Nonsecreted Insulin-like Growth Factor Binding Protein-3 (IGFBP-3) Can Induce Apoptosis in Human Prostate Cancer Cells by IGF-independent Mechanisms without Being Concentrated in the Nucleus*

Insulin-like growth factor binding protein-3 (IGFBP-3), a secreted protein, has the intrinsic ability to induce apoptosis directly without binding insulin-like growth factors. Previous studies suggested that IGFBP-3 must be secreted to exert its biological functions. IGFBP-3 contains a nuclear localization signal (NLS), and exogenous IGFBP-3 is translocated into the nucleus, suggesting that both secretion and nuclear localization may play important roles in IGFBP-3 action. To address these questions, we fused yellow fluorescent protein (YFP) to mature IGFBP-3 lacking its signal peptide so that it would remain intracellular and mutated the C-terminal NLS of IGFBP-3, 228KGRKR232, to MDGEA. Following transfection of PC-3 human prostate cancer cells with these constructs, Western blots indicated that YFP-IGFBP-3 lacking a signal peptide was cell-associated and not present in the extracellular media. Moreover, the fusion protein was not N-glycosylated, indicating that it had not entered the secretory pathway. Confocal imaging showed that intracellular YFP-MDGEA-IGFBP-3 was predominantly cytoplasmic. Transient transfection of nonsecreted YFP-wild-type IGFBP-3 decreased cell viability, as assessed by staining with annexin V followed by flow cytometry. Induction of cell death was caspase-dependent, indicative of apoptosis. Apoptosis also was induced by the nonsecreted NLS mutant (YFP-MDGEA-IGFBP-3) alone and when the IGF-binding site also had been mutated. These results indicate that IGFBP-3 can induce apoptosis in an IGF-independent manner without being secreted or concentrated in the nucleus.

Insulin-like growth factor binding protein-3 (IGFBP-3), 4 one of six IGFBPs, is an important modulator of IGF biological activity. It binds IGF-I and IGF-II with high affinity and determines the bioavailability of the IGFs to tissues by forming a ternary complex with IGF and an acid-labile subunit that is retained in plasma to provide a reservoir of IGFs for target tissues (1,2). IGFBP-3 also forms binary complexes with the IGFs that prevent them from activating IGF-I receptors on target cells to stimulate cell proliferation and survival (1,2).
IGFBP-3 negatively regulates the growth of many cancer cells. It inhibits proliferation and stimulates apoptosis in breast and prostate cancer cells (3,4). Elevated serum IGFBP-3 is associated with a decreased risk of prostate (5,6) and colorectal (7,8) cancer. Overexpression of IGFBP-3 decreases tumor formation in xenografts of non-small cell lung cancer cells (9) and M12 human prostate cancer cells (10), and combined treatment with IGFBP-3 and a synthetic ligand for the retinoid X receptor (RXR) decreases tumor growth and increases apoptosis in xenografts of LAPC-4 prostate cancer cells (11). IGFBP-3 expression is decreased in tumor samples from patients with hepatocellular and non-small cell lung carcinomas (12,13). A polymorphism in the IGFBP-3 promoter that decreases circulating levels of IGFBP-3 increases the risk of aggressive prostate cancers and metastasis (14,15). HoxC6, a transcription factor that is commonly overexpressed in prostate cancer (16), and the EWS/FLI-1 fusion protein that is frequently overexpressed in Ewing tumors (17) down-regulate IGFBP-3 transcription. In addition, IGFBP-3 is increased in senescence (18) and decreased in cells immortalized with the human papillomavirus type 16 oncoprotein E7 due to proteasomal degradation (19).
Until recently, the antiproliferative effects of IGFBP-3 have been thought to result solely from its ability to sequester the IGFs and inhibit their growth-promoting actions through the IGF-I receptor (1,2). In the past few years, however, it has become appreciated that IGFBP-3 also has an intrinsic ability to inhibit cell growth and stimulate apoptosis by direct mechanisms that do not involve interaction with the IGFs. This was initially suggested by studies showing that IGFBP-3 inhibited growth in cells that do not synthesize IGFs (3,32), growth stimulated by other polypeptide growth factors (33,34) or by IGF-I mutants that do not bind IGFBP-3 (35), and growth in cells that do not respond to IGFs (3,32) or lack IGF-I receptors (4,36). Strong confirmation for the concept of direct, IGF-independent growth inhibition by IGFBP-3 came from the demonstration that IGFBP-3 mutants with markedly reduced affinity for both IGF-I and IGF-II induced apoptosis as efficiently as wildtype IGFBP-3 (37,38), suggesting that the direct effects of IGFBP-3 might contribute significantly to its antiproliferative activity. Recently, overexpression of an IGF-nonbinding IGFBP-3 mutant has been shown to induce apoptosis and inhibit prostate tumor progression in a transgenic mouse model of prostate cancer (39).
IGFBP-3 is synthesized as a 264-amino acid mature protein with a 27-amino acid signal peptide (40) that targets it for secretion. Just as secreted IGFBP-3 binds to IGFs to prevent them from activating IGF-I receptors on the plasma membrane, it has been widely assumed that IGFBP-3 also must be secreted to exert its IGF-independent biological effects. Several observations support this premise. Exogenous IGFBP-3 exhibits direct, IGF-independent antiproliferative activity (3,4,32,37), and immunoneutralization of IGFBP-3 in the extracellular medium prevents the induction of apoptosis in PC-3 human prostate cancer cells by TGF-␤ (4) and TNF-␣ (27), cytokines that induce IGFBP-3 expression. Candidate IGFBP-3 receptors that might mediate the direct effects of IGFBP-3 have been described, including the ϳ500-kDa type V TGF-␤ receptor/ LRP-1 (low density lipoprotein receptor-related protein-1) (41) and an ϳ25 kDa transmembrane protein designated 4-33 (42). IGFBP-3 activates several signal transduction pathways (reviewed in Ref. 43), but a functional role only has been demonstrated for Stat1 (signal transducer and activator of transcription 1), which is required for IGFBP-3-induced apoptosis in rat chondroprogenitor cells (38).
The mechanisms by which IGFBP-3 induces apoptosis directly, independent of IGFs, are not well understood. The C-terminal region of IGFBP-3 (residues 215-232) contains a possible bipartite nuclear localization signal (44), and exogenous IGFBP-3 can be rapidly transported to the nucleus (45)(46)(47)(48)(49)(50)(51). Transport of IGFBP-3 to the nucleus in permeabilized cells occurs via direct interaction with importin (karyopherin) ␤ without the participation of the adaptor protein importin ␣ (49). IGFBP-3 binds to insoluble nuclear components when both the plasma membrane and nuclear membrane are permeabilized (49). It also binds to the nuclear retinoid X receptor (RXR-␣) and regulates transcription stimulated by RXR-RXR homodimers and RXR-RAR heterodimers (50). Together, these results raised the possibility that IGFBP-3 might act in the nucleus, as has been shown for other protein ligands (52).
The C-terminal IGFBP-3 amino acid residues 228 KGRKR 232 are necessary and sufficient for nuclear localization. Fusion of residues 215-232 to GFP was sufficient to direct translocation of the chimeric protein to the nucleus, and nuclear uptake was abolished when KGRKR was mutated to MDGEA, the corresponding sequence in IGFBP-1 that does not localize to the nucleus (49). The native 228 KGRKR 232 sequence also is necessary for nuclear localization of full-length IGFBP-3. Intracellular MDGEA-IGFBP-3 was predominantly localized to the cytoplasm in digitonin-permeabilized fibroblasts (49). The MDGEA mutation also abolished nuclear uptake of exogenous, fluorescently labeled IGFBP-3 (48), and overexpressed MDGEA-IGFBP-3 (53) in T47D breast cancer cells.
Because it does not translocate to the nucleus, MDGEA-IGF-BP-3 provides an opportunity to examine whether nuclear localization is required for the antiproliferative actions of IGFBP-3. Cell proliferation was inhibited and apoptosis was induced when MDGEA-IGFBP-3 was overexpressed in breast cancer cells (53), suggesting that nuclear localization of IGFBP-3 may not be required for these effects. In this study, however, IGFBP-3 levels were quite high in both the cell and the media, so that it was not possible to distinguish whether MDGEA-IGFBP-3 acted by an extracellular or intracellular mechanism. Although extracellular MDGEA-IGFBP-3 does not accumulate on the surface of the cells in which it is overexpressed (53,54), exogenous MDGEA-IGFBP-3 stimulated the phosphorylation of Smad2 and Smad3 in T47D breast cancer cells that expressed TGF-␤RII (55), suggesting that it is able to activate signaling pathways by transiently associating with cell surface receptors. Thus, without being localized to the nucleus, MDGEA-IGFBP-3 might induce apoptosis either by acting outside the cell to stimulate cell signaling via a plasma membrane receptor or intracellularly by nonnuclear mechanisms.
To distinguish between these possibilities, in the present study we have expressed the MDGEA mutant in a nonsecreted variant of full-length IGFBP-3 that lacks a signal peptide to examine whether nonsecreted, intracellular IGFBP-3 can induce apoptosis in human prostate cancer cells without being localized to the nucleus. We confirm that wild-type and mutant IGFBP-3 lacking a signal peptide are not secreted and that MDGEA-IGFBP-3 is predominantly localized to the cytoplasm. We demonstrate that nonsecreted wild-type and MDGEA-IGFBP-3 induce apoptosis in human prostate cancer cells in an IGF-independent manner, indicating that neither secretion nor concentration in the nucleus are required for the induction of apoptosis by IGFBP-3.

