The Low Density Lipoprotein Receptor-related Protein/ (cid:1) 2 Macroglobulin Receptor Is a Receptor for Connective Tissue Growth Factor*

Connective tissue growth factor (CTGF) expression is regulated by transforming growth factor- (cid:2) (TGF- (cid:2) ) and strong up-regulation occurs during wound healing; in situ hybridization data indicate that there are high levels of CTGF expression in fibrotic lesions. Recently the binding parameters of CTGF to both high and lower affinity cell surface binding components have been characterized. Affinity cross-linking and SDS-polyacrylamide gel electrophoresis analysis demonstrated the binding of CTGF to a cell surface protein with a mass of (cid:1) 620 kDa. We report here the purification of this protein by affinity chroma-tography on CTGF coupled to Sepharose and sequence information obtained by mass spectroscopy. The binding protein was identified as the multiligand receptor, low density lipoprotein receptor-related protein/ (cid:1) 2 -macro- globulin receptor (LRP). The identification of LRP as a receptor for CTGF was validated by several studies: 1) binding competition with many ligands that bind to LRP, including receptor-associated protein; 2) immunoprecipitation of CTGF-receptor complex with LRP antibodies; and 3) cells that are genetically deficient for LRP were unable to bind CTGF. Last, CTGF is rapidly internalized and degraded and this process is LRP-dependent. In sum-mary, our data indicate that LRP Internalization Experiments— To determine the kinetics of internal- ization of CTGF following binding to the cell surface, MG63 cells grown in 6-well dishes were affinity labeled at 4 °C with 0.2 n M 125 I-rhCTGF for 3 h inbinding buffer prepared with minimal essential medium containing 0.2% bovine serum albumin (MEM/BSA). Nonspecific bind- ing, internalization, and degradation were determined by the addition of 50 n M unlabeled CTGF and subtracted from experimental values. In some samples, 3 (cid:3) g/ml heparin or 250 n M lactoferrin was added to the cells during binding. After the binding period, the binding medium was removed and the cells were carefully rinsed twice with fresh, cold MEM/BSA. Pre-warmed MEM/BSA was then added to the cells and the cells were incubated at 37 °C in a CO 2 humidified incubator. At 0, 15, 30, 60, 120, and 180 min after transfer to 37 °C, the cells were har- vested. The medium was collected, treated with 10% trichloroacetic acid, and the precipitable fraction was separated by centrifugation. In these experiments, the supernatant resulting from trichloroacetic acid precipitation contained the degradation products of 125 I-rhCTGF that had been internalized, degraded, and returned to the extracellular space ( i.e. to the conditioned media). The internalized fraction was determined following trypsinization of the cell layer for 10 min. The resulting cell pellet was collected by centrifugation. The radioactivity associated with the trypsin-resistant pellet represented the internalized fraction of 125 I-rhCTGF. To demonstrate LRP-mediated endocyto- sis of CTGF, MG63 cells were affinity labeled with 0.2 n M 125 I-rhCTGF in MEM/BSA in the presence or absence of 50 n M RAP-GST. Nonspecific binding, internalization, and degradation of 125 I-rhCTGF were determined by the addition of 50 n M unlabeled CTGF and subtracted from experimental values. To inhibit lysosomal activity, 200

Connective tissue growth factor (CTGF) 1 (M r ϳ38,000 Da), is a member of the CCN family of growth factors, which is char-acterized by the presence and conserved spacing of some 38 cysteine residues. CTGF was initially identified in and subsequently purified from human umbilical vein endothelial cells conditioned media (1). Early studies demonstrated the strong induction of CTGF expression by transforming growth factor-␤ (TGF-␤), and the promoter region of the CTGF gene contains a unique TGF-␤ response element not shared by other members of the CCN family (2). In an anchorage-independent growth assay of TGF-␤, neutralization of CTGF activity with antibodies or inhibition of CTGF expression with antisense oligonucleotides reduced the ability of the cells to form colonies (3), suggesting that CTGF is a necessary part of the cascade for induction of anchorage-independent growth. More recently, CTGF was shown to be positively regulated by vascular endothelial growth factor (4,5), epidermal growth factor, fibroblast growth factor (4), plasma clotting factor VIIa (6, 7), thrombin (7,8), and by lysophosphatidic acid and serotinin activation of heptahelical receptors (9), but negatively regulated by tumor necrosis factor-␣ (10) and the Wilms tumor suppressor WT1 (11).
The initial discovery of CTGF in vascular endothelium (1), and the subsequent demonstration that CTGF is involved in the proliferation and migration of vascular endothelial cells (12), suggests that CTGF is also an angiogenic factor. The isolation of CTGF from uterine fluid and localization in embryonic and placental tissues suggests a role for CTGF in embryo implantation (13,14).
