Insulin-like growth factor binding protein-3 (IGFBP-3), the product of a tumor suppressor target gene, can modulate cell proliferation and apoptosis by IGF-I-dependent and IGF-I-independent mechanisms. IGFBP-3 controls the bioavailability of IGFs in the extracellular environment and is known to be subject to degradation by various extracellular proteases. Although nuclear localization and functions of IGFBP-3 have been described in the past, we show as the novel features of this study that the abundance of nuclear IGFBP-3 is directly regulated by ubiquitin/proteasome–dependent proteolysis. We show that IGFBP-3 degradation depends on an active ubiquitin-E1 ligase, specific 26S proteasome inhibitors can efficiently stabilize nuclear IGFBP-3, and the metabolic half-life of nuclear IGFBP-3 is strongly reduced relative to cytoplasmic IGFBP-3. Nuclear IGFBP-3 is highly polyubiquitinated at multiple lysine residues in its conserved COOH-terminal domain and stabilized through mutation of two COOH-terminal lysine residues. Moreover, we show that IGFBP-3, if ectopically expressed in the nucleus, can induce apoptotic cell death. These results suggest that ubiquitin/proteasome–mediated proteolysis of IGFBP-3 may contribute to down-regulation of apoptosis. (Cancer Res 2006; 66(6): 3024-33)
The insulin-like growth factor (IGF)/IGF-binding proteins (IGFBP) axis has been shown to influence the proliferation and survival of various tumors (1), including osteosarcoma, a malignant bone cancer, typically diagnosed in adolescents and young adults (2, 3). IGFBP-3 is a member of a protein family that can bind IGF-I and thereby regulate the mitogenic activity of IGF-I in the extracellular environment. The IGFBP-3 gene is transcriptionally activated by the tumor suppressor p53 (4), and it is assumed that increased expression of IGFBP-3 contributes to p53-dependent apoptosis (5). There is increasing evidence that IGF/IGF-receptor (IGF-R)–independent actions of IGFBP-3 play an important role in its antiproliferative and proapoptotic functions (1). IGFBP-3 has been shown to inhibit the proliferation of immortalized murine fibroblasts with a targeted disruption of IGF-I-R (6). IGFBP-3 can induce programmed cell death via IGF-I-independent pathways in immortalized IGF-R-I-negative mouse fibroblasts (5), and IGFBP-3 has been shown to accentuate apoptotic effects under conditions where IGF-I does not elicit a survival in Hs578T breast cancer cells (7). The IGF-independent proapoptotic activity of IGFBP-3 is underlined by IGFBP-3 mutants that do not bind IGFs but stimulate apoptosis in human prostate cancer (8) and breast cancer cells (9). The mechanisms underlying IGF-independent IGFBP-3 actions are presently not well understood but may involve regulation of gene expression via activation of cell surface receptor kinases or directly by nuclear IGFBP-3 (1).
IGFBP-3 contains a nuclear localization sequence (NLS) in its COOH-terminal domain, and importin-β-dependent import from the cytosol into the nucleus has been shown in vitro (10). Moreover, after addition to the cell culture supernatants, IGFBP-3 has been detected in the nuclei of opossum kidney cells (11), human lung, breast and prostate cancer cells (12–14), and human keratinocytes (15). These findings suggest that nuclear actions of IGFBP-3 may be important for its IGF-independent functions; however, the nuclear functions of IGFBP-3 and the mechanisms that control the activity of nuclear IGFBP-3 are very little understood.
IGFBP-3 has been shown to be subject to degradation by various extracellular proteases (16). We previously showed that the human papillomavirus type 16 (HPV-16) E7 oncoprotein can induce the proteolytic degradation of intracellular IGFBP-3, which can be inhibited by addition of a proteasome inhibitor (17); however, it remains to be determined if intracellular IGFBP-3 by itself is targeted to proteasome-dependent destruction independently of HPV-16 E7. The proteasome controls the metabolic stability of many cell proliferation and apoptosis-regulating proteins in response to multiubiquitination of the target proteins (18). Multiubiquitination-dependent proteolysis involves an enzymatic process, in which specific ubiquitin-conjugating enzymes ligate the ubiquitin polypeptide predominantly to lysine residues of target proteins, and the ubiquitin moiety of the resulting protein-ubiquitin conjugate can subsequently also be ubiquitinated at Lys48, leading to polyubiquitination, which acts as signal for proteasome-dependent degradation (19). The ubiquitin/proteasome pathway is frequently deregulated in tumorigenesis (18); however, no evidence exists at present for direct ubiquitination of IGFBP-3. In this study, the major aim was to investigate if IGFBP-3 is a target of ubiquitin/proteasome–dependent destruction and if IGFBP-3 can induce apoptotic cell death in U-2 OS osteosarcoma cells.
