Insulin-like Growth Factor (IGF)-binding Protein-3 Mutants That Do Not Bind IGF-I or IGF-II Stimulate Apoptosis in Human Prostate Cancer Cells*

Insulin-like growth factor (IGF)-binding protein-3 (IGFBP-3) can stimulate apoptosis and inhibit cell proliferation directly and independently of binding IGFs or indirectly by forming complexes with IGF-I and IGF-II that prevent them from activating the IGF-I receptor to stimulate cell survival and proliferation. To date, IGF-independent actions only have been demonstrated in a limited number of cells that do not synthesize or respond to IGFs. To assess the general importance of IGF-independent mechanisms, we have generated human IGFBP-3 mutants that cannot bind IGF-I or IGF-II by substituting alanine for six residues in the proposed IGF binding site, Ile56/Tyr57/Arg75/Leu77/Leu80/Leu81, and expressing the 6m-hIGFBP-3 mutant construct in Chinese hamster ovary cells. Binding of both IGF-I and IGF-II to 6m-hIGFBP-3 was reduced >80-fold. The nonbinding 6m-hIGFBP-3 mutant still was able to inhibit DNA synthesis in a mink lung epithelial cell line in which inhibition by wild-type hIGFBP-3 previously had been shown to be exclusively IGF-independent. 6m-hIGFBP-3 only can act by IGF-independent mechanisms since it is unable to form complexes with the IGFs that inhibit their action. We next compared the ability of wild-type and 6m-hIGFBP-3 to stimulate apoptosis in serum-deprived PC-3 human prostate cancer cells. PC-3 cells are known to synthesize and respond to IGF-II, so that IGFBP-3 could potentially act by either IGF-dependent or IGF-independent mechanisms. In fact, 6m-hIGFBP-3 stimulated PC-3 cell death and stimulated apoptosis-induced DNA fragmentation to the same extent and with the same concentration dependence as wild-type hIGFBP-3. These results indicate that IGF-independent mechanisms are major contributors to IGFBP-3-induced apoptosis in PC-3 cells and may play a wider role in the antiproliferative and antitumorigenic actions of IGFBP-3.

For many years, the antiproliferative effects of IGFBP-3 have been attributed to inhibition of IGF-I and IGF-II stimulation of cell proliferation and survival (reviewed in Ref. 28). IGFBP-3 binds IGF-I and IGF-II with high affinity, forming complexes that prevent the IGFs from binding to and activating IGF-I receptors. Recent studies suggest that IGFBP-3 also can act directly by a mechanism that is independent of binding IGFs (4, 8, 9, 12, 29 -36). The evidence supporting this hypothesis, however, only has been obtained under special circumstances: in cells that do not synthesize IGF-I or IGF-II, in serum-free medium so as not to introduce IGFs, or in cells that lack IGF-I receptors or do not respond to added IGFs. Since these conditions do not apply to the vast majority of normal and tumor cells that synthesize and respond to IGFs, the extent to which IGF-independent mechanisms contribute to the antiproliferative actions of IGFBP-3 is unknown.
In the present study, we have developed a novel and more generally applicable strategy to evaluate the contributions of IGF-independent mechanisms to growth inhibition by IGFBP-3. It is based on the recent identification of an IGF binding site in hIGFBP-5 (37) that is highly conserved in hIGFBP-3. We introduced mutations at six positions in the N-terminal region of hIGFBP-3 that abolished the binding of both IGF-I and IGF-II. The nonbinding hIGFBP-3 mutant inhibited DNA synthesis in a cell line in which inhibition by wild-type hIGFBP-3 previously had been shown to be exclusively IGF-independent (34). The hIGFBP-3 mutant then was used to determine the extent to which IGF-independent mechanisms contributed to the induction of apoptosis by IGFBP-3 in PC-3 human prostate carcinoma cells (38). Previous reports had suggested that IGFBP-3 might stimulate apoptosis or inhibit DNA synthesis in PC-3 cells by either IGF-independent (4) or IGF-dependent (26, 39 -41) mechanisms. The nonbinding hIGFBP-3 mutant was as effective as wild-type hIGFBP-3 in stimulating apoptosis in PC-3 cells, suggesting that direct IGFindependent mechanisms are important contributors to the antiproliferative actions of IGFBP-3.

EXPERIMENTAL PROCEDURES
Materials-Plasmids pRc/RSV and pcDNA3.1/His A, anti-Xpress antibody, and ProBond Resin were purchased from Invitrogen. A human IGFBP-3 cDNA clone (GenBank TM accession number M31159) was provided by William Wood (Genentech, South San Francisco, CA) (42). The QuikChange site-directed mutagenesis kit was obtained from Stratagene (La Jolla, CA). Recombinant hIGFBP-3 synthesized in NSO mouse myeloma cells was obtained from R & D Systems (Minneapolis, MN) and used as a reference standard. [Leu 60 ]IGF-I expressed in Escherichia coli was kindly provided by Celtrix Pharmaceuticals, Inc. (San Jose, CA) (34). Monoclonal antibodies to the N terminus (antibody 3; amino acids 1-97) or C terminus (antibody 1; residues 98 -264) of hIGFBP-3 (43) were purchased from Diagnostic Systems Laboratories (Webster, TX). 125 I-IGF-I and 125 I-IGF-II (2000 Ci/mmol), and the enhanced chemiluminescence (ECL) Western blotting detection reagent were purchased from Amersham Biosciences, Inc.; fetal calf serum was from Hyclone Laboratories, Inc. (Logan, UT); F12K nutrient mixture medium, Dulbecco's modified Eagle's medium (DMEM) (containing 4.5 g/liter D-glucose, pyridoxine hydrochloride, and sodium pyruvate), Li-pofectAMINE PLUS, and G418 (734 g/mg) were from Invitrogen. The BrdU Cell Proliferation ELISA, Apoptotic DNA Ladder, In Situ Cell Death Detection (TUNEL), and Cell Death Detection ELISA Plus assay kits and the fluorescent DNA-binding dye DAPI were purchased from Roche Molecular Biochemicals. Recombinant human EGF was obtained from Sigma.
