Inhibition of Endosomal Insulin-like Growth Factor-I Processing by Cysteine Proteinase Inhibitors Blocks Receptor-mediated Functions*

The receptor for the type 1 insulin-like growth factor (IGF-I) has been implicated in cellular transformation and the acquisition of an invasive/metastatic phenotype in various tumors. Following ligand binding, the IGF-I receptor is internalized, and the receptor·ligand complex dissociates as the ligand is degraded by endosomal proteinases. In the present study we show that the inhibition of endosomal IGF-I-degrading enzymes in human breast and murine lung carcinoma cells by the cysteine proteinase inhibitors, E-64 and CA074-methyl ester, profoundly altered receptor trafficking and signaling. In treated cells, intracellular ligand degradation was blocked, and although the receptor and two substrates, Shc and Insulin receptor substrate, were hyperphosphorylated on tyrosine, IGF-I-induced DNA synthesis, anchorage-independent growth, and matrix metalloproteinase synthesis were inhibited. The results suggest that ligand processing by endosomal proteinases is a key step in receptor signaling and function and a potential target for therapy.

The receptor for the type 1 insulin like growth factor (IGF-IR) 1 and its ligands IGF-I and IGF-II regulate normal somatic growth during embryogenesis and post-natal development (1). However, IGF-I and its receptor have also been implicated in the development of several human malignancies such as glioand neuroblastomas, breast, lung, colon, and prostate carcinomas (2). Furthermore, IGF-I has been shown to regulate the induction and maintenance of the transformed phenotype and the acquisition of the invasive/metastatic phenotype (3,4).
IGF-I binding to its cell surface receptor activates intrinsic tyrosine kinase activity generating docking sites for src homology or phosphotyrosine binding domain-containing proteins and leading to activation of downstream signaling cascades (5). This is coupled to ligand-mediated internalization of the receptor leading initially to attenuation of the IGF-I response through down-regulation of plasma membrane receptor levels and subsequently to receptor recycling to the cell surface (6). These processes of receptor internalization and recycling have long been regarded as key events in the regulation of receptor activity (7). In the acidic milieu of the endosomes, ligand dissociation from its receptor is thought to be required for receptor recycling to the cell surface, and abrogation of ligand dissociation leads to the intraendosomal accumulation of ligand⅐receptor complexes (8). In the case of insulin, ligandreceptor dissociation at acidic pH values is coupled to intraendosomal proteolysis by an as yet undefined proteinase (9). For other internalized ligands such as glucagon or epidermal growth factor (10,11), the cysteine proteinase cathepsin B has been implicated in endosomal ligand degradation. These studies define a tight coupling between receptor-ligand internalization, ligand dissociation and proteolysis, and receptor trafficking in the endosomal/lysosomal compartments.
Previously one of our laboratories reported that treatment of murine Lewis lung carcinoma subline H-59 cells with the cysteine proteinase inhibitor E-64 blocked their invasiveness in vitro and inhibited liver metastases formation in vivo (12). In the present work, we investigated further the molecular mechanism underlying the E-64 effects using H-59 and MCF-7 cells. We show here that E-64 and a second cathepsin B inhibitor, CA074-methyl ester (ME), inhibited endosomal degradation of IGF-I, and this led to an alteration of IGF-I receptor signaling. Under these conditions, the IGF-I receptor, as well as the signaling molecules Shc and IRS-1, were highly tyrosine-phosphorylated. On the other hand, IGF-I-induced DNA synthesis, anchorage-independent growth, and MMP-2 synthesis were all inhibited. Selective intervention at the internalization/degradation level may be of therapeutic relevance in abrogating IGF-I-mediated tumor formation and metastasis.