Materials and Methods
Plasmid constructions. The IGFBP-3 sequences were excised from pSF202, pSF210, pSF211, and pSF207 containing the sequences for IGFBP-3 wild type, IGFBP-3KED253-255RGD, IGFBP-3 228KGRKR232NLS228MDGEA232, and IGFBP-3 Δ185-264, respectively (20) by NheI/HindIII and inserted into the cytomegalovirus (CMV) promoter–driven expression vector pX using SpeI/HindIII to generate pXIGFBP-3, pXIGFBP-3KED253-255RGD, and pXIGFBP-3 228KGRKR232NLS228MDGEA232. pXΔls-IGFBP-3 plasmids series: (a) To generate a new ATG start-site for the IGFBP-3 open reading frames without signal peptide (Δls), the oligonucleotide 5′-AATTCCATATGGGCGCGAGCTCGATATCA-3′ (MWG Biotech, Ebersberg, Germany) was inserted into the plasmid pUC19 digested with EcoRI and HindIII. The resulting plasmid is referred to pUC19-ATGnew. (b.1) To generate pXΔls-IGFBP-3, pXΔls-IGFBP-3KED253-255RGD, and pXΔls-IGFBP-3228KGRKR232NLS228MDGEA232, the plasmids pSF202, pSF210, and pSF211 were digested with HindIII and partially digested with SacI, generating fragments coding for IGFBP-3, IGFBP-3KED253-255RGD, and IGFBP-3228KGRKR232NLS228MDGEA232 without leader sequence. (b.2) To generate pXΔls-IGFBP-3Δ185-264, the plasmid pSF207 was digested with SacI and HindIII, generating the fragment containing the coding sequence for IGFBP-3Δ185-264 without leader sequence. (c) pUC19-ATGnew was linearized by SacI/HindIII, and the Δls-IGFBP-3 fragments generated in (b.1) and (b.2) (Δls-IGFBP-3, Δls-IGFBP-3KED253-255RGD, Δls-IGFBP-3228KGRKR232NLS228MDGEA232, and Δls-IGFBP-3Δ185-264) were inserted, generating the pUC19-ATGnew-Δls-IGFBP-3 plasmids. (d) All Δls-IGFBP-3 sequences were excised from the pUC19-ATGnew-Δls-IGFBP-3 vectors as EcoRI/HindIII fragments and inserted into pX, generating pXΔls-IGFBP-3, pXΔls-IGFBP-3KED253-255RGD, pXΔls-IGFBP-3228KGRKR232NLS228MDGEA232, and pXΔls-IGFBP-3Δ185-264. The plasmids pCIneoHis-Δls-IGFBP-3, pCI-neoHis-Δls-IGFBP-3KED253-255RGD, pCI-neoHis-Δls-IGFBP-3228KGRKR232NLS228MDGEA232, and pCI-neoHis-Δls-IGFBP-3Δ185-264 were generated by NdeI/XhoI digestion of the plasmids of the pXΔls-IGFBP-3 series. The isolated fragments coding for Δls-IGFBP-3, Δls-IGFBP-3KED253-255RGD, Δls-IGFBP-3228KGRKR232NLS228MDGEA232, and Δls-IGFBP-3Δ185-264 were inserted into the bacterial expression vector pET-19b (Novagen, Vienna, Austria), which contains the His-tag-sequence, generating pET-19bHis-Δls-IGFBP-3, pET-19bHis-Δls-IGFBP-3KED253-255RGD, pET-19bHis-Δls-IGFBP-3228KGRKR232NLS228MDGEA232, and pET-19bHis-Δls-IGFBP-3Δ185-264. Finally, the four His-Δls-IGFBP-3 open reading frames were transferred into pCI-neo (CMV-promoter driven mammalian expression vector; Promega, Vienna, Austria). The plasmid pCI-neo-Igκ-His-Δls-IGFBP-3 was constructed using pET-19bHis-Δls-IGFBP-3, which was linearized by NruI and digested partially with NcoI. A 6,000-bp fragment was isolated, and the oligonucleotide coding for the Igκ leader sequence (5′-TCTAGAATGGAGACAGACACACTCCTGCTATGGGTACTGCTGCTCTGGGTTCCAGGTTCCACTGGTGACGAATTCCC-3′, MWG Biotech) was inserted. The Igκ-His-Δls-IGFBP-3 open reading frame was amplified by PCR (forward primer, 5′-CTAGTCTAGAATGGAGACAGACACACTCCTGCT-3′; reverse primer, 5′-ATAGTTTAGCGGCCGCCTACTTGCTCTGCATGCTGTAGCA-3′; MWG Biotech) and inserted into the XbaI/NotI sites of pCI-neo.
Cell culture. U-2 OS cells were cultured in DMEM/10% FCS as described (17). The cell lines ts20b and H.38.5 (provided by Dr. H.L. Ozer, Department of Microbiology and Molecular Genetics, University of Medicine and Dentistry of New Jersey, New Jersey Medical School and UMDNJ-Graduate School of Biomedical Sciences, Newark, NJ) were maintained at 33°C in DMEM/5%CS as described (21).
Indirect immunofluorescence. In all indirect immunofluorescence experiments, the cells were transfected with CMV-promoter driven pX expression vectors for IGFBP-3 and variants thereof. Twenty-four hours after transient transfection, the cells were fixed using 4%PFA and permeabilized by 0.2% Triton X-100. After incubation with primary antibodies (polyclonal goat anti-IGFBP-3 antibodies; DSL, Sinsheim, Germany) and secondary antibodies (TRITC-conjugated anti-goat IgG, Dianova, Hamburg, Germany) and/or the DNA stains 4′,6-diamidino-2-phenylindole (DAPI) or Sytox Green (Molecular Probes, Göttingen, Germany), the samples were viewed using confocal microscopy as described (17).