Cell Cultivation-Chinese hamster ovary (CHO)-K1 cells (44) were obtained from David Clemmons (University of North Carolina School of Medicine, Chapel Hill), mink lung epithelial cells (CCL64) were from Anita Roberts (NCI, National Institutes of Health) or the American Type Culture Collection (ATCC, Manassas, VA), and PC-3 human prostate adenocarcinoma cells (38) were from the ATCC. CHO-K1 and PC-3 cells were grown in F12K medium containing 10% fetal calf serum; CCL64 cells were grown in DMEM plus 10% fetal calf serum. All media contained penicillin (100 units/ml), streptomycin (100 g/ml), and fungizone (2.5 g/ml). Cells were grown at 37°C in a humidified environment with 5% CO 2 . Fresh cells were thawed at least every 2 months.
Construction of the Expression Plasmid Encoding Wild-type Human IGFBP-3 (pRSV-Sec-BP3)-Plasmid pRSV-Sec-BP-3 expresses a fusion gene encoding the signal peptide of the immunoglobulin chain and a peptide containing a His 6 /Xpress antibody recognition site/enterokinase C cleavage site upstream from the 795-nucleotide coding region of hIGFBP-3 cDNA. First, a double-stranded oligonucleotide (5Ј-AGCT ATG GAG ACA GAC ACA CTC CTG CTA TGG GTA CTG CTG CTC TGG GTT CCA GGT TCC ACT GGT GAC A-3Ј) encoding the IgG chain signal peptide (pSecTag2; Invitrogen) with HindIII sticky ends was introduced into the HindIII site (A/AGCTT) of pRc/RSV (Invitrogen) to form pRSV-Sec. The upstream HindIII site was destroyed by the single base change from AGCTT to AGCTA. Next, a double-stranded DNA fragment containing the His 6 /Xpress antibody/enterokinase C sequence fused to the coding region of mature hIGFBP-3 was prepared by overlapping PCR. The 5Ј-fragment containing the His 6 /Xpress antibody/enterokinase sequence (CAT CAT CAT CAT CAT CAT GGT ATG GCT AGC ATG ACT GGT GGA CAG CAA ATG GGT CGG GAT CTG TAC GAC GAT GAC GAT AAG) was amplified from pcDNA3.1/His A (Invitrogen); the 5Ј-end of the sense primer was extended by a HindIII sequence, and the 5Ј-end of the antisense primer was extended by the N-terminal 18 bp of hIGFBP-3 (GCC CCC CGA GCT CGC GCC). The 3Ј fragment contained the complete coding region of hIGFBP-3 (795 bp); the 5Ј-end of the sense primer was extended by the Xpress antibody/ enterokinase tag, and the 5Ј-end of the antisense primer was extended by an XbaI sequence. Following overlapping PCR of the two fragments using the HindIII and XbaI primers, the complete HindIII-His 6 -Xpress-EK-hIGFBP-3-XbaI fragment was ligated into the HindIII-XbaI gap of linearized pRSV-Sec to produce pRSV-Sec-BP3 (Fig. 1).
Construction of Plasmids Expressing hIGFBP-3 Mutants-Alanine substitution mutations were introduced into pRSV-Sec-BP-3 using the QuikChange site-directed mutagenesis kit (Stratagene) as described by the manufacturer. Complementary oligonucleotide primers to the same sequence containing the desired mutations were annealed to both strands of the double-stranded DNA vector (pRSV-Sec-BP3) and extended using PfuTurbo DNA polymerase to generate a mutated plasmid with staggered nicks. Following amplification, the parental DNA template was digested with DpnI endonuclease, and the DpnI-treated DNA was used to transform Epicurian Coli XL1-Blue supercompetent cells.
Transfection and Selection of Stable Cell Lines-CHO-K1 cells in 10-cm culture dishes were transfected with 4 g of plasmid DNA (wildtype or mutant pRSV-Sec-BP3 or empty vector pRSV-Sec), Lipo-fectAMINE (10 l), and PLUS reagent (15 l) in serum-free F12K medium according to the manufacturer's instructions. After 3 h, fetal calf serum was added to a final concentration of 10%, and the incubation continued for 24 h, following which G418 (1000 g/ml) was added to select the neomycin-resistant transfected cells. After 48 h, conditioned medium was examined for gene expression by immunoblotting with anti-Xpress antibody, and cells from positive transfections were replated at 1:1000 dilution in the same selection medium. After 7 days, ϳ85% of the cells had been killed. Single colonies were picked into 24-well dishes and grown to confluence in selection medium, and the medium was changed to serum-free medium containing G418. After 48 h, the conditioned media were examined by immunoblotting with monoclonal antibody to the N-terminal region of hIGFBP-3. Clones with the highest expression of transfected IGFBP-3 were selected and expanded.
Collection and Purification of Expressed hIGFBP-3-Stably transfected CHO-K1 cells expressing wild-type or mutant hIGFBP-3 were grown to confluence in 175-cm 2 flasks in 20 ml of F12K medium supplemented with 10% fetal calf serum and G418. The monolayer was washed with phosphate-buffered saline, the medium was changed to serum-free F12K medium containing G418, and the cells were cultured for another 2 days. The medium was harvested, serine protease inhibitors phenylmethylsulfonyl fluoride and 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride were added immediately (final concentration 0.1 mg/ml), and the medium was centrifuged to remove cell debris. The clarified medium was immediately concentrated ϳ10ϫ using Centriprep YM-10 filters (Millipore Corp.) and stored at Ϫ70°C. The cells were trypsinized and replated, and the process was repeated up to seven or eight times.
