb lines (Fig. 2a, S1). However, we found that the level of soluble c-Met in serum of ADAM10 ΔAlb ADAM17 ΔAlb was signi cantly reduced in comparison to Ctrls at all studied time points (Fig. 3d, S1).
Reduced c-Met levels in serum correspond to compromised shedding in ADAM10 ΔAlb ADAM17 ΔAlb mouse and may imply a higher level of remaining functional c-Met on hepatocyte surface leading to ampli ed HGF signalling. To address this hypothesis in vitro, we isolated primary hepatocytes from ADAM10 ΔAlb mice exhibit higher susceptibility to CCl 4− induced brosis As we identi ed alterations in signalling pathways during liver regeneration in ADAM10 ΔAlb ADAM17 ΔAlb mice, we were further interested how the de ciency of ADAM10 and ADAM17 may affect the development of liver injury. We therefore applied a model of 4-week CCl 4 -intoxication, which results in liver brosis.
As expected, markers of liver injury (serum activity of alanine aminotransferase (ALT) and aspartate aminotransferase (AST)), were elevated in all studied groups after 4 weeks of CCl 4 treatment. Strikingly the highest hepatocellular damage was found in ADAM10 ΔAlb animals ( Fig. 4a-b), while ADAM17 ΔAlb mice exhibited the same extent of damage as their Ctrl littermates. Interestingly, serum aminotransferases were not increased in ADAM10 ΔAlb ADAM17 ΔAlb in contrast to single ADAM10 ΔAlb mice. We further evaluated the development of brosis in liver by histopathological staining of brillary collagens, where the Sirius Red positive (SR+) area was determined by image analysis. Our results showed that higher levels of serum liver damage markers correlated with most extensive deposition of brillary collagens in ADAM10 ΔAlb mice when compared to other de cient lines and Ctrls (Fig. 4c). Sirius Red staining (Fig. 4d) showed that at this time point ADAM10 ΔAlb developed a more advanced bridging phase of brosis, in which collagen scars already interconnected central veins. The extent of collagen deposits of ADAM17 ΔAlb and also ADAM10 ΔAlb ADAM17 ΔAlb were comparable to Ctrls. Fibrotic scars in these lines stretched from the central vein but were not prominent enough to form bridges. Moreover, ki-67 staining revealed signi cantly higher hepatocyte proliferation (Fig. 4e-f) in the ADAM10 ΔAlb mouse line, which is most likely caused by a proliferative response to more severe damage caused by CCl 4 .
A prior study [23] has shown that ADAM10 downregulation decreased expression of several bile acid transporters, including Mrp2. As other members of Mrp family, Mrp3 and Mrp4, are involved in detoxi cation of CCl 4 , we analysed their expression at the mRNA level in our model using qPCR. In contrast, in our model ADAM10 ΔAlb mice displayed no signi cant differences in Mrp2 liver expression after 4-weeks of CCl 4 -intoxication, as well as in Mrp3 and Mrp4 (S2). Hence, the higher susceptibility of ADAM10 ΔAlb towards CCl 4 treatment is not a consequence of the difference in transporter proteins expression.

ADAM10 ΔAlb biliary epithelial cells are affected after CCl 4 intoxication
Serum analysis revealed that ADAM10 ΔAlb mice exhibit not only higher levels of hepatocellular damage markers (ALT and AST), but also markers of biliary damage, such as alkaline phosphatase (ALP). ALP activity was elevated in the serum of ADAM10 ΔAlb mice after 4 weeks of CCl 4 treatment in comparison with Ctrl, but not in ADAM17 ΔAlb or ADAM10 ΔAlb ADAM17 ΔAlb (Fig. 5a). To determine whether elevated ALP levels are connected to hyperproliferation of biliary epithelia, we stained liver sections with CK19, a marker of biliary epithelial cells. ADAM10 ΔAlb livers did not differ in the CK19 positive (CK19+) area from Ctrls, and neither did ADAM17 ΔAlb or ADAM10 ΔAlb ADAM17 ΔAlb (Fig. 5b). On the other hand, higher levels of CD44 protein, a molecule overexpressed in proliferating biliary epithelium and a substrate of ADAM10, were detected in lysates from ADAM10 ΔAlb livers than in Ctrls (Fig. 5c). A higher amount of CD44 was not determined in lysates from ADAM10 ΔAlb ADAM17 ΔAlb livers. The observed activation of biliary epithelial cells can be caused by a higher degree of overall liver damage, caused by chronic intoxication in ADAM10 ΔAlb mice. Importantly, ADAM17 de ciency overcomes the biliary damage enhanced by ablation of ADAM10.