Western blot analysis and SDS-PAGE system. Cells were extracted in 1.5× Laemmli sample buffer (22). Lysates were separated on a 12.5% SDS-PAGE using the SDS-PAGE System according to Laemmli (22) and transferred to a polyvinylidene difluoride membrane (NEN, Vienna, Austria) in transfer buffer (25 mmol/L Tris, 195 mmol/L glycine, 15% methanol). Nonspecific binding sites were blocked by incubating the membrane for 1 hour at room temperature in PBST [2.7 mmol/L KCl, 1.5 mmol/L KH2PO4, 137 mmol/L NaCl, Na2HPO4, 0.05% Tween 20 (pH 7.5)] containing 5% nonfat dry milk. Immunodetection was accomplished using primary antibodies as indicated and horseradish peroxidase–conjugated secondary antibodies with the enhanced chemiluminescence Western blotting detection system (NEN).
Protein half-life determination. U-2 OS cells were transiently transfected with pXIGFBP-3 expression vectors. The protein biosynthesis was blocked by incubation of the cells with 100 μg/mL fresh cycloheximide (Sigma, Vienna, Austria). After 20 minutes of preincubation, the cells were harvested at specific time intervals in 1.5× SDS Laemmli sample buffer. The IGFBP-3 expression levels were determined by Western blotting, using polyclonal goat antibodies to IGFBP-3 (DSL). The blot was developed using a chemiluminescence system, IGFBP-3 bands were scanned from low exposures and densitometrically analyzed. The half-life of the IGFBP-3 variants was calculated by half-logarithmic plotting of the intensity of the densitometrically analyzed bands against the incubation time. When indicated, the cells were preincubated for 30 minutes with 10 μg/mL Brefeldin A (Sigma) to block secretion or with LLnL (100 μmol/L) or epoxomicin (10 μmol/L) to block the proteasome.
In vivo ubiquitination assay. U-2 OS cells were transfected with the expression vectors pCI-neo-His-Δls-IGFBP-3, pCI-neo-His-Δls-IGFBP-3253KED255NLS253RGD255, pCI-neo-His-Δls-IGFBP-3228KGRKR232NLS228MDGEA232, pCI-neo-His-Δls-IGFBP-3Δ185-264, pCI-neo-Igκ-His-Δls-IGFBP-3, or pCI-neo, each together with pMT123 (23), an expression vector for HA1-tagged human ubiquitin. At 21 hours after transfection, cells were incubated for three hours with 100 μmol/L LLnL and harvested in 250 μL lysis buffer [1% SDS, 50 mmol/L Tris-HCl (pH 6.8)]. Samples were diluted to 25 mL in dilution buffer [50 mmol/L Tris-HCl (pH 7.4), 150 mmol/L NaCl, 0.2% NP40, 10% glycerin, 20 mmol/L imidazole, 0.2 mmol/L phenylmethylsulfonyl fluoride (PMSF), 0.1 mmol/L NaVO4, 10 mmol/L β-glycerophosphate, 1 mmol/L NaF]. Protein purification was done according to a standard His-tag purification protocol (Qiagen, Hilden, Germany). The eluates were separated on a SDS-PAGE, and Western blot analysis with polyclonal goat anti-IGFBP-3 antibodies and monoclonal mouse anti-HA1 antibodies (a gift from W. Nickel, Heidelberg, Germany) was done.
Apoptosis assay. U-2 OS cells were cultured in DMEM supplemented with 10% FCS in six-well plates until 60% confluence and transiently cotransfected with the plasmids pBB14 (24) coding for a membrane-bound green fluorescent protein (GFP) as marker of cells successfully transfected using the Effectene kit (Qiagen) and pX, pXIGFBP-3, pXΔls-IGFBP-3, and pXΔls-IGFBP-3228KGRKR232NLS228MDGEA232, respectively. The cells were grown in DMEM supplemented with 10% FCS for 16 hours. After trypsinization, one volume of FCS was added to inactivate the trypsin; the cells were washed twice in PBS, replated in DMEM without serum, and cultured for additional 72 hours. The cells were harvested by trypsinization and washed twice with PBS. Each sample was resuspended in 45 μL Annexin V binding buffer using the Annexin V-FITC Apoptosis kit II (BD PharMingen, San Diego, CA) and 5 μL Annexin V-Alexa Fluor 647 conjugate (Molecular Probes). After incubation for 15 minutes, 150 μL Annexin V binding buffer was added, and the cells positive for Annexin V and GFP were detected in a FACSCalibur flow cytometer (BD PharMingen). The broad-spectrum caspase inhibitor Z-VAD-FMK (40 μmol/L; Biomol, Hamburg, Germany) was used to analyze whether apoptosis is caspase dependent.
Cell fractionation. Cell pellets were lysed at 0°C in two packed cell volumes of lysis buffer [10 mmol/L HEPES-KOH (pH 7.5), 10 mmol/L KCl, 3 mmol/L MgCl2, 1 mmol/L EDTA, 0.2 mmol/L PMSF, 1 mmol/L DTT, 1 mmol/L NaF, 0.1 mmol/L NaVO3]. After addition of NP40 to a final concentration of 0.05% (v/v) and incubation for 10 minutes on ice, nuclei were pelleted by centrifugation at 1,000 × g, washed in lysis buffer, and extracted in high salt extraction buffer [10 mmol/L HEPES-KOH (pH 7.5), 500 mmol/L KCl, 5 mmol/L MgCl2, 0.5 mmol/L EDTA, 20% glycerin, 0.2 mmol/L PMSF, 1 mmol/L DTT, 1 mmol/L NaF, 0.1 mmol/L NaVO3]. Nuclear membrane debris were pelleted by centrifugation at 100,000 × g. Protein extracts from the cytoplasmic (1,000 × g supernatant) and nuclear fraction were analyzed by immunoblotting.
