Intracellular Association of a Mutant Insulin-like Growth Factor Receptor with Endogenous Receptors¹

The IGF-IR⁴ activated by its ligands is a powerful inhibitor of apoptosis induced by a number of agents (1, 2, 3). The variety of the procedures used to induce apoptosis suggests that the IGF-IR may have a widespread antiapoptotic effect. Impairment of IGF-IR function, by either antisense strategies (4, 5, 6, 7, 8, 9, 10, 11), dominant negative mutants (12, 13), or triple-helix formation (14), causes apoptosis of tumor cells in vitro and in vivo and abrogation of tumor growth and metastases in experimental animals (15, 16, 17). Targeting of the IGF-IR may therefore be of potential interest in the treatment of human cancer (18, 19), especially in view of the fact that when the IGF-IR function is impaired or abrogated, normal cells seem to be less affected than tumor cells (20).

Among the various strategies used to down-regulate IGF-IR function, the use of dominant negative mutants has received scarce attention. A number of mutants of the IGF-IR have been described that act as dominant negative mutants, but some of these, although they may inhibit colony formation in soft agar, are not effective against tumor cells in vivo (21). Two dominant negative mutants of the IGF-IR have been shown to be effective against rodent and/or human tumors in experimental animals: (a) 952/STOP (12, 22), which includes the α subunit and part of the β subunit up to the juxtamembrane region, thus being anchored to the membrane; and (b) 486/STOP (13), which has a frameshift mutation resulting in a stop codon at residue 486 (not including the signal peptide). Because 486/STOP does not have a transmembrane domain, it can be secreted in the medium, where it causes apoptosis, inhibits tumor growth, and prevents human breast cancer metastases, also in a nude mouse model (13, 17, 23).

Although the proapoptotic function of 486/STOP is well established, very little is known about its mechanism of action. 486/STOP has a bystander effect in experimental tumors (23), and CM containing 486/STOP has an inhibitory effect on cells in culture (13, 17). An obvious explanation would be that it competes with the endogenous IGF-IR for ligands, as dominant negative mutants of growth factor receptors are wanted to do. However, there are theoretical considerations as well as experimental data (see below), indicating that the mechanism of action of 486/STOP may be complex. In the first place, 486/STOP is also found in lysates of cells (23), and, in some cases, it is more abundant in cell lysates than in the medium. Secondly, the secretion of 486/STOP in the medium is highest in R− cells (13), suggesting that in these cells, the absence of an IGF-IR (24) allows increased secretion in the medium. Thus, it is possible that one of the mechanisms of action of 486/STOP may be intracellular. On the other hand, its bystander effect in experimental tumors (23) also indicates an effect from outside the cell.

We show in this study that 486/STOP associates intracellularly with endogenous IGF-IR. In addition, cells treated with CM containing 486/STOP take up the polypeptide into the cell, where it again interacts with endogenous IGF-IR. The presence of an IGF-IR is necessary for the apoptotic effect of 486/STOP, despite the fact that only a fraction of the endogenous receptors actually associates with 486/STOP. These results help us not only to understand the mechanism of action of 486/STOP but also suggest that dominant negative mutants of the IGF-IR and possibly of other growth factor receptors (25) may have complex mechanisms of inhibition.

The retroviral vectors MSCV.pac and MSCV.neoEB were kindly provided by Dr. R. G. Hawley (University of Toronto, Toronto, Canada) and are described elsewhere (26). The SIN version of the MSCV.pac retroviral vector was engineered by deleting 299 bp in the 3′-LTR. The deleted region contains sequences that encode for the enhancer and other promoter functions. The retroviral vector has an active 5′-LTR in the proviral form. The 5′-LTR is then inactivated during the process of cell infection (27). The Drosophila melanogaster HSP70 promoter was excised from plasmid pSpTK⁺ (28) and inserted into the polylinker of the SIN MSCV.pac retroviral vector, which is downstream of the packaging signal (Ψ; Ref. 26). The soluble IGF-IR was engineered by fusing in frame the IGF-IR cDNA encoding for the first 516 amino acid residues (486 residues, after the removal of the signal peptide) with the FLAG sequence (Kodak) at the 3′-end. The numbering of the IGF-IR amino acid residues follows the system proposed by Ullrich et al. (29), in which the 30 amino acids of the signal peptide are omitted. The IGF-IR sequence fused in frame with the FLAG epitope at the 3′-end was produced by PCR. The detailed methodology for the construction of the retroviral vector expressing 486/STOP is described by Romano et al. (30). The SIN retroviral vector contains the Drosophila HSP70 promoter driving the expression of the 486/FLAG gene and was named pGR228. The polypeptide product of this construct is the same as that of the original plasmid (13, 23), except for the presence of the FLAG epitope. To avoid confusion, the term 486/STOP is used throughout this article.