Construction of Plasmids Expressing Yellow Fluorescent Protein (YFP)-IGFBP-3 Fusion Proteins
Nonsecreted YFP-IGFBP-3-Wild-type human IGFBP-3 (residues 1-264) (40) without its 27-amino acid signal peptide was fused to the C terminus of YFP in the YFP-C1 vector (BD Biosciences) by PCR amplification of the open reading frame of the mature protein from the corresponding expression plasmid (Fig.  1). The sense oligonucleotide contained a SalI site (underlined) followed by the start of the mature IGFBP-3 coding sequence: 5Ј-ACGTATGTCGACGGCGCGAGCTCGGGGGGCTTGGGT-3Ј. The antisense oligonucleotide contained a BamHI site (underlined) followed by a stop codon and the last 8 amino acids of the IGFBP-3 coding sequence: 5Ј-ACGTATGGATCCCTACTTGC-TCTGCATGCTGTAGCAGTG-3Ј. The PCR-amplified products were digested with SalI and BamHI and ligated with YFP-C1 that had been digested with SalI and BamHI. Fusion constructs of YFP and IGFBP-3 mutants were constructed similarly after PCR amplification of mutant IGFBP-3 from the appropriate pRSV-Sec-BP3 construct (37). For the mutation in the nuclear localization signal, KGRKR (AAAGGCAGGAAGCGG) was mutated to MDGEA (AtgGatgGGgAGgcG). The double mutant (6m/MDGEA-IGFBP-3) was constructed by introducing the 6m mutation in MDGEA-IGFBP-3 as previously described (37).
Secreted IGFBP-3-YFP-Wild-type IGFBP-3 with its natural signal peptide (40) was fused to the N terminus of YFP in the YFP-N1 plasmid (BD Biosciences) to form Pre-IGFBP-3-YFP ( Fig.  1) using a similar amplification strategy. The sense oligonucleotide contained an EcoRI site (underlined) followed by TG and the beginning of the IGFBP-3 signal peptide amino acid sequence: 5Ј-ACGTATGAATTCTGATGCAGCGGGCGCGACCCACG-CTC-3Ј. The antisense oligonucleotide contained a BamHI site (underlined) followed by two nucleotides and the last 8 amino acids of the IGFBP-3 coding sequence without its stop codon: To construct Pre-MDGEA-IGFBP-3-YFP, Pre-WT-IGFBP-3-YFP DNA was digested with EcoRI and BamHI (ϳ875 bp) and then with SphI (which cuts after nucleotide 538 relative to 1 ATG). The larger EcoRI-SphI fragment encodes the N-terminal fragment of IGFBP-3, including the signal peptide. Similarly, digestion of YFP-MDGEA DNA with SalI and BamHI followed by SphI generated a smaller SphI-BamHI insert encoding a C-terminal fragment containing the MDGEA mutation. Following ligation of these two inserts, the full-length Pre-MDGEA-IGFBP-3 insert was ligated into YFP-N1 DNA predigested with EcoRI-BamHI to form Pre-MDGEA-IGFBP-3-YFP.
To construct Pre-6m-IGFBP-3-YFP and Pre-6m/MDGEA-IGFBP-3-YFP, SalI-BamHI insert fragments were prepared from the corresponding YFP-IGFBP-3 constructs and digested with SacI (nucleotide 91), and the large SacI-BamHI fragment containing the two mutations was isolated. Pre-WT-IGFBP-3-YFP DNA was digested first with EcoRI and BamHI and then with SacI. The small EcoRI-SacI fragment (encoding the signal peptide and the first 3 amino acids of the mature protein) was then ligated separately to the SacI-BamHI fragments to create full-length Pre-6m-IGFBP-3 or Pre-6m/MDGEA-IGFBP-3 inserts. These inserts were ligated separately into EcoRI-BamHI-digested YFP-N1 DNA to generate the corresponding fusion constructs. Recombinants were confirmed by DNA sequencing and expression of the fusion protein in the media from transfected cells.