N-glycosidase F digestions. IGFBP-3 (500 ng) was digested with N-glycosidase F as described in the manufacturer's introduction (Roche, Vienna, Austria).
Targeting of IGFBP-3 to the cytosol and nucleus in U-2 OS cells. Although nuclear localization of IGFBP-3 has been described in several cell types (10–15), nuclear IGFBP-3 is hardly detectable in the dedifferentiated human U-2 OS osteosarcoma cells (17). Similarly, it has been shown that even if large quantities of recombinant IGFBP-3 are added to intact T47D breast carcinoma cells, only a very minor fraction of IGFBP-3 is finally detectable in the nuclei of these cells (13). These observations imply that IGFBP-3 uptake might be inefficient in the cells tested or, alternatively, that IGFBP-3 is taken up with reasonable efficiency, but intracellular accumulation of the protein is prevented by its degradation within the cell, which could in principle occur either in the cytoplasmic or nuclear compartment.
To address if the accumulation of IGFBP-3 in the nucleus and/or cytosol is prevented by its rapid degradation, we designed an experiment where IGFBP-3 is quantitatively directed to the nuclear or cytosolic compartment, respectively, and hence excluded from the secretory pathway. To this end, the nucleotides coding for the signal peptide were deleted from the IGFBP-3 cDNA to generate Δls-IGFBP-3 (Fig. 1A), a protein of otherwise unchanged amino acid sequence that lacks the leader sequence required for endoplasmic reticulum (ER) localization and subsequent secretion. The subcellular localization of Δls-IGFBP-3 in U-2 OS cells was studied by indirect immunofluorescence and confocal microscopy. When Δls-IGFBP-3 was expressed in these cells, which have very low levels of endogenous IGFBP-3 (Fig. 2A, empty vector), this protein was quantitatively retained within the cells (Fig. 2B,, top) and found to accumulate predominantly within the nucleus, as corroborated by DAPI-costaining (Fig. 2A, Δls-IGFBP-3). This result is in line with the finding that intracellular IGFBP-3 can be actively transported into the nucleus in intact cells (13). Using permeabilized cells, it was shown before that nuclear uptake of IGFBP-3 depends on a NLS sequence in the COOH-terminal half of the protein and the nuclear import factor importin-β (10). To test if this sequence is functional also in intact cells, we employed IGFBP-3 mutants, in which the signal peptide had been deleted (Fig. 1A). The mutation of five amino acids in the presumptive NLS, as in construct Δls-IGFBP-3228KGRKR232NLS228MDGEA232 (Fig. 1A), led to a loss of nuclear accumulation of IGFBP-3 (Fig. 2A, Δls-IGFBP-3228KGRKR232NLS228MDGEA232), which instead was predominantly redirected into the cytosol. In contrast, a distinct alteration in the COOH-terminal adjacent acid region (KED to RGD), generating an integrin-binding site that failed to affect nuclear localization in permeabilized cells (13), as in construct Δls-IGFBP-3KED253-255RGD (Fig. 1A), did not affect nuclear localization of IGFBP-3 in the experimental system used here, as revealed by costaining with DAPI (Fig. 2A, Δls-IGFBP-3KED253-255RGD). These findings suggest that nuclear import of IGFBP-3 from the cytosol requires the NLS in living cells and nuclear import of IGFBP-3 does not seem to require secretion. The deletion of the leader sequence in the Δls-IGFBP-3 variants leads to complete loss of secretion for these proteins (Fig. 2B,, top), suggesting that these proteins can not be N-glycosylated in their central domain. To determine the glycosylation level, IGFBP-3 wild type was incubated with N-glycosidase F. We found that glycosidase-digested IGFBP-3 has the same molecular weight as Δls-IGFBP-3 (i.e., 32 kDa; Fig. 2B , bottom), the size of an unglycosylated IGFBP-3 full-length protein.
When the leader sequence containing IGFBP-3 cDNA (Fig. 1A) was ectopically expressed in U-2 OS cells, large amounts of secreted IGFBP-3 could be shown (Fig. 2B,, middle), but the accumulation of nuclear IGFBP-3 was not detected (Fig. 2C,, top), in agreement with previous findings (17). The leader peptide containing IGFBP-3 mutants IGFBP-3228KGRKR232NLS228MDGEA232 and IGFBP-3KED253255RGD (Fig. 1A) are both predominantly secreted (Fig. 2B,, middle), and the confocal immunofluorescence microscopy analysis shows that very little, if any, of these IGFBP-3 variants is detectable in the nucleus of the U-2 OS cells (Fig. 2C). To confirm these observations, nuclear and cytoplasmic extracts of IGFBP-3 wild type expressing U-2 OS cells were prepared. As expected, the nuclear protein lamin B was retrieved in the nuclear fraction, whereas the Golgi marker GM130 and the ER-specific protein calnexin were found exclusively in the cytoplasmic fraction, suggesting that a clean separation of nuclear and cytoplasmic proteins had been achieved (Fig. 2D). In these experiments, a very little proportion of IGFBP-3 is retained in the nucleus, suggesting that nuclear IGFBP-3 is scarcely detectable in U-2 OS cells. Together, these results clearly establish that IGFBP-3 can be forced to accumulate intracellularly by removing the signal peptide from the IGFBP-3 cDNA. Hence, the IGFBP-3 variants described here provide a suitable experimental system to assess the metabolic stability of IGFBP-3 in defined cellular compartments.
