Protease profiling of liver fibrosis reveals the ADAM metallopeptidase with thrombospondin type 1 motif, 1 as a central activator of transforming growth factor beta


  • Potential conflict of interest: Nothing to report.

  • This work was supported by the Institut National de la Santé et de la Recherche Médicale, the Ligue Nationale Contre le Cancer, the Association pour la Recherche contre le Cancer, and the Région Bretagne (PRIR no.: 3193). K.B. and A.L were recipients of a postdoctoral fellowship from the Région Bretagne and a Ph.D. fellowship from the Ligue Nationale Contre le Cancer, respectively.


During chronic liver disease, tissue remodeling leads to dramatic changes and accumulation of matrix components. Matrix metalloproteases and their inhibitors have been involved in the regulation of matrix degradation. However, the role of other proteases remains incompletely defined. We undertook a gene-expression screen of human liver fibrosis samples using a dedicated gene array selected for relevance to protease activities, identifying the ADAMTS1 (A Disintegrin And Metalloproteinase [ADAM] with thrombospondin type 1 motif, 1) gene as an important node of the protease network. Up-regulation of ADAMTS1 in fibrosis was found to be associated with hepatic stellate cell (HSC) activation. ADAMTS1 is synthesized as 110-kDa latent forms and is processed by HSCs to accumulate as 87-kDa mature forms in fibrotic tissues. Structural evidence has suggested that the thrombospondin motif-containing domain from ADAMTS1 may be involved in interactions with, and activation of, the major fibrogenic cytokine, transforming growth factor beta (TGF-β). Indeed, we observed direct interactions between ADAMTS1 and latency-associated peptide-TGF-β (LAP-TGF-β). ADAMTS1 induces TGF-β activation through the interaction of the ADAMTS1 KTFR peptide with the LAP-TGF-β LKSL peptide. Down-regulation of ADAMTS1 in HSCs decreases the release of TGF-β competent for transcriptional activation, and KTFR competitor peptides directed against ADAMTS1 block the HSC-mediated release of active TGF-β. Using a mouse liver fibrosis model, we show that carbon tetrachloride treatment induces ADAMTS1 expression in parallel to that of type I collagen. Importantly, concurrent injection of the KTFR peptide prevents liver damage.


Our results indicate that up-regulation of ADAMTS1 in HSCs constitutes a new mechanism for control of TGF-β activation in chronic liver disease. (HEPATOLOGY 2011)

Liver fibrosis is a wound-healing response to chronic liver injuries, including viral infection, alcohol consumption, and metabolic diseases.1 Persistent regeneration stimuli lead to an excessive accumulation of extracellular matrix (ECM) and disorganized liver architecture. As the main cellular source of ECM, hepatic stellate cells (HSCs) play a critical role in hepatic fibrosis and, after injury, undergo an “activation” process that consists of the transition from quiescent vitamin A–rich cells in the healthy liver to proliferating, fibrogenic, and contractile myofibroblasts.2 HSCs also drive ECM remodeling by providing matrix metalloproteinases (MMPs) and tissue inhibitors of MMPs (TIMPs). MMPs have been implicated in the breakdown of normal matrix during the early steps of fibrosis, facilitating its replacement by scar matrix, whereas an increase in the synthesis of TIMPs blocks collagenase activities in advanced stages of fibrosis.3 More recently, we and others have reported the altered expression of other metallopeptidases, including members of the A Disintegrin And Metalloprotease (ADAM) protein family and the related proteins with thrombospondin mombospondin motifs (ADAMTSs), thereby leading to a more complex view of metalloprotease involvement in fibrosis.4-8

ADAMs constitute a family of cell-surface proteins involved in ectodomain shedding, cell adhesion, and cell signaling. ADAMs share a multidomain organization that includes metalloprotease, disintegrin, cystein, transmembrane, and cytoplasmic domains9 and have been implicated in diverse biological processes, including spermatogenesis/fertilization, neurogenesis, inflammatory responses, and cancer.10 ADAMTSs, and their related forms, ADAMTSLs (ADAMTS-like molecules that lack proteolytic activity), are characterized by an ancillary domain containing one or more thrombospondin type 1 repeat.11 Unlike mammalian ADAMs that are, with the exception of variant forms of ADAM-12 and -28, transmembrane proteins, ADAMTSs are secreted molecules that associate with ECM components. ADAMTS proteases are involved in the maturation of procollagen and von Willebrand factor, as well as in ECM proteolysis relating to morphogenesis, angiogenesis, fertility, arthritis, and cancer.

Metallopeptidases are the most diverse class of human proteases. Their expression is altered in various pathologies12, 13 and they constitute promising therapeutic targets. However, the development of successful therapies has been underwhelming, most likely because of redundancy, substrate diversity, and complex regulation of activities. Future design of specific inhibitors, some of which might possibly target extracatalytic sites or adaptor proteins,14, 15 hence requires more studies to define cellular expression profiles and molecular mechanisms involved in their activities.