Transfection of PC-3 Cells
PC-3 cells (1.5 ϫ 10 5 ) were grown to 70 -80% confluence in 6-well (35-mm) plates (Corning Glass) in F12K medium containing 10% fetal bovine serum. Immediately before transfection, the cells were rinsed twice with serum-free F12K medium. The cells were transfected (37°C, 3 h) in 1 ml of serum-free medium containing 500 ng of different plasmid constructs using Lipofectamine Plus (Invitrogen) according to the manufacturer's instructions. In some experiments, 20 ng of pRSVluciferase was cotransfected with the indicated YFP-IGFBP-3 construct DNA to normalize transfection data. The cells were allowed to recover by adding 1.7 ml of F12K medium plus 0.3 ml of fetal bovine serum and continuing the incubation at 37°C for 48 h unless otherwise specified.

Western Blotting of Whole Cell Extracts and Media from Transfected Cells
PC-3 cells were transfected with the indicated YFP-IGFBP-3 plasmids as described above and allowed to recover by incubation in F12K medium containing 10% fetal bovine serum for 24 h. The cells were rinsed with serum-free F12K medium, and the incubation continued for another 24 h in serum-free medium. The conditioned media were collected, protease inhibitors (1ϫ protease inhibitor set I; Calbiochem) were added, and the media were centrifuged (4°C, 5 min, 2,000 rpm) to remove any contaminating cells. The media were concentrated 10-fold by centrifugation using Centricon YM-10 (Millipore Corp., Bedford, MA) centrifugal filter devices (2,000 rpm, 4°C, ϳ1 h). Protease inhibitors were replenished, and the concentrated media were stored at Ϫ70°C until use.
To prepare whole cell extracts, cells were washed with phosphate-buffered saline (PBS) and were lysed (100 l per 20-l cell pellet) in whole cell extract buffer (10 mM HEPES, pH 7.4, 10% glycerol, 250 mM NaCl, 1 mM EDTA, 0.1% Nonidet P-40, 1 mM dithiothreitol, and 1ϫ protease inhibitor mixture) by rotating the plate at 4°C for 20 min. Cell lysates were transferred to tubes using a cell scraper (Sarstedt, Newton, NC). Cells were vortexed at high speed for 15 s and placed in ice for another 5-10 min. After centrifugation (12,000 rpm, 5 min, 4°C), whole cell extracts (supernatants) were stored at Ϫ70°C.
Proteins in whole cell extracts and media were fractionated using 4 -20% SDS-PAGE (Bio-Rad) under reducing conditions. Protein concentrations were measured using the DC protein assay (Bio-Rad). For cells, 50 g of protein was loaded per lane; for media, an approximately equal aliquot (25 g/lane) was loaded. Following transfer of the fractionated proteins to nitrocellulose membranes, specific proteins were identified by Western blotting using goat anti-IGFBP-3 antibody, mouse anti-GFP monoclonal antibody, and anti-IB. Bound antibody was identified following incubation with second antibody using SuperSignal West Pico Chemiluminescent Substrate (Pierce).
In some experiments (results not shown), nuclear and cytoplasmic fractions were isolated biochemically and processed for Western blotting. Fractionation was performed using the NE-PER kit (Pierce) and by the method described by Schreiber et al. (56). Purity of the fractions was confirmed by reblotting with anti-poly-ADP-ribose polymerase (nuclear) and anti-IB (cytoplasmic). Equal amounts of protein were loaded in each lane. Scanning results were adjusted to equivalent aliquots.

In Vitro Deglycosylation
PC-3 cells transfected with YFP-WT-IGFBP-3, Pre-WT-IGFBP-3-YFP, YFP-MDGEA-IGFBP-3, and Pre-MDGEA-IGFBP-3-YFP DNA for 3 h as described above were incubated in serum-containing medium for 24 h. The medium was changed to serum-free medium, and the incubation continued for another 24 h. Media and whole cell extracts were prepared as previously described. Samples (5 g of protein) were deglycosylated according to the manufacturer's instructions. The proteins were denatured (0.1% SDS, 50 mM ␤-mercaptoethanol) by boiling for 5 min. Nonidet P-40 solution (final concentration 0.75%) and 5 milliunits of N-glycanase (Prozyme, San Leandro, CA) were added, and the incubation continued overnight at 37°C. Sample buffer was added, and the samples were fractionated on a 4 -20% SDS-polyacrylamide gel. The Western blot was first probed with a polyclonal goat antibody raised against hIGFBP-3. Then the blot was stripped with Restore stripping buffer (Pierce) and was incubated with anti-mouse IgG and horseradish peroxidase substrate to verify that no residual bands were observed. The blot was probed with a mouse monoclonal antibody raised against GFP, again stripped with verification, and reprobed with a mouse monoclonal antibody specific for N-terminal amino acids 1-97 of human IGFBP-3.

Imaging and Confocal Microscopy
Imaging of different YFP-IGFBP-3 constructs was performed as previously described (57). PC-3 cells (1.5 ϫ 10 5 cells/well in a 6-well plate) were seeded on a Corning 22-mm 2 coverglass (Corning, NY) and were grown for 48 h prior to transfection. Cells were transfected with either YFP-N1 (empty vector) or different YFP-IGFBP-3 constructs for 3 h at 37°C. Cells were allowed to recover in medium containing 10% fetal bovine serum for 24 h and fixed in 4% paraformaldehyde in 1ϫ PBS (room temperature, 20 min). The cells were washed twice with PBS at room temperature and permeabilized with 0.2% Triton X-100 in PBS (5 min on ice). The nuclear DNA was stained with 4Ј,6-diamidino-2-phenylindole dihydrochloride (DAPI; Invitrogen; 300 nM in PBS) and mounted on slides using Prolong-Gold mounting media (Invitrogen). Cells were imaged on a Carl Zeiss 510 Meta laser-scanning confocal microscope equipped with an argon laser and ϫ63 objective, numerical aperture 1.4.

