Identification of Endoglin in Rat Hepatic Stellate Cells NEW INSIGHTS INTO TRANSFORMING GROWTH FACTOR (cid:1) RECEPTOR SIGNALING* □ S

Transforming growth factor (cid:1) (TGF- (cid:1) ) signaling is me-diated by the cell surface TGF- (cid:1) type I (ALK5), type II, and the accessory type III receptors endoglin and betaglycan. Hepatic stellate cells (HSC), the most profibro-genic cell type in the liver, express ALK5, T (cid:1) RII, and betaglycan. To monitor the expression of betaglycan in HSC, we used the commercially available antibody sc-6199 in Western blot analysis. This antibody, raised against a peptide mapping at the carboxyl terminus of the human betaglycan, is claimed to be specific for betaglycan, although it is known that the C-terminal domain is highly conserved in type III receptors. Proteins recognized in HSC by sc-6199 did not match the charac-teristic migration pattern of betaglycan. Moreover, the determined molecular weight ( M r 160) and the observed reductant sensitivity after treatment with dithiothreitol resemble those of a closely related type III receptor, endoglin (CD105). Endoglin, a disulfide-linked homo-dimer, is an accessory component of the TGF- (cid:1) receptor complex and mainly expressed on endothelial cells. The presence of endoglin in HSC of rat liver was confirmed by molecular cloning of the endoglin cDNA and immu-nocytochemistry. The reactivity of sc-6199

The transforming growth factor family of growth factors regulates a diverse set of physiologic processes including pro-liferation, cellular differentiation, apoptosis, and expression of extracellular matrix genes (1)(2)(3). Signaling by the three identified mammalian TGF-␤ 1 isoforms TGF-␤1, TGF-␤2, and TGF-␤3 is initiated by binding to the high affinity cell surface receptors, the accessory TGF-␤ type III receptor (T␤RIII) (4), the type II receptor (T␤RII) (5), and type I (6) receptor (T␤RI), of which T␤RII and T␤RI possess intracellular serine/threonine kinase domains. Upon binding of TGF-␤1 or TGF-␤3 to the receptor type II, the liganded receptor dimer associates with and its constitutive active kinase transphosphorylates the type I receptor at the GS domain (7). The active type I receptor in turn phosphorylates and thereby activates the intracellular signal transducers represented by the R-Smad family members 2 and 3 (2,7,8).
Because TGF-␤2 has a lower affinity for the type II receptor (9,10), cells sense this ligand by either expressing a type II receptor splice variant T␤RII-B (11)(12)(13) or employing an accessory receptor of the type III family (i.e. betaglycan) (14 -16). Nevertheless, betaglycan has a high affinity for all three TGF-␤ isoforms (17)(18)(19), which has been exploited to counteract TGF-␤ signaling by sequestering the ligand using the soluble extracellular ligand binding domain (20,21). Depending on the cell type, betaglycan either enhances (14,15,22) or inhibits TGF-␤ responses (23). As the ligand specificity suggests, T␤RIII is not only required for TGF-␤2 but also for TGF-␤1 signaling (4). Although the C terminus of betaglycan lacks a protein kinase domain (24), it does not only present ligand to T␤RII (22) but comprises protein-protein interaction domains, which govern internalization of betaglycan and down-regulation of TGF-␤ signaling by binding to ␤-arrestin 2 on its phosphorylated cytoplasmic domain (25). Membrane localization of betaglycan is modulated by GIPC, a protein that interacts with the class I PDZ domain located in the cytoplasmic domain of the type III receptor, resulting in regulation of expression of the type III receptor (26). Such a class I PDZ domain is also found in a second member of the type III receptor family (27) (i.e. endoglin (CD105), which shares 71% sequence identity at the protein level in its C terminus with betaglycan) (24,28). Endoglin is a disulfide-linked, homodimeric transmembrane glycoprotein highly expressed in cells of the vascular system (29), fibroblasts (30), macrophages (31), and vascular smooth muscle cells (32). From the phenotype of endoglin-deficient * This work was supported by a grant from the Deutsche Forschungsgemeinschaft. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Likewise, HSC and MFB show a different cellular response pattern when exposed to TGF-␤ (52). In primary cultures of HSC, TGF-␤ inhibits proliferation (52,53). In contrast, activated HSC are not growth-inhibited by TGF-␤ and synthesize excessive quantities of extracellular matrix components accompanied by a strong up-regulation of inhibitors for matrix degrading enzymes (for a review, see Refs. 54 and 55). Here, we describe for the first time the identification of endoglin in HSC, which is expressed in the course of their activation. This receptor may represent or be included in an alternative TGF-␤ signaling branch in HSC that might account for the different cellular behavior in response to TGF-␤ during activation and the process of fibrogenesis.