Wild-type and mutant His 6 -hIGFBP-3 were purified by affinity chromatography using ProBond resin, which contains immobilized nickel divalent cations. First, individual samples were immunoblotted with monoclonal antibody to the N terminus of hIGFBP-3 to exclude samples containing 30-kDa hIGFBP-3 fragments. Then the column (5 ml) was loaded with an equal volume of concentrated conditioned media at 0°C. It was washed with 20 mM sodium phosphate, 0.5 M NaCl buffer (pH 7.8) and then with sodium phosphate-NaCl, pH 6.0, followed by the same buffer containing 50 mM imidazole. The column was eluted successively with sodium phosphate-NaCl, pH 6.0, buffer containing 200 mM imidazole and 350 mM imidazole. The combined eluates were concentrated 10ϫ using Centriprep YM-10 filters and desalted using a PD-10 column (Sephadex G25M; Amersham Biosciences, Inc.) equilibrated with phosphate-buffered saline. Serine protease inhibitors were added again, and the desalted purified samples were stored at Ϫ70°C.
Quantification of Affinity-purified hIGFBP-3 Samples-The concentration of hIGFBP-3 present in the affinity-purified preparations was determined by quantitative immunoblotting using N-terminal and Cterminal monoclonal antibodies to hIGFBP-3. Samples were tested at three or four concentrations and compared with a standard curve generated using recombinant glycosylated hIGFBP-3 (R & D Systems). The resulting autoradiographs were scanned, and the signal was quantified using the NIH Image program as described below. The concentration of hIGFBP-3 in the samples was determined from the linear portion of the standard curve in four assays. Results using N-terminal and C-terminal antibodies were not significantly different and were combined.
Control medium was collected in parallel from CHO-K1 cells stably transfected with pRSV-Sec empty vector and subjected to the same concentration and affinity purification. The amount of empty vector control is given as equivalents of wild-type CHO-hIGFBP-3 obtained from the same volume of conditioned medium in a parallel purification.
Immunoblotting-IGFBP samples were mixed with 2ϫ Laemmli loading buffer without dithiothreitol (Bio-Rad) and were heated at 95°C for 5 min. The samples were separated on a 10 -20% gradient SDS-PAGE, and proteins were transferred onto a nitrocellulose mem-brane. After blocking with 1ϫ phosphate-buffered saline, 10% nonfat dry milk (4°C, overnight), the membrane was incubated with a 1:10,000 dilution of monoclonal antibody to the N terminus or C terminus of hIGFBP-3 for 2 h. The membrane was washed three times with phosphate-buffered saline plus 0.1% Tween 20, incubated with antimouse IgG-horseradish peroxidase (1:5000 dilution; Santa Cruz Biotechnology, Inc., Santa Cruz, CA), and processed for detection using an enhanced chemiluminescence detection system. The membrane was sealed in a plastic bag and exposed to high sensitivity x-ray film. The resulting autoradiograph was scanned using an ArcusII scanner and Foto Look 2.07.02 software. Signal intensities were analyzed using the NIH Image program.
Ligand Blotting-Following immunoblotting, the membranes were washed with 10 mM Tris-Cl (pH 7.4), 0.15 M NaCl, 3% Nonidet P-40, 0.5 mg/ml sodium azide (22°C, 1 h) and then incubated in 5 ml of the same buffer containing 400,000 cpm 125 I-IGF-I or 125 I-IGF-II (3 h, room temperature) (45,46). After washing three times (15 min each) with the same buffer without radioligand, the membrane was exposed to high sensitivity film at Ϫ70°C. Ligand blotting confirmed that the 24-kDa IGFBP-4 that was present in medium from nontransfected CHO-K1 cells had been removed by affinity chromatography.
Binding of 125 I-IGF-I and 125 I-IGF-II in Solution-125 I-IGF-I or 125 I-IGF-II (ϳ25,000 cpm) was incubated overnight at 4°C with different concentrations of recombinant hIGFBP-3 standard (R & D Systems) or purified wild-type or mutant CHO-hIGFBP-3 in 0.4 ml of phosphatebuffered saline supplemented with 0.2% fatty acid-free bovine serum albumin. Following the addition of 0.5 ml of a 5% suspension of activated charcoal (Sigma) to adsorb unbound IGF tracer, the samples were centrifuged. Radioactivity bound to hIGFBP-3 remained in the charcoal supernatant and was quantified in a ␥ counter (46).
DNA Synthesis-DNA synthesis was measured in CCL64 cells by the incorporation of the thymidine analog BrdUrd into newly synthesized as previously described (34). Quiescent cells in serum-free medium were stimulated to synthesize DNA by adding EGF. EGF was used to stimulate proliferation instead of serum to avoid introducing IGFs. In brief, the cells were plated in 96-well microtiter plates (30,000 cells/ well) in 0.2 ml of DMEM containing 10% fetal calf serum and incubated for 3 h at 37°C. The medium was replaced with serum-free DMEM supplemented with 0.5% bovine serum albumin (radioimmunoassay grade; Sigma), and the incubation continued for another 3 h. EGF (20 ng/ml) and the indicated concentrations of hIGFBP-3 were added, and the incubation continued overnight. BrdUrd (10 M) was added for 2-3 h, the cells were fixed, and BrdUrd incorporation was quantified by an immunocolorimetric assay using monoclonal antibody to BrdUrd conjugated to peroxidase. Triplicate points were examined. The absorbance at 450 nm was measured in a scanning multiwell spectrophotometer.
Trypan Blue Staining of Nonviable Cells-PC-3 cells were plated in 12-well culture dishes (50 -130,000 cells/well) and grown to confluence (24 h) in F12K medium supplemented with 10% fetal calf serum. The medium was changed to serum-free medium for 24 h and then replaced with fresh serum-free medium containing wild-type CHO-hIGFBP-3 (1 g/ml), 6m-hIGFBP-3 (1 g/ml), or protein purified from pRSV-Sec empty vector transfectants (equivalent to 2 g/ml wild-type CHO-hIGFBP-3); [Leu 60 ]IGF-I was added where indicated. After 24-, 48-, or 72-h incubation, floating cells in the medium were sedimented and resuspended; adherent cells were dissociated with trypsin and resuspended. Trypan blue (0.4%) was added to the suspensions of floating and attached cells and incubated for 10 min. The total number of cells and the number of nonviable cells stained with trypan blue were counted in a hemocytometer.