Flow Cytometric Assay for Apoptosis
PC-3 cells were transfected for 3 h and allowed to recover by incubation in F12K medium containing 10% fetal bovine serum for 48 h. For analysis of apoptosis, attached cells were trypsinized, combined with cells floating in the media, and centrifuged (2,000 rpm, 5 min). The cell pellets were washed with 1ϫ binding buffer (10 mM HEPES, pH 7.4, 140 mM NaCl, 2.5 mM CaCl 2 ; BD Pharmingen). The cells were stained using 300 -500 l of binding buffer containing 5 l each of annexin V-APC and 7-AAD at room temperature for 15-30 min. Cells were analyzed using a CyAn LX flow cytometer equipped with Summit software (DakoCytomation, Fort Collins, CO). 7-AAD, a vital dye that binds to nucleic acids, is normally excluded from cells and only can penetrate the membranes of dead or damaged cells. Annexin V-APC binds to phospholipids that are inaccessible in viable cells, since they are located in the inner leaflet of the plasma membrane, but become accessible early in apoptosis when they are translocated to the outer leaflet of the plasma membrane or later when membrane integrity is lost (58). Following transfection of PC-3 cells with YFP-IGFBP-3, staining with annexin V-APC only was observed at late times coincident with the appearance of 7-AAD staining (result not shown), so that it reflects loss of plasma membrane integrity. Almost all of the annexin V ϩ cells were also 7-AAD ϩ . The loss of cell viability was caspase-dependent, establishing that it was due to apoptosis.

Cell Sorting
After transfection with plasmids expressing YFP-IGFBP-3 fusion proteins and recovery for 24 h in serum-containing F12K medium, green fluorescent cells were separated from nongreen cells using a FACSVantage SE sorting flow cytometer (BD Biosciences) at 4°C. Only the viable population of cells within a standard forward and side scatter gate were sorted into green fluorescent (YFP ϩ ) and non-green fluorescent (YFP Ϫ ) fractions. The sorted YFP ϩ population contained Ͼ95% green fluorescent cells. Sorted YFP ϩ and YFP Ϫ cells were replated and grown for another 24 h before staining with annexin V-APC and 7-AAD for cell death analysis using flow cytometry or examination of cell lysates by Western blotting.

Determination of Cell Viability Using Trypan Blue Staining
Nonviable cells were identified by trypan blue staining as previously described (37). PC-3 cells were plated in 12-well culture dishes (100,000 cells/well) and grown to ϳ80% confluence in serumsupplemented F12K medium. The medium was changed to serum-free medium for 24 h and then replaced with 0.4 ml of fresh serum-free F12K medium containing 0.1% bovine serum albumin, the indicated concentrations of TRAIL, recombinant hIGFBP-3 standard, His 6 -WT-hIGFBP-3, or His 6 -6m-hIGFBP-3 purified from transfected CHO-K1 cells or purified control media from CHO-K1 cells transfected with pRSV-Sec empty vector, with or without a 20 M concentration of the indicated caspase inhibitors (Z-VADfmk, Z-LEHD-fmk, or Z-IETD-fmk). After a 72-h incubation, detached cells were collected and resuspended in 0.4 ml of medium; adherent cells were dissociated with trypsin and resuspended in the same volume of medium. Trypan blue (0.4%) was added to the suspensions and incubated for 10 min. Total cells and nonviable cells stained with trypan blue in both adherent and detached fractions were counted in a hemocytometer.

