Nuclear translocation of insulin receptor substrate-1 by oncogenes and Igf-I. Effect on ribosomal RNA synthesis.

The insulin receptor substrate-1 (IRS-1) is one of the major substrates of both the insulin and IGF-I receptors and is generally localized in the cytosol/membrane fraction of the cell. We show here that a substantial fraction of IRS-1 is translocated to the nucleus in mouse embryo fibroblasts (MEF) expressing the simian virus 40 T antigen. Nuclear translocation of IRS-1 occurs also in MEF stimulated with IGF-I or in MEF expressing the oncogene v-src. Nuclear translocation of IRS-1 can be demonstrated by confocal microscopy, immunohistochemistry, or subcellular fractionation. An antibody to IRS-1 immunoprecipitates from nuclear fractions (but not from cytosolic fractions) the upstream binding factor, which is a key regulator of RNA polymerase I activity and ribosomal RNA (rRNA) synthesis. In agreement with this finding, in 32D murine hemopoietic cells, nuclear translocation of IRS-1 correlates with a markedly increased rRNA synthesis. Our experiments suggest that nuclear IRS-1 may play a specialized role in rRNA synthesis and/or processing.

The insulin receptor substrate (IRS) 1 proteins are a family of docking proteins, which include IRS-1-4, Gab-1, and p62 dok (1). IRS-1 was the first to be identified as a docking protein for both the insulin and the IGF-I receptors. It transmits a signal from both receptors by interacting with a number of partners, including phosphatidylinositol 3-kinase, SHP2, Grb2, Crk, and others (2,3). Tyrosine kinase activity of the receptors and phosphorylation of IRS-1 are essential steps in signal transduction. Of the downstream signals generated by IRS-1, the best studied is the phosphatidylinositol 3-kinase signaling pathway, which plays a major role in a number of biological responses to growth factors (1,4). IRS-1 interacts directly with both the insulin and the IGF-I receptors, and the domains required for their interaction have also been identified (5). Because of its direct interaction with the receptors, its size, and its downstream signaling, it has been generally assumed that IRS-1 is an exclusively cytosolic (or plasma membrane) protein (6,7). However, IRS-1 is known to interact with the SV40 T antigen (8,9) and nucleolin (10). T antigen and nucleolin are predominantly nuclear proteins, although minor fractions of either protein can be found in the cytosol (11)(12)(13). It has been therefore tacitly assumed that IRS-1, anchored to the receptor, was interacting with the minor cytosolic fractions of T antigen and nucleolin. There is evidence, however, that signal-transduction molecules can translocate to the nucleus. They include mitogen-activated protein kinase (14), p70 S6K /TOR (15,16), the STAT proteins (17,18), Akt (19), ␤-catenin (20), the epidermal growth factor receptor (21), phosphatases (22), IRS-3 (23), and a cleaved ErbB-4 receptor (24). Indeed, Jans and Hassan (25) have summarized in a review the evidence that growth factors and their receptors can accumulate in the nuclei of cells. IGF-I, IGF-IR, and IRS-1 are not mentioned in that review, but insulin is (26).