EXPERIMENTAL PROCEDURES
Cell Lines and Tissues-H-59 is a highly metastatic subline of the Lewis lung carcinoma with metastatic predilection for the liver, developed by one of our laboratories (13). Human breast carcinoma cell line MCF-7 (14) was a gift from Dr. Mader (Dept of Biochemistry, University of Montreal, Quebec, Canada). Endosomal fractions were prepared from livers of male Harlan Sprague-Dawley rats after an 18-h period of fasting. The livers were homogenized, and the endosomal fractions were isolated by discontinuous sucrose gradient centrifugation and collected at the 0.25 to 1.0 M sucrose interface (9, 10, 15). The soluble extract (ENs) from the endosomal fractions was isolated by freeze/thawing in 5 mM sodium phosphate, pH 7.4, and disrupted in the same hypotonic medium using a small Dounce homogenizer (15 strokes with the tight Type A pestle) followed by centrifugation at 300,000 ϫ g for 30 min as described previously (9,10,15).
Reagents and Antibodies-E-64 (trans-epoxysuccinyl-L-leucylamido (4-guanidino)-butane), protein A-Sepharose beads, and MTT (3-(4,5dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; thiazolyl blue) were purchased from Sigma. CA074-ME (N-(L-3trans-propylcarbamoyloxirane-2-carbonyl)-L-isoleucyl-L-proline), a pro-inhibitor of intracellular cathepsin B (16), was from Peptides International (Louisville, KY). [ 3 H]thymidine (2.0 Ci/mmol) was from DuPont. Recombinant Human IGF-I used for all the biological assays was purchased from Intergen (Purchase, NY). 125 I-labeled IGF-I (2000 Ci/mmol) used for the ligand binding assay was obtained from Amersham Pharmacia Biotech. Human recombinant IGF-I used for the IGF-I proteolysis assay was radioiodinated by the lactoperoxidase method as described previously for insulin (15) to specific activities of 350 -500 Ci/mmol and purified by gel filtration on Sephadex G-50. A 1.1-kb type IV collagenase cDNA fragment was kindly provided by Dr. W. Stetler-Stevenson (NIH, Bethesda, MD). A 700-bp IGF-IR cDNA fragment was a kind gift from Dr. M. Pollak (Lady Davis Research Institute, Montreal, PQ, Canada). The following antibodies were used: rabbit antiserum to MMP-2 (Ab-45), a kind gift from Dr. William Stetler-Stevenson (NIH); rabbit antisera against IRS-1 were either obtained as a kind gift from Dr. Barry I. Posner (McGill University) or purchased from Calbiochem; anti-phosphotyrosine mAb PT-66 was from Sigma; horseradish peroxidase (HRP)-conjugated mAb RC20-H was from Transduction Laboratories (Mississauga, Ontario, Canada); mAb C-20 to the murine IGF-IR ␤ subunit was from Santa Cruz Biotechnology (Santa Cruz, CA); mAb ␣ IR3 to human IGF-IR was from Calbiochem; HRP-conjugated goat anti-mouse and goat anti-rabbit IgG antibodies were from Bio-Rad Top, Northern blot analysis was performed on total RNA extracted from H-59 cells cultured for 48 h in the presence of serum, with or without 10 g/ml E-64. The blots were probed consecutively with 32 P-labeled MMP-2 and cyclophilin cDNA probes, and the bands were analyzed by laser densitometry. Results of the densitometry are shown in the bar graph and are expressed as MMP-2: cyclophiline ratios. Bottom, proteins in the concentrated media (ϫ 60) collected from H-59 cells cultivated for 48 h with or without 10 g/ml E-64 and in the presence or absence of 10 ng/ml IGF-I were analyzed by immunoblotting using mAb Ab-45 to MMP-2 (I) and by gelatin zymography (II). a, control (untreated); b, E-64treated; c, untreated, IGF-I stimulated; and d, E-64 treated, IGF-I stimulated H-59 cells.