YFP-MDGEA-IGFBP-3 Is Intracellular and Not Secreted-
Wild-type IGFBP-3 and IGFBP-3 mutated in its nuclear localization signal (MDGEA-IGFBP-3) were fused to YFP to facilitate analysis of the subcellular localization of the proteins. Mutation of the IGF-binding site also was introduced (YFP-6m-IGFBP-3 and YFP-6m/MDGEA-IGFBP-3) to establish that the induction of apoptosis was IGF-independent. Constructs were prepared using mature, full-length IGFBP-3 without a signal peptide (YFP-IGFBP-3) that would not be secreted so the fused proteins only could act intracellularly. For comparison, the corresponding constructs were prepared using IGFBP-3 with its native signal peptide so that they would be secreted (Pre-IGFBP-3-YFP) (Fig. 1).
Western blots of whole cell extracts from transfected PC-3 cells showed that the ϳ67-kDa YFP-IGFBP-3 fusion protein (Fig. 2, upper left panel, lane 3) was readily distinguished from 27-kDa YFP (lane 2). As expected, wild-type YFP-IGFBP-3 and all of the mutant YFP-IGFBP-3 fusion proteins (6m, MDGEA, and 6m/MDGEA) synthesized from constructs lacking a signal peptide were detected only in cell lysates and not in conditioned media (Fig. 2, upper right panel, lanes [5][6][7][8][9][10][11][12], consistent with their remaining inside the cell and not being secreted. By contrast, in cells transfected with Pre-IGFBP-3-YFP, the IGFBP-3 fusion protein was found in the media (lane 14) but was almost undetectable in cell extracts (lane 13). IGFBP-3 was not detected in cells or media from mock-transfected cells or cells transfected with YFP empty vector (lanes 1-4), although endogenous IGFBP-3 could be detected by an immunoradiometric assay in the media of cells transfected with YFP empty vector (results not shown). Stripping the Western blot and rep- Top bars, IGFBP-3 was fused to YFP to generate fusion proteins that could or could not be secreted. Nonsecreted variants were generated by fusing mature human IGFBP-3 lacking a signal peptide (amino acids 1-264) to the C terminus of YFP (YFP-IGFBP-3). Secreted Pre-IGFBP-3-YFP was generated by fusing IGFBP-3 with its natural 27-amino acid signal prepeptide (40) to the N terminus of YFP. (The nonsecreted counterpart of Pre-IGFBP-3-YFP, IGFBP-3-YFP, also was constructed. Its properties were indistinguishable from those of nonsecreted YFP-IGFBP-3 (results not shown).) Bottom bar, mutations were introduced in the N-terminal IGF-binding site of IGFBP-3, the C-terminal NLS (49), or both. Replacing 6 residues in the IGF-binding site (Ile 56 , Tyr 57 , Arg 75 , Leu 77 , Leu 80 , and Leu 81 ) with alanine to generate the 6m-IGFBP-3 mutant greatly reduced its affinity for IGF-I and IGF-II (37). The C-terminal residues 228 KGRKR 232 in the nuclear localization signal were mutated to 228 MDGEA 232 , the corresponding sequence in IGFBP-1. IGFBP-1 is not translocated to the nucleus (48).  11 and 12), or Pre-WT-IGFBP-3-YFP (lanes 13 and 14) were identified using goat antibody to human IGFBP-3. By immunoradiometric assay (IRMA; results not shown), intracellular IGFBP-3 was 17-37-fold greater in cells expressing YFP-WT-IGFBP-3 and YFP-MDGEA-IGFBP-3 than in cells transfected with YFP empty vector. Endogenous IGFBP-3 was detected in the media of cells transfected with YFP empty vector by IRMA, although it was not seen by Western blotting. Lower right panel, the blot shown in the upper right panel was stripped and reprobed with antibody to the intracellular marker protein, IB. IB was seen in all cell extracts but was not detected in any of the media, indicating that leakage of cell contents had not occurred.
robing it with antibody to the intracellular protein marker, IB, confirmed that the cells were intact and that no leakage of cellular contents had occurred.
To demonstrate that the intracellular YFP-IGFBP-3 fusion proteins had never been secreted as opposed to having been secreted and then reentering the cells, we determined whether they were N-glycosylated. The rationale for this experiment is that secreted native IGFBP-3 contains three N-glycosylation motifs (Asn-X-Ser/Thr) and is N-glycosylated (59), and the enzymes responsible for transferring the preassembled N-glycan to the nascent protein are located within the lumen of the endoplasmic reticulum (ER) (60). YFP-IGFBP-3, which lacks a signal peptide and presumably is synthesized in the cytosol, should not be translocated into the ER lumen and consequently would not be expected to be N-glycosylated. Lysates from cells expressing the YFP-IGFBP-3 constructs and media from cells transfected with Pre-IGFBP-3-YFP constructs were incubated with N-glycanase in vitro and examined by Western blotting using antibodies to IGFBP-3, GFP, or N-terminal IGFBP-3 ( Fig.  3 and results not shown). The electrophoretic mobilities of secreted Pre-WT-IGFBP-3-YFP and Pre-MDGEA-IGFBP-3-YFP, like recombinant N-glycosylated hIGFBP-3 (data not shown), increased following incubation with N-glycanase, consistent with a decrease in the molecular mass of IGFBP-3 due to removal of the N-linked oligosaccharides (59). By contrast, the mobilities of nonsecreted YFP-WT-IGFBP-3 and YFP-MDGEA-IGFBP-3 were unchanged, indicating that the YFP-IGFBP-3 fusion proteins were not N-glycosylated. This provides strong independent evidence that the YFP-IGF-BP-3 fusion proteins had not been translocated into the ER lumen, N-glycosylated, and secreted, indicating that they had remained exclusively intracellular after synthesis.
Nonsecreted YFP-WT-IGFBP-3 Induces Apoptosis in PC-3 Cells-Having established that the YFP-IGFBP-3 fusion proteins are exclusively intracellular, we next determined whether nonsecreted wild-type IGFBP-3 could induce apoptosis in PC-3 cells as we had previously shown for extracellular exogenous IGFBP-3 (37). Cells were transfected with plasmids expressing YFP empty vector (YFP) or YFP-WT-IGFBP-3 for 3 h, incubated in serum-containing medium for 48 h, and stained with annexin V-APC to identify nonviable cells by flow cytometry (Fig. 4). FACS analysis indicates that 40 -55% of the transfected cells exhibited green fluorescence (right quadrants), providing a minimum estimate of transfection efficiency. Cell death (indicated by increased annexin V-APC staining in the upper quadrants) was 7-fold greater in cells transfected with YFP-WT-IGFBP-3 compared with YFP empty vector.
The loss of cell viability induced by YFP-WT-IGFBP-3 was caspase-dependent, indicating that cell death was due to apoptosis (Fig. 4). Caspases are cysteine-rich proteases that cleave key proteins essential for maintaining cell integrity after aspartate residues (61), generating the characteristic biochemical and morphological features of apoptosis (62). Treatment with the broad specificity caspase inhibitor Z-VAD-fmk during the post-transfection incubation decreased cell death in cells transfected with YFP-IGFBP-3 by 83% but had no effect in cells transfected with YFP (results not shown). This is the first demonstration that nonsecreted IGFBP-3 can induce apoptosis by intracrine intracellular mechanisms that do not involve interaction with the plasma membrane.
Unexpectedly, although the induction of apoptosis was IGFBP-3-dependent, since it was observed with YFP-IGFBP-3 but not with YFP empty vector, most of the apoptotic cells did not exhibit green YFP fluorescence. The non-green fluorescent apoptotic cells could be cells that had been transfected with YFP-IGF-BP-3 and became YFP ϩ but lost their green YFP fluorescence during post-transfection incubation, or nontransfected YFP Ϫ  7 and 8). 5-g aliquots of each whole cell extract or medium were incubated with (ϩ, even-numbered lanes) or without (Ϫ, odd-numbered lanes) 5 milliunits of N-glycanase for 16 h. Treated and untreated samples were fractionated by SDS-PAGE and examined by Western blotting using polyclonal goat antibody against IGFBP-3 (upper panel), mouse monoclonal antibody against GFP (lower panel), and mouse monoclonal antibody against an N-terminal fragment of human IGFBP-3 (amino acids 1-97; not shown). Molecular mass markers are indicated. The mobilities of the nonsecreted fusion proteins are appropriate for their molecular mass, but the mobility of Pre-IGFBP-3-YFP is more rapid in this experiment and Fig. 2. Pre-IGFBP-3-YFP reacts with antibodies to the N terminus of IGFBP-3 (results not shown) and C-terminal YFP, indicating that the fusion protein is probably intact. Nonsecreted IGFBP-3-YFP and nonsecreted YFP-IGFBP-3 have the same mobility (results not shown), indicating that the mobility difference does not reflect a conformational change resulting from coupling YFP to the N-or C termini of IGFBP-3. The reasons for the mobility differences are not clear but may be due to using conditioned media for secreted Pre-IGFBP-3-YFP and cell lysates for nonsecreted YFP-IGFBP-3 (and IGFBP-3-YFP).
bystander cells in which apoptosis was induced by the transfected cells although the fusion protein itself was not secreted. To distinguish between these possibilities, transfected cells were sorted into YFP ϩ and YFP Ϫ populations after a 24-h incubation, replated for another 24 h, and analyzed by flow cytometry (Fig. 5A).
Approximately 60% of the sorted YFP ϩ cells that had been transfected with YFP-WT-IGFBP-3 or YFP-6m-IGFBP-3 became YFP Ϫ after replating (Fig. 5B), contrasted with only 14% of sorted YFP ϩ cells that had been transfected with YFP empty vector. This indicates that a large number of the YFP-IGFBP-3-transfected cells that were YFP ϩ at the time of sorting became YFP Ϫ after replating and that the loss of green fluorescence depended on IGFBP-3 expression. The decreased fluorescence appears to result from a corresponding decrease in the abundance of the fusion protein during replating of YFP ϩ sorted cells (Fig. 5C). This decrease was selective, since the abundance of YFP and ␤-actin was unchanged.
Not only did cells transfected with YFP-WT-IGFBP-3 or YFP-6m-IGFBP-3 that were YFP ϩ at the time of sorting become YFP Ϫ during replating, but they became apoptotic (Fig. 5B). Four times as many cells transfected with the YFP-IGFBP-3 fusion proteins became apoptotic as cells transfected with YFP empty vector, indicating that the induction of apoptosis was IGFBP-3-dependent. A similar IGFBP-3-dependent induction of apoptosis was observed in sorted YFP ϩ cells that remained YFP ϩ after replating and in sorted YFP Ϫ cells (results not shown). These results suggest that cells transfected with YFP-IGFBP-3 constructs become apoptotic and lose their green YFP ϩ fluorescence at different times relative to cell sorting into YFP ϩ and YFP Ϫ populations at 24 h. Some sorted YFP ϩ cells became apoptotic during replating but remained YFP ϩ . Others may have lost their green fluorescence and turned YFP Ϫ prior to sorting but only became apoptotic during replating.