The first observation that IRS-1 can be translocated to the nuclei should be credited to Lassak et al. (27), who used medulloblastoma cells expressing the human JCV virus. The nuclear translocation of IRS-1 in association with T antigen (this time, the SV40 T antigen) was confirmed by Prisco et al. (28). In this communication, we have extended the observations of Lassak et al. (27) and Prisco et al. (28) to other cell lines and have attempted to obtain information on the biological significance of the nuclear translocation. Our results confirm that IRS-1 is translocated to the nuclei and especially to the nucleoli of cells in culture. Oncogenes like SV40 T antigen and v-src, as well as the activated IGF-IR can induce nuclear/nucleolar translocation of IRS-1. In addition, our data indicate that nuclear IRS-1 (but not cytosolic IRS-1) interacts with the upstream binding factor (UBF), a protein that regulates RNA polymerase I activity and, therefore, ribosomal RNA (rRNA) synthesis (29,30). In agreement with this interaction, cells with nuclear/nucleolar IRS-1 have a markedly increased rRNA synthesis at levels superior to those found in cells with cytosolic localization of IRS-1. Our findings, coupled with previous findings that IRS-1 increases cell size (see "Discussion"), suggest that nuclear IRS-1 may be involved in sustained cell growth, especially through the production of rRNA. with a plasmid expressing the simian virus 40 (SV40) T antigen (32). RϪ/v-src cells are RϪ cells expressing the v-src oncogene (34). RϪ/IRS/ FLAG cells and Rϩ/IRS/FLAG cells were generated, respectively, from RϪ and Rϩ cells by transduction with a retroviral vector expressing mouse IRS-1 (28) fused in frame with a FLAG epitope at its 3Ј end. Viral transduction was performed as previously described (35). Selection was carried out with 1 g/ml puromycin (Invitrogen). All cell lines are mixed populations. Two cell lines, which do not express IRS-1, 32D cells (36), and LNCaP cells (37) were used as negative controls for anti-IRS1 staining. For experiments on rRNA synthesis, 32D IGF-IR, 32D/IRS-1, and 32D IGF-IR/IRS1 cells were used. These cells are described in detail in Zhou-Li et al. (9) and Valentinis et al. (38). The cells used for the experiments shown in Fig. 9 are LNCaP cells stably expressing a wild-type mouse IRS-1 (37).
Plasmid-pIRS/FLAG was generated from pGR159 MSCV.pac retroviral vector (28) by fusing in-frame the wild-type mouse IRS-1 sequence with the FLAG sequence (Eastman Kodak Co.) at the 3Ј end. The IRS-1 sequence fused in frame with the FLAG epitope at the 3Ј end was produced by PCR. The detailed methodology for the construction of the retroviral vector has been already described (35).
Immunofluorescence/Confocal Microscopy-Cells plated on glass coverslips were washed with PBS and fixed with 3.0% paraformaldehyde in PBS for 20 min at room temperature, followed by permeabilization with 0.2% Triton X-100 in PBS for 2 min at room temperature. Coverslips were washed with PBS, and nonspecific binding of IgG was blocked with 10% normal donkey serum (sc-2044, Santa Cruz Biotechnology) in PBS for 20 min at room temperature. The cells were then stained with the antibodies described below, as indicated. After incubation with primary antibodies, coverslips were washed with PBS three times. The cells were subsequently stained with fluorescein isothiocyanate-and rhodamine-conjugated secondary antibodies (sc-2078 and sc-2095, Santa Cruz) for 1 h at room temperature. Finally, coverslips were washed with PBS three times and mounted on glass slides with Vectashield mounting medium (H-1000, Vector Laboratories Inc.). Fluorescent images were collected on a Zeiss Axiovert 100 confocal microscope using a Zeiss 40ϫ objective. In one experiment, propidium iodide was used to stain the nuclei. In this case, the cells were digested with RNase A (1 mg/ml) for 30 min before staining with propidium iodide (2.5 g/ml, Molecular Probes P-3566) for 5 min.
Immunohistochemistry-32D and 32D-derived cells were washed three times with Hanks' buffer and seeded at a density of 5 ϫ 10 4 /2 ml of RPMI 1640 medium supplemented with 10% fetal bovine serum plus or minus IGF-I (50 ng/ml). Cells were harvested after 16 h, and cytospins were prepared. After fixing in 3.7% formaldehyde solution in PBS and permeabilization with 0.2% Triton X-100 in PBS, the immunostaining was carried out using the Histomouse SP kit (Zymed Laboratories Inc. 95-9541) following the manufacturer's protocol. The magnifications for the figures presented are 1000ϫ.