Reduced steady-state level of nuclear IGFBP-3. We first determined the expression level of the IGFBP-3 variants in the specific subcellular compartments. For quantification, constant amounts of extracts from cells expressing the IGFBP-3s were run on a SDS-PAGE, and the relative abundance of the different IGFBP-3 variants was determined by Western blotting, scanning of the bands, and densitometric analysis (Fig. 3A). The signals obtained for variants devoid of a leader sequence are consistently much weaker than the signals obtained with the signal peptide containing variants, although a large proportion of the latter molecules are permanently secreted (see Fig. 2B,, middle) and hence absent from the cellular extract. All IGFBP-3 expression vectors used here gave rise to a similar level of IGFBP-3 mRNA (data not shown), suggesting that the expression of ectopic IGFBP-3 is not substantially regulated at the transcriptional level under these conditions. These findings suggest that the intracellular environment has an influence on the stability of IGFBP-3 and its variants. Strikingly, the decreased abundance of a given Δls-IGFBP-3 variant (compared with the corresponding variant containing the signal peptide) is much more pronounced for variants that are retained in the nucleus (Fig. 3A,, lane 2 versus lane 3 and lane 4 versus lane 5) relative to the IGFBP-3 variant that accumulates in the cytoplasm (Fig. 3A , lane 6 versus lane 7). The differences in the steady-state levels of the IGFBP-3 variants could be due to altered translation efficiency or, alternatively, to a difference in metabolic stability. Collectively, the data suggest that nuclear and, to a minor extent, cytosolic IGFBP-3 is target for posttranscriptional regulation that is not observed for those IGFBP-3 molecules that are confined to the secretory pathway and extracellular space, respectively.
Decreased abundance of nuclear IGFBP-3 is caused by decreased protein stability. To further analyze the mechanism underlying the decrease in the steady-state level of nuclear and cyctoplasmic IGFBP-3, we determined the half-lives of the IGFBP-3 variants and investigated if their abundance can be increased by specific proteasome inhibitors (Fig. 3B-D). To do this, U-2 OS cells were transiently transfected with the corresponding IGFBP-3 expression vectors. First, we determined the metabolic half-life of luminal IGFBP-3. To rule out confounding factors due to IGFBP-3 secretion, the cells were preincubated for 30 minutes with Brefeldin A, which blocks protein transport in the secretory pathway from the ER to the Golgi complex. To rule out any influence of altered protein synthesis on the observed IGFBP-3 steady-state level and metabolic half-life, cellular protein biosynthesis was blocked by addition of the 80 S ribosome inhibitor cycloheximide. Cells were harvested and lysed, and IGFBP-3 expression was analyzed by SDS-PAGE and Western blotting using polyclonal anti-IGFBP-3 antibodies (Fig. 3B,, top). IGFBP-3 bands were densitometrically analyzed. The half-life of the IGFBP-3 variants was calculated by half-logarithmic plotting of the intensity of the bands against the incubation time. A plot of the means from three independent experiments is shown in Fig. 3B (bottom). Under such conditions, the half-life of the IGFBP-3 wild type protein was ∼109.51 ± 8.36 minutes (n = 3; Fig. 3B,, bottom), suggesting that IGFBP-3 is a rather stable protein when retained in the secretory pathway. Next, we determined the half-life of Δls-IGFBP-3, which is predominantly nuclear and not secreted (see Fig. 2A and B). When U-2 OS cells were transiently transfected with an expression vector for Δls-IGFBP-3, the metabolic half-life of nuclear IGFBP-3 was determined to be 46.4 ± 5.3 minutes (n = 4; Fig. 3C), irrespective of the presence of Brefeldin A (data not shown). This finding shows that the stability of nuclear IGFBP-3 is strongly decreased compared with the half-life of vesicular IGFBP-3. The half-life of the cytosolic point mutant Δls-IGFBP-3228KGRKR232NLS228MDGEA232 was 78.3 ± 9.6 minutes (n = 4; P < 0.001 relative to Δls-IGFBP-3; P < 0.001 relative to IGFBP-3; Fig. 3D), which is significantly higher than the value determined for nuclear IGFBP-3 but significantly lower relative to luminal IGFBP-3. These findings suggest that proteolytic degradation of intracellular IGFBP-3 not only occurs preferentially in the nucleus but also, to a lower extent, in the cytosol. These results are in agreement with the steady-state measurements described above (Fig. 3A). To determine the role of the proteasome for IGFBP-3 proteolysis, the stability of IGFBP-3 was also analyzed in the presence of two different proteasome inhibitors LLnL and epoxomicin. Although the stability of nuclear IGFBP-3 was strongly increased by both proteasome inhibitors (Fig. 3C), cytosolic IGFBP-3 was stabilized to a lesser extend (Fig. 3D). Luminal IGFBP-3 was also stabilized to some extent by the proteasome inhibitors (Fig. 3B), as shown previously (17). These results suggest that proteasome-dependent proteolysis considerably accounts for the destruction of both nuclear and cytoplasmic IGFBP-3.