Intracellular YFP-MDGEA-IGFBP-3 Is Predominantly Localized to the
Cytoplasm-Having established that intracellular YFP-IGFBP-3 can induce apoptosis without being secreted, we analyzed the subcellular localization of nonsecreted wild-type and mutant YFP-IGFBP-3 using laser-scanning confocal microscopy of transfected PC-3 cells (Fig. 6). Strong intracellular YFP fluorescence was visualized with all constructs. YFP-WT-IGFBP-3 and the IGF-non-    1-6 show cells prepared for SDS-PAGE immediately after sorting. Lanes 7-12 show sorted cells that were replated and incubated in serumcontaining medium for another 24 h before being prepared for SDS-PAGE. Equal aliquots of cell lysate from each sample were electrophoresed on a 4 -20% polyacrylamide gel. The blot was probed with mouse anti-GFP monoclonal antibody and then stripped and probed with antibody to ␤-actin. Open arrowhead, YFP; solid arrowhead, YFP-WT-IGFBP-3 and YFP-6m-IGFBP-3. The decreased abundance of the YFP-IGFBP-3 fusion proteins after replating may be due to degradation. Release of the fusion proteins to the extracellular media was not observed after a 48-h incubation following transfection (Fig. 2).

. YFP-IGFBP-3-induced cell death in PC-3 cells is due to caspase-dependent apoptosis. PC-3 cells were transfected with YFP empty vector (left panel) or YFP-WT-IGFBP-3 (center and right panels) for 3 h and then incubated for 48 h in serum
binding mutant, YFP-6m-IGFBP-3, mainly localized to the nucleus. By contrast, IGFBP-3 constructs in which the presumed nuclear localization signal had been mutated, YFP-MDGEA-IGFBP-3 and the double mutant YFP-6m/MDGEA-IGFBP-3, predominantly localized to the cytoplasm. YFP expressed by the empty vector is present throughout the cell, because it is small enough (27 kDa) to diffuse through the nuclear pore and does not contain a nuclear localization signal. Similar results were obtained following biochemical fractionation of transfected cells. Western blotting of nuclear and cytoplasmic fractions indicated that a much lower percentage of YFP-MDGEA-IGFBP-3 was in the nucleus compared with YFP-WT-IGFBP-3 (results not shown). These results indicate that nonsecreted MDGEA-IGFBP-3 is not concentrated in the nucleus of PC-3 cells and establish that 228 KGRKR 232 is required for the nuclear localization of intracellular full-length IGFBP-3 in intact cells. They support and extend previous results obtained using digitonin-permeabilized cells (49) and in intact cells with exogenous IGFBP-3 (48) or expression of secreted IGFBP-3 (53). In the latter two studies, however, the decreased nuclear uptake may result from decreased ability of MDGEA-IGFBP-3 to interact with the plasma membrane and be internalized (48,53,54) rather than impaired translocation across the nuclear membrane.
Confocal imaging also was performed with the secreted Pre-IGFBP-3-YFP constructs to determine the effect of the MDGEA mutation on nuclear translocation of secreted fusion protein that had reentered the cell (Fig. 7). The subcellular distribution of Pre-IGFBP-3-YFP and Pre-MDGEA-YFP could not be evaluated, because the uptake of the secreted proteins from the extracellular media to the cells was too low to be detected. Internalization was observed, however, in Pre-IGFBP-3 fusion proteins containing the 6m mutation in the IGF-binding site. This difference may be explained by the fact that free IGFBP-3, but not IGF⅐IGFBP-3 complexes, can bind to the cell surface (32,51,(63)(64)(65) or proposed IGFBP-3 receptors (41). Internalized Pre-6m-IGFBP-3-YFP localized predominantly to the nucleus, whereas Pre-6m/MDGEA-IGFBP-3-YFP localized mainly to the cytoplasm. Thus, although neither secreted nor nonsecreted fusion proteins containing the MDGEA mutation are concentrated in the nucleus, differences in their extracellular versus intracellular distribution may have important implications for their mechanism of action.

Nonsecreted YFP-MDGEA-IGFBP-3 and YFP-6m/MDGEA-IGFBP-3 Induce Apoptosis in PC-3 Cells-Transfection
with nonsecreted, predominantly cytoplasmic YFP-MDGEA-IGFBP-3 increased the number of apoptotic cells staining with annexin V-APC or 7-AAD more than 5-fold relative to cells transfected with YFP empty vector (Fig. 8). Comparable levels of stimulation were seen with YFP-6m-IGFBP-3 and YFP-6m/ MDGEA-IGFBP-3 double mutant, indicating that the induction of cell death by the nonsecreted nuclear localization signal (NLS) mutant is IGF-independent. These results indicate that YFP-MDGEA-IGFBP-3 induces apoptosis by IGF-independent mechanisms that do not require the fusion protein to be concentrated in the nucleus, to be secreted, or to interact with the plasma membrane. Apoptosis also was induced by secreted Pre-MDGEA-IGFBP-3-YFP and Pre-6m/MDGEA-IGFBP-3-YFP (Fig. 9). Pre-6m/-MDGEA-IGFBP-3-YFP may be internalized after secretion, so that apoptosis could be induced by either extracellular or intracellular IGF-independent mechanisms. Pre-MDGEA-IGFBP-3-YFP, on the other hand, is not appreciably internalized after secretion, so that the induction of apoptosis by the extracellular fusion protein probably occurs after interaction with the plasma membrane.
Initiator Caspases 8 and 9 Are Involved in the Induction of Apoptosis by Both Nonsecreted and Exogenous IGFBP-3-Apoptosis is the end result of the proteolysis of proteins necessary to maintain cell integrity by effector caspases (caspases 3, 6, and 7) that are activated by proteolysis of inactive precursors by one of two main initiator caspases, caspase 8 or caspase 9 (66,67). Caspase 8 is activated in a plasma membrane complex that is composed of one of six death receptors (which include Fas, the TNF-␣ receptor, and two receptors for TRAIL), a common adaptor protein, FADD (Fas-associated death domain), and the caspase 8 precursor, procaspase 8. Initiator caspase 9 is activated from its precursor, procaspase 9, in a cytosolic complex that is formed following the release of cytochrome c from the intermembrane mitochondrial space. Results with IGFBP-3 have been conflicting, indicating that either or both of the initiator caspase pathways may be involved (11,68,69).
We addressed this question by examining the effect of selectively inhibiting caspase 9 (using Z-LEHD-fmk) or caspase 8 (using Z-IETD-fmk) on the induction of apoptosis in PC-3 cells by exogenous IGFBP-3 (Fig. 10). Coincubation with the general caspase inhibitor (Z-VAD-fmk) or either of the selective initiator caspase inhibitors decreased cell death induced by exogenous TRAIL or wild-type or 6m-IGFBP-3 by Ͼ80%. These results indicate that both initiator caspases 8 and 9 are involved in the induction of apoptosis by exogenous IGFBP-3 and can be explained by cross-talk between the two initiator caspase pathways. Similar results were obtained in cells expressing nonsecreted YFP-IGFBP-3 constructs (Table 1). Apoptosis induced by wild-type YFP-IGFBP-3 or YFP-6m-IGFBP-3 was inhibited significantly by both Z-IETD-fmk and Z-LEHD-fmk. Thus, both caspase 8 and caspase 9 participate in the induction of apoptosis by endogenous nonsecreted YFP-IGFBP-3 as well as by exogenous recombinant IGFBP-3.