Subcellular Fractionation-For cell fractionation, cell monolayers were trypsinized and collected in serum-containing medium. Cells were washed with ice-cold PBS and taken up in buffer A (10 mM N-2hydroxyethylpiperazine NЈ-2-ethanesulfonic acid (HEPES), PH 7.9, 0.1 mM EDTA, 0.1 mM EGTA, 10 mM KCl, 1 mM dithiothreitol and 0.5 mM phenylmethylsulfonyl fluoride). After swelling on ice for 10 min, plasma membranes were disrupted by adding 0.1% Nonidet P-40 and mixing for 10 s. Cell breakage was examined under the microscope. The nuclei were pelleted by centrifugation at 6000 rpm for 45 s at 4°C, and a cytoplasmic fraction (supernatant) was recovered. The pellet was then washed in 1 ml of ice-cold sucrose buffer (0.32 M sucrose, 3 mM CaCl 2 , 2 mM magnesium acetate, 0.1 mM EDTA, 10 mM Tris-HCl PH 8.0, 1 mM dithiothreitol, and 0.5 mM phenylmethylsulfonyl fluoride) and centrifuged at 3000 rpm for 5 min at 4°C. The washing of nuclei was repeated three times. The pellet was subsequently resuspended in buffer C (20 mM HEPES PH 7.9, 1 mM EDTA, 1 mM EGTA, 400 mM NaCl, 1 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride). After rocking for 20 min at 4°C, the samples were centrifuged at 12,000 rpm for 15 min at 4°C to recover a nuclear fraction (supernatant).
Western Blots-Aliquots of either cytoplasmic or nuclear extracts (50 g/sample) were resolved by SDS-4 -15% polyacrilamide gels and transferred to nitrocellulose membranes. They were then probed with the antibodies indicated in the figures and developed with the ECL system (Amersham Biosciences). Equal amounts of cell lysates (500 g/sample, unless otherwise stated) determined by Bio-Rad protein assay, were immunoprecipitated with an anti-IRS-1 antibody (UBI) and 25 l of protein AϩG beads (Amersham Biosciences). After washing, precipitates were directly subjected to Western blot analysis.
Metabolic Labeling of rRNA-32D IGF-IR, 32D/IRS-1, and 32D IGF-IR/IRS-1 cells (38) were seeded at a density of 5 ϫ 10 4 cells/ml in RPMI 1640 medium supplemented with heat-inactivated 10% fetal bovine serum and 50 ng/ml of IGF-I (Life Technologies) for the indicated times. In one experiment, 32D IGF-IR/IRS1 cells were pre-treated with rapamycin (Sigma, 10 ng/ml) for 48 h. The cells were then labeled for 4 h with [ 32 P]orthophosphate at a final concentration of 250 Ci/ml (ICN Biochemicals) in phosphate-free RPMI 1640 medium (Life Technologies). After labeling, the cells were washed and incubated in fresh medium for 2 h. Total RNA was isolated using RNeasy MiniKit (Qiagen) and separated by electrophoresis on 1% agarose formaldehyde gels. After drying, the 32 P-labeled rRNA was visualized by autoradiography. The bands were also counted in a liquid scintillation counter.

Co-localization of IRS-1 and SV40 T Antigen in the Nuclei of RϪ Derived
Cells-In previous papers, we showed that IRS-1 could be co-immunoprecipitated with the SV40 T antigen, either in RϪ cells (8) or in 32D cells (9). We have investigated the localization of IRS-1 by confocal microscopy, using the appropriate antibodies described under "Experimental Procedures." Fig. 1A shows a confocal microscopy picture of RϪ/T cells, growing in fetal bovine serum. The cells were stained for IRS-1 (rhodamine, left) or SV40 T antigen (FITC, center), and the right panel gives the merged picture. T antigen is, as expected, almost exclusively localized in the nucleus, but so is IRS-1. The merged picture gives a clear co-localization (there are tiny specks of both proteins in the surrounding cytoplasm, but both T antigen and IRS-1 are mostly nuclear). The specificity of the immunohistochemistry for IRS-1 was monitored in several ways (see also below in the section on Rϩ cells). First, we determined the localization of IRS-1 in the parental RϪ cells by confocal microscopy. RϪ cells have no IGF-I receptors, but high levels of IRS-1, at least in comparison to other 3T3 cells. Fig. 1B shows RϪ cells stained with rhodamine for IRS-1 and with FITC for nucleolin. RϪ cells had been stimulated with IGF-I for 8 h, but the same results were obtained with RϪ in SFM or stimulated with serum (data not shown). In RϪ cells, IRS-1 (as detected by this antibody) is essentially limited to the cytoplasm. The nuclei appear as dark centers, dotted by the nucleoli stained with the anti-nucleolin antibody. The same antibody was completely negative when used on two cell lines that do not express IRS-1, prostatic human cancer cells LNCaP, and murine hemopoietic 32D cells (data not shown).