CTGF is expressed at high levels during granulation tissue deposition in normal healing wounds (15,16). Expression of the extracellular matrix proteins fibronectin, ␣ 5 -integrin, and type I collagen is regulated by CTGF (16 -18). Co-administration of CTGF with TGF-␤ produced a persistent fibrotic reaction that lasted for 14 days, but had resolved and was absent by 7 days in animals treated with TGF-␤ alone (19). The overexpression of TGF-␤ in fibrotic lesions is well documented (reviewed in Refs. 20 and 21), and now many reports indicate that CTGF, too, is overexpressed in many fibrotic lesions (Refs. [22][23][24][25][26][27][28]. The emerging understanding that CTGF is actively involved in the induction and/or maintenance of persistent fibrosis has provided a target for the modulation of matrix overproduction in fibrotic disease. The low density lipoprotein receptor-related protein (LRP)/ ␣ 2 -macroglobulin receptor (␣ 2 MR) (hereto referred to as LRP), is a member of the family of low density lipoprotein (LDL) receptors (30). The LDL receptor family includes two subfamilies: one containing "small" receptor members of ϳ120 kDa, including the LDL receptor (LDLR), apoE receptor-2 (apoE-R2), and very low density lipoprotein receptor; and one containing "large" receptor members of ϳ600 kDa, including LRP, epithelial glycoprotein 330/megalin, and a related protein, sorLA. LDL family receptors bind multiple ligands. At least two of these protein ligands, apolipoprotein E (apoE) and a 39-kDa receptor-associated protein (RAP), bind to all LDL receptor family members (30). A number of functionally and structurally distinct ligands bind LRP and the diversity of these ligands suggests that it may function in a variety of distinct physiological processes, such as lipoprotein metabolism, protease regulation, tissue repair and remodeling, and embryonic development (reviewed in Ref. 30).
Here we have further characterized the major CTGF-binding protein. This protein was affinity purified and identified as LRP. Competitive inhibition of CTGF binding using other LRP ligands and immunoprecipitation of CTGF receptor complexes with LRP antibodies has confirmed that LRP is a receptor for CTGF. Additionally, this report shows that CTGF is rapidly internalized and degraded by cells through an LRP-dependent pathway. This is the first identification of direct interaction of a growth factor with LRP. Our findings suggest that LRP has a regulatory role in the biology of CTGF.

EXPERIMENTAL PROCEDURES
Materials-A baculovirus expression system was used for the preparation of recombinant human CTGF (rhCTGF) and the protein was purified as described. 2 Apolipoprotein E (apoE), LDL, lactoferrin, ␣ 2 M, and chloroquine were purchased from Sigma. Membrane grade Triton TM X-100 was purchased from Roche Molecular Biochemicals Corp. (Indianapolis, IN). Heparin was purchased from Life Technologies, Inc. (Bethesda, MD). A panel of receptor-grade detergents, including Nonidet TM P-40 (Nonidet P-40), Triton TM X-100, Tween TM 20, digitonin, dodecyl-␤-D-maltoside, octylglucoside, deoxycholic acid, and CHAPS was purchased from Roche Molecular Biochemicals.
Cell Lines-The cell lines MG63, MEF, PEA10, and PEA13 were purchased from the American Type Culture Collection (ATCC). Cells were maintained as recommended by the ATCC. The murine bone marrow stromal cell line BMS 2 was graciously provided by Dr. Jeff Gimble (formerly of the Oklahoma Medical Center, currently at Artecel Sciences, Inc., Durham, NC). BMS 2 cells were maintained in Dulbecco's modified Eagle's medium, containing pyruvate, 55 M ␤-mercaptoethanol, and 10% fetal bovine serum.
Binding of CTGF to Monolayer Cultures-rhCTGF was iodinated with chloramine-T (Sigma) by the procedures described. For binding analysis, cells were plated at 2-5 ϫ 10 4 cells/cm 2 in 24-well dishes, 16 -24 h prior to a binding experiment. The cells were washed twice with binding buffer (Dulbecco's phosphate-buffered saline (PBS) (Life Technologies, Inc.) containing 0.2% bovine serum albumin and 0.2% sodium azide) at 4°C. Binding experiments were performed by incubating monolayers of cells with various concentrations of 125 I-rhCTGF in binding buffer for 4 h at 4°C with gentle rocking. Duplicate wells were incubated with at least 100-fold excess of unlabeled rhCTGF for the determination of nonspecific binding.