DNA Ladder-PC-3 cells were plated in a 10-cm culture dish (600,000 cells/well) in serum-supplemented F12K medium and grown to confluence (24 h). The medium was changed to serum-free medium for 24 h and was replaced with fresh serum-free medium containing wildtype CHO-hIGFBP-3 (1 g/ml), 6m-hIGFBP-3 (1 g/ml), or protein from pRSV-Sec empty vector transfectants (equivalent to 2 g/ml wild-type CHO-hIGFBP-3). After 72-h incubation, the cells were lysed, and the DNA was purified and analyzed on 1% agarose gels containing ethidium bromide using the Apoptotic DNA Ladder Kit according to the manufacturer's instructions. In brief, the cells were lysed with 3 M guanidine hydrochloride, 5 mM urea, 10% Triton X-100 and extracted with isopropyl alcohol. The extract was applied to filter tubes containing a glass fiber fleece and centrifuged. After washing, the nucleic acids bound to the glass fibers were eluted with 10 mM Tris-HCl, pH 8.5, prewarmed to 70°C, and analyzed by agarose gel electrophoresis and UV photography. A ladder pattern of multiples of 180-bp nucleosomal subunits is generated in apoptotic cells.
DAPI Staining of Nuclear DNA-PC-3 cells were plated in a six-well culture dish (200,000 cells/dish) and grown to 60% confluence in F12K medium containing 10% fetal calf serum. The medium was changed to serum-free medium for 24 h and was replaced with fresh serum-free medium containing wild-type CHO-hIGFBP-3 (1 g/ml), 6m-hIGFBP-3 (1 g/ml), or protein from pRSV-Sec empty vector transfectants (equivalent to 2 g/ml wild-type CHO-hIGFBP-3) for 72 h. The cells were washed once with 1 g/ml DAPI-methanol, incubated with DAPI-methanol (15 min, 37°C), washed with methanol, and examined by fluorescence microscopy. DAPI is a fluorescent dye that binds selectively to DNA. Nuclear condensation and fragmentation is characteristic of apoptotic cells. TUNEL Assay-Cleavage of genomic DNA into oligonucleosomes during apoptosis was identified in individual cells by labeling 3Ј-OH termini using terminal deoxynucleotidyl transferase and fluoresceinlabeled dUTP substrate (TUNEL assay). PC-3 cells (30,000 cells) were plated on eight-well chamber slides in serum-supplemented F12K medium and grown to 80% confluence (24 h). The medium was changed to serum-free medium, and after 24 h it was replaced with fresh serumfree medium containing wild-type CHO-hIGFBP-3 (1 g/ml), 6m-hIGFBP-3 (1 g/ml), or purified medium from cells transfected with pRSV-Sec empty vector (equivalent to 2 g/ml wild-type CHO-hIGFBP-3). After a 72-h incubation, the cells were fixed with 2% paraformaldehyde, washed three times with phosphate-buffered saline, permeabilized with 0.1% Triton X-100 on ice, and incubated with the TUNEL reaction mixture in a humidified chamber (1 h, 37°C). Cells incorporating labeled dUTP were identified by fluorescence microscopy and photographed.
ELISA Assay of Histone-associated DNA Fragments-This assay measures histone-bound DNA fragments generated by internucleosomal cleavage in the cytosol of apoptotic cells. PC-3 cells (10,000 cells/ well) were grown to 80% confluence in 96-well culture plates in serumsupplemented F12K medium. After a 24-h incubation in serum-free medium, the indicated hIGFBP-3 preparations were added at different concentrations for 72 h. The cell membranes were lysed according to the manufacturer's instructions, and the supernatants were added to streptavidin-coated microplates. Biotin-labeled anti-histone (to bind the histone component of the nucleosomes and fix the complex to the plate) and anti-DNA peroxidase (to bind to nucleosomal DNA) were added, and the incubation continued for 2 h. After washing, the amount of nucleosome DNA was determined photometrically after the addition of 2,2Ј-azinodi(3-ethyl-benzthiazoline-sulfonate) peroxidase substrate for 30 min. Absorbance was determined at 405 and 490 nm (substrate blank).

Construction and Characterization of hIGFBP-3 Mutants
That Do Not Bind IGF-I or IGF-II-Candidate mutations that might decrease the binding of IGF-I and IGF-II to hIGFBP-3 were designed based on the studies of Kalus et al. (37) with hIGFBP-5. NMR spectroscopy showed that an N-terminal peptide fragment of hIGFBP-5 (residues 40 -92) that binds IGF-II has a compact globular structure and contains a surface hydrophobic patch that includes residues Val 49 , Leu 70 , and Leu 74 . Val 49 -Tyr 50 -Pro 62 and residues 68 -74 (Lys-Pro-Leu-His-Ala-Leu-Leu) showed the greatest shift in position after binding IGF-II. Residues 50 -83 of hIGFBP-3 correspond to residues 43-76 of hIGFBP-5 ( Fig. 1) and are highly conserved. A synthetic rat IGFBP-3 peptide that corresponds to residues 51-91 of hIGFBP-3 also binds IGF-II (47).
Based on this structural information, we substituted alanine for the native amino acids at six positions in hIGFBP-3 (Ile 56 / Tyr 57 /Arg 75 /Leu 77 /Leu 80 /Leu 81 ). These residues correspond to Val 49 /Tyr 50 /Lys 68 /Leu 70 /Leu 73 /Leu 74 of hIGFBP-5 and constitute six of the 10 amino acids that change position after binding IGF-II. Ile 56 , Leu 77 , and Leu 81 correspond to residues in the hydrophobic patch. Leu 77 and Leu 80 are conserved in all six IGFBPs, and basic amino acids corresponding to Arg 75 are present in IGFBP-4 and IGFBP-5.