To further confirm the results obtained by confocal microscopy, we carried out a subcellular fractionation of RϪ and RϪ/T cells (Fig. 2). After subcellular fractionation (see "Experimental Procedures"), IRS-1 can be detected in the nuclear fraction of RϪ/T but not of RϪ cells (lanes 3 and 4). In both cell lines, there is still IRS-1 in the cytosol. The fractions are reasonably pure, as Grb2 is detectable in the cytosol but not in the nuclear fraction of both cell lines. On the contrary, the T antigen can only be detected in the nuclear fraction of RϪ/T cells (Fig. 2C). In Fig. 2D, we have taken the cytosolic and nuclear fractions of both cell lines and immunoprecipitated them with an antibody to IRS-1. T antigen was detected only in the nuclear fraction of RϪ/T cells.
By confocal microscopy, it would seem that IRS-1 is, if not exclusively, largely localized to the nucleus of RϪ/T cells, while, in Western blots, the majority of IRS-1 is cytosolic, even in RϪ/T cells. We will discuss this discrepancy in a later section. Fig. 1) or by subcellular fractionation (Fig. 2), IRS-1 is detectable in RϪ cells only in the cytosol. This is true regardless of the growth conditions (SFM, IGF-I, or fetal bovine serum). We then asked whether IRS-1 would be detectable in the nuclei of Rϩ cells (33), which are RϪ cells expressing a wild-type human IGF-IR. Rϩ cells were stained with antibodies to IRS-1 (rhodamine) or nucleolin (FITC). Fig. 1C is a confocal microscope image of Rϩ cells kept in SFM for 48 h (quiescent cells). Fig. 1D, shows the same cells 8 h after stimulation with IGF-I (50 ng/ml). In quiescent cells, IRS-1 is mostly, but not exclusively, localized in the cytoplasm. By comparing Fig. 1, C and D, three things are noticeable. In stimulated cells, there is a marked increase in the amount of IRS-1 translocated to the nucleus, the nucleoli are larger, and some of the IRS-1 co-localizes with nucleolin in the nucleoli. These observations were highly reproducible.

IRS-1 Localizes in the Nuclei and Nucleoli of Rϩ Cells Stimulated with IGF-I-Either by confocal microscopy (
Subcellular Fractionation of Rϩ Cells-We followed the same methodology used for RϪ and RϪ/T cells. Fig. 3 shows that a substantial fraction of IRS-1 is found in the nuclei of Rϩ cells stimulated with IGF-I. The nuclear fraction is apparently pure, because Grb2 is not detectable. There is much PCNA in the cytosol fraction, which can only be due to leakage. The presence of PCNA in the cytosol does not invalidate the conclusion that IRS-1 is present in the nucleus. On the contrary, it suggests that its nuclear localization may be underestimated by subcellular fractionation. There is a small amount of IRS-1 in the nuclei of quiescent Rϩ cells in SFM (the original shows a detectable albeit very weak band), which is in agreement with the findings obtained by confocal microscopy. The nuclear presence of minute amounts of IRS-1 in the nuclei of quiescent Rϩ cells is explained in the "Discussion." Also in the "Discussion" will be given the other controls used to validate the specificity of the antibodies used to detect IRS-1.
Localization of IRS-1 in the Nuclei of RϪ v-src Cells-IRS-1 has also been shown to interact with v-src (34,39), and we predicted that IRS-1 would also be translocated to the nuclei in RϪv-src cells. RϪ v-src cells were originally described by Valentinis et al. (34). In monolayer cultures, they grow in serum-free medium and form foci in 10% serum. They also form colonies in soft agar, v-src being one of two among several oncogenes tested that can transform RϪ cells (reviewed in Ref. 40). Fig. 4 shows a confocal microscopy picture of RϪ v-src cells in SFM for 48 h. There is staining for IRS-1 in the nuclei of these cells, although the cytoplasm also stains (rhodamine). In these cells, nucleolin stains the nucleoli, but also gives a diffuse

IRS-1 Translocation and rRNA Synthesis
staining of the nucleoplasm (FITC). The merged picture confirms the localization of IRS-1 in the nuclei, and the nucleoli, of RϪ v-src cells.