We determined the kinetics of binding of 125 I-rhCTGF to cells as follows. Supernatants were collected from the cells and directly counted on a Beckman Gamma 5500B counter. Radioactivity measured from this was called the "free" fraction. The cells were then washed four times with cold binding buffer, and lysed with 1% Triton TM X-100. The lysis fraction was counted as above, and contained the "total bound" fraction. The nonspecific bound fraction was also determined and subtracted from each point to yield specific bound fractions. The specific bound and free fractions were analyzed for one-or two-site binding using nonlinear regression (GraphPad Prism, version 3.0, GraphPad Software, Inc., San Diego, CA). The same protocol and analysis was used for competitive binding experiments.
Affinity Labeling and Cross-linking of Monolayer Cultures-Cells to be affinity labeled were washed twice with cold binding buffer, then incubated with 50 -100 pM 125 I-rhCTGF in binding buffer at 4°C for 3-4 h. The binding medium was removed and replaced with binding buffer containing 0.5 mM bis-succinimidyl suberate (BS 3 , Pierce Chemical Co., Rockville, IL). Alternatively, 0.1 mM ethylene glycobis(sulfosuccinimidyl succinate) (Sulfo-EGS, Pierce Chemical Co.) was also used for cross-linking. Cross-linking proceeded at room temperature for 15 min (BS 3 ) or at 4°C for 30 min (Sulfo-EGS) and was terminated by removing the medium and washing the cells several times with buffer containing 250 mM sucrose, 10 mM Tris, pH 7.4, and 10 mM EDTA at 4°C. The complexes were collected in the soluble fraction following cell lysis with 1% Triton TM X-100 (unless indicated otherwise) in PBS with a mixture of protease inhibitors (Calbiochem TM , San Diego, CA). In some experiments, 0.5% Nonidet P-40 was used for cell lysis.
Preparation of Crude Cell Membranes-Membranes were prepared from BMS2 cells by a modification of the procedure described by Atkinson (31). The cells were grown to confluence in roller bottles, and then dissociated with 5 mM EDTA in Dulbecco's PBS lacking calcium and magnesium. Cell pellets were then collected by centrifugation and washed. The cells were then suspended in hypotonic phosphate buffer (7.5 mM NaPO 4 , pH 7.2) and incubated 10 min on ice. The membranes were disrupted by sonication, and the nuclei were stabilized in buffer containing 10 mM NaPO 4 , pH 7.2, 10 mM NaCl, and 3 mM MgCl 2 . The nuclei and whole cells were removed by centrifugation for 5 min at 800 ϫ g, and the supernatant was collected. The supernatant was centrifuged over a cushion of 45% sucrose in Dulbecco's PBS, pH 7.2, for 1 h at 24,000 ϫ g. The membrane fraction located at the sucrose/PBS interface was carefully collected, diluted, and concentrated by centrifugation at 100,000 ϫ g for 15 min. The membrane pellet was resuspended in Dulbecco's PBS, and protein content estimated with the Pierce BCA reagent against an albumin standard (Pierce Chemical Co.). From a preparation of an estimated 10 9 cells, 10 -20 mg of crude membrane protein was frequently obtained.
Affinity Chromatography-One milligram of rhCTGF prepared from baculovirus was immobilized on Reacti-Gel TM GF-2000 (Pierce Chemical Co.) according to the manufacturer's specifications. The membranes (10 -15 mg) were solubilized in 0.2% Triton TM X-100, 20% glycerol, in Dulbecco's PBS (Buffer A) containing a mixture of protease inhibitors, centrifuged (14,000 ϫ g) for 10 min at 4°C to remove insoluble material, and applied to the CTGF affinity matrix. The sample was re-circulated over the column 5-10 times. The flow through was collected and the column was washed with 20 column volumes of Buffer A. The bound sample was eluted with a gradient of 0.135-2 M NaCl in buffer containing 0.2% Triton X-100, 20% glycerol, and PBS. The CTGF receptorcontaining fractions were identified using a solution binding assay (described below). Further purification was achieved by electrophoresis of the fractions in 5% SDS-PAGE under nonreducing conditions. Solution Binding Assay-Preparations of membrane proteins (5 g each) were solubilized in 0.2% Triton TM X-100, 20% glycerol in Dulbecco's PBS (Buffer A) and the insoluble material was removed by centrifugation (14,000 ϫ g) for 10 min at 4°C. Fractions from the affinity column were used directly in the solution binding assay (10 l/fraction). The samples were incubated with 0.2 nM 125 I-rhCTGF for 3-4 h. The cross-linker, BS 3 , was added to a final concentration of 0.5 mM, and the reaction proceeded at room temperature for 15 min. Gel sample buffer was added to each sample, the samples were then heated for 2 min at 100°C, and applied to 5% SDS-PAGE. Following electrophoresis, the gels were dried and analyzed by autoradiography.