DISCUSSION
IGFBP-3 can induce apoptosis in prostate cancer cells in an IGF-independent manner (4, 37), but its mechanism of action remains poorly understood. It has been suggested that since IGFBP-3 contains a nuclear localization signal (44) and exogenous IGFBP-3 can be transported to the nucleus (45-51), nuclear localization may play an important role in its IGF-independent action. To test this hypothesis, we mutated residues 228 KGRKR 232 in the C-terminal nuclear localization signal (49) to MDGEA in a full-length IGFBP-3 construct that lacks a signal peptide so that it would not be secreted. We show that nonsecreted MDGEA-IGFBP-3 is predominantly cytoplasmic and that it can induce apoptosis in an IGF-independent manner.
Western blotting of cell lysates and conditioned media indicated that YFP-WT-IGFBP-3 and YFP-MDGEA-IGFBP-3 fusion proteins synthesized without a signal peptide are intracellular and not detectable in the media, whereas Pre-IGFBP-3-YFP, which contains a signal peptide, is secreted. Secreted   . Inhibitors of caspase 8 and caspase 9 decrease the induction of apoptosis by exogenous IGFBP-3 or TRAIL. Subconfluent PC-3 cells that had been equilibrated with serum-free F12K medium for 24 h were incubated with serum-free medium containing recombinant hIGFBP-3 Standard (2 g/ml), purified wild-type CHO-hIGFBP-3 (0.9 g/ml), purified 6m-CHO-hIGFBP-3 (1.7 g/ml), TRAIL (1 g/ml), or an equivalent amount of purified media from CHO-K1 cells transfected with pRSV-Sec empty vector (Control) as previously described (37). Cells were incubated without or with a 20 M concentration of the general caspase inhibitor, Z-VAD-fmk; the caspase 9 inhibitor, Z-LEHD-fmk; or the caspase 8 inhibitor, Z-IETD-fmk as indicated. After a 72-h incubation, adherent and detached cells were harvested, the cell suspensions were stained with trypan blue to identify nonviable cells, and total and dead (blue) cells in both pools were counted. Results from the attached and detached pools are combined. The percentage of total dead cells is plotted. Results of a representative experiment are shown.  AUGUST 25, 2006 • VOLUME 281 • NUMBER 34 IGFBP-3 is N-glycosylated at three sites (59). N-Glycosylation occurs within the lumen of the endoplasmic reticulum (60) by transfer of preassembled oligosaccharide chains to the amido N of asparagines in the nascent polypeptide chains. We found that secreted Pre-IGFBP-3-YFP was N-glycosylated, whereas intracellular YFP-IGFBP-3 was not. Most likely the YFP-IGFBP-3 fusion proteins that lack a signal peptide are synthesized on free ribosomes in the cytosol rather than in the rough endoplasmic reticulum and never enter the ER/Golgi secretion pathway where they would have been N-glycosylated, indicating that they remained intracellular after synthesis. Confocal imaging indicated that nonsecreted YFP-WT-IGFBP-3 was predominantly located in the nucleus, whereas YFP-MDGEA-IGFBP-3 was predominantly cytoplasmic. Biochemical fractionation and Western blotting confirmed that the percentage of YFP-MDGEA-IGFBP-3 in the nucleus was much lower than the percentage of nuclear YFP-WT-IGFBP-3 (results not shown). Mutation of the IGF-binding site did not affect the distribution of either wild-type or NLS-mutant fusion proteins; nonsecreted YFP-6m-IGFBP-3 was predominantly nuclear, and nonsecreted YFP-6m/MDGEA-IGFBP-3 was predominantly localized in the cytoplasm.
Like IGFBP-3, other proteins also can act by intracrine mechanisms that occur entirely within the cell of synthesis without the protein being secreted or interacting with the plasma membrane (70,71). Intracrine action has been reported for nonsecreted isoforms of parathyroid hormone-related protein (72) and fibroblast growth factor 2 (73, 74) as well as for parathyroid hormone-related protein (75) and recombinant angiotensin II (76) expressed without their signal peptides. The intracellular isoforms of parathyroid hormone-related protein and fibroblast growth factor 2 localized to the nucleus (72,74,75) and exhibited different functions than the secreted isoforms (75,77). In the case of IGFBP-3, secreted Pre-IGFBP-3-YFP and nonsecreted YFP-IGFBP-3 both induce apoptosis in PC-3 cells, suggesting that the secreted and nonsecreted proteins may be functionally equivalent.
In order to induce apoptosis, secreted Pre-IGFBP-3-YFP fusion proteins first must interact with the plasma membrane, presumably binding to an IGFBP-3 receptor to activate a signal transduction pathway, be internalized by endocytosis, or both (Fig. 11). Candidate IGFBP-3 receptors have been proposed, including TGF-␤RV/LRP-1, a large endocytic receptor that binds multiple ligands including ␣ 2 -macroglobulin (41,78) and a small receptor designated 4-33 (42). IGFBP-3-induced inhibi-tion of DNA synthesis in mink lung epithelial cells is decreased by incubation with a receptor-associated protein antagonist of ligand binding to LRP-1, and inhibition of DNA synthesis in H1299 lung cancer cells by IGFBP-3 requires LRP-1 expression (41). Although these results suggest that LRP-1 might mediate some of the antiproliferative activities of IGFBP-3, this is unlikely to be the sole mechanism of IGFBP-3 action, since IGFBP-3 induces apoptosis in MCF-7 breast cancer cells (79) that do not appear to express LRP-1 (80).
IGFBP-3 activates several signal transduction pathways, but none of them have been linked to activation of the putative plasma membrane receptors, LRP-1 or 4-33. They include induction and activation of Stat1 (38), phosphorylation of Smad2 and -3 (55,81), and stimulation of phosphotyrosine phosphatase or phosphatidylinositol 3-kinase and increasing intracellular Ca 2ϩ (reviewed in Ref. 43). Antisense oligonucleotide to Stat1 blocked IGFBP-3-induced apoptosis in a rat chon-