Shedding of TNF RI is affected by de ciency of ADAM17 and ADAM10 Previous studies have shown that CCl 4 -induced liver damage is connected with elevated levels of soluble TNF RI (sTNF RI) and TNF RII [32]. As TNF receptors are well-described substrates of ADAM17, we were interested how ablation of ADAM17 and ADAM10 in hepatocytes affects levels of sTNF RI in our models.
In consistency with previous reports, we observed an increase of sTNF RI serum levels after 4 weeks of CCl 4 treatment in all studied groups in comparison to untreated animals. As expected, ADAM17 ΔAlb displayed signi cantly lower serum levels of sTNF RI than Ctrl littermates (Fig. 6a). Interestingly, ADAM10 ΔAlb ADAM17 ΔAlb mice did not show signi cant differences to Ctrls. Surprisingly, ADAM10 ΔAlb exhibited the opposite situation with higher levels of TNFRI in serum than Ctrl littermates (Fig. 6a).
Increase of sTNF RI was found also in the serum of unchallenged ADAM10 ΔAlb animals (Ctrl = 740 pg/ml, ADAM10 ΔAlb = 1165pg/ml, p = 0.0012), as well as ADAM10 ΔAlb primary hepatocytes without any challenge (Fig. 6b). To exclude that identi ed differences originated from a changed TNF RI expression, we performed qPCR using RNA from whole liver tissue. No differences in Tnfr1 transcript were found between individual mouse lines (Fig. 6c), indicating that observed differences resulted from alteration in sTNF RI release.
While ADAM17 ΔAlb mice exhibit reduced shedding of sTNF RI and ADAM10 ΔAlb mice exhibit an increased release of sTNF RI into serum, the situation in ADAM10 ΔAlb ADAM17 ΔAlb was dependent on the type of the liver challenge. Under physiological conditions, we determined higher levels of sTNF RI ADAM10 ΔAlb ADAM17 ΔAlb compared to Ctrl (Fig. 6d), similar to ADAM10 ΔAlb . However, after partial hepatectomy, ADAM10 ΔAlb ADAM17 ΔAlb resembled ADAM17 ΔAlb with hampered release of sTNF RI (Fig. 6d). ADAM10 ΔAlb ADAM17 ΔAlb primary hepatocytes exhibit inhibited shedding of TNF RI, similarly to ADAM17 ΔAlb . As isolation of primary hepatocytes is linked to their activation due to the treatment with digestive enzymes, we can consider their state closer to the conditions in hepatectomized rather than unchallenged liver. A regulatory role of sTNF RI in liver pathological states is emerging [33] [34] [35]. Our data suggest that both ADAM10 and ADAM17 in uence its release.

Discussion
ADAM10 and 17 are metalloproteinases that modulate signalling in many physiological processes. For the liver regeneration, EGFR and c-Met pathways were identi ed as the most important ones. Several recent articles showed that only concurrent inactivation of both pathways causes inability of liver to recover from the partial hepatectomy and higher necrosis after CCl 4 intoxication [11] [36] [7]. In this study, we demonstrate that both pathways are modulated by ADAM proteases in a pathological liver.
Simultaneous hepatocyte-speci c de ciency of both ADAM10 and ADAM17 led to lower activation of EGFR as well as to the lower shedding of a c-Met receptor. Thus, initial slower regenerative response after partial hepatectomy, caused by impaired shedding of EGFR activating ligands, was compensated by stronger HGF signalling through c-Met receptor. Furthermore we show, that both proteases are also involved in TNF RI release. ADAM17 directly cleaves the receptor, while ADAM10 de ciency causes elevation of TNF RI serum levels through an unidenti ed process.
Several EGFR ligands promote the proliferation of hepatocytes. It has been described that expression of HB-EGF, TGF-α and AREG is elevated after partial hepatectomy, even though they are not completely fungible [37]. ADAM17 has been described as the main protease in the release of TGF-α and AREG [27].