Here, we investigated protease involvement in chronic liver diseases by using a protease-related gene array. Sixty-eight genes were significantly deregulated in liver fibrosis, and an integrative data-mining study of overexpressed genes identified ADAMTS1 as a new component of this protease-related network. Up-regulation of ADAMTS1 was associated with HSC activation. Interaction of ADAMTS1 with the latent form of transforming growth factor beta (TGF-β), latency-associated peptide-TGF-β (LAP-TGF-β), led to TGF-β activation, suggesting a pivotal role for ADAMTS1 in promoting TGF-β activity in liver fibrosis. In line with this conclusion, we show that induction of hepatic damage in a mouse liver fibrosis model is inhibited by treatment with the ADAMTS1 KTFR peptide that is implicated in TGF-β activation.


ADAM, A Disintegrin And Metalloprotease; ADAMTS, ADAM metallopeptidase with trombospondin type 1 motif; alpha-SMA, α-smooth muscle actin; ALT, alanine aminotransferase; AST, aspartate aminotransferase; CCl4, carbon tetrachloride; ECM, extracellular matrix; HBV, hepatitis B virus; HCV, hepatitis C virus; HSC, hepatic stellate cell; LAP-TGF-β, latency-associated peptide-TGF-β; MMP, matrix metalloproteinase; qRT-PCR, quantitative reverse-transcriptase polymerase chain reaction; scr, scrambled; SHG, second harmonic generation; TGF-β, transforming growth factor-beta; TIMP, tissue inhibitor of MMP; TPEF, two-photon excitation fluorescence; TSP1, thrombospondin type 1 motif.

Materials and Methods

Sample Tissues.

Matching nontumor liver samples (n = 32) were obtained from patients undergoing surgical hepatectomy or liver transplantation for hepatocellular carcinoma, as previously described.16 Controls were obtained from nontumor liver samples complicated with colorectal metastases (n = 10). Histological stages of fibrosis were graded according to the METAVIR score: F1, portal fibrosis without septa; F2, portal fibrosis with rare septa; F3, numerous septa without cirrhosis; and F4, cirrhosis. Access to this material was in agreement with French regulations and satisfied the requirements of the local ethics committee.

Other Methods.

Animal models, cell culture and transfections, DNA microarray experiments, messenger RNA (mRNA) quantification by quantitative reverse-transcriptase polymerase chain reaction (RT-qPCR), western blotting and immunoprecipitation, immunostaining and imaging, transcriptional reporter assays, TGF-β, collagen quantification, and bioinformatics tools are described in Supporting Materials and Methods.


ADAMTS1 is a Major Node of Up-regulated Protease Gene Networks in Fibrotic Livers.

To explore the degradosome associated with chronic liver disease, we performed gene-expression profiling for 32 patients with hepatic fibrosis associated with hepatitis C virus (HCV) infection (n = 16) or alcohol abuse (n = 16). Using a dedicated gene microarray, we identified 69 deregulated genes, including 10 metallopeptidases, 3 TIMPs, and 9 members of the serine protease inhibitor family, related to protease activities in fibrosis tissues, compared to a pool of 10 histologically healthy liver samples (Supporting Table 1). This approach, which yields a listing of candidate genes, was complemented by the integration of both DNA microarray data and array-independent literature mining. Forty-two genes were clustered after prefiltering for genes connected with at least two members of the input set, according to PubMed abstracts (Fig. 1; see Supporting Information for details). The network graph of gene connections showed two major nodes, MMP2 and ADAMTS1. MMP2 is a well-known MMP secreted by activated HSCs and associated with the fibrosis process,17 and we recently demonstrated its involvement in CX3CL1 processing during chronic liver injury.8 In contrast, ADAMTS1 expression in the liver has been poorly documented and its role in fibrogenesis has never been investigated.

Figure 1.

Gene connection networks as a framework for degradosome studies. Relationships from literature databases between 69 up-regulated genes in fibrotic livers were extracted and analyzed using Bibliosphere software (Genomatix Software GmbH, Munich, Germany). Forty-two genes were documented: The source-target direction of a relationship is indicated by a filled or blocked arrowhead for activation and inhibition, respectively. Circles indicate the source of annotation of the connection: molecular connection experts (orange), Genomatix experts (blue), or both (orange surrounded by blue line). Other connections have no expert curated-annotations.

Up-regulation of ADAMTS1 Is Associated With Fibrogenesis.