Functional Assays for IGF-IR-Thymidine incorporation and soft agar cloning assays were performed as follows: semiconfluent cultures of H-59 or MCF-7 were cultured in serum free-medium for 24 h with or without E-64, dispersed, seeded onto 96-well polystyrene plates (Falcon), and incubated with different concentrations of IGF-I with or without E-64 for 54 h prior to pulsing with 0.1 mCi/ml of [ 3 H]thymidine for 18 h. For soft agar cloning (17) the tumor cells were mixed with a solution of 0.8% agar added to an equal volume of a 2ϫ concentrated RPMI-fetal calf serum medium with or without 10 g/ml of E-64, plated on solidified 2% agar at a concentration of 10 4 cells/plate, and supplemented with 1 ml of RPMIfetal calf serum containing or not 10 g/ml of E-64. This medium was replenished on alternate days for 12 days. MMP-2 synthesis was analyzed by Northern and Western blotting and by gelatin zymography, which were performed as previously described (18). For Northern blotting, 32 Plabeled 1.1-kb human MMP-2 and 800-bp rat cyclophilin (19) cDNA frag-ments were used as hybridization probes. Western blotting and gelatin zymography were performed on concentrated (60ϫ) serum-free media conditioned by H-59 cells for 48 h in the presence or absence of IGF-I with or without 10 g/ml E-64. Blots were probed with a 1:500 dilution of mAb Ab-45 to MMP-2 and an alkaline phosphatase-conjugated affinity purified goat anti-rabbit IgG, diluted 1:2000.
Measurement of Cell Surface IGF-I Receptors-The ligand binding assay and fluorocytometry were used to measure cell surface IGF-I receptors on the murine H-59 and human MCF-7 cells, respectively. 2-day-old H-59 cultures were replenished with fresh medium containing or not 10 g/ml E-64, and the binding assay was performed 24 h later using 8 -1500 pM 125 I-labeled IGF-I with or without graded concentrations of unlabeled IGF-I. Incubation was for 1 h at 37°C following which the cells were rinsed and lysed in 0.01 N NaOH containing 0.1% Triton X-100 and 0.1% SDS, and the radioactivity was measured. The number of cells/plate at the time of the assay was determined from triplicate control wells, which were manipulated in a similar manner. The Ligand program (20, 21) was used to calculate the number of IGF-I binding sites per cell. IGF-I receptors on MCF-7 cells were immunofluorescence-labeled using 5 g/ml mAb ␣ IR3 and a fluorescein isothiocyanate-conjugated goat anti-mouse IgG (diluted 1:50). Prior to labeling, the cells were cultured for 24 h in serum free-RPMI with or without 10 g/ml E-64 then dispersed, reseeded at a density of 10 5 cells/well into 96-well plates, stimulated with 10 ng/ml of IGF-I for 10 min, and incubated for an additional 30 min at 37°C. Labeled cells were fixed in phosphate-buffered saline containing 1% formalin and analyzed using a FACS Calibur system (Becton-Dickinson, San Jose, CA).
Ligand Proteolysis Assays-Proteolysis of IGF-I was measured using the soluble endosomal extract prepared from rat liver parenchyma (1 ng), and cell lysates (3-15 mg) were derived from H-59 and MCF-7 cells cultured for 24 h with or without 10 g/ml E-64, lysed by incubation in 50 mM phosphate buffer, pH 7.4, containing 0.5% Triton X-100, 0.5% deoxycholate, and 0.2 M NaCl for 30 min at 4°C and then clarified by centrifugation at 30,000 ϫ g for 30 min. These preparations were incubated for various lengths of time at 37°C with 10 Ϫ6 M unlabeled or 50,000 cpm [ 125 I]-labeled IGF-I in 200 or 400 l of 50 mM citratephosphate, pH 5, respectively. The integrity of the radiolabeled ligand was assessed by precipitation with 10% trichloroacetic acid (9,10,15). To measure proteolysis of the unlabeled IGF-I, the samples were acidified with acetic acid (15%) and immediately loaded onto a reversephase HPLC column. Reverse-phase HPLC was performed on a Waters model 600 liquid chromatograph equipped with a model U6K sample injector fitted with a 500-ml loop and a mBondapak C18 column (0.39 ϫ 30 cm, 10 mm particle size; Waters). Samples were chromatographed using as eluent a mixture of 0.1% trifluoroacetic acid in water (Solvent A) and 0.1% trifluoroacetic acid in acetonitrile (Solvent B) with a flow rate of 1 ml/min. Elution was carried out using two sequential linear gradients followed by the following isocratic elution: an initial gradient of 0 -20% Solvent B (30 min); a second gradient of 20 -39% Solvent B (15 min); and a third isocratic elution of 39% Solvent B (15 min). Eluates were monitored on-line for absorbance at 214 nm with a liquid chromatography spectrophotometer.