-YFP (A) and nonsecreted YFP-IGFBP-3 (B) can induce apoptosis.
A, Pre-YFP-IGFBP-3 is synthesized on ribosomes in the rough endoplasmic reticulum. The signal peptide is cotranslationally removed as the fusion protein is translocated into the lumen of the ER, where it is N-glycosylated (indicated by an asterisk) and from which it is ultimately secreted. To induce apoptosis, IGFBP-3-YFP in the media presumably interacts with a plasma membrane IGFBP-3 receptor, either to initiate an intracellular signal transduction pathway, to be internalized within an endocytic vesicle, or both. Vesicular IGFBP-3-YFP can activate a signaling pathway or be retrotranslocated across the ER membrane by unknown mechanisms into the cytosol and possibly translocated into the nucleus. B, nonsecreted IGFBP-3-YFP is presumably synthesized on free ribosomes in the cytosol. It does not enter the ER lumen, is not N-glycosylated or secreted from the cell, and does not interact with the extracellular domain of a possible IGFBP-3 receptor to activate signaling from the plasma membrane. Cytosolic IGFBP-3-YFP can activate intracellular signaling pathways or be translocated into the nucleus to regulate transcription. drogenic cell line, suggesting that Stat1 may play an essential role in IGFBP-3 action (38).
Exogenous IGFBP-3 can be internalized (63,65) by clathrindependent endocytosis of IGFBP-3-transferrin complexes bound to the transferrin receptor (51) and by clathrin-independent endocytosis via lipid rafts (51,82). Based on studies with other ligands and their receptors, endocytosis may have diverse outcomes. For example, TGF-␤ receptors localize to clathrin-dependent early endosomes that are enriched in the Smad2 anchor protein SARA and promote signal transduction, and to caveolin-1-positive vesicles derived from lipid raft membrane microdomains that are enriched in inhibitory Smad7-Smurf2 ubiquitin ligase complexes that target the receptor for degradation (83,84). Protein ligands (typically endocytosed together with their receptors) can signal from within endocytic vesicles (85). In addition, some ligands can translocate across the vesicle membrane into the cytosol by unknown mechanisms, gaining access to importins and Ran GTPase so that they may be actively translocated into the nucleus (52).
YFP-IGFBP-3 and YFP-MDGEA-IGFBP-3 fusion proteins, on the other hand, are not secreted, so they cannot interact with the extracellular domain of plasma membrane receptors to initiate signaling or be internalized in endocytic vesicles (Fig. 11). Instead, they can activate intracellular signaling pathways from the cytosol. These may or may not be the same signal transduction pathways that are activated by extracellular Pre-IGFBP-3-YFP acting through the IGFBP-3 receptor or from internalized endocytic vesicles. Cytosolic wild-type YFP-IGFBP-3, like endocytosed Pre-WT-IGFBP-3-YFP that had been translocated to the cytosol, also can translocate to the nucleus to regulate transcription. Although YFP-MDGEA-IGFBP-3 is not concentrated in the nucleus, the mutant protein still might regulate transcription by indirect mechanisms.
Fusion proteins in which the nuclear localization signal was mutated, alone or together with the 6m mutation in the IGF-binding site, induced apoptosis effectively in PC-3 cells. In the case of Pre-6m/MDGEA-IGFBP-3-YFP, the fusion protein was secreted and, to some extent, able to reenter the cell, making it impossible to distinguish whether the extracellular protein or secreted protein that had reentered the cell was responsible for the induction of apoptosis. Similar ambiguity was inherent in the studies of human breast cancer cells in which MDGEA-IGFBP-3 induced apoptosis and was expressed at high levels in both cells and media (53). The fact that nonsecreted YFP-MDGEA-IGFBP-3, which is exclusively intracellular, and secreted Pre-MDGEA-IGFBP-3-YFP, which is predominantly extracellular, both induce apoptosis indicates that MDGEA-IGFBP-3 can induce apoptosis by either intracellular or extracellular mechanisms without being concentrated in the nucleus.
Finally, we showed that the two major initiator caspase pathways, caspases 8 and 9 (66,67), both were involved in the induction of apoptosis in PC-3 cells by IGFBP-3. Caspase 8 is typically activated by extrinsic ligands via death receptors and mediated by the adaptor FADD, whereas caspase 9 is typically activated by an intrinsic pathway involving perme-abilization of mitochondrial membranes, release of cytochrome c, and activation of procaspase 9. Previous studies with IGFBP-3 have yielded contradictory results. Induction of IGFBP-3 in stably transfected MCF-7 breast cancer cells only activated caspase 8 (68), whereas exogenous IGFBP-3 activated effector caspases 3 and 7 but did not activate caspase 8 in 22RV1 human prostate cancer cells (11), implying that an intrinsic mitochondrial pathway possibly including caspase 9 was involved. In PC-3 cells, we observed that apoptosis induced by either exogenous IGFBP-3 or expression of nonsecreted IGFBP-3 was decreased by inhibition of either caspase 8 or caspase 9, indicating that both initiator caspase pathways are involved. Similar results recently were reported in MDA-MB231 human breast cancer cells in which apoptosis was induced by adenovirus-mediated expression of IGFBP-3 (69). Activation of both initiator caspase pathways can be explained by cross-talk between the two pathways (86). Which of the two initiator caspase pathways triggers apoptosis and whether the same initiator caspase triggers apoptosis induced by exogenous and nonsecreted IGFBP-3 remain to be determined.
In summary, we have shown that the MDGEA mutation is sufficient to prevent the nuclear concentration of nonsecreted YFP-MDGEA-IGFBP-3 as well as IGFBP-3 that has been secreted and then internalized (Pre-6m/MDGEA-IGFBP-3-YFP), further establishing that it is part of the nuclear localization signal. We show for the first time that nonsecreted IGFBP-3 can induce apoptosis in PC-3 cells by intracrine mechanisms, in which the protein remains inside the cell in which it was synthesized. Nonsecreted YFP-MDGEA and the double mutant, YFP-6m/MDGEA, that does not bind IGFs also induce apoptosis in an IGF-independent manner, although they are not concentrated in the nucleus. It remains to be determined whether internalized MDGEA-IGFBP-3-YFP, presumably localized in endocytic vesicles in the cytoplasm, and nonsecreted, cytosolic, nonvesicular YFP-MDGEA-IGFBP-3 utilize the same signaling pathways and trigger the same apoptotic pathways to induce apoptosis in PC-3 prostate cancer cells.