Wild-type hIGFBP-3 and mutant hIGFBP-3 containing the six alanine substitutions (I56A/Y57A/R75A/L77A/L80A/L81A; 6m-hIGFBP-3) were expressed in CHO-K1 cells as secreted proteins containing an N-terminal polyhistidine tag to allow purification by nickel cation affinity chromatography. Different amounts of the purified proteins and recombinant hIGFBP-3 standard were fractionated using SDS-PAGE and examined by immunoblotting with monoclonal antibodies to the N-and Cterminal domains of hIGFBP-3 and by ligand blotting with 125 I-IGF-I or 125 I-IGF-II (Fig. 2). The three proteins were recognized by monoclonal antibodies to the N-terminal (upper panel) and C-terminal (results not shown) epitopes.
The 6m-, 4m-, and 2m-hIGFBP-3 mutant proteins also were unable to bind 125 I-IGF-I or 125 I-IGF-II in a solution binding assay that did not expose them to denaturing conditions (Fig.  4). Dose-dependent binding of 125 I-IGF-I or 125 I-IGF-II was observed with recombinant hIGFBP-3 standard or wild-type CHO-hIGFBP-3, reaching a maximum of 70 -80% of input radioactivity bound. By contrast, only negligible binding was observed with any of the three mutants at concentrations as high as 200 ng/ml, less than the binding observed to 80-fold lower concentrations of wild-type CHO-hIGFBP-3. Thus, the mutant hIGFBP-3 molecules have profoundly decreased ability to bind IGF-I and IGF-II.
hIGFBP-3 Mutants That do Not Bind IGFs Still Inhibit DNA Synthesis in Mink Lung Epithelial Cells-Nonglycosylated recombinant hIGFBP-3 expressed in E. coli inhibited DNA synthesis in CCL64 mink lung epithelial cells in serum-free medium (34). The inhibition was considered to be IGFindependent, since CCL64 cells do not synthesize functionally significant levels of IGF-I or IGF-II, and IGF-I does not stimulate CCL64 DNA synthesis. Dose-dependent inhibition of DNA synthesis was observed not only with glycosylated recombinant hIGFBP-3 reference standard and wild-type CHO-hIGFBP-3, but also with the nonbinding hIGFBP-3 mutant proteins containing two, four, or six mutations (Fig. 5). No inhibition was observed with equivalent amounts of conditioned medium purified from nontransfected CHO-K1 cells. Thus, the hIGFBP-3 mutants retain the ability to inhibit DNA synthesis in mink lung epithelial cells although they do not bind IGFs. These results provide strong independent confirmation of our previous conclusion that inhibition of CCL64 DNA synthesis by wild-type hIGFBP-3 is IGF-independent (34).
Free hIGFBP-3 inhibits CCL64 DNA synthesis, but hIGFBP-3 complexed to IGF-I does not (34), presumably because IGF-I induces a conformational change in IGFBP-3 when it binds to it. Since 6m-hIGFBP-3 cannot bind IGF-I, coincubation with IGF-I should not affect its ability to inhibit CCL64 cell DNA synthesis. As in the previous study, we used [Leu 60 ]IGF-I, an IGF-I analogue in which leucine is substituted for tyrosine at position 60 (48), instead of native IGF-I since the analogue binds to hIGFBP-3 but has low affinity for and does not activate the IGF-I receptor. As expected, coincubation with [Leu 60 ]IGF-I (at 0.5 or 2 g/ml) abolished the inhibition of DNA synthesis caused by 2 g/ml wild-type CHO-hIGFBP-3 but did not decrease the inhibition induced by 6m-hIGFBP-3 (Fig. 6). These results demonstrate directly that [Leu 60 ]IGF-I must bind to hIGFBP-3 to decrease its ability to inhibit CCL64 DNA synthesis.
6m-hIGFBP-3 Induces Apoptosis in PC-3 Human Prostate Cancer Cells-Unlike mink lung epithelial cells, most cells synthesize IGF-I, IGF-II, or both proteins and respond to the IGFs, making it difficult to determine the relative contributions of IGF-dependent and IGF-independent mechanisms to the antiproliferative actions of IGFBP-3. The hIGFBP-3 mutants that do not bind IGF-I or IGF-II provide an opportunity to examine the relative importance of IGF-independent mechanisms in the more typical situation in which IGF-dependent mechanisms also are possible. We chose to study the induction of apoptosis by IGFBP-3 in the PC-3 human prostate cancer cell line. Rajah et al. (4) proposed that the stimulation of PC-3 cell apoptosis by IGFBP-3 was IGF-independent. PC-3 cells, however, synthesize IGF-II, which can act as an autocrine growth factor and stimulate their proliferation in serum-free medium (26, 39 -41). Exogenous IGFBP-3 inhibited PC-3 cell proliferation, which was attributed to preventing IGF-II from activating the IGF-I receptor (26).
The ability of wild-type CHO-hIGFBP-3, 6m-hIGFBP-3, or media from CHO-K1 cells transfected with empty vector to kill serum-deprived PC-3 cells was examined (Fig. 7, upper panel). After 72 h, ϳ50% of the cells recovered after incubation with wild-type or 6m-CHO-hIGFBP-3 had detached from the monolayer, whereas Ͻ0.1% of the cells recovered after incubation with media from empty vector transfectants were floating. Over 86% of the floating cells from the wild-type or 6m-CHO-hIGFBP-3 incubations were nonviable (i.e. stained with trypan blue), whereas Ͻ20% of cells that remained attached to the culture dish were dead whether or not they had been incubated with hIGFBP-3. Thus, incubating serum-deprived PC-3 cells with either wild-type CHO-hIGFBP-3 or 6m-hIGFBP-3 promoted the detachment of cells from the monolayer and greatly increased the percentage of nonviable cells.