Mass Spectroscopy-The gel band of interest (migrating above the 220-kDa marker) was excised with a fresh razor band, destained, and subjected to trypsin digestion (32). The recovered peptide fragments were analyzed by liquid chromatography-mass spectrometry. Microelectrospray columns of 360 m outer diameter ϫ 100 m inner diameter fused silica capillary were packed with 10 -12-cm POROS 10R2, a reversed phase packing material (PerSeptive Biosystems, Framingham, MA) (33). The flow from the high performance liquid chromatography pump (typically 150 l/min) was split pre-column to achieve a flow rate of 500 nl/min. The mobile phase for the gradient elution consisted of (A) 0.5% acetic acid and (B) acetonitrile/water 80:20 (v/v) containing 0.5% acetic acid. The gradient was linear from 0 to 60% B in 30 min. Mass spectra were recorded on an LCQ ion trap mass spectrometer (Finnigan MAT, San Jose, CA) equipped with a microelectrospray ionization source (33). Tandem mass spectra were acquired during the entire gradient automatically as previously described (34). Protein sequence data bases were searched with the tandem mass spectra using the computer program SEQUEST (35). SEQUEST correlates tandem mass spectra of peptides with amino acid sequences from protein and nucleotide data bases (15 public data bases available). The FBSC Non-Redundant Protein Data base (NRP) was obtained as an ASCII file in the FASTA format from Frederick Biomedical Supercomputing Center (ncbi.nlm.nih.gov in/pub/nrdb) by anonymous ftp. Each sequence produced by SEQUEST was verified by manually inspecting the fit of the amino acid sequence to the corresponding tandem mass spectrum.
Purification of RAP-Plasmid DNA containing the human RAP pro-tein sequence was purchased from the ATCC as the I.M.A.G.E. Consortium Clone ID 511113 (36). The RAP cDNA sequence was excised from the vector with the restriction enzymes BamHI and XhoI, and the 1-kilobase fragment was gel purified. The pGEX-4T-1 vector (Amersham Pharmacia Biotech) was cut with BamHI and XhoI, into which the RAP cDNA was ligated. Expression of RAP from this expression construct results in the formation of a fusion protein of RAP and glutathione S-transferase. Correct RAP cDNA insertion was verified by sequence analysis. The RAP-GST fusion protein and the GST protein alone (control vector) were expressed in Escherichia coli and purified as described (37). Antibodies and Immunoprecipitation-A mouse monoclonal antibody reactive to the COOH terminus of LRP was prepared from IgG-11H4 hybridoma supernatant (ATCC) and purified by Protein G-Sepharose (Amersham Pharmacia Biotech). Monoclonal antibodies reactive with the extracellular domains of the ␣ chain and ␤ chain of LRP were purchased from American Diagnostica (catalog numbers 3402 and 3501, respectively).
Affinity labeled cells were extracted with 1% Triton TM X-100 or 0.5% Nonidet P-40 in PBS containing a mixture of protease inhibitors. The extracts were incubated for 2.5 h at 4°C with 1 g of antibody to LRP. Protein G-Sepharose was added to the sample and the incubation continued at 4°C for 1 h. The sample that bound to the suspension was collected by centrifugation, washed four times with PBS, then eluted by boiling with Laemmli gel buffer. The eluted proteins were applied to 5% SDS-PAGE, and analyzed by autoradiography.

FIG. 1. Purification of a CTGF receptor.
A, the solubilized receptor binds CTGF. Crude membranes were prepared from BMS2 cells and solubilized in 1% Triton in PBS with 10% glycerol, or 20% glycerol, and a mixture of protease inhibitors. Intact membranes were used as a binding control (far left panel). The membranes were affinity labeled with 0.2 nM 125 I-rhCTGF and cross-linked. Half the samples also contained a 200-fold excess of unlabeled CTGF. The cross-linked samples were separated on 5% SDS-PAGE under nonreducing conditions. The gels were dried and visualized by autoradiography. Molecular size markers are indicated at the left in kDa, and an arrow indicates the CTGF-receptor complex. The far right panel demonstrates 125 I-rhCTGF cross-linked to BMS2 cells in monolayer culture. B, binding assay of solubilized membrane proteins separated by CTGF affinity chromatography. Crude membrane preparations were solubilized in 1% Triton X-100 in PBS with 20% glycerol and protease inhibitors. The sample was applied to a CTGF affinity matrix and washed through with Buffer A. The bound material was eluted with a NaCl gradient. An aliquot of the loaded sample, the flow through, and each fraction collected during wash and elution were analyzed by affinity labeling, cross-linking, and separation on nonreducing 5% SDS-PAGE. Peak fractions are indicated with a bracket. An arrowhead indicates the CTGF-receptor complex. C, Coomassie staining of affinity purified receptor. The peak fractions were pooled and separated on nonreducing 5% SDS-PAGE. The upper band (long arrow) migrates at the expected position for the previously identified complex. Short arrows indicate 2 other proteins co-purifying on the column. The three indicated bands were excised for further analysis.