To explore the possible role(s) of ADAMTS1, we analyzed its expression in an independent set of 22 samples. Patients were 20 men and 2 women with a median age of 60.9 + 9.6 years; 3 were positive for HCV and 6 for hepatitis B virus (HBV). Steady-state ADAMTS1 mRNA levels in fibrotic tissues and control livers were measured by real-time PCR. ADAMTS1 mRNA levels were significantly increased in fibrotic liver samples, compared with healthy livers, and were correlated with grade of fibrosis: ADAMTS1 mRNA levels were significantly induced in cirrhotic (F4) livers, compared with F1-F3 livers (Fig. 2A). Moreover, up-regulation of ADAMTS1 was correlated with the known induction of MMP2 expression in chronic liver disease. To identify the cellular source of ADAMTS1 in the liver, we analyzed its expression in isolated hepatic cells. ADAMTS1 was highly expressed in activated HSCs, compared to hepatocytes and enriched Kuppfer cell fractions (Fig. 2B).

Figure 2.

ADAMTS1 expression is up-regulated in liver fibrosis and associated with the activation of HSCs. (A) ADAMTS1 mRNA levels were measured by RT-qPCR in 10 histologically healthy liver controls (C), fibrotic (F1-F3), and cirrhotic (F4) liver tissues. ADAMTS1 expression was correlated with MMP2 expression. (B) ADAMTS1 mRNA expression was analyzed in isolated HSCs (n = 4), Kupffer cell (KC)-enriched fraction (n = 2), and cultured (48 hours) human hepatocytes (HH; n = 3). Results were normalized relative to a value of 1 for expression in healthy liver (A). (C) ADAMTS1 and thrombospondin mRNA levels during HSC transdifferentiation (d: day of primary culture; P: passages during secondary culture). Obtained values were normalized as described above. (D) Schematic structure of ADAMTS1 showing the major structural domains of the unprocessed (p110) and processed forms, including the active p87 and the degradation product, p65 (TSP, thrombospondin domain; Cys, cystein-rich domain). (E) Western-blot analysis of ADAMTS1 in cellular extracts (CE) of HSCs and of HSC-conditioned media (CM). Molecular weights (kDa) of the different ADAMTS1 forms are shown to the right. The asterisk denotes a processing intermediate between p110/87 and p65. (F) Western-blot analysis of ADAMTS1 in normal and fibrotic liver tissue samples.

We further investigated ADAMTS1 expression during HSC activation, which reflects the transition from a quiescent to a myofibroblastic-like phenotype, a change that can be mimicked by culturing freshly isolated HSCs in uncoated tissue-culture plastic plates. qPCR analyses were performed on total RNA extracts from 1- to 11-day-old cultures and after 1-6 cell passages. The quiescent and activated status of HSCs was confirmed by analysis of the expression of specific markers, including peroxisome proliferator-activated receptors (PPAR), alpha-smooth muscle actin (α-SMA), and type I collagen (COL1A2) (Supporting Fig. 1). In agreement with previous reports,18-21 the three PPAR isoforms were expressed in isolated HSCs over the first 4 days, with a maximum increase of PPARβ at day 4. At day 11, PPARδ mRNA levels were undetectable and PPARβ/γ levels were strongly diminished, whereas α-SMA and COL1A2 mRNA levels increased. In fully activated cells, expression of PPAR isoforms was not detected, whereas expression of α-SMA and COL1A2 dramatically increased. Similar to the progression in the expression pattern of these genes and consistent with activated HSCs being a major source of ADAMTS1, ADAMTS1 mRNA expression was undetectable in quiescent HSCs (days 1-4), enhanced in cultured HSCs (10±0.75-fold increase at day 11) and strongly increased after 4 passages (150- to 200-fold increase; Fig. 2C). In contrast, thrombospondin, previously reported to be present in isolated rat HSCs,22 was expressed early in culture and reached an increase of 11.58±0.24-fold at day 11, but its levels were strongly diminished in myofibroblast-like cells. To avoid interference from thrombospondin activity, human HCSs were routinely used between passages 4 and 10 in all subsequent experiments. Up-regulation of ADAMTS1 expression was confirmed by western blotting in both activated HSCs and fibrotic liver tissues relative to normal livers. The processed 87-kDa ADAMTS1 active form (see Fig. 2D) was recovered mainly in cell extracts and HSC-conditioned media (Fig. 2E) and was clearly induced in liver fibrosis (Fig. 2F). The 110-kDa unprocessed form was only present in cell extracts, and the 65-kDa shorter form was recovered only in conditioned media (Fig. 2E).

The TSP1-Containing Domain of ADAMTS1 Mediates Interactions With the Latent Form of TGF-β, LAP-TGF-β.