HB-EGF is shed mainly by ADAM17 and ADAM12, and betacellulin is a substrate of ADAM10 [27]. In addition, EGF [38] supplied to the liver from the duodenum, was identi ed as a substrate of ADAM10 [39]. We show that combined deletion of ADAM10 and ADAM17 in hepatocytes leads to lower levels of EGFR activating factors in the serum from peripheral blood after 2/3 partial hepatectomy. This was demonstrated by lower levels of phosphorylated AKT in ADAM10 ΔAlb ADAM17 ΔAlb mice 6 hours post hepatectomy, which was not observed in single de cient mice. Moreover, decreased phosphorylation of EGFR was observed in the cultured cells treated by mouse serum originating from ADAM10 ΔAlb ADAM17 ΔAlb mice after hepatectomy. Altogether this indicates that hepatocytes alone are a signi cant source of growth factors which modulate their cell cycle. Similar results were described by McMahan et al. [33] in hepatocyte-speci c ADAM17-de cient mice, in which dramatically reduced levels of AREG were observed. However, AREG levels in their study were measured in liver tissue lysates and therefore represented the expression of a protein, not its shed pool. In our model we did not observe differences of expression in AREG protein but we clearly showed that the deletion of ADAM proteases in uenced release of EGFR ligands to the serum. Our data indicate that not only ADAM17, but also ADAM10 is a regulator of EGFR ligand shedding after partial hepatectomy. Of note is that factors released from hepatocytes can potentially in uence also other liver cell types, e.g. hepatic stellate cells.
This may have a negative impact in conditions such as liver brosis when enhanced HSC proliferation is not desired.
Similarly to EGFR, HGF signalling through c-Met receptor is another independent pathway leading to the proliferation of hepatocytes [7] [40]. In our model, serum HGF was strongly increased after partial hepatectomy. On the other hand, we showed that shedding of c-Met was also elevated, especially in the rst hours after the hepatectomy, which is probably part of a regulatory mechanism of HGF signalling. We observed that isolated primary ADAM10 and ADAM17 double-de cient hepatocytes had a more prominent response to HGF treatment. We concluded that these differences in HGF response were caused by higher availability of c-Met receptor on cell surface due to its limited shedding in the combined absence of ADAM10 and ADAM17. Shedding of c-Met did not signi cantly differ in the absence of ADAM10 or ADAM17 individually. The question remains whether shedding of the c-Met receptor results only in lower availability of functional receptor on the cell surface, or if there is an additional contribution of a solubilized part of the receptor. Released cleft fragment of c-Met could bind HGF and prevent binding of HGF to functional surface-bound receptors. We did not identify this decoy function of a c-Met fragment in our in vitro conditions, but further investigation is desired.
In summary, the deletion of ADAM10 and ADAM17 driven by the albumin promoter leads to hampered EGFR signalling, but enhanced HGF/c-Met signalling in regenerating hepatocytes. As a result, ADAM10 ΔAlb ADAM17 ΔAlb mice did not show any alteration from Ctrl littermates 40 hours post hepatectomy, hepatocytes proceeded towards proliferation and liver weights increased normally.
We have reported previously that liver-speci c ADAM10-de cient mice, with stronger expression of Cre (Alb-Cre homozygotes or α-fetoprotein-Cre hemizygotes), developed spontaneous brosis [23]. This was not the case in mice used in this study (Alb-Cre hemizygotes). ADAM10 ΔAlb mice were monitored up to 33 weeks of age, and none of the brotic markers were increased in unchallenged animals. The lower expression levels of Cre in hemizygote animals resulted in sustained expression of ADAM10 in a negligible proportion of hepatocytes, which was su cient to protect from development of spontaneous brosis. However, ADAM10 ΔAlb mice were much more susceptible to liver damage than control littermates when exposed to the toxic agent. Difference was not apparent after acute intoxication (single dose of CCl 4 ) but was clearly distinct after 4 weeks of chronic CCl 4 treatment. The reason for the higher susceptibility of ADAM10-de cient mice towards CCl 4 intoxication is still not clear. It was shown that ADAM10 directly in uences the expression of bile acid transporters, e.g. MRP2 [23], which play an important role in removal of toxic products generated in hepatocytes after CCl 4 treatment. However, we did not nd the difference in Mrp2 mRNA expression, nor in the expression of Mrp3 and Mrp4 which are involved in CCl 4 detoxication [41]. The difference in Mrp2 expression between two studies may have arisen due to the differing nature of the injuries evoked in the studies, as we used toxin-induced injury, while model of Muller et al. [23] resulted from an inner imbalance of liver homeostasis.