EXPERIMENTAL PROCEDURES
Cell Culture-HSC were isolated from male Sprague-Dawley rats by the Pronase-collagenase method followed by a density centrifugation on a Nycodenz gradient as described before (56,57). Briefly, livers were perfused with Hanks' buffered saline solution (HBSS) (PAA Laboratories, Linz, Austria) and subsequently with 0.35% (w/v) Pronase E and 0.015% (w/v) collagenase, both in HBSS. Thereafter, the livers were dissected in HBSS supplemented with DNase type II (100 g/ml) (Roche Applied Science) and filtered through nylon mesh. The resulting cell suspension was centrifuged, and cells were washed in ice-cold HBSS containing 0.25% (w/v) bovine serum albumin. HSC were further purified on 8.25% (w/v) Nycodenz (Nyegaard Co. AS, Oslo, Norway) gradients as described previously (56). The mean purity estimated by vitamin A autofluorescence was higher than 95%, and the yield ranged from 20 to 50 ϫ 10 6 cells/liver. Cells were seeded in growth medium consisting of HEPES-buffered Dulbecco's modified Eagle's medium (BioWhittaker, Verviers, Belgium) supplemented with L-glutamine (4 mmol/ liter), penicillin (100 IU/ml), streptomycin (100 g/ml), and 10% FCS (v/v) (Invitrogen) in a humidified atmosphere containing 5% CO 2 . Human MFB were isolated as outgrowth from human liver tissues and subcultured for several passages.
TGF-␤ Stimulation Experiment-HSC were isolated as mentioned before and cultured in 10% FCS (v/v). 12 h before the experiment was started, the serum content was reduced to 1% FCS (v/v) for 16 h and then to 0.5% (v/v) or 0.2% (v/v) supplemented with 5 ng/ml TGF-␤. The cells were harvested after 4 or 7 days, while the medium containing TGF-␤ was renewed every 24 h. The cells were extracted in lysis buffer containing the Complete™ mixture of proteinase inhibitors (Roche Applied Science) and, when indicated, the phosphatase inhibitor mixture set II (Sigma). Transient Transfections-COS-7 cells (CRL 1651; American Type Culture Collection, Manassas, VA) were cultured in growth medium and subcultured when 80% confluent. For transfection, cells were plated into 6-well dishes and transfected with 2 g of the expression plasmids pcDNA-endoglin (AY562420) or pCMV-HA-rT␤RIII (5) using the FUGENE 6™ method (Roche Applied Science). After 24 h, the medium was renewed, and cells were extracted in lysis buffer (50 mmol/liter Tris/HCl (pH 7.2), 250 mmol/liter NaCl, 2% (v/v) Nonidet P-40, 0.1% (w/v) SDS, 0.5% (w/v) sodium deoxycholate, 2.5 mM EDTA) for SDS-PAGE or fixed for 10 min in 4% paraformaldehyde for immunocytochemistry.