Proteolysis of intracellular IGFBP-3 depends on an active ubiquitin-activating enzyme E1. Degradation of cellular proteins by the proteasome frequently depends on the modification of a target protein by polyubiquitination. However, there are precedents for proteasomal degradation that do not depend on target polyubiquitination (25), and a role for polyubiquitination in the degradation of IGFBP-3 has not been described before. To address this question, we used the mouse BALB/c 3T3 cell line ts20b, which expresses a temperature-sensitive mutant of the ubiquitin-activating enzyme E1 (21). E1 is invariably required for all intracellular ubiquitination reactions, and target proteins of the ubiquitin-proteasome pathway accumulate when these cells are shifted to the restrictive temperature because inactivation of the E1 enzyme blocks ubiquitination. In a control cell line (H.38.5), generated by reintroduction of sequences encoding the wild-type human ubiquitin-activating enzyme E1 into the genome of ts20b cells, the temperature-sensitive defect is corrected; therefore, the accumulation of proteasomal target proteins is prevented. We compared the stability of transiently expressed wild-type IGFBP-3 in both cell lines at permissive and restrictive temperatures (Fig. 4A). The steady-state level of IGFBP-3, as analyzed by Western blotting, was rather low in ts20b cells at permissive temperature (33°C; Fig. 4A,, left). After the shift to the restrictive temperature (39°C), we obtained a significant increase in the IGFBP-3 expression level. In H.38.5 cells, which overexpress the wild-type E1 enzyme, the IGFBP-3 levels were low at both temperatures (Fig. 4A,, right). To analyze whether nuclear IGFBP-3 contributes to the partial increase in the IGFBP-3 level after inactivation of the E1 enzyme, the nuclear Δls-IGFBP-3 was transiently expressed in ts20b and H.38.5 cells. Δls-IGFBP-3 was neither detectable in the H.38.5 cells nor in the ts20b cells at 33°C but well detectable in ts20b cells at 39°C (Fig. 4B). These findings suggest that the degradation of IGFBP-3, and especially of nuclear IGFBP-3, depends on an active ubiquitin-activating enzyme E1.
Polyubiquitination of nuclear IGFBP-3. To establish IGFBP-3 as direct target for ubiquitin-mediated degradation, we set out to show polyubiquitinated IGFBP-3 intermediates, using an in vivo polyubiquitination assay (23). U-2 OS cells were transiently cotransfected with expression vectors for nuclear His-tagged IGFBP-3 (Fig. 1B) and HA1-tagged ubiquitin, incubated with LLnL and lysed in a stabilizing buffer containing 1% SDS. The His-tagged IGFBP-3 protein and its derivatives were purified from the lysates, normalized for the presence of identical amounts of IGFBP-3 by Western blot using anti-IGFBP-3 polyclonal sera (Fig. 5, right), and analyzed for polyubiquitination by Western blotting using anti-HA1 antibodies (Fig. 5, left). Under these conditions, we detected abundant higher molecular weight ubiquitin bands for nuclear IGFBP-3 (Fig. 5, left), showing that polyubiquitination of IGFBP-3 occurs in the nucleus in living cells. This is in agreement with the low steady-state level (Fig. 3A), short half-life (Fig. 3C), and stabilization of nuclear IGFBP-3 by the proteasome inhibitors (Fig. 3C). A similar high polyubiquitination grade was found for the nuclear mutant Δls-IGFBP-3KED253-255RGD (compare Fig. 5, left, lane 2 versus lane 3), underlining that nuclear IGFBP-3 is efficiently targeted to ubiquitin/proteasomal degradation. The grade of multiubiquitination was strongly reduced in the COOH-terminal point mutant IGFBP-3228KGRKR232NLS228MDGEA232 (Fig. 5, left), and this correlates with the increased metabolic stability of this predominantly cytosolic mutant relative to nuclear IGFBP-3 (compare Fig. 3D versus C). The low multiubiquitination grade of IGFBP-3228KGRKR232NLS228MDGEA232 correlated with its removal from the nucleus and conceivably results from its physical separation from the nuclear ubiquitinating machinery. The residual ubiquitination observed for this variant is probably due to inefficient exclusion from the nucleus because ∼10% to 20% of this protein still localizes to the nucleus, as shown by confocal microscopy (Fig. 2A). Alternatively, one can not fully rule out that polyubiquitination of IGFBP-3 also occurs in the cytosol, albeit with strongly reduced efficiency.
In most instances, covalent attachment of the polyubiquitin chains to the target protein requires lysine residues (19). IGFBP-3 contains a total of 19 lysine residues, 7 of which are located in the nonconserved central region and 12 in the conserved COOH-terminal domain of the mature protein (see Fig. 1), and each of these lysine residues could be a potential anchor for polyubiquitination. To determine the role of the COOH-terminal lysine residues for multiubiquitination, we constructed an additional mutant, Δls-IGFBP-3Δ185-264 (Fig. 1B), in which the entire conserved COOH-terminal domain, including 12 lysine residues, is deleted. As expected, this IGFBP-3 variant, when expressed without leader peptide sequence, localizes predominantly to the cytosol (data not shown), because it lacks the entire NLS. Multiubiquitination was very low for Δls-IGFBP-3Δ185-264 (Fig. 5, left), amounting to <5% relative to nuclear IGFBP-3 (Fig. 5, left). This correlates with an increased metabolic stability of Δls-IGFBP-3Δ185-264, whose half-life was determined to 142.6 ± 16.4 minutes (n = 3; P < 0.0001 relative to Δls-IGFBP-3). For the other cytosolic variant Δls-IGFBP-3228KGRKR232NLS228MDGEA232, in which only two lysine residues at position 228 and 231 in the conserved COOH terminus are mutated, the multiubiquitination grade is significantly higher and the half-life twice as much shorter, suggesting that in fact, multiple lysine residues in IGFBP-3 function as polyubiquitin anchors. However, because the polyubiquitination grade of both cytosolic IGFBP-3 variants is strongly reduced and their half-lives are clearly longer relative to nuclear IGFBP-3, we hypothesize that efficient polyubiquitination of IGFBP-3 requires its nuclear localization.