Immunoprecipitation and Western Blot Analysis-MCF-7 and H-59 cells were treated with 10 ng of IGF-I for 5 min following or not pretreatment with E-64 or CA074-ME as described above. Cells were then washed with phosphate-buffered saline, solubilized in 30 mM Hepes, pH 7.4, 150 mM NaCl, 1% Triton X-100 and spun at maximal speed in a microfuge for 15 min at 4°C. Cell lysates (1 to 3 mg) were then immunoprecipitated, respectively, with anti-Shc, anti-IRS-1 (from B. I. Posner), or anti-IGF-IR overnight at 4°C. Immunoprecipitates were collected by addition of protein A-Sepharose beads, washed three times with lysis buffer, and resuspended in Laemmli sample buffer (22). Immunoprecipitates were resolved by SDS polyacrylamide gel electrophoresis and transferred onto nitrocellulose membranes followed by either immunoblotting with anti-phosphotyrosine antibodies (Sigma) or RC-20 (Transduction Laboratories) conjugated to HRP or with anti-IRS-1 (Calbiochem), anti-Shc, or anti-IGF-IR antibodies. The blots were revealed by enhanced chemiluminescence followed by radioautography on Kodak X-Omat AR films.

Abrogation of IGF-IR Functions by the Cysteine Proteinase
Inhibitor E-64-DNA synthesis, anchorage-independent growth, and production of the matrix metalloproteinase MMP-2 are three IGF-I-regulated cellular functions that are critical for the expression of the malignant phenotype (4,18). Treatment of MCF-7 and H-59 cells with the cysteine proteinase inhibitor E-64 at the non-toxic concentration of 10 g/ml (12) reduced by factors of 7 and 10, respectively, the cloning efficiency of these cells in semi-solid agar (Fig. 1A) and abrogated [ 3 H]thymidine incorporation in response to IGF-I in both cell lines (Fig. 1B). Furthermore, MMP-2 mRNA synthesis in serum-containing medium, which was previously shown to depend on IGF-IR expression in these cells (23), was reduced by a factor of 2, and this was reflected in reduced levels of MMP-2 protein levels in H-59 conditioned medium analyzed by immunoblotting and gelatin zymography (Fig. 1C). IGF-I-induced MMP-2 production and activity measured following addition of 10 ng/ml IGF-I to serum-starved cells were likewise reduced (Fig. 1C). dopeptidases such as cathepsin B and L are inhibited by E-64 and have been implicated in the processing of receptor⅐ligand complexes (9,10). We postulated that IGF-I receptor-mediated cellular functions in E-64-treated cells were altered as a consequence of perturbed endosomal processing of the internalized receptor-bound IGF-I. Changes in IGF-I proteolysis were therefore investigated in lysates of E-64-treated tumor cells, as well as in isolated liver parenchymal endosomal fractions, which were incubated with exogenous IGF-I at acidic pH values. Reversephase HPLC analysis revealed that IGF-I degradation products, which were detectable in the untreated preparations, were absent following E-64 treatment (Fig. 2, A and C). This was subsequently confirmed when cell lysates and endosomal fractions were incubated with radioiodinated IGF-I for 1 h, and trichloroacetic acid precipitation was used to monitor ligand integrity. An increase in trichloroacetic acid-soluble radioactivity over time was evident in the untreated preparation, but this was completely abolished by E-64 pretreatment (Fig. 2, B and D) indicating that IGF-I proteolysis was dramatically reduced.