Western Blot Analysis-After treatment with or without NAC, cells were washed in ice-cold HBSS and extracted with lysis buffer containing the Complete™-mixture of proteinase inhibitors (Roche Applied Science) and, when indicated, the phosphatase inhibitor mixture set II (Sigma). Equal amounts of protein lysates were diluted with Nu-PAGE™ LDS electrophoresis sample buffer (Invitrogen) supplemented with or without reducing agents, heated at 70°C for 10 min, and separated in a 4 -12% bis-Tris gradient gels (Invitrogen) using MOPS-SDS running buffer containing 50 mmol/liter MOPS, 50 mmol/liter Tris-HCl (pH 7.7), 3.47 mmol/liter SDS, and 1.025 mmol/liter EDTA. Proteins were electroblotted onto nitrocellulose membranes (Schleicher & Schuell), and equal loading was monitored by Ponceau S staining. Unspecific binding sites were blocked in TBST (10 mM Tris/HCl, 150 mM NaCl, 0.1% (v/v) Tween 20, pH 7.6) containing 5% (w/v) nonfat milk powder. Primary antibodies were directed against T␤RIII or endoglin (for details, see Table I) and detected with horseradish-peroxidase (HRP)-conjugated secondary antibodies (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) and the Supersignal™ chemiluminescent substrate (Perbio Science, Bonn, Germany). The signal of the biotinylated antibody BAF-242 was detected by using a streptavidin-HRP conjugate (DAKO). In some cases, the signal was enhanced with a secondary biotinylated antibody prior to incubation with the streptavidin-HRP conjugate. All antibodies were diluted in 2.5% (w/v) nonfat milk powder in TBST.
Affinity Labeling of TGF-␤ Receptors-125 I-TGF-␤1 affinity labeling and cross-linking experiments were performed as described previously (53). Briefly, confluent monolayers of HSC were cultured in medium containing 10% FCS for 4 days. After a preincubation without FCS, . Thereafter, proteins were extracted in lysis buffer including proteinase inhibitors and subjected to immunoprecipitation with receptor-specific antibodies. Precipitated proteins were resolved by SDS-PAGE, and gels were dried and exposed for indicated times at Ϫ80°C to X-Omat AR films (Eastman Kodak Co.).

RNA Isolation and Northern
Blot Analysis-Total RNA was isolated using the RNeasy Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. RNA samples (5 g) were separated in 1.2% (w/v) denaturing agarose gel, transferred to Hybond-N membrane (Amersham Biosciences), and fixed by baking for 2 h at 80°C. Hybridization probes specific for rat endoglin (nt 1-792; accession number AY562420), rat T␤RIII (nt 1-819; accession number NM 017256), and human endoglin (nt 1384 -1933; accession number BT007558) were prepared using the random prime labeling system (Amersham Biosciences). Hybridization and washing of membranes was carried out as previously described (58). For internal standardization, the blots were rehybridized with a cDNA specific for glyceraldehyde-3-phosphate dehydrogenase.
RT-PCR-For RT-PCR experiments, purified samples of total RNA (1 g) was reverse transcribed at 42°C for 60 min using the Superscript II reverse transcriptase (Invitrogen) and random hexamer primer according to the manufacturer's instructions. Aliquots of first strand cDNAs were subjected to PCR with 2 M forward and reverse primer, dNTPs (each 10 mM dATP, dCTP, dGTP, dTTP) in 1ϫ PCR buffer, and 2.5 units of Taq DNA polymerase (Roche Applied Science). The coding region of rat endoglin cDNA was amplified with primers SM012 5Јd(CTC CCG GGT GGA CAG C)-3Ј and SM017 5Ј-d(AGG CTC CAG GCT GGG T)-3Ј (initial denaturation for 5 min at 95°C; 40 cycles at 95°C for 50 s, 56°C for 50 s, 72°C for 2 min; final elongation at 72°C for 10 min). A fragment of human endoglin cDNA was amplified from human MFB using primers SM014 5Ј-d(CCA AGG CTG CCA CTT GG)-3Ј and SM011 5Ј-d(CGA TGC TGT GGT TGG TAC)-3Ј, specific for rat endoglin. Cycle conditions were set to the following: initial denaturation for 5 min at 94°C, 40 cycles at 94°C for 1 min, 50°C for 1 min, 72°C for 3 min, final elongation at 72°C for 10 min. PCR products were gel-purified, cloned into pGEM-T® Easy vector (Promega, Madison, WI), and sequenced using the the ABI PRISM BigDye® termination reaction kit (PerkinElmer Life Sciences) as described elsewhere (58).