Not only levels of ALT and AST were increased in ADAM10 ΔAlb , but also ALP. ALP in liver damage is connected with activated biliary epithelial cell. This is in line with previously published data which suggest that ADAM10-de cient mice have increased ductular reaction [23]. Consistently with this, we found that ADAM10 ΔAlb had increased protein levels of CD44 in liver lysates. Biliary epithelial cell are a prominent source of CD44 in liver [42], and its expression is connected to the proliferation of biliary epithelia. The identi ed increase of CD44 levels could be directly linked to ADAM10 de ciency, as ADAM10 is the primary sheddase for CD44 [43].
Interestingly, the hepatocyte damage in ADAM10 ΔAlb ADAM17 ΔAlb and ADAM17 ΔAlb mice was comparable to Ctrl littermates. Thus, ADAM17 elimination was not protective in CCl 4 induced liver brosis when functional ADAM10 was present, but alleviated aggravating impact of ADAM10 de ciency on brosis development. Similarly, we have shown previously [44] that ADAM17 inhibition by ursodeoxycholic acid is protective in model of cholestatic injury. In contrast, recent work from Sundaram et al. [33], showed that impaired ADAM17 maturation exacerbated bile duct obstruction induced brosis. The main difference between their work and our present ndings is the cell type affected by genetic models. Our model leads to speci c ablation of ADAM17 in hepatocytes and biliary epithelial cells. Sundaram et al. studied mice with whole body de ciency of iRhom2 that leads to the inhibition of ADAM17 in cell types, in which iRhom2 governs maturation of ADAM17. They explained their phenotype through the suppression of ADAM17 in hepatic stellate cells, a cell type which was not affected in our genetic model. This shows that the modulation of ADAM17 activity on different cell types within the liver can lead to opposite pathological consequences.
In summary, ADAM10 ΔAlb are more prone to chronic intoxication, but subsequent proliferation of hepatocytes is not affected. Interestingly, ADAM17 de ciency counteracted the negative effect of ADAM10 de ciency in liver brosis development. ADAM10 ΔAlb ADAM17 ΔAlb animals were the same as Ctrl littermates in all studied parameters during brosis development. Further deciphering of molecular mechanism behind this effect represent interesting area for future research.
Role of ADAM17 in TNF-α shedding is very well-known. However, growing number of evidence suggest the importance of TNF receptors shedding in the regulation of TNF-α signalling [32]. Shedding of TNF RI or TNF RII not only disables functionality of this receptor on the cell surface but released part -soluble receptor, can bind and limit the availability of TNF-α. Role of TNF RI in liver injury is not easy to interpret. Inhibition of TNF RI led to lower hepatocellular damage after CCl 4 treatment [45]. On the other hand, the different report described the importance of TNF RI or liver regeneration after CCl 4 intoxication, as mice lacking TNF RI showed inhibited hepatocyte DNA synthesis [46]. A recent study showed that TNF RI de ciency led to the increased cholestatic injury in the model of chronic hepatitis, mediated through higher in ltration of immune cells [47]. In our study, we found that ADAM17 de ciency in hepatocytes leads to lower levels of TNFRI in serum in both studied liver challenges, as expected. Surprisingly, ADAM10-de cient animals consistently show higher levels of TNF RI in serum than their Ctrl littermates. Same results were con rmed in vitro on primary hepatocytes, suggesting elevated serum levels of sTNF RI in vivo had a source in ADAM10-de cient hepatocytes. This effect is most likely not caused by compensational activation of ADAM17, as other substrates of ADAM17, TNF-α and AREG, were not elevated in the medium of ADAM10-de cient cells. Moreover, ADAM10 ΔAlb ADAM17 ΔAlb mice show an interesting pattern in levels of TNF RI in serum. In an unchallenged state, ADAM10 ΔAlb ADAM17 ΔAlb exhibit increased shedding of TNF RI similarly to ADAM10 ΔAlb , while during the liver challenge, double de cient mice were comparable to ADAM17 ΔAlb or Ctrl littermates. This nicely shows that the activity of ADAM17 is induced in liver pathological states and the lack of this induction becomes evident in ADAM10 ΔAlb ADAM17 ΔAlb after the challenge, while unchallenged animals resemble more ADAM10 ΔAlb in respect to TNF RI shedding. Besides proteolytic shedding, soluble TNF RI can be released into serum by exosome-like vesicles [48]. It would be worth to investigate whether ADAM10 de ciency alters this export.
Altogether our data show, that hepatocyte-derived ADAM10 and ADAM17 contribute to the regulation of several distinct pathways, i.e. EGFR, HGF and TNF RI, during liver injury and regeneration. Potential therapies based on inhibition of ADAM10 and ADAM17 should carefully consider various roles of these proteases in different cell types.