Reduced Cell Surface Levels of IGF-IR in E-64-treated Cells-One possible consequence of ligand proteolysis blockade is the endosomal trapping of receptor⅐ligand complexes leading to a decreased availability of free receptor for recycling at the cell surface. We measured the effect of E-64 treatment on cell surface IGF-I receptor levels on H-59 and MCF-7 cells. Ligand binding analysis revealed that the number of IGF-I binding sites as measured after the addition of 125 I-IGF-I to H-59 cells was reduced by more than 2-fold, from 3.9 ϫ 10 4 sites/cell on untreated to 1.8 ϫ 10 4 sites/cell on E-64-treated cells (Fig. 3A). Flow cytometric analysis with a monoclonal antibody (mAb ␣ IR3) to the ␣ subunit of the human IGF-I receptor revealed that 40 min after the addition of ligand to serum-starved MCF-7 cells, there was a reduction of 45% in the number of immunolabeled cells with the mean intensity of fluorescence declining from 255 to 82 (Fig. 3B). In neither of these cell types did E-64 treatment cause a reduction in IGF-IR mRNA levels (Fig. 3C) or in the total level of immunoprecipitable receptor (Fig. 3D). These experiments suggested that the reduction of IGF-IR expression at the cell surface was not because of a change in receptor transcription or translation.
Increased Levels of Tyrosine-phosphorylated IGF-IR and Substrates in Cells Treated with Cysteine Proteinase Inhibitors-One of the earliest molecular events in IGF-IR ligand-induced signaling is the autophosphorylation of tyrosine residues on the receptor ␤ subunit and the subsequent phosphorylation of downstream substrates such as IRS-1 and Shc. We first measured ligandinduced tyrosine phosphorylation of the receptor in E-64-and ME-treated cells by immunoprecipitation with anti-IGF-IR antibodies followed by immunoblotting with either anti-phosphotyrosine antibodies or anti-IGF-IR antibodies. Three experiments were performed for each condition. In all the experiments, we found an increase in tyrosine-phosphorylated receptor ␤ subunit in inhibitor-treated cells. For ME-treated cells the means of the increases were 3.75 (p Ͻ 0.0007, n ϭ 3)-and 2.08 (p Ͻ 0.001, n ϭ 3)-fold, and for E-64-treated cells they were 2.1 (p Ͻ 0.002, n ϭ 3)and 1.33 (p Ͻ 0.005, n ϭ 3)-fold compared with IGF-I-treated H-59 and MCF-7 cells, respectively (Fig. 4, A and B). The different responses to the two inhibitors may reflect a higher efficiency of intracellular uptake of ME because of its increased membrane permeability. Two major substrates of the IGF-IR are Shc and IRS-1. Their interactions with the activated IGF-IR have been mapped to the Tyr 950 residue suggesting that they could potentially compete for binding (24). In MCF-7 cells, a 1.7-fold increase in tyrosine-phosphorylated IRS-1 was noted in ME-treated cells stimulated with IGF-I as compared with cells stimulated with IGF-I only (Fig. 4D, n ϭ 3, p Ͻ 0.0008), but no significant change in p52 Shc tyrosine phosphorylation was noted (data not shown). In H-59 cells, on the other hand, p52 Shc tyrosine phosphorylation increased by 1.9-fold (n ϭ 2) in response to ligand binding in E-64-or ME-pretreated cells (Fig. 4C, upper and middle panels), whereas Shc expression per se did not change (Fig. 4C, bottom  panel, n ϭ 3). In these cells, no detectable levels of IRS-1 were observed by Western blotting or immunoprecipitation using experimental conditions that identified this protein in MCF-7 cells. It appears therefore that in these cells, differential expression levels of the adaptor proteins Shc and IRS-1 may determine their preferential IGF-IR binding and phosphorylation in response to IGF-I. DISCUSSION Our results show that inhibition of cysteine proteinase activity by E-64 resulted in reduced cell surface IGF-IR expression levels and abrogation of cellular responses to IGF-I. In an apparent paradox, however, treatment with this or a second cathepsin B inhibitor, CAO74-ME, also caused an increase in tyrosine phosphorylation of the IGF-IR ␤ subunit, IRS-1, and Shc.