A major feature of ADAMTS1 is the presence of three thrombospondin type 1 motifs (TSP1), with the proximal TSP1 being separated from the two carboxy-end TSP1 motifs by a “spacer sequence” rich in cysteine residues (Fig. 2D). Next to the proximal TSP1 sequence, we identified a KTFR motif that aligns with the active KRFK sequence of the human thrombospondin TSP1 repeats previously shown to be involved in the interaction with TGF-β (Fig. 3A).23 A tryptophan-rich peptide (WxxW), described as a docking site that promotes the interaction of KRFK sequences with LAP-TGF-β, is also present in the proximal TSP1 motif of ADAMTS1. The WxxW and KxFx motifs are not present in the two carboxy-end TSP1 motifs of ADAMTS1 (not shown). Because the proximal TSP1-containing domain of ADAMTS1 resembles that of thrombospondin, we asked whether it might display a structural organization, allowing for interactions with TGF-β (Fig. 3B). An “hhsearch” against the Protein Data Bank (see Supporting Information) identified the following candidate structural templates for ADAMTS1 TSP-like and thrombospondin domains (P values < 10−15): ADAMTS23 (PDB:3ghm); ADAMTS5 (PDB:2rjq); the TSP1 type 1 repeat (PDB:1lsl); F-spondin (PDB:1vex, 1szl); the thrombospondin anonymous protein, Trap (PDB:2bbx); VAP1 (PDB:2ero); Properdin (PDB:1w0r); Catrocollastatin (PDB:2dw0), and the Vitelline membrane outer layer protein I (PDB:1vmo). Except for 2dw0, 2ero, and 1vmo, all matching structural templates have a triple-strand organization, suggesting that this TSP1-like structure is shared by both the TSP1 repeat from thrombospondin and the motif found in ADAMTS1 (Fig. 3C). Over 100 models of this region of ADAMTS1 were built by combining different templates (see Supporting Information for details). The upper panel of Fig. 3D represents one such model that matches region 554-627 of the full ADAMTS1 structure. The WxxW (yellow) and KTFR (red) motifs are located at the surface of the domain, suggesting an availability for molecular interactions. The flexibility of the 612-627 region, which allows for structural adaptation to interactors, is also fully compatible with our hypothesis of an interaction with TGF-β similar to that of the TSP1-containing domain of thrombospondin (Fig. 3D, lower panel). In this case, KTFR/LSKL interactions would lead to the unfolding of the LAP-TGF-β structure, making it accessible for processing into its active form.

Figure 3.

ADAMTS1 interacts with LAP-TGF-β1. (A) Sequence alignment of the proximal TSP1 motif-containing domain of human ADAMTS1 and human thrombospondin. (B) Hypothetical representation of the interaction of ADAMTS1 with LAP-TGF-β (adapted from Hugo, 200343). TGF-β is synthesized as an inactive homodimeric large precursor molecule consisting of a self-inhibiting propeptide, the LAP linked to the active form of TGF-β. Pro-TGF-β is cleaved by furin-type enzymes to generate mature TGF-β (black box), which remains noncovalently associated with LAP as the small latent complex (SLC). Similar to thrombospondin/TGF-β interactions, we propose that the proximal TSP1-containing domain of mature ADAMTS1 interacts via its KTFR and WGPW peptides with LSKL and RKPK motifs in the LAP-TGF-β molecule to promote the release of active TGF-β. This process most likely involves a simple molecular competition mechanism and does not require the protease activity of ADAMTS1. The organization of the other ADAMTS1 domains is not known. (C) Representative structural templates for the TSP1 motifs of ADAMTS1 using the “hhsearch” protein structure prediction tool against the Protein Data Bank: ADAMTS23 (gray); the TSP1 type 1 repeat (green); F-spondin (magenta and cyan, two structures in PDB); and the thrombospondin anonymous protein (Trap, pink), Properdin (yellow, PDB structure completed using the SABBAC tool44). No structure is available in PDB for the ADAMTS5 TSP1-like region (not shown). (D) Representative models for the TSP1-containing domain for ADAMTS1 (upper panel) and thrombospondin (lower panel). WxxW and KXFX peptides are colored in yellow and red, respectively. (E) HSC-conditioned medium was immunoprecipitated with anti-LAP-TGF-β (LAP) and immunoblotted with anti-ADAMTS1. (F) Localization of ADAMTS1 and LAP-TGF-β (LAP) in HSCs. ADAMTS1 was visualized with an Alexa Fluor 488–labeled antibody (green), and LAP-TGF-β was visualized with an Alexa Fluor 555–labeled antibody (red). Two representative fields are shown, colocalization results in yellow cellular staining.

Does ADAMTS1 interact with LAP-TGF-β in activated HSCs? Immunoprecipitation of endogenous LAP-TGF-β, highly expressed in HSCs, demonstrates its interaction with ADAMTS1 (Fig. 3E) and the two proteins also exhibit colocalization in these cells (Fig. 3F). We next asked whether the KTFR motif would play a role in this interaction. HSC-conditioned media were incubated with peptide competitors, including KTFR and LSKL, its predicted complementary site on LAP-TGF-β: LSKL was previously shown to interact with KFRK motifs in human thrombospondin.24 LAP-TGF-β was then immunoprecipitated and complexes with ADAMTS1 were analyzed as described above. Both peptides diminished the interaction between ADAMTS1 and LAP-TGF-β (Fig. 4A,B), suggesting that the KTFR motif of ADAMTS1 and the LSKL motif of LAP-TGF-β are directly implicated in mediating the interaction between the two proteins.