Immunocytochemistry-Approximately 3 ϫ 10 5 cells were plated on coverslips mounted in 6-well dishes and incubated for 24 h at 37°C. Thereafter, cells were transfected with selected expression plasmids. 24 h post-transfection, the medium was exchanged, and after incubation for an additional 24 h, the cells were fixed for 15 min in 4% (w/v) paraformaldehyde buffered in phosphate-buffered saline (PBS) (pH 7.4). After permeabilization (2 min on ice) in 0.1% (w/v) sodium citrate, 0.1% (v/v) Triton X-100, endogenous biotin was blocked by a biotin blocking system (DAKO). Nonspecific binding was further prevented by preincubation in 1% (w/v) bovine serum albumin in PBS for 30 min at 37°C and subsequent incubation in 0.5% bovine serum albumin plus 0.1% (w/v) fish gelatin in PBS. Then the primary antibodies sc-6199 (COS-7) or P4A4 (human MFB) as well as the corresponding normal control IgG were applied for 1 h at 37°C in 1% (w/v) bovine serum albumin in PBS followed by incubation with an anti-goat (sc-6199) or anti-mouse (P4A4) biotin conjugate. The detection of immunocomplexes was accomplished using the streptavidin-fluorescein isothiocyanate flourophore. Cells were mounted in antifade for laser-scanning microscopy (LS 410 invert; Zeiss) using a standard objective (40 ϫ 1.3 oil) and an external argon laser at 488 nm. For detection of heterologously expressed endoglin, the incubation with the streptavidin-fluorescein isothiocyanate fluorophore was followed by an anti-fluorescein isothiocyanate AP conjugate and subsequent Fast-Red substrate application. The chromogenic stain was documented by light microscopy.
Data Base Analysis-Screening of nonredundant and expressed sequence tag as well as the Swissprot data bases was done at the National Center for Biotechnology Information using the Blast algorithm (59).

Rat HSC Express a Homolog of TGF-␤ Type III Family
Receptor Betaglycan-TGF-␤ is a growth factor regulating key aspects of cellular activation and transdifferentiation of HSC (55). The ligand, TGF-␤, is bound by a heterooligomeric membrane receptor complex, consisting of the signaling receptors type I and II and at least one accessory receptor betaglycan. In a previous unpublished work of our laboratory, 2 we could show that TGF-␤ signaling in cultured HSC could be abrogated by reducing agents like N-acetylcysteine (NAC). As one possible mechanism, we found that NAC reduces ligand binding to a high molecular weight receptor complex present in HSC, most likely representing a TGF-␤ type III receptor. To test the hypothesis that NAC acts directly on the T␤RIII polypeptide in a similar manner as has been described for DTT by Philip et al. (60), we performed Western blot analysis using the commercially available T␤RIII-specific antibody sc-6199, directed against a C-terminal epitope (Fig. 1A, left). In our experiments, the antibody recognizes a protein of 144 kDa under nonreducing conditions that is shifted to 74.5 kDa under reducing conditions. This migration pattern was somewhat unexpected in regard to the published molecular weight of betaglycan (61). Therefore, we decided to test a second independent antibody (AF-242-PB) directed against the extracellular domain of T␤RIII (Fig. 1A, right). This antibody detects a protein of ϳ90 kDa in size, only slightly shifting its apparent molecular weight under reducing conditions, corresponding to the unmodified core protein of T␤RIII. The faint bands around ϳ200 kDa most likely represent glycosylated species of T␤RIII. Immunoprecipitation of protein extracts from COS-7 cells transfected with a T␤RIII expression plasmid (Fig. 1B, right) as well as immunoprecipitation of 125 I-TGF-␤1-labeled T␤RIII from HSC (Supplement 1) confirmed the specificity of AF-242-PB toward T␤RIII. In contrast, sc-6199 obviously recognizes T␤RIII when overexpressed in COS-7 cells (Fig. 1B, left) but not in extracts taken from HSC (cf. Fig. 1A, left). Therefore, HSC express a protein that shares high homology with T␤RIII with respect to the sc-6199 epitope, and this protein is expressed at a much higher level compared with T␤RIII. Furthermore, NAC treatment of COS-7 cells transiently overexpressing T␤RIII revealed that the T␤RIII betaglycan is not reductant-sensitive (Supplement 2).