Other Controls for IRS-1-The nuclear translocation of IRS-1 has been confirmed by using a FLAG antibody to detect an IRS-1 expressing a FLAG epitope. For this purpose, we generated cell lines from both RϪ and Rϩ cells transfected with a plasmid in which the mouse IRS-1 (41) had been tagged with a FLAG epitope (28). In Rϩ IRS-1/FLAG cells stimulated with IGF-I, the FLAG antibody gave FLAG-positive nuclei. RϪ cells were also transfected with the IRS-1/FLAG plasmid. In these cells in SFM, the FLAG-stainable material was in the cytoplasm and IGF-I did not cause nuclear translocation (data not shown). The results were the same as in 32D cells transfected with this construct (28). Other controls include the following. 1) We used three different commercially available antibodies to IRS-1. All of them gave the same results. 2) One of these antibodies came with the peptide used for immunization. Competition with this peptide completely abrogated the staining for IRS-1, both in immunohistochemistry and in confocal mucroscopy. 3) None of the antibodies used gave a reaction with cells not expressing IRS-1, such as LNCaP cells (37) and parental 32D cells (36,38). This was true by confocal microscopy and by Western blots. The experimental data are not shown, but are available on request.
Time Course of IRS-1 Translocation in Rϩ Cells-We have looked at IRS-1 nuclear translocation in Rϩ cells at various times after stimulation with IGF-I (50 ng/ml). The results are shown in Fig. 5A, where, for convenience, only the merged pictures are given. In all instances, IRS-1 is predominantly nuclear and especially nucleolar. The nucleoli become more prominent after IGF-I stimulation, and the increase in nucleolar size is accompanied by a co-localization of nucleolin and IRS-1 (yellow staining of nucleoli). The increase in nucleolar size as visualized by the anti-nucleolin antibody is dramatic, as confirmed in Fig. 5B. Compare for instance the size of the nucleoli in IGF-I stimulated Rϩ cells versus the same cells in serum free medium, or the RϪ cells stimulated with IGF-I.
IRS-1 Co-precipitates UBF from Nuclear Extract-A legitimate question at this point is whether nuclear IRS-1 shows different biological effects than cytosolic IRS-1. The confocal microscopy pictures show strong evidence that IRS-1 localizes not just to the nuclei, but also to the nucleoli. We have also mentioned above that the nucleoli markedly increase in size in Rϩ cells stimulated with IGF-I, and that IRS-1 often co-localizes with nucleolin in the nucleoli. The interaction of IRS-1 with the nucleoli prompted us to ask whether nuclear/nucleolar IRS-1 plays a role in rRNA synthesis and/or processing. We addressed this question by two different approaches: 1) interaction (by immuno-coprecipitation) of IRS-1 with nucleolar proteins and 2) effect of subcellular localization of IRS-1 on rRNA synthesis.
We first asked whether IRS-1 would interact with UBF, a key regulator of the rDNA promoter, and therefore of rRNA synthesis (29,30). It localizes exclusively to the nucleolus and stimulates RNA polymerase I activity (29). To test the interaction of IRS-1 with UBF, whole-cell lysates and cytosolic or nuclear fractions of RϪ and RϪ/T cells were examined. The cells were growing in 10% serum. In Western blots of whole-cell lysates, UBF was detectable in both RϪ and RϪ/T cells (Fig. 6). When cytosolic and nuclear fractions of these MEF were immunoprecipitated with an antibody to IRS-1, UBF was immunoprecipitated in the nuclei but not in the cytosol of both cell lines (RϪ cells are growing in serum). In RϪ/T cells, an antibody to IRS-1 co-precipitates both UBF1 and UBF2 (last lane of Fig. 6), but UBF1 is by far the most abundant. This is an important finding, as UBF1 is the active form (29). The reverse is also true, i.e. an antibody to UBF co-precipitates IRS-1, although this time IRS-1 is detectable only in the nuclei of RϪ/T cells (lane 2 of lower row). This is probably due to the lower amount of IRS-1 (and UBF) in the growing RϪ cells. The purity of the fractions was monitored as usual (data not shown). In this experiment, we used cells in which IRS-1 was localized to the nuclei. To verify the result, we repeated the experiment on 32D-derived cells. As already mentioned, 32D IGF-IRS1 cells grow exponentially in IGF-I (42). 32D IGF-IR cells expressing mutant IRS-1 proteins have been described in a previous paper (28). We have chosen the 32D IGF-IR cells expressing the ␦PTB IRS-1, an IRS-1 protein with a deletion of the PTB domain (37). In these cells, IRS-1 remains cytoplasmic (28) and the cells do not survive in IGF-I (40). Fig. 6B shows that UBF is co-precipitated by an antibody to IRS-1 in 32D IGF-IR IRS-1 cells (nuclear IRS-1) but not in the cells expressing the ␦PTB mutant of IRS-1 (cytoplasmic). The lane marked

IRS-1 Translocation and rRNA Synthesis
whole lysates shows that UBF is present in the cells with the mutant IRS-1, whereas the parental 32D cells are used as the usual control, because they do not express IRS-1.