The increased death of PC-3 cells incubated with wild-type or 6m-CHO-hIGFBP-3 reflects increased apoptosis (Fig. 8). This was demonstrated using several indices of apoptosis-induced DNA fragmentation. Agarose gel electrophoresis of DNA preparations from cells incubated with wild-type or 6m-CHO-hIGFBP-3, but not from control cells, revealed a ladder of DNA fragments of different sizes that represent oligonucleosomes containing different numbers of nucleosomes (Fig. 8A). Nuclear staining of individual cells with the fluorescent dye DAPI revealed DNA fragmentation and condensation characteristic of apoptosis in cells treated with wild-type CHO-hIGFBP-3 or 6m-hIGFBP-3 (Fig. 8B). Apoptosis also was seen in individual cells using the TUNEL assay, in which terminal deoxynucleotidyl transferase catalyzes the addition of fluorescein-dUTP to the free 3Ј-OH ends of DNA fragments generated by apoptosis (Fig. 8C). Numerous cells incorporating the fluorescent nucleotide were evident by fluorescent microscopy of cells treated with wild-type CHO-hIGFBP-3 or 6m-hIGFBP-3 but not with media from empty vector transfectants. Finally, using a quantitative ELISA assay, the abundance of cytosolic histone-bound DNA fragments was increased ϳ10-fold in cells incubated with 1 g/ml wild-type CHO-hIGFBP-3 or 6m-hIGFBP-3 compared with media from empty vector transfectants (Fig. 8D). Stimulation was observed at 30 ng/ml, and the dose-response curves with the native and mutant proteins were superimposable. Thus, the stimulation of PC-3 cell apoptosis by 6m-hIGFBP-3 and wild-type CHO-hIGFBP-3 is similar in magnitude and concentration dependence, suggesting that IGF-independent mechanisms are major contributors to the induction of apoptosis in PC-3 cells by IGFBP-3. DISCUSSION IGFBP-3 can stimulate apoptosis and inhibit cell proliferation directly in an IGF-independent manner or indirectly by 1-4; 5, 10, 15, and 20 ng), affinity-purified wild-type CHO-hIGFBP-3 (lanes 5-8; WT; 13.4 ng/l; 1, 2, 3, and 4 l), or 6m-hIGFBP-3 (lanes 9 -12; 6m; 13.2 ng/l; 1, 2, 3, and 4 l). Proteins were blotted to nitrocellulose and examined by immunoblotting with monoclonal antibody specific for the N terminus (upper panel) or C terminus (not shown) of hIGFBP-3. The same blots were incubated with 125 I-IGF-I (middle panel) or 125 I-IGF-II (lower panel) and autoradiographed. Wild-type and 6m-CHO-hIGFBP-3 migrate more slowly than the hIGFBP-3 reference standard, in part due to the His 6 -Xpress-EK tag. They also appear as doublets due to differences in N-glycosylation. preventing the stimulation of cell survival and proliferation by IGF-I and IGF-II (33). IGF-independent actions only have been demonstrated under special circumstances: in cells that do not synthesize IGF-I and IGF-II, lack signaling IGF-I receptors, or do not respond to IGFs (4, 30, 32, 34 -36). To assess the general importance of IGF-independent mechanisms to the antiproliferative actions of IGFBP-3, we have generated hIGFBP-3 mutants that do not bind either IGF-I or IGF-II. The mutants retain their ability to inhibit DNA synthesis in an IGF-independent manner. We examined the ability of one of the nonbinding hIGFBP-3 mutants to stimulate apoptosis in the PC-3 human prostate cancer cell line, in which both IGF-independent (4) and IGF-dependent (26) actions of hIGFBP-3 have been reported. The nonbinding hIGFBP-3 mutant stimulated apoptosis in PC-3 cells as effectively as wild-type hIGFBP-3, establishing that hIGFBP-3 can induce apoptosis without binding IGFs and suggesting that direct IGF-independent mechanisms could be major contributors to the induction of apoptosis by hIGFBP-3.

FIG. 2. Characterization of wild-type CHO-hIGFBP-3 and 6m-hIGFBP-3 by immunoblotting and ligand blotting. SDS-PAGE of recombinant hIGFBP-3 standard (lanes
The nonbinding hIGFBP-3 mutant, 6m-hIGFBP-3, in which alanine was substituted for the natural amino acid at six positions (Ile 56 /Tyr 57 /Arg 75 /Leu 77 /Leu 80 /Leu 81 ), was designed based on the IGF binding site recently identified in a peptide from the N-terminal region of hIGFBP-5 (37) that is strongly conserved in hIGFBP-3. The mutant protein did not bind either IGF-I or IGF-II under denaturing or nondenaturing conditions, satisfying the first requirement for use in our proposed studies. While our experiments were in progress, Imai et al. (49) re-ported that a hIGFBP-3 mutant with five substitutions in the same region (R75S/P76A/L77S/L80Q/L81G) also exhibited greatly reduced binding of IGF-I; IGF-II was not examined in their study. Their mutant failed to inhibit two IGF-dependent actions, IGF-I-stimulated DNA synthesis in and migration of smooth muscle cells. IGF-I and IGF-II also bind to C-terminal fragments of hIGFBP-3 (50,51), and mutations in the C-terminal regions of IGFBP-3 (52), IGFBP-4 (53), and IGFBP-5 (54) decrease IGF binding. Although IGFs can bind to the C-terminal region of hIGFBP-3, our results together with those of Imai et al. (49) demonstrate that the N-terminal binding site is essential for high affinity binding of IGF-I and IGF-II to intact hIGFBP-3.