Figure 4.

KTFR and LSKL peptides inhibit the association between ADAMTS1 and LAP-TGF-β. HSCs were cultured for 48 hours in the presence of increasing concentrations (0.1-10 μM) of KTFR (A) or LSKL (B) and scrambled Scr1 (TKFR) or Scr2 (SLLK) peptides, respectively. HSC-conditioned media were immunoprecipitated with anti-LAP antibodies and immunoblotted using anti-ADAMTS1 antibodies.

ADAMTS1 Induces Activation of TGF-β.

One effect of the interaction between ADAMTS1 and LAP-TGF-β might be on TGF-β activation. To test this, Chinese hamster ovary (CHO) cells, which provide a useful overexpression system to assay the effects of ADAMTS1, LAP-TGF-β, and mutant forms thereof, were transfected with LAP-TGF-β with or without ADAMTS1. Activation of TGF-β was assayed by enzyme-linked immunosorbent assay (ELISA) to measure active and total (after acid activation) TGF-β in the supernatant. Overexpression of ADAMTS1, indeed, induced the release of active TGF-β (Fig. 5A). ADAMTS1 is a proteolytic enzyme, and TGF-β activation likely requires its catalytic activity. Quite unexpectedly, expression of ADAMTS1-E386Q, a mutant lacking protease activity, enhanced the release of active TGF-β to an extent similar to that of wild-type ADAMTS1, demonstrating that TGF-β activation occurs through a protease-independent mechanism.

Figure 5.

Expression of ADAMTS1 induces the activation of TGF-β. CHO cells were transfected with LAP-TGF-β and ADAMTS1 or mutants thereof, as indicated. Active and total TGF-β levels were quantified by ELISA in supernatants 48 hours post-transfection. (A) Overexpression of wild-type ADAMTS1 (ADAMTS1-WT) or the protease-deficient mutant, ADAMTS1-E386Q, induces the release of active TGF-β, compared to empty vector (control). (B) Deletion mutants of LSKL (ΔLSKL LAP-TGF-β) or RKPK (LAP-TGF-β ΔRKPK) motifs, but not the LSA56L point mutant (LSA56L-LAP-TGF-β), prevent the ADAMTS1-dependent release of active TGF-β. Results are expressed as the mean ± standard deviation (SD) (error bars) of triplicates from three independent experiments.

We performed the reverse experiment by measuring the release of active TGF-β from LAP-TGF-β mutants. The mutations we tested affect the LSKL and the RKPK motifs previously shown to interact with the KRFK and WxxW sequences in thrombospondin24, 25: complete deletions of the LSKL sequence (ΔLSKL LAP-TGF-β); an alanine substitution for lysine 56 (LSA56L LAP-TGF-β); and a complete deletion of the RKPK peptide (LAP-TGF-β ΔRKPK). Release of active TGF-β was diminished in cells overexpressing ADAMTS1 and transfected with the ΔLSKL LAP-TGF-β and LAP-TGF-β ΔRKPK deletion mutants. In contrast, the A56L substitution within the LSKL peptide did not prevent the release of active TGF-β that was induced (Fig. 5B). In all cases, overexpression of ADAMTS1 did not affect the secretion of total TGF-β, but, in agreement with the known impaired secretion of LAP-TGF-β variants,25 we found reduced levels of total TGF-β in supernatants of cells transfected with these constructs.

ADAMTS1 Promotes TGF-β-Dependent Transcriptional Activity in HSCs.

Activated HSCs express ADAMTS1 and LAP-TGF-β at high levels (Fig. 2). This physiological model serves to demonstrate the binding of endogenous ADAMTS1 to endogenous LAP-TGF-β. We asked next, by performing ADAMTS1 small interfering RNA (siRNA) knockdowns, whether this interaction, indeed, would lead to the activation and release of mature TGF-β. Robust RNA interference efficiency led to significantly reduced steady-state levels of ADAMTS1 mRNA and proteins within 48 hours (Fig. 6A), and we used HEK 293T cells as a sensitive luciferase assay system to measure TGF-β-mediated transcriptional activation from treated or untreated HSCs. HEK 293T cells transfected with the TGF-β-responsive 3TPE-luciferase reporter gene were incubated with the conditioned media of ADAMTS1-depleted HSCs for 18 hours, and luciferace activity was then measured in cell extracts. Direct stimulation of HEK cells by TGF-β was used as an internal positive control. Depletion of ADAMTS1 in HSCs clearly affected TGF-β-dependent transcriptional activity in conditioned media, compared to control siRNAs (Fig. 6B). In agreement with these observations, the mRNA levels of two other metalloproteinases, MMP2 and ADAM12, also induced by TGF-β in HSCs,5 were significantly reduced in cells silenced for ADAMTS1 (Fig. 6B, insert). The implication of ADAMTS1 in the activation of TGF-β in HSCs was confirmed by incubating HSCs with the KTFR peptide. This significantly reduced the activity of TGF-β in conditioned media, as shown by decreased luciferase activity (Fig. 6C). Moreover, coincubation with the broad-spectrum BB94 metalloproteinase inhibitor did not affect the peptide-induced inhibition of TGF-β activity. Taken together, these observations strengthen our conclusion that proteolytic activities are not involved in the activation of TGF-β by ADAMTS1.