Cloning of Endoglin Transcripts from Rat HSC and MFB-
By means of Western blot analysis, we verified that the epitope of the antibody sc-6199 maps to the T␤RIII cytoplasmic domain (Supplement 3). To identify proteins with homology to the carboxyl terminus of T␤RIII, we screened the Swissprot data base and found that murine TGF-␤ receptor endoglin matches the respective region spanning the T␤RIII epitope. To elucidate whether endoglin is expressed in HSC, we next performed RT-PCR. To generate specific primers for our analysis, we screened a rat expressed sequence tag library using the murine endoglin cDNA as the query. We identified six individual rat expressed sequence tags covering the whole coding sequence of the rat cDNA ( Fig. 2A). Based on this information, we designed primers flanking the start and stop codons and amplified the complete coding region (ϳ1950 nt) of rat endoglin. The open reading frame contains 650 amino acids, predicts a protein with a calculated molecular mass of 69.9 kDa, and proves the assumed high degree of sequence identity between endoglin and betaglycan at the carboxyl terminus (Figs. 2B and 3). Rat endoglin shares 69 and 82% sequence identity with the human and murine orthologues (62,63). The amino terminus is preceded by a putative signalpeptide (amino acids 1-25) following the von Heijne prediction (64). In contrast to human endoglin, which contains an RGD tripeptide in the extracellular domain (62), both rat and mouse endoglin lack this integrin binding domain (63). Amino acids 549 -573 constitute the membranespanning domain conserved between rats and mice. The carboxyl terminus of endoglin is about 47 amino acids and shares a potential PDZ class I domain with betaglycan (26).
Expression of Type III Family Receptor mRNA during Transdifferentiation of HSC-To monitor the transcript levels of betaglycan and endoglin during transdifferentiation, we performed Northern blot analysis. For this purpose, we analyzed total RNA isolated from HSC cultured for 2 and 7 days and MFB cultured for 4 days after the first passage. Hybridization with an endoglin-specific probe resulted in the detection of two transcripts of 3 and 3.6 kb in size (Fig. 4A). The presence of two different transcripts originate most likely from the usage of two alternative polyadenylation signals rather than Endoglin in Hepatic Stellate Cells 3082 the occurrence of differential splicing, because accompanying PCR experiments (not shown) gave no experimental hints for the expression of different splice variants previously reported in humans (65). During transdifferentiation of HSC into MFB, the amount of endoglin mRNA remains constant. In contrast, the expression of the 6-kb betaglycan messenger markedly declines in the course of transdifferentiation (Fig. 4B), confirming previous reports showing that the expression of different TGF-␤ receptors can be modified during fibrogenesis (66 -68).

Detection of the Heterologously Expressed Endoglin Confirms the Specificity of the Antibody sc-6199 toward Both Type III
Receptors-To confirm the hypothesis that antibody sc-6199 crossreacts with endoglin, we transiently expressed the rat endoglin cDNA in COS-7 cells and examined respective protein lysates by Western blot analysis under reducing and nonreducing conditions (Fig. 5). In this analysis, sc-6199 detects the expressed endoglin, unambiguously demonstrating that the antibody is not specific for betaglycan (Fig. 5A). Under nonreducing conditions, the antibody detects a protein of 130 kDa, representing the receptor dimer, as well as oligomeric, high molecular weight complexes. This migration pattern is in agreement with those previously observed for endoglin (38,62,65). The addition of reducing substances (e.g. 50 mM DTT) results in disintegration into receptor monomers possessing a relative molecular mass of 68 kDa. The slightly higher molecular mass of endogenous endoglin (cf. Fig. 1A) in HSC might be attributable to a different posttranslational receptor processing in HSC and COS-7. Furthermore, immunocytochemical staining of COS-7 cells, overexpressing endoglin, also demonstrate that antibody sc-6199 reacts with endoglin (Fig. 5B).