Internalization Experiments-To determine the kinetics of internalization of CTGF following binding to the cell surface, MG63 cells grown in 6-well dishes were affinity labeled at 4°C with 0.2 nM 125 I-rhCTGF for 3 h in binding buffer prepared with minimal essential medium containing 0.2% bovine serum albumin (MEM/BSA). Nonspecific binding, internalization, and degradation were determined by the addition of 50 nM unlabeled CTGF and subtracted from experimental values. In some samples, 3 g/ml heparin or 250 nM lactoferrin was added to the cells during binding. After the binding period, the binding medium was removed and the cells were carefully rinsed twice with fresh, cold MEM/BSA. Pre-warmed MEM/BSA was then added to the cells and the cells were incubated at 37°C in a CO 2 humidified incubator. At 0, 15, 30, 60, 120, and 180 min after transfer to 37°C, the cells were harvested. The medium was collected, treated with 10% trichloroacetic acid, and the precipitable fraction was separated by centrifugation. In these experiments, the supernatant resulting from trichloroacetic acid precipitation contained the degradation products of 125 I-rhCTGF that had been internalized, degraded, and returned to the extracellular space (i.e. to the conditioned media). The internalized fraction was determined following trypsinization of the cell layer for 10 min. The resulting cell pellet was collected by centrifugation. The radioactivity associated with the trypsin-resistant pellet represented the internalized fraction of 125 I-rhCTGF. To demonstrate LRP-mediated endocytosis of CTGF, MG63 cells were affinity labeled with 0.2 nM 125 I-rhCTGF in MEM/BSA in the presence or absence of 50 nM RAP-GST. Nonspecific binding, internalization, and degradation of 125 I-rhCTGF were determined by the addition of 50 nM unlabeled CTGF and subtracted from experimental values. To inhibit lysosomal activity, 200 M chloroquine was added to some of the cultures. After a 2-h incubation period at 37°C, the degraded and internalized fractions were determined as described above.

RESULTS
Solubilized Receptor Assay-The bone marrow stromal cell (BMS2) line was chosen as a source for purification of the receptor due to their high level of CTGF-binding to a low affinity receptor. The ease of growth and subculture combined with the high level of CTGF binding were factors involved in choosing these cells for receptor purification. Prior to affinity purification, we optimized the binding and solubilization conditions that would allow for solubilization of cell membrane proteins as well as retain the capacity to bind CTGF. Crude membranes were prepared from BMS2 cells by homogenization and differential centrifugation. A panel of receptor-grade detergents was examined individually in order to solubilize the membrane proteins. Solubilized membranes were incubated with 0.2 nM 125 I-rhCTGF in the presence or absence of 200-fold excess unlabeled CTGF. The samples were cross-linked with BS 3 and separated by 5% SDS-PAGE. Competitive cross-linking and binding was achieved using 1% Triton X-100 containing glycerol to help stabilize the membrane proteins (Fig. 1A). The left panel demonstrates CTGF binding and cross-linking to intact membrane fragments; the addition of 10 and 20% glycerol to the solubilized membranes gave similar binding results, with the inclusion of 20% glycerol providing more favorable results (middle and right panels, Fig. 1A). These conditions were used for the affinity purification of the receptor.
Affinity Purification-Membranes prepared from BMS2 cells were solubilized and applied to a column of CTGF coupled to Sepharose, washed with the same solvent, and bound proteins were eluted with a salt gradient as described above. Fractions were collected and incubated with 125 I-rhCTGF to assay for CTGF binding activity (Fig. 1B). The peak CTGF binding fractions eluted between 50 and 75% buffer B (1.15-1.65 M NaCl). Peak fractions were pooled and separated by SDS-PAGE (Fig.  1C). A band migrating above the highest M r standard (220,000) and estimated as Ͼ400,000 was observed in the Coomassiestained gel. We occasionally observed additional proteins migrating at M r ϭ 200,000 and 150,000 (as shown in Fig. 1C). These bands were excised and analyzed by mass spectroscopy.
The Ͼ400-kDa band contained the peptides listed in Table I and data base analysis identified the protein as LRP. Repeat purification and mass spectroscopy analysis confirmed the result. Identification of the M r 200,000 and 150,000 proteins was not achieved.