Figure 6.

ADAMTS1 mediates TGF-β-dependent transcriptional activity in human HSCs via the KTFR motif. HSCs were transfected with three different ADAMTS1 siRNAs (Si-1, -2, and -3) or scrambled siRNAs (Scr) and further cultured for 48 hours. (A) Real-time PCR (left panel, white for Scr, grey for Si1 and 2, and black for Si2 and 3) and western blotting (right panel) analyses confirm the efficiency of RNA interference. Results are expressed as the mean ± SD (error bars) of triplicates from three independent experiments. (B) HEK 293T cells transfected with the TGF-β-responsive 3TPE luciferase reporter gene were treated for 18 hours with conditioned media from HSCs transfected with ADAMTS1 siRNA or Src. HEK 293T cells treated for 18 hours with TGF-β (1 ng/mL) were used as an internal control. Insert shows the decrease in ADAM12 and MMP2 expression in Si-1 and -2 (grey) and Si-2 and -3 (black) ADAMTS1-depleted cells. (C) HEK 293T cells were transfected with 3TP-lux and were stimulated for 18 hours with conditioned media from HSCs treated with scrambled peptides (black square), KTFR (white circle), or KTFR plus the metalloprotease inhibitor, BB95 (white triangle), at the indicated concentrations. Luciferase activity was determined and normalized relative to untreated controls. Results are expressed as the mean ± SD (error bars) of triplicates from three independent experiments.

ADAMTS1 Is Implicated in a Carbon-tetrachloride–Induced Liver Fibrosis Model in Mice.

To evaluate the physiological relevance of the involvement of ADAMTS1 in liver fibrosis, we investigated the dynamics of ADAMTS1 expression in the carbon tetrachloride (CCl4)-induced fibrotic mouse model that we have previously described.26 As shown in Fig. 7A and in full agreement with our observations in human fibrosis tissues, levels of mouse ADAMTS1 and type I collagen transcripts were increased in mice given an oral administration of CCl4 for 1 week (∼3-fold) or 12 weeks (∼7-fold for ADAMTS1 and ∼40-fold for type I collagen). Immunohistochemical staining of ADAMTS1 in CCl4-treated mice showed increased ADAMTS1 labeling, compared to control oil-treated mice (Supporting Fig. 2). Because the KTFR peptide inhibits the ADAMTS1-dependent activation of TGF-β in HSCs (Fig. 6), we then asked whether this peptide might also prevent the progression of hepatic damage in the mouse model. We assayed blood levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) in C57Bl/6 mice treated with CCl4 and the KTFR peptide for 1 week and sacrificed 48 hours after the last CCl4 administration. Fig. 7B shows that the increase in plasma AST and ALT levels observed upon CCl4 administration was significantly reduced when mice were simultaneously treated with the KTFR peptide, compared to the TKFR scrambled peptide (1.49-fold versus 5.07-fold and 3.89-fold versus 52.4-fold for AST and ALT, respectively). We also compared the amounts of collagen deposited in hepatic tissues after oral administration of CCl4 for 1 week, using second harmonic generation (SHG) analysis, which allows for the direct detection and quantification, without staining, of fibrillar collagen. The increase in collagen deposits observed upon CCl4 administration was significantly reduced when mice were simultaneously treated with the KTFR peptide, compared to the TKFR scrambled peptide (Fig. 8; P < 0.001). Similar results were observed with the LSKL peptide that was previously shown to reduce liver fibrosis after several weeks of CCl4 administration in rats.27 Taken together, these results support our conclusion that ADAMTS1 is implicated in the early events of liver fibrosis, and suggest that the KTFR peptide helps prevent hepatic fibrosis induced by CCl4.

Figure 7.