Expression of Type III Family Receptor Proteins during Transdifferentiation of HSC-To examine the protein expression of betaglycan and endoglin during transdifferentiation, we performed Western blot analysis (Fig. 6). We found that the content of betaglycan, especially its glycosylated form, increases during primary culture and drops at later stages (Fig.  6A). In contrast, the level of endoglin expression is constant during transdifferentiation (Fig. 6B), confirming the results of our Northern blot analysis (cf. Fig. 4). Furthermore, we found that both receptors are not expressed in hepatocytes.
Endoglin Expression and Localization in Human Myofibroblasts-We analyzed endoglin expression in human MFB ob-

FIG. 4. Expression of T␤RIII and endoglin in rat HSC/MFB. A
and B, Northern blot analysis was performed using 5 g of total RNAs isolated from the following sources: HSC at day 2, HSC at day 7, and MFB at day 4. The blots were hybridized with 32 P-labeled probes specific for endoglin (A) and T␤RIII (B). As internal loading controls, the 18 S rRNA, 28 S rRNA, and signals obtained after hybridization with a glyceraldehyde-3-phosphate dehydrogenase (GAPDH)-specific probe are shown. The estimated sizes of hybridization signals are given on the right. Blots were exposed for 1 h (endoglin) or 2 weeks (T␤RIII), respectively. tained from liver outgrowth by RT-PCR and Northern blot analysis (Fig. 7, A and B). The identity of the RT-PCR product was verified by sequencing (not shown), and the observed transcript size of 3.4 kb estimated by Northern blot exactly corresponds to those reported previously (62), suggesting that, in contrast to mouse 3 and rat (cf. Fig. 4A) expressing two transcripts, there is only one distinct transcript present in humans. We next performed Western blot experiments under reducing and nonreducing conditions (Fig. 7C). As a control, we used protein extracts taken from human umbilical vein endothelial cells (HUVEC) known to express high amounts of endoglin (29). In both lysates, the endoglin-specific antibody sc-20632 recognizes a protein of ϳ140 kDa and high molecular weight complexes under nonreducing conditions (Fig. 7C, left). In line with our hypothesis, the molecular mass of the detected protein was lowered to 73 kDa in MFB and 76 kDa in HUVEC after treatment with DTT. An identical migration pattern was obtained using antibody sc-6199 (Fig. 7C, right). To determine the subcellular localization of endoglin in human MFB, we performed immunocytochemical staining using a monoclonal antibody (P4A4) raised against human endoglin (Fig. 8). Confocal microscopy revealed a significant staining in the proximity of the plasma membrane and in areas surrounding the nucleus, which might represent receptors in the endoplasmatic reticulum.
Endoglin Is Expressed in the Plasma Membrane of HSC and MFB-Having established the expression of endoglin in HSC and MFB we next analyzed if endoglin is localized in the plasma membrane, an essential prerequisite for an integral membrane receptor. Because previous reports demonstrated that only a very small fraction of endogenous, membrane localized endoglin becomes labeled in cross-linking experiments (28), even when overexpressed in culture cells (38), we used a different strategy to analyze membrane insertion. In this approach, proteins exposed at the surface of HSC and MFB were first labeled with a membrane-impermeable biotinylation agent, and rat endoglin was subsequently immunoprecipitated from lysates of biotinylated HSC (Fig. 9A, left) and MFB (Fig.  9A, right). Precipitated proteins were then subjected to SDS-PAGE under reducing and nonreducing conditions, and surface receptors were detected by the streptavidin-HRP conjugate. In both HSC and MFB we were able to label and precipitate endoglin, demonstrating that endoglin is exposed at the cell surface. Consistently, under the same experimental conditions,

FIG. 7. RT-PCR, Northern blot, and Western blot analysis for the detection of endoglin expression in human MFB.