LRP Ligands Compete for CTGF Binding-That LRP was a binding protein for CTGF was further confirmed. A number of commercially available LRP ligands were tested for their ability to compete with CTGF for binding to cells (Fig. 2). All ligands tested inhibited 125 I-rhCTGF binding to cells ( Fig. 2A) albeit with 5-10-fold lower affinity than that of CTGF. Additionally, RAP which is able to displace all LRP ligands, prevented CTGF binding (Fig. 2B). Cross-linking analyses confirmed that the inhibition was due to lack of binding to the high M r receptor protein (Fig. 2C). These data support the identity of LRP as a CTGF-binding protein.
Binding of CTGF to LRP-deficient Cells-The availability of cells isolated from LRP gene deletion mouse embryos provided an opportunity to study CTGF binding on cells genetically lacking LRP. The cell lines examined include a homozygous LRP-deficient mouse embryo fibroblast cell line, PEA 13 (Ϫ/Ϫ), a heterozygous LRP-deficient cell line, PEA 10 (ϩ/Ϫ), and a wild type mouse embryo fibroblast, MEF1 (ϩ/ϩ).
Each of these cell lines was examined for CTGF binding in a CTGF-binding assay. The binding parameters obtained (determined by nonlinear regression analysis) are summarized in Table II. MEF1 (ϩ/ϩ) and PEA10 (ϩ/Ϫ) cells bound 125 I-rhCTGF with single site binding kinetics, while the LRPdeficient cells, PEA13, did not bind 125 I-rhCTGF in this assay. The heterozygous cell line, PEA10, appeared to have approximately one-fifth the number of CTGF-binding sites as observed in wild type MEF1 cells, yet had the same K d for binding. These data are suggestive of a gene dosage affect on LRP protein expression levels, but not on the kinetics of ligand association/dissociation.
Cross-linking analysis was performed using these LRP-deficient cells to confirm that the observed binding of CTGF to these cells involved the high M r protein (Fig. 3). Both MEF (ϩ/ϩ) and PEA10 (ϩ/Ϫ) cells, but not PEA 13 (Ϫ/Ϫ) cells, bound and could be cross-linked to CTGF to the high M r protein. The binding of CTGF to the LRP-expressing cells, but not to the LRP-deficient cells, supports the identification that LRP is a CTGF-binding protein.

TABLE I
The gel band of interest (migrating above the highest M r standard of 220,000) was excised with a clean razor blade, destained, and subjected to trypsinization The recovered peptide fragments were analyzed by light chromatography/mass spectrometry and mass measurements were analyzed using the SEQUEST analysis program.
First round of mass spectroscopy yielded the following peptides (R)AALSGANVLTLIEKDIR (K)NAVVQGLEQPHGLVVHPLR (R)SERPPIFEIR (K)TVLWPNGLSLDIPAGR (R)TTLLAGDIEHPR (R)YVVISQGLDKPR Second round of mass spectroscopy yielded the following peptides

Immunoprecipitation of CTGF/High M r Complexes-Fur-
ther confirmation that LRP is the high M r CTGF-binding protein was demonstrated using LRP antibodies. Detergent lysates were prepared from 2 cell types, BMS2 and MG63, that had been affinity labeled with 125 I-rhCTGF, cross-linked, and immunoprecipitated with antibodies directed to LRP (Fig. 4). Antibodies specific to the extracellular epitopes in both the ␣ and ␤ subunit chains of LRP, as well as an antibody that recognizes a cytoplasmic domain of LRP, immunoprecipitated the CTGF-containing complex, while control normal murine IgG did not. CTGF, migrating at the front of the gel, was poorly immunoprecipitated with these LRP antibodies. Notably, the antibody recognizing the LRP ␣ chain was most active for MG63 (human origin) cells while the antibody recognizing the LRP ␤ chain was most active for BMS2 (murine origin) cells. The LRP ␣ chain antibody is immunoreactive with the human protein only. While the ␤ chain antibody is reactive with both human and rodent protein, the difference in recognition could be due to different detergent extraction conditions.