The parental cell lines used were MEFs and include p6 cells (31), R− cells (24), R600 cells (32), R−/v-src cells (33) and R−/IR cells (34). All cell lines were cultured in DMEM (Life Technologies, Inc., Grand Island, NY) supplemented with 1% BEM vitamins, 50 units/ml penicillin, 50 ng/ml streptomycin, and 10% FBS (Sigma, St. Louis, MO). Semiconfluent cultures were transduced with retroviral vectors according to the methodology described previously (30). R−/vsrc cells were stably transfected with pIGFIRsol plasmid (13, 23) or with pCVN empty vector in the presence of Transfectam Reagent, as recommended by the manufacturer’s (Promega) protocol. In all other experiments, the cells were transduced with the new retroviral vector expressing 486/STOP with the FLAG tag. Selection of neomycin-resistant cells was carried out with 1 mg/ml of G418 (Life Technologies, Inc.).

The methodology described previously was followed (23). Briefly, R−/v-src cells expressing either the 486/STOP (R−vsrc/486/STOP) or the empty vector (R−vsrc/E) were plated at 2 × 10³ cells/well in EMEM containing 10% FBS and 0.2% agarose (with 0.4% agarose underlay). The number of colonies larger than 125 μm in diameter was determined at 2 weeks after plating.

To compare growth in monolayer of different clones of R−/v-src cells expressing either the 486/STOP (R−vsrc/486/STOP) or the empty vector (R−vsrc/E), cells were plated at 5 × 10⁴ cells/35-mm dish and cultured for 120 h in the presence of 10% FBS. Cell number was determined by counting cells in a bright light hemocytometer at time 0, i.e., 24 h after plating, and at 72 and 120 h after time 0.

Protein levels for the IGF-IR (α and β subunits) and for 486/STOP were examined in all cell lines used in this study. Cells were lysed on ice with 400 μl of lysis buffer [50 mm HEPES (pH 7.5), 150 mm NaCl, 1.5 mm MgCl₂, 1 mm EGTA, 10% glycerol, 1% Triton X-100, 1% phenylmethylsulfonyl fluoride, 0.2 mm sodium orthovanadate, and 1% aprotinin]. Protein concentration was determined by a Bio-Rad Protein Assay (Bio-Rad, Hercules, CA). Samples of 50 μg of total proteins were separated on a 4–15% gradient SDS-PAGE (Bio-Rad) and transferred into nitrocellulose membranes. Blots were blocked with 5% nonfat dry milk in 10 mm Tris-HCl (pH 7.5), 150 mm NaCl, and 0.1% Tween 20 and probed with rabbit anti-IGF-IRα (N-20), anti-IGF-IRβ (C-20), and anti-insulin receptor antibodies (C-19; Santa Cruz Biotechnology, Santa Cruz, CA). The secondary antibody used was an antirabbit IgG conjugated to horseradish peroxidase (Santa Cruz Biotechnology), which was subsequently visualized by enhanced chemiluminescence detection reagents (Amersham). N-20 antibody was also used to detect 486/STOP protein because it recognizes amino acids 31–50 mapping at the NH₂ terminus of human IGF-IRα subunit. Finally, anti-FLAG-M2 agarose affinity gel (Sigma) was used to immunoprecipitate the 486/STOP/FLAG fusion protein from either cell lysates or conditioned media.

To collect CM containing the 486/STOP product, R−/486/STOP cells were allowed to reach 70% confluence in DMEM supplemented with 10% FBS and 1 μg/ml puromycin. Cells were washed three times with PBS and cultured in SFM at 39°C for an additional 24 h. The CM (486/CM) was collected, and aliquots were frozen at −80°C. C/CM was prepared in a similar way, by using R−/E/HSP cells, which contain empty retroviral vector, instead of R−/486/STOP cells. Collected aliquots of the CM were concentrated about 10-fold by the Centriprep 50 centrifugal filters (Amicon), and aliquots were used in both in vitro and in vivo experiments.