A hIGFBP-3 mutant with alanine substitutions at only two of the six positions, Arg 75 and Leu 77 , also exhibited markedly decreased binding of both IGF-I and IGF-II. Leu 77 is conserved in all six IGFBPs and is part of a hydrophobic patch on the surface of the IGF binding region of hIGFBP-5 (37). Basic residues corresponding to Arg 75 are present in IGFBP-4 and IGFBP-5 but not in the other hIGFBPs (1) or rat IGFBP-3 (47). The conservation of Leu 77 suggests that it may be the more critical of the two residues for high affinity binding of IGF-I and IGF-II. This inference is supported by the recent report of the 2.1-Å crystal structure of the complex of IGF-I and the Nterminal IGF-binding domain of hIGFBP-5 (55). The principal and 4m-hIGFBP-3 (4m, solid squares). In the absence of added hIGFBP-3, EGF stimulated BrdUrd incorporation into newly synthesized DNA by 6 -7-fold. BrdUrd incorporation in the presence of EGF alone was taken as 100% after subtracting the absorbance in the absence of EGF. Mean Ϯ S.D. of triplicate determinations in a representative experiment is plotted. The slightly less complete inhibition by 6m-hIGFBP-3 compared with wild type CHO-hIGFBP-3 was not observed in other experiments.
interaction of one of the protruding IGF-I side chains is with a cluster of hydrophobic residues (Val 49 , Leu 70 , and Leu 74 ) in a solvent-exposed cleft of mini-IGFBP-5. These residues are analogous to Ile 56 , Leu 77 , and Leu 81 of hIGFBP-3 (Fig. 1).
We previously reported that nonglycosylated recombinant hIGFBP-3 inhibited DNA synthesis in a mink lung epithelial cell line exclusively by IGF-independent mechanisms (34). This conclusion is strongly supported by the present findings that the three hIGFBP-3 mutants with markedly reduced binding of IGF-I and IGF-II still inhibited DNA synthesis in CCL64 cells as effectively as glycosylated recombinant hIGFBP-3 standard or wild-type CHO-hIGFBP-3. The nonbinding hIGFBP-3 mutants only can act by IGF-independent mechanisms, since they are unable to form complexes with the IGFs that would inhibit their actions. Inhibition of CCL64 DNA synthesis by wild-type hIGFBP-3 was abolished when it bound [Leu 60 ]IGF-I, an IGF-I analogue that binds poorly to IGF-I receptors (34), whereas coincubation with [Leu 60 ]IGF-I did not affect inhibition by 6m-hIGFBP-3. These results support the concept that [Leu 60 ]IGF-I must bind to hIGFBP-3 to induce a conformational change that prevents it from directly inhibiting DNA synthesis.
The critical test remained, to use the nonbinding 6m-hIGFBP-3 mutant to examine the relative contributions of IGF-independent and IGF-dependent mechanisms to the antiproliferative actions of IGFBP-3 in a cell line that expressed and responded to IGFs so that both IGF-dependent and IGFindependent mechanisms were possible. The induction of apoptosis by hIGFBP-3 in PC-3 human prostate carcinoma cells provided such a test. Rajah et al. (4) showed that hIGFBP-3 stimulated apoptosis in PC-3 cells and suggested that this stimulation was IGF-independent because IGFBP-3-induced apoptosis was decreased about 50% by coincubation with IGF-I but not by coincubation with LongR3-IGF-I, an IGF-I analogue that binds with high affinity to IGF-I receptors but poorly to IGFBP-3. Other investigators, on the other hand, had reported that PC-3 cells synthesize IGF-II (40,41,56) and that endogenous IGF-II acts as an autocrine growth factor by stimulating the IGF-I receptor. Spontaneous growth of PC-3 cells in serumfree medium was inhibited by an antisense oligonucleotide to the IGF-I receptor (39) and by monoclonal antibodies to the IGF-I receptor (40,41) or to IGF-II (26,40). IGFBP-3 also inhibited PC-3 cell proliferation, presumably by forming complexes with IGF-II that could not activate the IGF-I receptor (26).
The nonbinding 6m-hIGFBP-3 mutant induced apoptosis in serum-deprived PC-3 cells as effectively as wild-type hIGF-BP-3. Incubation of PC-3 cells with wild-type or 6m-hIGFBP-3 for 72 h increased the detachment of cells from the monolayer, with most of the floating cells being dead. Cell death and detachment appear to be separate processes. In Fig. 7 (lower  panel), coincubation of PC-3 cells with wild-type hIGFBP-3 and [Leu 60 ]IGF-I did not reduce the percentage of recovered cells that were floating compared with cells treated with hIGFBP-3 alone, but it markedly decreased the percentage of floating cells that were nonviable. Thus, cells did not need to die to detach, and cell detachment was not sufficient to trigger cell death. The induction of cell death by 6m-hIGFBP-3 is independent of binding IGFs and is not prevented by coincubation with [Leu 60 ]IGF-I, because it cannot bind to the hIGFBP-3 mutant.
The induction of PC-3 cell death by wild-type CHO-hIGFBP-3 or 6m-hIGFBP-3 results at least in part from apoptosis. Apoptosis-induced DNA fragmentation was demonstrated by agarose gel electrophoresis of cell lysates showing a ladder of DNA fragments, by DAPI staining showing condensed and fragmented nuclei in individual cells, by TUNEL staining DNA fragments with free 3Ј-OH ends in individual cells, and by an ELISA sandwich assay, which measures cytosolic histonebound DNA fragments. In each assay, apoptosis was stimulated to the same extent by a high concentration of 6m-hIGFBP-3 and wild-type CHO-hIGFBP-3. Moreover, in the quantitative ELISA assay, wild-type CHO-hIGFBP-3 and 6m-hIGFBP-3 induced apoptosis with the same concentration dependence, strongly suggesting that IGF-independent mechanisms are major contributors to the induction of apoptosis by hIGFBP-3 in PC-3 cells. The IGF-independent anti-proliferative actions of hIGFBP-3 in PC-3 cells are not limited to apoptosis. Although endogenous IGF-II can stimulate PC-3 cell proliferation via the IGF-I receptor (26, 39 -41), 6m-hIGFBP-3 inhibited PC-3 cell DNA synthesis in serum-free medium, 2 indicating that hIGFBP-3 also can inhibit proliferation by IGFindependent mechanisms.