Internalization of CTGF-The internalization kinetics of CTGF following binding was examined. For these experiments, the MG63 cell line was utilized because the conditions often promoted lifting of BMS2 cells from monolayers whereas the MG63 cultures remained intact. Monolayer cultures of MG63 cells were incubated at 4°C with 0.2 nM 125 I-rhCTGF in me-   dium. The cells were then transferred to 37°C and internalization and degradation of 125 I-rhCTGF were followed at specific time points (Fig. 5). In one-third of the cultures, an LRP ligand, lactoferrin, was included to compete for LRP binding. To the other third of the cultures, heparin was added to inhibit binding of CTGF to the cell surface. The results demonstrated that CTGF is rapidly internalized by cells, occurring within 30 min following the temperature shift to 37°C. Degradation of CTGF, determined by the release of label into the medium, was detected after 30 min, and continued throughout the time course of the experiment. Lactoferrin moderately reduced internalization and degradation, suggesting LRP-mediated internalization. An LRP internalization pathway transports ligands through an endosomal compartment, wherein the acidic lumen of the endosome promotes ligand dissociation, and the ligand is subsequently degraded (38). This process is sensitive to the drug, chloroquine, which raises the pH of the endosome and thus inhibiting the pH-dependent ligand dissociation. We examined whether CTGF is internalized by LRP and degraded through the same pathway (Fig. 6). In these experiments, cultures of MG63 cells were incubated at 37°C with 125 I-rhCTGF in the presence or absence of the LRP antagonist, RAP, or with chloroquine. MG63 cells treated with chloroquine showed in an increased level of intracellular 125 I-rhCTGF. Degraded 125 I-rhCTGF was barely detectable in these chloroquine-treated cell cultures, whereas the control cultures had high levels of degraded CTGF. These results support an endosomal-dependent pathway for the LRP-mediated internalization of CTGF, with ligand dissociation required for its degradation. RAP addition to the cultures eliminated CTGF internalization and degradation. These results provide evidence that CTGF uptake and degradation by cells are LRP dependent. DISCUSSION In this study, we have presented evidence that the major cell membrane protein to which CTGF binds is a high M r protein, identified previously as LRP. This study provides the first report of a growth factor that binds directly to LRP, although recently the Wnt family of secreted molecules has been demonstrated to bind to other members of the LRP family (39). Previous suggestion of growth factor interaction with LRP has been through complexes with ␣ 2 M as a clearance molecule (40).
We were unable to obtain sequence information for two lower M r proteins that were occasionally affinity co-purified with LRP (estimated M r of 150,000 and 200,000). These proteins were not detected in the cross-linking/binding assay. CTGF, and the structurally related protein, Cyr61, bind to a number of integrins, including ␣ v ␤ 3 , ␣ IIb ␤ 3 , and ␣ 6 ␤ 1 (41)(42)(43)(44). Recently it has been demonstrated that CTGF-dependent cell adhesion induces cell signaling through integrin-mediated pathways (42). While the identity of the co-purifying bands is unclear, the M r estimates are suggestive for some of the integrin subunits. The possibility exists that these bands represented breakdown products of the large LRP protein. Further purification and analysis will be necessary for unambiguous sequence identity of these proteins.
Our data indicate that at least part of the CTGF-binding site on LRP is similar or common to the binding site utilized for many of the LRP ligands, as these ligands competed for CTGF binding, although with lower affinity (Fig. 2). LRP contains multiple copies of cysteine-rich repeats known as LDLR class A repeats (complement-type repeats) arranged in 4 clusters, forming ligand-binding sites (45)(46)(47)(48)(49). The binding site for many LRP ligands has been sublocalized (50), and NMR solution structure of the complement-like repeat CR3 from LRP has recently been determined (51). It will be of interest to sublocalize the binding site for CTGF within the extracellular domain of LRP.
With the exception of RAP, which binds to LRP with very high affinity (K d ϭ 3 nM (52)), most LRP ligands bind LRP with a moderate affinity. For example, hepatic lyase binds LRP with a K d of 52 nM (53); PEA binds LRP with a K d of 14 nM (54); and, coagulation factor VIII binds LRP with a K d of ϳ50 nM (55). Urokinase-type plasminogen activator (uPA) is synthesized as a single chain zymogen, pro-urokinase, and binds LRP with a K d of 45 nM; the active two-chain enzyme (tc-uPA) binds LRP with a K d of 60 nM. Complex formation with plasminogen activator inhibitor type I increases the affinity of uPA to LRP, such that the uPA/plasminogen activator inhibitor type I complex binds LRP with a K d of 1-3 nM (56,57).
Many LRP ligands are polyanionic proteins that bind heparin, suggesting that their binding to cell surface heparin sulfate proteoglycan may be required for interaction with LRP (58). heparin sulfate proteoglycan may serve to concentrate the ligands at the cell surface, promoting binding with the relatively low affinity interaction with LRP. The observed K d for CTGF binding to LRP is 0.5-1 nM, 2 which is a much higher binding affinity than reported for other LRP ligands. In Nesbitt et al., 2 we showed that, while CTGF is a heparin-binding protein, CTGF is capable of binding to the surface of cells depleted of GAGs. Perhaps the higher affinity of CTGF for LRP obviates concentration of CTGF on the cell surface by heparan sulfate proteoglycans. Nevertheless, the significance of the high affinity binding of CTGF to LRP is unclear at this time, but suggests an important role for LRP in CTGF biology.