To determine these parameters, we essentially followed the methodology described by Reiss et al. (35). Briefly, quiescent cells were detached from a culture dish with 0.02% EDTA (disodium EDTA) and seeded on dishes coated with polyHEMA (Aldrich, Milwaukee, WI). Cells were seeded in SFM at 1 × 10⁴ cells/cm² and treated either with 10% FBS or 50 ng/ml IGF-I or left untreated. Twenty-four h later, cell suspensions were collected and dissociated with 0.25% trypsin, and cells were counted in a Brightline Hemocytometer. For the terminal deoxynucleotidyl transferase-mediated nick end labeling method to detect apoptotic cells, we also followed the description of Reiss et al. (35), which is based on the manufacturer’s instructions (Boehringer Mannheim). Results represent an average of three experiments and are expressed as a percentage of apoptotic cells.

To investigate the presence of the 486/STOP/FLAG fusion protein within cells, R− and R600 cells were incubated with 486/CM (see above) for 6 h. Cells were washed three times with PBS, fixed in 10% methanol-free formaldehyde, washed again, and stained with anti-FLAG FITC-conjugated antibody according to recommendations given by the manufacturer’s protocol (Sigma). Fluorescent images of cells were analyzed using a fluorescence microscope. Additionally, to ensure the intracellular localization of the 486/STOP/FLAG protein, a series of pictures from 0.5-μm-thick optical sections were taken, using confocal microscopy.

The effect of 486/STOP on tumor cells and its proapoptotic effect have been documented previously (13, 17, 23). This study focuses on its mechanism(s) of action. Although we have inserted the sequence of 486/STOP in another vector, the polypeptide product remains identical to the one described previously.

The first question we have addressed is whether 486/STOP is a specific inhibitor of cells expressing the IGF-IR or is only a polypeptide toxic to cells in general. A reasonable way to answer this question is to determine its effect on cells devoid of IGF-IR. We have developed in our laboratory a cell line of MEFs designated R− cells (24, 36), generated by a 3T3-like protocol from mouse embryos with a targeted disruption of the IGF-IR genes (37, 38). R− cells, however, do not respond to IGF-I and do not form colonies in soft agar (24, 36); therefore, they cannot be tested for growth inhibition by 486/STOP. They cannot be tested in 10% serum, either, because serum contains growth factors (such as progranulin, for instance) that bypass the IGF-IR (25, 39). In a previous study, we described a cell line derived from R− cells by stable transfection with a plasmid expressing v-src (33). R−/v-src cells grow in monolayer cultures in SFM and form colonies in soft agar. If the product of 486/STOP were to be nonspecifically toxic to cells, it should inhibit the growth of R−/v-scr cells.

For this purpose, we transfected R−/v-src cells with a plasmid expressing 486/STOP, and we selected a number of clones. Fig. 1,A shows that we obtained four clones expressing 486/STOP. Three of the clones expressed substantial amounts of 486/STOP, whereas in clone 9, only a faint band could be detected. No band indicative of 486/STOP is detectable in two clones (C1 and C2) of R−/v-scr cells transfected with the empty vector (see “Materials and Methods”). Four of these clones were tested for colony formation in soft agar, and the number of colonies formed after 2 weeks is given in Fig. 1,B. Expression of 486/STOP has no effect on R−/v-scr cells in terms of colony formation in soft agar. There are variations among the clones, but these variations seem to be more clone specific than due to the presence or absence of 486/STOP. The same clones were also tested for growth in monolayer in SFM. Fig. 1,C shows that the expression of 486/STOP has little effect on the growth of R−/v-scr cells. If anything, the clones expressing 486/STOP grow better than the clone transfected with the empty vector. The cell numbers of Fig. 1 were determined at 72 and 120 h, but the trend was already evident at 24 and 48 h (data not shown). Incidentally, the ease with which we obtained clones expressing substantial amounts of 486/STOP (see below) was in itself an indication that R−/v-src cells are not affected by its expression. These results show that the growth of cells devoid of IGF-IR is not affected by 486/STOP, indicating that the inhibitory effect of 486/STOP requires the presence of an IGF-IR.