Mouse models
Mouse lines ADAM10 ox on 129/C57Bl/6 background [49] (kindly provided by Dr. Andreas Ludwig, RWTH Aachen University) and ADAM17 ox C57Bl/6J [50] (kindly provided by Dr. Carl Blobel, HSS Research institute) were rst backcrossed to C57Bl/6N. N4 generations were crossed together to create ADAM10 ox ADAM17 ox line. All three oxed stains were crossed to line expressing Cre under the control of albumin promoter, Alb-Cre (B6.Cg-Tg(Alb-cre)21Mgn/J; The Jackson Laboratory). For liver challenges 9-14 weeks old males were used. Animals used in all experiments were homozygous for ox alleles and hemizygous for Cre recombinase and were termed ADAM10 ΔAlb , ADAM17 ΔAlb , ADAM10 ΔAlb ADAM17 ΔAlb . Littermates without Cre gene were used as control group (Ctrl) in each line.
Model of 2/3 partial hepatectomy was performed as described previously [51] with slight modi cations. Brie y, mice were anesthetized with mixture of Zoletil (Virbac, Carros, France) and Rometar (Bioveta, Ivanovice, Czech Republic), frontal lobes and left lateral lobe are ligated with silk suture and subsequently removed. For sample harvesting mice were anesthetized at indicated time points after the surgery, blood was collected from orbital plexus, animals were subsequently euthanized by cervical dislocation and liver samples were taken from de ned areas of remaining lobes -one piece was xed for histological analysis (see below) and several smaller pieces were snap frozen and kept in -80°C for RNA and protein isolation.
To induce brogenesis, mice were injected intraperitoneally with 1µl/g of CCl 4 (Sigma-Aldrich, Darmstadt, Germany) diluted 1:3 with olive oil twice a week for 4 weeks. Liver and blood samples were collected 48 hours after the last injection as described for partial hepatectomy model.
The animals were handled in accordance with the Guide for the Care and Use of Laboratory Animals, approved by the Animal Care and Use Committee of the Academy of Sciences of the Czech Republic.
Mice were kept under standard laboratory conditions with free access to food and water.

Primary hepatocyte isolation and treatment
Hepatocytes were isolated by perfusion with collagenase I as described [52] [53]. Hepatocytes were plated on collagen I (0.3 mg/mL) and let to attach in DMEM containing 10% fetal bovine serum (FBS), 1% penicillin-streptomycin and 0.08 U/mL insulin (culture medium) for 24 hours before proceeding with experiments.
For shedding assay, cells were maintained in fresh culture medium for 6 hours. Harvested medium was cleared from debris by centrifugation (10,000 x g/10 min) and analysed by ELISA. For stimulation of primary hepatocytes with mouse serum, cells were kept in culture media without FBS for 6 hours, then treated for 30 min with DMEM medium supplemented with 10% of mouse sera and subsequently lysed with RIPA buffer. Samples were further processed for immunoblotting and ELISA.

Histological Analysis
Specimens were xed in 4% buffered formaldehyde, embedded in para n, sectioned, and stained with hematoxylin and eosin (H&E), Sirius Red [54], or processed for immunohistochemistry. To visualize proliferating cells, sections were stained with antibodies against ki-67 (Dako, Santa Clara, CA) and PCNA (Santa Cruz Biotechnology, Dallas, TX). For the analysis, 10 optical elds (original magni cation 400 x) were taken per animal and positively stained nuclei were counted. Sirius red staining was quanti ed using ImageJ software [55].

Analysis of Serum Parameters
Collected blood was allowed to cloth for 1 hour in 37°C, spun down (10,000 g /20 min), and cleared serum was kept frozen before further analysis. Serum levels of IL-6, KC, soluble c-Met and soluble TNFreceptor I were assayed by ELISA using antibodies from R&D Systems (Minneapolis, MN). Activities of alanine aminotransferase (ALT), aspartate aminotransferase (AST) and alkaline phosphatase (ALP) were assessed using kits from Roche Diagnostics (Basel, Switzerland).

Immunoblotting
Frozen pieces of liver tissues were homogenized in RIPA buffer using TissueLyzer II (Qiagen, Hilden, Germany). Equal amounts of protein as assessed by BCA assay were resolved on SDS polyacrylamide gels and transferred to nitrocellulose membranes; proteins were detected using antibodies against