Experiments aimed at creating mice lacking LRP by genetic deletion of the LRP allele failed to produce viable embryos (59), suggesting an important role for LRP during early development. CTGF is highly expressed in uterine epithelium and in decidualizing endometrial stromal cells during early pregnancy, as well as in embryonic ectoderm, endoderm, and at the ectoplacental cone after implantation (13). LRP is localized to invading trophoblastic cells, in decidualizing tissue, and at the ectoplacental cone after implantation (61,62). The similar developmental localization patterns support possible interaction between CTGF and LRP during embryonic development.
We have demonstrated that CTGF is rapidly internalized and degraded by cells, and that this represents an LRP-mediated process. We have experienced a rapid disappearance of CTGF from culture medium after addition of CTGF to cell cultures. A rapid internalization and degradation pathway would account for the inability to detect CTGF in conditioned medium. We suggest that one function of LRP in CTGF biology is to modulate the concentration of free CTGF in the extracellular space or at the cell surface.
Whether or not LRP serves as a signaling receptor for CTGF bioactivity remains to be determined. LRP binds and internalizes numerous ligands that would necessitate separate ligandspecific signaling mechanisms; yet, signaling involving LRP has been suggested in at least some systems. Most recently, signaling by the Wnt family of proteins requires the LRP-5 or LRP-6 members of the LRP family to function as co-receptors with the Frizzled family of receptors (39,63,64). In a different system, LRP mediates long-term potentiation in hippocampal neurons (65) and activation with activated ␣ 2 M promotes calcium influx in neuronal cells (66). Other members of the LDLR family have recently been demonstrated to transduce extracellular signals. For instance, very low density lipoprotein receptor and apoE-R2 function as receptors in the Reelin/disabledmediated neuronal migration pathway (67)(68)(69)(70). The short cytoplasmic domain of LRP contains multiple potential endocytosis motifs including 2 NPXY motifs; but, a recent study demonstrated that a YXXL motif serves as the dominant signal for LRP endocytosis (71), leaving the NPXY motifs available for interaction with other signaling or adaptor proteins. Disabled-1 (dab-1) interacts with the NPXY motifs of LRP in neuronal cells (69,70,72,73). When tyrosine phosphorylated, mDab1 binds nonreceptor tyrosine kinases, such as src, fyn, and abl (74). Another member of this family, mDab2/p96/ DOC-2, has ϳ50% sequence conservation with the amino-terminal sequence of mDab1, but is expressed in a wider variety of cells (75). Dab2/p96/DOC-2 has been shown to compete with SOS to bind Grb2, suggesting a negative regulatory role in the Ras signaling cascade (76) and may negatively regulate mitogenesis (77). Whether Dab2/p96/DOC-2 or an unidentified member of the Dab family binds to LRP remains unknown and is of much interest to investigate.
We have been unable to detect tyrosine phosphorylation of LRP in cells treated with CTGF (data not shown). In addition, we have examined LRP purified by CTGF affinity chromatography in kinase assays. Although an associated kinase activity toward casein co-purified with LRP, this kinase activity was found not from LRP itself, but rather from an associating protein. Most importantly, the kinase activity did not respond to CTGF stimulation. 3 Therefore it remains unclear whether LRP phosphorylation plays any role in mediating biological functions of CTGF.
LRP and many of its ligands have been localized to the senile 3 D. Li, unpublished observations. plaques of Alzheimer's disease (78 -80), in plaque-associated activated astrocytes, but not in resting astrocytes (80). Recent biochemical data support a role for LRP in the pathogenesis of Alzheimer's disease (81,82). There is an increase in LRP expression in monocytes from patients with coronary heart disease (83), and LRP is abundantly expressed by smooth muscles cells and macrophages in human atherosclerotic lesions (84). In the renal ablation model of experimental kidney fibrosis, LRP expression is significantly increased in the glomeruli and interstitium with preferential localization to fibrotic lesions (60). While to date CTGF has not been documented in Alzheimer's lesions, it is highly expressed in fibrotic kidney disease (22) and atherosclerotic plaques (24). With co-localization of CTGF and LRP in these diseases, it will be important to understand the functional role of the high affinity interaction of these proteins.
It will be of interest to examine whether altered LRP expression is a general marker of disease characterized by impaired wound healing or in fibroproliferative disorders. Future work examining the role of LRP in CTGF signaling by itself or as a co-receptor to a yet unidentified CTGF receptor, and the effect of the modulation of LRP on CTGF activity will be important studies to probe the biology of CTGF and for the design of therapeutics for intervention in fibrotic disease.