Since the generation of 486/STOP, we had noticed that it was quite difficult to obtain clones expressing it in detectable amounts (13, 23). Although we could obtain many antibiotic-resistant clones, only a few expressed 486/STOP, and those that did often expressed it only in small amounts. In addition, we had noticed that the levels of 486/STOP polypeptide in the CM were always higher in R− cells (13) than in cells with endogenous IGF-IR. We hypothesized that R− cells were not inhibited by the expression of 486/STOP, a hypothesis confirmed by the above-described experiment with R−/v-src cells (Fig. 1). Experiments with clones that were antibiotic resistant but did not express 486/STOP indicated that expression of 486/STOP had been silenced because its expression was induced by treatment with 5-azacytidine (30). We thought that this silencing could be circumvented by transducing the cells with a retroviral vector in which 486/STOP would be under the control of an inducible promoter. For this purpose, we used a retroviral vector with a SIN LTR (27), where 486/STOP was under the control of the Drosophila HSP70 promoter (28, 40). To facilitate its recognition, 486/STOP in this retroviral vector also carried a FLAG epitope (see “Materials and Methods”). Fig. 2,A shows a diagram of this construct with the cloning sites for the antibiotic (pkg) and the 486/STOP plasmid. A more complete description of this construct and its inducibility has been given by Romano et al. (30). R− cells were transduced with this retroviral vector, and mixed populations were selected in G418. Lysates were prepared and immunoprecipitated with anti-FLAG antibody, and the blots were developed with an antibody to the α subunit of the IGF-IR. Fig. 2,B, Lanes 3 and 4, shows that the HSP70 promoter is indeed inducible because 486/STOP is strongly expressed at the temperature of 39°C, but not at the temperature of 34°C (Lanes 1 and 2). The bands of nonspecifically reacting proteins show that the amounts of proteins in the various lanes were comparable. The use of the construct diagrammed in Fig. 2 A has allowed us to obtain quickly mixed populations of cells that could be induced to express substantial amounts of 486/STOP by a simple shift in temperature. This new construct was designated as GR228, whereas the retroviral vector designated as GR219 is the empty retroviral vector used as control. In subsequent experiments, we used the GR228 retroviral vector for transduction. Whereas the plasmid is designated as GR228, its polypeptide product still retains the name of 486/STOP because it is identical to the one produced by the previously described plasmid (13).

In previous studies (13, 17, 23), 486/STOP could be detected in lysates from cells, although it was also detectable in the CM, especially that of R− cells (13). The presence of 486/STOP in cell lysates was confirmed in the experiments of Fig. 2. It raised the question of whether part of the inhibitory effect of 486/STOP could be due to an intracellular association with the endogenous IGF-IRs. For this purpose, we used p6 cells transduced with GR228 (see Fig. 3,A), incubated at either 34°C or 39°C. The lysates were immunoprecipitated with an anti-FLAG antibody, and the immunoprecipitates were subjected to a Western blot with antibodies to either the α (Fig. 3,A) or the β subunit (Fig. 3,B) of the IGF-IR. Fig. 3,A shows that the anti-FLAG antibody coprecipitates the IGF-IR, as evidenced by the bands for the α subunit and the proreceptor. These bands appear only at 39°C (Lane 1), together with the 486/STOP band, and are totally absent at 34°C (Lane 2), where the 486/STOP is also nondetectable. The other lanes (Lanes 4–6) are control lysates of p6 cells transfected with the empty vector (E) and parental p6 cells that were also immunoprecipitated with an anti-FLAG antibody. Fig. 3,B confirms that immunoprecipitation of 486/STOP with an anti-FLAG antibody coprecipitates the IGF-IR because two bands are detectable with an antibody to the β subunit of the IGF-IR. One band is the β subunit itself, and the upper band is the proreceptor. Again, the two bands are visible only in lysates of cells incubated at 39°C and are not detectable in immunoprecipitates of parental p6 cells (last two lanes) or p6 cells transfected with the empty retroviral vector (E). An additional experiment was carried out in R− cells expressing the insulin receptor, roughly in the same number as IGF-IR in p6 cells (34). These R−/IR cells were transduced with GR228, a mixed population was selected, and lysates were immunoprecipitated with an anti-FLAG antibody. No insulin receptor is detectable in these immunoprecipitates (Fig. 3 C). Protein amounts in each lane were monitored by the levels of proteins nonspecifically recognized by the antibody.