A, 1 g of total RNA isolated from human MFB (hMFB) was subjected to RT-PCR using primers specific for endoglin. B, 12 g of total RNA isolated from human MFB outgrowth were analyzed by the Northern blotting using a 32 P-labeled probe specific for human endoglin. The blot was exposed for 3 days. C, protein extracts (15.5 g of protein per lane) of hMFB and HUVEC were tested under nonreducing and reducing conditions for endoglin expression in Western blot using antibody sc-20632 (left) and sc-6199 (right). The positions of protein size markers are given. Note that the observed size of proteins is redox-dependent.
we precipitated endoglin using the monoclonal antibody P4A4 from human MFB (Fig. 9B, left). To further demonstrate that the detected protein represents human endoglin, the blot membrane was stripped and reprobed with the endoglin-specific antibody sc-6199 (Fig. 9B, right).

Type III Receptor Expression in HSC and MFB Is Not
Induced by TGF-␤-We attempted to analyze the influence of TGF-␤ on betaglycan and endoglin expression in HSC and MFB by Western blotting (Fig. 10). Under the chosen conditions, which are commonly used in these kinds of experiments, we found that (i) the expression of both receptors is decreased during periods of serum starvation and (ii) TGF-␤ has no influence on the expression of betaglycan and endoglin in HSC. A down-regulation of TGF-␤ receptors following serum starvation was previously demonstrated in rat MFB (69), most likely representing a general phenomenon. Furthermore, we found that lowering the serum content from 10 to 1% (data not shown) or 0.5 and 0.2% for 7 days resulted in a near absence of betaglycan (Fig. 10A) and endoglin (Fig. 10B) in HSC. DISCUSSION In the present study, we first demonstrate that rat and human HSC/MFB express the accessory TGF-␤ receptor endoglin. Endoglin is a member of the TGF-␤ type III receptor family (70), which includes a second receptor known as betaglycan FIG. 8. Immunohistochemistry for analysis of endoglin expression in human MFB. The monoclonal antibody P4A4 directed against human endoglin (A-C) was used to determine the distribution of endoglin in human MFB. A respective control serum (D) was taken as a negative control. Nuclei were counterstained with propidium iodide, and cells were monitored for regional localization by confocal laserscanning microscopy. Original magnification was ϫ800. also expressed in HSC (53,66,67). Although both receptors share a conserved domain structure, they could be clearly distinguished by their ligand binding ability and specificity (17-19, 28, 37, 38). Therefore, it is assumed that both receptors fulfill different physiological functions. The human endoglin cDNA has been cloned (62), and mutations in the endoglin gene have been linked to the human disease HHT1 (43). In line, corresponding mouse models disrupted for the endoglin gene confirmed the importance of this receptor especially in TGF-␤ signaling (33)(34)(35)(36). Here, we have cloned the rat endoglin cDNA from HSC, the cell type that, upon TGF-␤ stimulation, is mainly responsible in liver for excessive formation of extracellular matrix components, resulting in hepatic fibrosis (55,71). In contrast to endothelial cells, expressing two endoglin transcript variants, denoted L-and S-endoglin, we found that rat HSC and MFB express two variants of L-endoglin most likely provoked by usage of different polyadenylation sites. The Svariant encodes a protein, which is truncated by 40 amino acids compared with the L-variant (65), missing at least the mutual class I PDZ domain (27). The deduced amino acid sequence of rat endoglin is 69 and 82% identical to the corresponding orthologs of humans and mice (62,63) and shares the typical domain structure excluding an RGD domain found in the human counterpart (62). The carboxyl terminus of endoglin is highly homologous to betaglycan and does not possess a kinase domain (4, 24, 72) but contains a potential interface for interaction with ␤-arrestin 2 and GIPC (25,26). Whether endoglin also interacts with ␤-arrestin 2 or GIPC is currently not known, and the expression of these components in HSC has not been analyzed. Recently, the LIM domain-containing protein zyxin has been shown to specifically interact with the C terminus of endoglin in HUVEC cells, leading to a modulation of their migration behavior (73). In liver, we have recently shown that the LIM domain containing protein CRP2 is exclusively expressed in HSC and is transiently up-regulated during the activation phase of HSC (74). Therefore, CRP2 is a putative candidate interaction partner for endoglin in activated HSC.