ADAMTS1 expression is induced in mouse liver fibrosis. (A) ADAMTS1 and COL1A2 mRNA levels were measured in the livers of C57Bl/6 mice given oral CCl4 administration for 1 and 12 weeks and sacrificed 0, 4, 24, 48, and 72 hours or 7 days (7d) after the last CCl4 dose. Results are normalized relative to a value of 1 for expression at 0 hours and expressed as the mean ± SD from 5 mice (*P < 0.05; **P < 0.01). (B) Blood levels of ALT and AST in C57Bl/6 mice given oral CCl4 or vehicle-only administration, together with intraperitoneal injections of the indicated peptides for 1 week, and were sacrificed 48 hours after the last CCl4 dose. Results are expressed as the mean ± SD from 3 mice, and CCl4/vehicle ratios are shown in the inserts.

Figure 8.

Extent of CCl4-induced hepatic collagen deposition is reduced in KTFR-treated mice. (A) Representative simultaneous two-photon excitation fluorescence (TPEF) (liver tissue autofluorescence, red) and SHG (collagen, blue) imaging of liver sections from mice treated with oil, CCl4, and CCl4+peptides. Laser excitation: 288 mW with an 820-nm wavelength; enlargement: 10×; scale bar: 250 μm. (B) Collagen deposition (% of area) was quantified as previously described using the ImageJ application.41 Results are expressed as the mean ± standard error of the mean (error bars) of 10-20 independent images. *P < 0.05; **P < 0.01; ***P < 0.001, according to the Student's t-test.


ECM remodeling is pivotal to liver fibrosis and is associated with increases in the synthesis of ECM components as well as proteases. In this study, we undertook a screening approach combined with integrated array-data analysis to create a meaningful landscape of proteases that are deregulated in chronic liver disease. We first generated an interactive graph, and the resulting network was used as a framework for interpreting gene contribution to remodeling. Unexpectedly, the aggrecanase, ADAMTS1, emerged as a central node, together with MMP2, a well-known protease involved in chronic liver diseases, suggesting that ADAMTS1 might play a key role in the fibrosis process.

Whereas MMPs have been widely implicated in liver fibrosis remodeling,28 growing interest has come more recently from the observation that members of the ADAM family are up-regulated in chronic liver diseases.29 In addition to ADAMTS1, we also observed the up-regulation of ADAMTS2 and ADAMTS13, two metallopeptidases recently implicated in liver fibrosis. ADAMTS13 is required for the proteolysis of von Willebrand factor and has been observed to be present in HSCs.30 Inactivation of ADAMTS2, a procollagen aminopeptidase, has been shown to reduce liver fibrosis in CCl4-induced mouse models.31 Unlike the down-regulation initially reported in endothelial cells from cirrhotic rats,4 the up-regulation of ADAMTS1 was more recently reported in human chronic liver disease,7 and we now provide evidence for its central role in the up-regulated protease network in liver fibrosis. We demonstrate here that HSCs are the major source of ADAMTS1, whose expression is increased nearly 250-fold upon full activation. MMP2 has been similarly associated with HSC activation during chronic liver injury, and, accordingly, we establish a clear correlation of ADAMTS1 and MMP2 expression in fibrotic liver samples. Taken together, our data identify ADAMTS1 as a new hub of the protease network that also contains the well-known MMP2 and is associated with liver fibrosis.

The major regulatory step for all metalloprotease activity in vivo occurs at the protein level and requires a primary proteolysis of the N-terminal prodomain. ADAMTS1 has been shown to undergo a second cleavage at the C-terminal end, leading to a shorter form that lacks the two carboxy-terminal TSP1 repeats and has a reduced ability to bind to the ECM.32 Here, we describe, for the first time, the role of HSCs in the synthesis of a full-length 110-kDa unprocessed polypeptide secreted as the p87 active form. We also detected the shorter 65-kDa form that has been suggested to reflect an inactivation pathway for p87. In addition, we show that only the 87-kDa active form is detected during chronic liver injury, suggesting that the p65-kDa form does not accumulate within liver tissue. Similar observations have been reported in non-small-cell lung carcinomas.33 However, characterization of ADAMTS1 forms within human tissues remains poorly documented, and their contribution to the onset and development of disease is still unclear.

ADAMTS1 Interacts With and Releases Active TGF-β.

A mechanistic understanding of the effect of ADAMTS1 during liver fibrosis may be deduced from its catalytic activity against matrix components, such as aggrecan, versican, and nidogen. However, metallopeptidase activities are highly redundant, and genetic inactivation of many metallopeptidases leads to minimal phenotypes. Moreover, no alteration of aggrecan turnover was found in ADAMTS1 knockout mice.34 In contrast, loss-of-function ADAMTS1 studies have shown severe embryonic and perinatal lethality, suggesting an implication in development35, 36 that may be related to its noncatalytic functions that depend on interactions with growth factors, such as vascular endothelial growth factor and fibroblast growth factor-2.37, 38 We now demonstrate that ADAMTS1 also interacts with the profibrotic cytokine, TGF-β, leading to its release from its latent to active forms. Increased ADAMTS1 expression during chronic liver injury contributes to TGF-β-dependent transcriptional activity and, hence, to liver fibrosis. This interpretation is in line with the recent report of the implication of ADAMTS1 in the stimulation of the stromal reaction in lung cancer, including induction of TGF-β and collagen.39

Thrombospondin-Like, Proteolysis-Independent Activation of TGF-β by ADAMTS1 in Liver Fibrosis.