IGFBP-3 has received considerable recent attention as a negative risk factor for several common human cancers (57)(58)(59)(60). The combination of high serum IGF-I and low IGFBP-3 was associated with the greatest cancer risk, with the effect of low IGFBP-3 generally being attributed to increasing the bioavailability of IGF-I. In prostate cancer (57), colorectal cancer (59) and childhood leukemia (61), however, high serum IGF-BP-3 is associated with decreased cancer risk irrespective of the IGF-I level, suggesting that the anti-tumorigenic effect of IGFBP-3 may be independent of binding IGF-I. Together with our results in PC-3 cells, these data suggest that IGF-independent mechanisms may be more widespread and important contributors to the antiproliferative and anti-tumorigenic actions of IGFBP-3 than previously anticipated.
IGFBP-3 alone can induce apoptosis in serum-deprived PC-3 cells and T47D breast cancer cells (7), two cell lines which lack the wild-type p53 tumor suppressor. Even in cells in which hIGFBP-3 by itself does not decrease cell survival, hIGFBP-3 can enhance apoptosis induced by other apoptotic stimuli: for example, ceramide (32) or the mitochondrial respiratory chain inhibitor antimycin A (62) in Hs578T breast cancer cells, UV irradiation in esophageal carcinoma (35), and ionizing radiation in colon cancer (63) and T47D breast cancer (7) cells. The potentiation in colon cancer cells was dependent on p53 (63). Although most of these cells express and respond to IGFs, the nonbinding 6m-hIGFBP-3 mutant should make it possible to assess the contribution of IGF-independent mechanisms to the proapoptotic effects of IGFBP-3.
The receptor and signaling pathway that mediate the IGFindependent anti-proliferative actions of IGFBP-3 remain unclear. Candidate IGFBP-3 receptors include the type V TGF-␤ receptor or a related serine-threonine kinase receptor (34,64), and other proteins in cell lysates and membranes that bind IGFBP-3 (4,30,36,65). IGFBP-3 signaling has been proposed to involve phosphorylation of Smad2 and Smad3 (11), nuclear localization (6, 66 -70), and proteolysis (31,71,72). Smad2 and Smad3, signaling molecules downstream from the TGF-␤I receptor that function as transcription comodulators (73), were increased in anti-phosphoserine immunoprecipitates of lysates from IGFBP-3-treated T47D breast cancer cells (11). IGFBP-3 did not stimulate 32 P-incorporation into Smad2 and Smad3, however, in mink lung epithelial cells (74). Nuclear localization and nuclear transport of IGFBP-3 have been reported in kidney cells and keratinocytes, and in lung, breast and prostate cancer cells (6, 66 -70), and can occur by direct binding to importin-␤ (70). IGFBP-3 binds to the retinoid X receptor-␣ nuclear receptor, and stimulates apoptosis in F9 embryonal carcinoma cells that express retinoid X receptor-␣ but not in those that lack the nuclear receptor (6), raising the possibility that association of IGFBP-3 with retinoid X receptor-␣ in the nucleus may be involved in the stimulation of apoptosis. Availability of the nonbinding 6m-hIGFBP-3 mutant should help clarify the mechanisms by which IGFBP-3 exerts its direct antiproliferative actions since it avoids possible confusion resulting from 2 J. Hong, unpublished results. FIG. 8. 6m-hIGFBP-3 stimulates apoptosis in PC-3 cells as effectively as wild-type CHO-hIGFBP-3. PC-3 cells in serum-free medium were incubated for 72 h with wild-type CHO-hIGFBP-3 (1 g/ml), 6m-hIGFBP-3 (1 g/ml), or purified medium from empty vector transfectants (Vector; equivalent to 2 g/ml wild-type CHO-hIGFBP-3) unless otherwise specified. Results from representative experiments are presented. A, DNA ladder. Cell lysates were fractionated by electrophoresis on 1% agarose gels in the presence of ethidium bromide. The DNA was visualized with a UV light source and photographed. Lane 1, markers; lane 2, empty vector (pRSV-Sec); lane 3, positive control (DNA from U937 cells treated with camptothecin); lane 4, wildtype CHO-hIGFBP-3; lane 5, 6m-hIGFBP-3. B, DAPI staining of nuclear DNA. Cells were fixed with methanol, stained with DAPI, and examined by fluorescence microscopy. Left panel, vector; center panel, wild-type CHO-hIGFBP-3; right panel, 6m-hIGFBP-3. The characteristic fragmentation and condensation of apoptotic nuclei is evident in cells incubated with wild-type CHO-hIGFBP-3 or 6m-hIGFBP-3. C, TUNEL assay. PC-3 cells were fixed, permeabilized, and incubated with terminal deoxynucleotidyl transferase and fluorescein-labeled dUTP to label DNA fragments with 3Ј-OH termini. The same microscopic fields were examined by fluorescence microscopy (upper panel) and phase microscopy (lower panel). Left panel, vector; center panel, wild-type CHO-hIGFBP-3; right panel, 6m-hIGFBP-3. D, cell death detection ELISA. PC-3 cells in serum-free medium were incubated with the indicated concentrations of recombinant hIGFBP-3 standard (open circles), wild-type CHO-hIGFBP-3 (WT; solid triangles), 6m-hIGFBP-3 (6m; solid circles), or medium from empty vector transfectants (Vector, open squares; plotted as wild-type CHO-hIGFBP-3 equivalents) for 72 h. Histone-bound DNA in cell lysates was quantified using a sandwich ELISA in which biotin-labeled anti-histone coupled the nucleosomal histones to streptavidin-coated microplates, and peroxidase-conjugated anti-DNA was used to measure the bound nucleosomal DNA. Following the addition of peroxidase substrate, the absorbance at 405 nm was determined and plotted for each IGFBP-3 concentration after correction for the absorbance of the substrate reagent blank (A 490 ). inhibition of IGF-stimulated IGF-I receptor signaling.
In summary, the present results suggest that IGF-independent mechanisms of IGFBP-3 action may be important contributors to the antiproliferative and anti-tumorigenic effects of IGFBP-3. The nonbinding hIGFBP-3 mutants may facilitate the identification of the signal transduction pathway responsible for IGF-independent induction of apoptosis by IGFBP-3, making it possible to develop therapeutic agents that can selectively activate this antiproliferative pathway in tumor cells.