The comparison of the signals obtained in Western blot analysis using lysates of HUVEC cells and human myofibroblasts reveals that MFB express relatively high amounts of endoglin (cf. Figs. 6B and 7C). One potential factor that is able to increase the expression of endoglin is the receptor ligand TGF-␤ itself. In endothelial cells, monocytic cells, human mesangial cells, and cultured cell lines, TGF-␤ is able to upregulate promoter activity and endoglin expression (75)(76)(77)(78)(79). Because it is well documented that activated HSC and MFB produce large amounts of TGF-␤ (52, 80), endoglin expression could be maintained during transdifferentiation by persistent autocrine stimulation by TGF-␤. However, under the setting of our TGF-␤ stimulation experiments, there was no significant increase in endoglin protein expression detectable (Fig. 10B). At least in late stage HSC and MFB, this might be explained by the inactivation of the underlying signaling pathway(s). It is postulated that the transcriptional activity of the endoglin gene is facilitated through binding of Smad3 and Smad4 to regulatory sequences within the endoglin promoter (77). In cultureactivated HSC, it was demonstrated that signaling via Smad2 and Smad3 does not occur (69). Therefore, if TGF-␤-dependent transcription of the endoglin gene depends on these Smad proteins, there should be at least a strong diminished effect of TGF-␤ on endoglin transcription.
However, concerning the function of endoglin in HSC and MFB, we can only speculate at this time. The current view of TGF-␤ signaling in HSC comprises the signaling receptors type I (ALK5) and type II and betaglycan as well as the intracellular mediators R-Smad2 and -3 (53, 66 -68). Several studies have been addressed to link TGF-␤ signaling to the process of HSC activation and fibrogenesis. Whereas the growth of quiescent and early activated HSC is inhibited by TGF-␤, fully activated HSC and MFB are released from this inhibitory effect (52,53). On the other hand, TGF-␤ mediates extracellular matrix gene expression in activated HSC and to a high extent in MFB through autocrine stimulation (82). These effects evoked by TGF-␤ have been assigned to the above mentioned simple signaling cascade, involving ALK5, type II receptor, betaglycan, and R-Smad2 and -3 as central signaling elements (53,69). In more detail, it was shown that Smad2 is primarily involved in early HSC, accounting for growth inhibition. Further, Smad3 is activated in early and fully activated HSC, thereby initiating the onset of extracellular matrix gene expression (83). What mechanism governs the postulated autocrine extracellular matrix gene stimulation through TGF-␤ in MFB is presently not known. The messenger RNA for type II receptor and betaglycan are down-regulated in the course of fibrogenesis (66 -68). In line, we observed a down-regulation of betaglycan expression, whereas the endoglin mRNA was almost constant during the transdifferentiation process (Fig. 4B), and in parallel the activation of the R-Smad2 and -3 as well as a Smad3-responsive reporter construct CAGA-luciferase could not be detected in MFB (69).
However, it has recently been shown that TGF-␤ is able to activate an alternative signaling pathway involving a different type I receptor (ALK1) and the intracellular mediators of the R-Smad family 1, 5, and 8, formerly known to be activated by BMPs (8, 48 -51). In endothelial cells, ALK1 signaling seems to be involved in the activation phase, leading to a motile and proliferative phenotype (50), a phenotype resembling transdifferentiated MFB. Current data (e.g. the human diseases HHT1 and HHT2 (43,44) and their respective mouse knockout models of ALK1 (49) and endoglin (33)(34)(35)(36) as well as the observed direct interaction of endoglin and ALK1 (40)) support the hypothesis that endoglin is a component of the ALK1 signaling branch. As a first prerequisite, we have demonstrated the presence of endoglin throughout the process of transdifferentiation by Northern blot (Fig. 4A), Western blot (Fig. 6B), and immunoprecipitation experiments (Fig. 9). The expression of the other elements of this putative signaling pathway in HSC/ MFB is the focus of our current study. Based on data showing the expression of Smad1 (81) as well as our own unpublished data showing ALK1 expression in HSC/MFB, we are confident that this TGF-␤ signaling pathway is active in HSC/MFB.