Sequence- and structure-predicted interactions between ADAMTS1 and TGF-β predict that direct interactions occur between the highly conserved KTFR motif of ADAMTS1 and the LSKL motif of TGF-β. This was suggested by the known interaction with the KRTK motif of thrombospondin, the major activator of TGF-β in vivo,40 leading us to propose that the KTFR sequence could play a similar role in ADAMTS1 (Fig. 3). The inhibition of ADAMTS1-mediated activation of TGF-β by KTFR peptides indicates that the mechanism of interaction is similar—if not identical—to that reported for thrombospondin-mediated activation.24 In this case, a “conformational” mechanism is in full agreement with the results of our experiments using proteolysis-deficient ADAMTS1 mutants and protease inhibitors, which show that TGF-β activation by ADAMTS1 is independent of the proteolytic activity of the latter. Although it might occur via similar interactions, activation of TGF-β by ADAMTS1 and thrombospondin are unlikely to overlap in vivo. Thrombospondin is expressed in freshly isolated human HSCs,22 but we show that their activation induces a decrease in thrombospondin expression counterbalanced by a dramatic increase in ADAMTS1 expression. The physiological relevance of this activation process is fully supported by our finding that depleting ADAMTS1 in HSCs strongly diminishes the release of TGF-β-dependent transcriptional activity.

The KTFR Peptide Antagonizes Induced Mouse Liver Fibrosis.

A conformational model of TGF-β activation by ADAMTS1 predicts that interfering with KTFR/SLKL interactions should lead to a decrease in available active TGF-β. We demonstrate that such a mechanism is, indeed, at play in vivo, using a murine model of induced liver fibrosis. In full agreement with the activation pathway described above, injection of the KTFR peptide in CCl4-treated mice that develop liver fibrosis reduces the levels of biological markers associated with hepatic damage (Fig. 7). This conclusion is further borne out by the demonstration, using highly sensitive SHG analysis,41 that concomitant injection of KTFR dramatically reduces collagen deposition associated with the early onset of fibrosis (Fig. 8). The full conservation of KTFR and LSKL motifs between humans and mice suggests that this important proof of concept can be extrapolated to humans, identifying ADAMTS1 as a new therapeutic target during chronic liver injury.

Many factors have been implicated in TGF-β activation, including thrombospondin, proteases (e.g., plasmin, thrombin, and MMP), and integrins, but also heat, acid, reactive oxygen species, and mechanical force. All these components build up an environmental network, in which the role of each one is obviously part of the sum and depends on dynamic tissue changes, especially as the liver proceeds from a healthy to a fibrotic state. In agreement with this view, serine protease inhibitors,42 TGF-β peptide inhibitors,27 and ADAMTS1 peptide inhibitors (this work) reduce liver fibrosis, suggesting that therapeutic developments should integrate mechanisms of TGF-β activation as a complex system, taking into account diverse molecules and their spatiotemporal organization. As a newly identified partner in the TGF-β activation network that is specifically expressed in HSCs during chronic liver injury, we propose that ADAMTS1 is a key player in the dynamic interplay that helps regulate TGF-β activity.


The authors thank the Rennes Biological Resources Center (CHRU Pontchaillou, IFR 140) for its contribution to human tissue sampling. We acknowledge the excellent support of the Nice-Sophia Antipolis Transcriptome Platform of the Marseille-Nice Genopole, in which the microarray experiments were carried out. Special thanks are due to Virginie Magnone and Géraldine Rios for microarray production. The authors thank Dr. J.E. Murphy-Ullrich (University of Alabama at Birmingham, Birmingham, AL) and Dr. D. Cataldo (University of Liège, Liège, Belgium) for providing the LAP-TGF-β and ADAMTS1 constructs, respectively. The authors thank Dr. M. Baudy-Floc'h (University of Rennes, ICMV, UMR CNRS 6226, Rennes, France) for peptide synthesis, Dr. C. Piquet-Pellorce (University of Rennes, SeRAIC EA4427) for animal experimentation, Dr. C. Lucas (Service Biochimie, CHU Rennes) for enzyme measurements, and Dr. E. Schaub for SHG analyses (PIXEL facilities, University of Rennes 1). The authors thank Dr. E. Käs (LBME, CNRS/Université Paul Sabatier) for useful discussions and a critical reading of the manuscript.