The fraction of IGF-IR that coprecipitates with 486/STOP is very small. Considering that p6 cells have about 5 × 10⁵ receptors/cell, our densitometric measurements of blots under different conditions indicate that the fraction of IGF-IR coprecipitating with 486/STOP is less than 5%. Clearly, it could be higher in cells expressing fewer receptors, but the fact remains that in p6 cells, the fraction is very low. However, 486/STOP causes apoptosis of p6 cells (see below). For the moment, we can state that these experiments show that 486/STOP coprecipitates in cell lysates with the endogenous IGF-IR (but not with the insulin receptor), indicating an intracellular association between the two.

We and others have reported previously (13, 17, 23) that 486/STOP is secreted into the medium. We had also observed that 486/STOP is more abundant in the CM of cells that do not express the IGF-IR (R− cells and R−/v-src cells) or express it poorly (LNCaP cells) than in cells with higher numbers of endogenous IGF-IR. It suggested that some of 486/STOP is trapped inside the cells in association with endogenous IGF-IR, as also indicated by the experiments in Fig. 3. If this hypothesis is correct, the ratio of 486/STOP levels in lysates should be higher than that in the medium in cells with abundant IGF-IR. Conversely, the levels of 486/STOP should be higher in the medium than in the lysates of cells without IGF-IR. This prediction was verified by the experiments in Fig. 4. Fig. 4 A shows that the amount of 486/STOP is much higher in lysates of p6 cells (5 × 10⁵ IGF-IRs/cell) than in lysates of R− cells. No 486/STOP is detectable at 34°C.

Another prediction can be made on the basis of our hypothesis. In R− cells, 486/STOP may be found in lysates after the first few hours of incubation at 39°C. However, if the cells are now downshifted to 34°C (which represses the expression of 486/STOP), the levels should decrease in the lysates and increase in the medium. This is indeed what we observed (Fig. 4 B). The amount of 486/STOP in the lysates declines after downshifting the cells to 34°C, whereas it increases in the medium. In the medium, incidentally, two bands of 486/STOP are detectable. At the present time, we do not have a satisfactory explanation for the presence of two bands in the medium.

We have mentioned above that a very small fraction of IGF-IR in p6 cells is coprecipitated by 486/STOP. However, as we show in Fig. 5, 486/STOP has a dramatic effect on p6 cells. We have tested this effect in two ways. In the first approach, we seeded p6 cells on polyHEMA plates (see “Materials and Methods”). On such plates, cells are denied attachment to a substratum and can undergo a form of apoptosis that has been called anoikis (41, 42). On polyHEMA plates, p6 cells expressing 486/STOP undergo apoptosis when incubated at 39°C (Fig. 5,A), as measured by the terminal deoxynucleotidyl transferase-mediated nick end labeling method. The same cells survive very well at 34°C (Fig. 5,C). Fig. 5, B and D, is a phase-contrast picture of the cells shown in Fig. 5, A and C. This effect was quantitated (with parental p6 cells as controls) in Fig. 5 E, where all cells were plated on polyHEMA plates. Parental p6 cells are not affected by the shift in temperature, and the fraction of apoptotic cells is very low in both cases. When p6 cells are stably transfected with plasmid GR228, the fraction of apoptotic cells is still low at 34°C, but it increases dramatically at 39°C. As an additional control, we incubated parental p6 cells with CM containing the 486/STOP polypeptide (see “Materials and Methods” for its preparation). The cells were incubated at 37°C and underwent apoptosis, almost as much as p6/486/STOP cells at 39°C. This dramatic effect of 486/STOP on p6 cells, despite the very low fraction of receptor coprecipitated, suggests an explanation that will be presented in the “Discussion.”

The association of 486/STOP with endogenous receptors within the cells could account for a direct inhibitory effect of 486/STOP in cells expressing IGF-IR. However, as already mentioned, previous evidence indicated that 486/STOP can also inhibit the growth of cells treated with CM from R− cells expressing 486/STOP. In trying to understand the mechanism by which 486/STOP in the medium inhibits cell growth, we tried several possibilities without success. At last, we asked whether the secreted 486/STOP could exhibit its bystander effect because it is taken up by the cells and returned to the intracellular pool. We took advantage of the fact that our construct of Fig. 2 has a FLAG epitope. We collected from R−/486/STOP cells a considerable amount of medium containing 486/STOP, and we incubated with it R600 cells. R600 cells are R− cells expressing about 3 × 10⁴ IGF-IRs/cell (32). Fig. 6,C shows that R600 cells treated with CM containing 486/STOP stain intensely with an anti-FLAG antibody, whereas they are completely negative when incubated with a C/CM (no 486/STOP, Fig. 6,A). As an additional control, we stained R− cells stably transduced with GR228 with an anti-FLAG antibody, and they are positive (Fig. 6,E). Fig. 6 G is a control antibody, which is negative.