To investigate if lumenal expressed IGFBP-3 is multiubiquitinated, we constructed a secreted His-tagged form of IGFBP-3, Igκ-His-IGFBP-3, which contains an IgK-leader sequence and His-tag (Fig. 1B). IgK-His-IGFBP-3 was expressed in U-2 OS cells, and as expected, this variant was very efficiently imported into the lumen of the secretory pathway and secreted (data not shown). Very little, if any lumenal expressed IGFBP-3 was found multiubiquitinated (Fig. 5, left). Although the expression level of luminal IgK-His-IGFBP-3 is higher relative to the nuclear expressed variant, it was found at least 20 times less multiubiquitinated (Fig. 5), underlining that ubiquitin-dependent proteolysis of IGFBP-3 predominantly occurs in the nucleus.
Nuclear IGFBP-3 induces apoptotic cell death in U-2 OS cells. The expression of IGFBP-3 caused a significant increase in the percentage of Annexin V positive cells under serum deprivation, relative to mock-transfected cells (Fig. 6, left), suggesting that transiently, cellular expressed IGFBP-3 can induce apoptosis in U-2 OS cells. When the predominantly nuclear Δls-IGFBP-3 (Fig. 2A) was transiently expressed in the U-2 OS cells under the same conditions, we found a significant increase in the percentage of cells undergoing apoptosis, relative to the mock-transfected cells (Fig. 6, left). The proportion of apoptotic cells was similar to the results obtained with IGFBP-3 wild type, suggesting that IGFBP-3 can readily induce apoptosis in osteosarcoma cells, when restricted to the nucleus. We found that the transient expression of the predominantly cytosolic mutant Δls-IGFBP-3228KGRKR232NLS228MDGEA232 led also to a significant increase in the percentage of apoptotic cells (Fig. 6, left). Because Δls-IGFBP-3228KGRKR232NLS228MDGEA232 (Fig. 2A) is partially nuclear, it can, however, not be excluded that this apoptotic signal originates from the nuclear fraction of this protein. The proapoptotic activity of the IGFBP-3 variants could be inhibited by the addition of the specific caspase inhibitor z-VAD-FMK (Fig. 6, right). These findings suggest that apoptosis induced by secreted and nuclear IGFBP-3 depends on the activity of caspases, as recently shown for secreted IGFBP-3 in MCF-7 breast cancer cells (9).
In this study, we show for the first time that the abundance of nuclear IGFBP-3 is regulated by direct multiubiquitination and proteasome-dependent degradation in U-2 OS cells. If ectopic IGFBP-3 is targeted to the nucleus in these cells, this results in apoptotic cell death, suggesting that the accumulation of nuclear IGFBP-3 is not compatible with osteosarcoma cell survival. These results suggest that destruction of nuclear IGFBP-3 by the ubiquitin/proteasome pathway may contribute to down-regulation of the apoptotic response to IGFBP-3 in bone cancer cells, consistent with the model that IGFBP-3 plays an important role in tumor suppression (1).
Functions of extracellular and intracellular IGFBP-3. IGFBP-3 has been shown to control the bioavailability of IGFs in the circulation and in the cellular microenvironment. Several studies suggest that IGFBP-3 has additional IGF-independent effects, including its ability to trigger apoptosis (1). Within a cell, IGFBP-3 has been found in the secretory compartments before its excretion (1). In addition, it has been shown that extracellular added IGFBP-3 can be internalized in T47D breast cancer cells (13), and that IGFBP-3 secreting PC-3 prostate cancer cells can reinternalize the protein, which is subsequently transported into the nucleus (14). This process has been shown to involve the interaction of IGFBP-3 with transferrin/transferrin receptor complexes and caveolin-1 (14). Nuclear localization of IGFBP-5, closely related to IGFBP-3 both in sequence and domain structure, has been observed in osteosarcoma cells (26), suggesting that the pathway for nuclear transport is intact in these cells. Whereas IGFBP-3 is abundantly expressed in human osteoblasts (3), no data are available about its subcellular localization and intracellular functions in these cells. We show here that endogenous nuclear IGFBP-3 is detectable only in very low level, if any, in U-2 OS osteosarcoma cells. Its low nuclear abundance may be, at least in part, caused by its rapid proteolytic turnover in this cellular compartment.
IGFBP-3 as inducer of apoptosis and tumor suppressor. The gene coding for IGFBP-3 is induced by the tumor suppressor p53 (4), and IGFBP-3 was identified as a mediator of p53-dependent apoptosis in colon carcinoma cells (27). Moreover, the IGFBP-3 expression in cancer cells is induced by the tumor suppressor PTEN (28) and several proapoptotic and growth inhibitory factors, like transforming growth factor-β (5) and retinoic acid (29). These observations suggest that IGFBP-3 can act as tumor suppressor, presumably via its ability to halt proliferation and induce apoptosis in cancer cells. We show here that nuclear IGFBP-3 can induce apoptosis in osteosarcoma cells. The broad-spectrum caspase inhibitor Z-VAD-FMK abolished the induction of apoptosis by nuclear IGFBP-3, suggesting that the antiproliferative effect of nuclear IGFBP-3 results mainly from the induction of caspase-dependent apoptosis, similar to recent findings in breast cancer cells (9). Our finding that nuclear IGFBP-3 can induce apoptosis is in line with one previous study showing that IGFBP-3 interacts with the nuclear receptor retinoid X receptor-α (30), an interaction that was found essential for mediating the effects of IGFBP-3-induced apoptosis in prostate cancer cells (30). Both studies support the notion that nuclear IGFBP-3 functions in the induction of programmed cell death.