When taken together with our findings that IGF-I proteolysis was blocked in E-64-treated liver parenchymal endosomes and in tumor cell lysates, our results indicate that inhibition of processing of the IGF-IR⅐IGF-I complex has the following two FIG. 4. Increased levels of tyrosine-phosphorylated IGF-I receptor in E-64-treated tumor cells. Serum-starved H-59 (A) or MCF-7 (B) cells were treated overnight with 10 g/ml E-64 or CA-074 ME and then stimulated for 10 min with 10 ng/ml IGF-I. The cells were lysed, and the proteins were immunoprecipitated using 10 g of mAb ␣ IR3 (for MLF-7) or mAb c-20 (for H-59) and then immunoblotted with anti-phosphotyrosine (PY) mAb PT-66 or anti-IGF-IR mAb C-20 using an HRP-conjugated goat anti-mouse IgG antiserum as a secondary antibody. Each of these experiments were done three times. Cell lysates were also immunoprecipitated with anti-Shc or anti-IRS-1 antibodies, and the immunoprecipitates were resolved by SDS polyacrylamide gel electrophoresis and immunoblotted with either anti-phosphotyrosine antibodies or antibodies to Shc or IRS-1. In H-59 cells (n ϭ 2), Shc was found to be tyrosine-phosphorylated (C) whereas in MCF-7 cells (n ϭ 3), IRS-1 was tyrosine-phosphorylated (D). major consequences: (i) receptor recycling to the plasma membrane is decreased, and (ii) the IGF-IR ␤ subunit and the IRS-1/Shc substrates remain hyperphosphorylated, and this attenuates rather than activates IGF-IR-mediated biological functions such as the induction of DNA synthesis and MMP-2 production. A direct effect of cysteine proteinase inhibitors on the synthesis or activity of tyrosine phosphatases have not, to our knowledge, been documented. Our data may therefore be best reconciled in a model whereby inhibition of IGF-I proteolysis "traps" the receptor⅐ligand complex in a subcellular compartment that retains the receptor in a hyperphosphorylated state (as represented in Fig. 5). This, in turn, may result in retention of phosphorylated receptor substrates in the same compartment, preventing them from accessing normal signaling pathways in the cytoplasm. Alternatively, receptor retention in the endosomes may trigger an alternate signal transduction cascade that leads to inhibition rather than activation of DNA synthesis and MMP-2 production. Receptor phosphorylation and signaling within the endosomal compartment have been documented for the insulin receptor kinase (15,25), and epidermal growth factor receptor activation and Shc recruitment within endosomes have also been observed (26,27). Reports have linked receptor internalization to the activation of the Shc/mitogen-activated protein kinase pathway (28,29). On the other hand, inappropriate localization of signaling molecules after ligand-induced activation of tyrosine kinase receptors may also lead to the generation of abnormal signaling pathways (30,31).
It has been clearly shown that ligand degradation is a key event in receptor recycling (11,15) and signaling (27). Recently it was reported that anti-p185/HER2 antibody-mediated targeting of a cysteine proteinase inhibitor to a cathepsin Bcontaining intracellular compartment resulted in growth inhibition of two breast carcinoma cell lines including MCF-7 (32). Our model offers mechanistic insight into these observations, suggesting that growth impairment in these cells was related to defective ligand processing in the endosomes. Collectively, the results identify the endocytic apparatus as a critical component of growth factor receptor signaling that can be accessible and sensitive to specific proteinase inhibitors.