Localization of 486/STOP to R600 cells could be due to internalization or to binding to the cell membrane. We therefore carried out confocal microscopy on the R600 cells of Fig. 6,C, and the results are shown in Fig. 7. It is clear that the FLAG-stained 486/STOP of the CM has been taken up and internalized by the cells. These experiments indicate that the 486/STOP in the medium is taken up avidly by cells and internalized, where it could act to form hybrid receptors, as in the case of cell lysates.

This is indeed the case, as shown in Fig. 8. In these experiments, p6 cells were incubated for 6 h with CM containing 486/STOP as described for Fig. 7. Cell lysates were made and immunoprecipitated with anti-FLAG antibody, and the blots were developed with an antibody to the β subunit of the IGF-IR. Lanes 2 and 3 show that the β subunit of the IGF-IR coprecipitates with an anti-FLAG antibody, and the amount precipitated is proportional to the amount of 486/STOP in the CM (dilution with medium). Lane 1 is a control, showing the position of the IGF-IR β subunit in Western blots of lysates from exponentially growing p6 cells. Lane 4 is another control, in which lysates of p6 cells not incubated with 486/STOP were immunoprecipitated with an anti-FLAG antibody. No IGF-IR is detectable.

The novel findings in these experiments can be summarized as follows: (a) 486/STOP is not toxic to cells that do not express IGF-IR; (b) 486/STOP interacts intracellularly with endogenous IGF-IR, but not with insulin receptors; (c) secreted 486/STOP is taken up by the cells and internalized, where it interacts again with the endogenous receptors; and (d) although the fraction of endogenous IGF-IR that associates with 486/STOP is small, the latter still sends a strong proapoptotic signal. These findings suggest a mechanism of action for 486/STOP, which would not be the typical dominant negative effect (competition with endogenous receptors for the same ligands).

It is generally agreed that the IGF-IR plays a substantial role in tumor growth. Its importance in human tumors has been summarized in two reviews by Macaulay (18) and by Khandwala et al. (19). The importance of the IGF-IR in the survival and growth of tumor cells is supported by several experimental findings, which include: (a) MEFs with a targeted disruption of the IGF-IR genes are refractory to transformation by a number of viral and cellular oncogenes (20); (b) down-regulation of the IGF-IR results in apoptosis (3), especially when the cells are in anchorage-independent conditions (1); and (c) conversely, the IGF-IR activated by its ligands protects from a variety of apoptotic injuries (3, 20). Targeting of the IGF-IR, resulting in apoptosis of tumor cells, could therefore be an attractive model for therapeutic approaches against cancer.

Targeting of the IGF-IR or its function has been achieved using four different modalities: (a) antisense strategies (4, 5, 6, 7, 8, 9, 10, 43); (b) dominant negative mutants of the receptor (12, 13, 17, 22, 23); (c) triple helix formation (14); and (d) antibodies to the extracellular region of the receptor (44, 45). All of these approaches have achieved a measure of success in experimental animals as well as in vitro. The use of dominant negative mutants of the IGF-IR has been successful in inducing apoptosis and inhibition of tumor growth and metastases in experimental animals (12, 13, 22, 23). Thus far, we have referred to 486/STOP as a dominant negative mutant of the IGF-IR, but previous results as well as the present findings suggest that 486/STOP may not act as a simple dominant negative but by a different and more complex mechanism.