Potential physiologic consequences of extracellular and nuclear IGFBP-3-proteolysis. Many extracellular IGFBP-3-proteases, which function to release IGFs from IGFBP-3/IGF complexes by the destruction of extracellular IGFBP-3 have been described (16). The existence of several extracellular IGFBP-3 proteases underlines the complexity and importance of the protease systems, which regulate the turnover of extracellular IGFBP-3 and hence the bioavailability of IGFs. High-affinity interactions between IGFs and IGFBP-3 antagonize the binding of IGF to the type 1 IGF receptor or stabilize IGF. Thus, proteolysis of extracellular IGFBP-3 directly modulates the first step in IGF receptor signaling and thereby modulates cell survival, proliferation (1), and life span (31). Little is known about the regulation of intracellular IGFBP-3 by proteolysis. Evidence was presented, which suggests that the lysosomal cysteine protease cathepsin L may act as endosomal and lysosomal IGFBP-3 degrading enzyme (32) and cathepsin D as lysosomal IGFBP-3 protease (33), suggesting that intracellular degradation of IGFBP-3 occurs also along the endocytic pathways and in lysosomes. We now show that intracellular IGFBP-3 is regulated by ubiquitin/proteasome–dependent degradation, but the molecular mechanisms underlying its proteasome-mediated destabilization are not understood. According to the current model, proteasomes are abundantly found in both nuclei and cytosol of eukaryotic cells (34) but virtually absent from the lumen of membranous organelles such as the ER (35), and this raises the question of how IGFBP-3 can be degraded in a proteasome-dependent fashion. A possible solution to this paradox comes from the observations that extracellular IGFBP-3 can be taken up by cells, and this is correlated with nuclear localization of exogenously added IGFBP-3 (10–15). These findings suggest a possible scenario where IGFBP-3 might be transiently exposed to cytosolic and/or nuclear proteasomes, which might then be able to degrade it, in response to a signal(s) that is presently unknown. We found that mainly nuclear IGFBP-3 was targeted by ubiquitin/proteasome–dependent destruction; however, polyubiquitination was very low for IGFBP-3 expressed in the lumen of the membranous organelles. These findings suggest that ubiquitin/proteasome–dependent destruction of nuclear IGFBP-3 may play an important role for the control of cell proliferation and apoptosis by IGFBP-3 (36).
Role of polyubiquitination for the proteolysis of intracellular IGFBP-3. Polyubiquitination involves three types of enzymes: E1, which activates ubiquitin, the ubiquitin-conjugating E2 enzymes and the E3 ligases, which are involved in the specific recognition of substrates for ubiquitination (19). Evidence for the involvement of the ubiquitin/proteasome system in the regulation of the IGFBP-3 protein level was obtained through the use of ts20b cells, containing a temperature-sensitive mutant of the E1 ubiquitin-activating enzyme (22). When ts20b cells were cultured at the permissive temperature, the detected IGFBP-3 levels were low, whereas a shift to the restrictive temperature produced inactivation of the E1 ubiquitin-activating enzyme, and this led to a significant accumulation of IGFBP-3. This data, together with the demonstration that nuclear IGFBP-3 can efficiently form covalent higher molecular weight complexes with polyubiquitin chains in living cells, formally establishes nuclear IGFBP-3 as a direct target for ubiquitination/proteasome–dependent destruction. This is underlined by its stabilization through LLnL, a tripeptidyl aldehyde 26S proteasome inhibitor (37), and epoxomicin, a member of the epoxy-β-aminoketone group, which selectively inhibits the 26S proteasomal chymotrypsin-like activity (38). To distinguish between the mechanisms underlying the recognition and degradation of IGFBP-3, it will be important to analyze which out of the 19 lysine residues in IGFBP-3 are essential for the anchoring of polyubiquitin chains. The finding that the multiubiquitination grade for Δls-IGFBP-3228KGRKR232NLS228MDGEA232, in which two lysine residues in the conserved COOH terminus are mutated, is significantly higher as in Δls-IGFBP-3Δ185-264, in which 12 COOH-terminal lysine residues are deleted suggest that multiple lysine residues in IGFBP-3 function as polyubiquitin anchors. More work is necessary to precisely determine the structural domain important for the recognition of IGFBP-3 by the ubiquitinating machinery and to identify the cellular E2 and E3 ubiquitination enzymes specific for IGFBP-3.
Note: F. Santer and N. Bacher contributed equally to this work.
Grant support: Austrian Science Funds/FWF projects P15383 (W. Zwerschke), 14288 (P. Jansen-Dürr), and SFB021 (P. Jansen-Dürr); Austrian Cancer Society-Tyrol (W. Zwerschke); European Union INKA project LSHC-CT-2005-018704 (W. Zwerschke), CELLAGE project QLK6-CT-2001-00616 (P. Jansen-Dürr).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank L.W. Enquist for the plasmid pBB14.