The relevant findings that must be considered for an explanation of its mechanism of action are: (a) 486/STOP is effective only on cells that express endogenous IGF-IR (Ref. 13 and this study); (b) 486/STOP associates with endogenous receptors, as evidenced by immune coprecipitation; the interaction is specific for the IGF-IR because no association is detectable with the insulin receptor; (c) the interaction is intracellular and occurs whether 486/STOP is synthesized inside the cell or presented to the cells in the medium; (d) in p6 cells, the amount of endogenous IGF-IR coprecipitating with 486/STOP is only a small fraction (<5%); however, under these conditions, 486/STOP sends a strong proapoptotic signal (Refs. 13 and 17 and this study); and (e) IGF-IR signaling is not detectably impaired by the presence of 486/STOP (13). Clearly, these characteristics of 486/STOP are not those usually associated with a dominant negative mutant, whose expression should be in excess of the endogenous receptors and should cause impairment of signaling. We would like at this point to formulate a hypothesis on the unusual mechanism of action of 486/STOP. This hypothesis is based on the reports that the COOH terminus of the IGF-IR sends a proapoptotic signal, despite the fact that the whole receptor is strongly antiapoptotic. Thus, a truncated IGF-IR (deletion of the last 92 amino acids) is more effective in protecting cells from apoptosis than the wild-type receptor (46). Expression of the IGF-IR COOH terminus causes apoptosis in a variety of cell types (47, 48), and a peptide sequence from the COOH terminus of the IGF-IR also causes apoptosis when introduced into cells (35). We would like to propose that the IGF-IR associated with 486/STOP sends a proapoptotic signal independently of IGF-IR function. This signal could be due to a change in conformation of the endogenous receptor or alternatively to the fact that the endogenous receptor COOH terminus would not be part of a dimer. In this respect, both the COOH terminus peptide sequence and the products of the COOH terminus expression plasmids that cause apoptosis (47, 48) are present in the cells as free peptides (35).

Whether or not this explanation is correct, an important observation in these experiments is the coimmunoprecipitation of 486/STOP with the endogenous IGF-IR. In the first place, we show unequivocally that 486/STOP is secreted very slowly, and in fact, a substantial amount remains associated with the cell. Within the cells, 486/STOP interacts with the endogenous receptor in an exquisitely specific fashion. It does not interact with the insulin receptor (70% homology), although the number of insulin receptors in the cells we used was very high. At the moment, we can only speculate on the mechanism of this association. Perhaps the presence of disulfide isomerases in the endoplasmic reticulum (49) or the disulfide isomerase activity of IGF-I (50) could serve as the basis for additional studies. 486/STOP contains at least 29 cysteine residues (29). A similar mechanism was described for the platelet-derived growth factor β receptor by Ueno et al. (25). These authors reported that expression of a truncated receptor formed complexes with the wild-type platelet-derived growth factor receptors.

However, 486/STOP also acts from outside the cell. We already had a clue from previous studies (17, 23), but the present experiments are conclusive. 486/STOP in the medium is taken up by cells and internalized. The uptake is substantial, and 486/STOP is truly within the cell, as confocal microscopy clearly shows. Once it is back into the cell, 486/STOP interacts again with the endogenous IGF-IR.

The mutant receptor used in these experiments is essentially the same as the one used successfully by D’Ambrosio et al. (13), Dunn et al. (17), and Reiss et al. (23). The final product is the same, a polypeptide comprising the first 516 amino acids of the proreceptor, which become 486 amino acids after cleavage of the signal peptide. The construct, however, is different. First of all, instead of a plasmid, we used a retroviral vector; secondly, the insert is truncated directly after the codon for amino acid 486 and is completed by a FLAG epitope. Finally, with the exception of the experiment in Fig. 1, all other experiments were done with a construct in which the sequence of 486/STOP was under the control of the Drosophila HSP70 promoter (40). In this way, we have obtained a construct in which 486/STOP is inducible by a simple shift in temperature. This change is very beneficial because it is difficult to obtain clones expressing 486/STOP in substantial amounts (13, 23). We have evidence that the noninducible promoters of 486/STOP are often methylated because expression can be restored by treating the cells with 5-azacytidine (30). With the new construct, the cells can be passaged at 34°C, at which temperature expression of 486/STOP cannot be detected. The cells therefore grow very well in serum, and when they are shifted to 39°C, they express substantial amounts of 486/STOP. The originality and flexibility of this retroviral vector have already been discussed in a previous study (30).

In conclusion, we have made further progress in elucidating the mechanisms by which 486/STOP induces apoptosis and inhibits tumor growth in nude mice (13, 17, 23). The mechanism is somewhat unusual, and it seems to be based essentially on its intracellular interaction with endogenous IGF-IR. The ability to act from outside the cell in an efficient manner makes this mutant receptor an attractive model for additional studies on its mechanism and its therapeutic potential.