The interaction of IRS-1 with UBF is important for two reasons. It confirms the nucleolar localization of IRS-1 and places IRS-1 in close contact with a major regulator of rRNA synthesis. A corollary of this observation is that nuclear IRS-1 ought to increase rRNA synthesis. This corollary is supported by three different experiments, described below.
Subcellular Localization of IRS-1 and rRNA Synthesis-RϪ cells and RϪ derived cells express substantial amounts of IRS-1. In the circumstances, we thought it preferable to examine the effect of IRS-1 on rRNA synthesis in 32D-derived cells. Parental 32D cells do not express IRS-1 (36,38). We have previously described two 32D-derived cell lines, one expressing only IRS-1 (9) and the other expressing IRS-1 and an increased level of IGF-IR (38). The former cell line, because of the lower levels of IGF-IR, is not IL-3-independent. After IL-3 withdrawal and IGF-I supplementation, 32D/IRS-1 cells survive somewhat longer than parental cells, but eventually die (9). 32D IGF-IR/IRS-1 cells, instead, are IL-3-independent and even form tumors in mice (42) . Fig. 7, A-C shows an immunohistochemistry of these two cell lines plus the parental 32D cells. All cells were stained with an antibody to IRS-1 and with hematoxylin to stain the nuclei. Parental 32D cells, as expected (36,38), do not stain at all for IRS-1. 32D/IRS-1 cells stain strongly for IRS-1, but the localization is essentially cytoplasmic. In 32D IGF-IR/IRS-1 cells, most of the cells have both nuclear and cytoplasmic IRS-1, with the nuclear localization being predominant (note the different color of the nuclei, when the cell has little or no IRS-1). These two cell lines were labeled with 32 P at 16 h after shifting from IL-3 to IGF-I. The incorporation of 32 P into rRNA was determined as detailed under "Experimental Procedures." Fig. 7D shows rRNA synthesis in 32D/IRS-1 and 32D/IGF-IR/IRS-1 cells. Both cell lines were stimulated with IGF-I for 16 h. At this time (16 h after shifting to IGF-I) both cell lines survive in the absence of IL-3, but rRNA synthesis is 10 times higher in the cells with nuclear IRS-1 than in the cells with cytosolic IRS-1 (confirmed by counting the radioactivity in the bands).
To confirm that nuclear localization of IRS-1 increases rRNA synthesis, we carried out another experiment. Rapamycin spe- cifically inhibits mTOR (43) and, therefore blocks (albeit not always completely) IRS-1 signaling at the level of p70 S6K activation (3). In 32D IGF-IR/IRS-1 cells, rapamycin inhibits transformation and causes differentiation (42). In 32D IGF-IR/ IRS-1 cells treated with rapamycin, IRS-1 is exclusively localized to the cytosol (Fig. 8B). In this experiment, the nuclei were stained with propidium iodide (red) and IRS-1 with FITC (green). There is no overlap between the two stains in the merged picture. Fig. 8A shows that rapamycin markedly inhibits rRNA synthesis in these cells, thus confirming the importance of nuclear IRS-1 in increasing rRNA synthesis.
Effect of IRS-1 on rRNA Synthesis-We then compared two cell lines that are identical, except that one expresses IRS-1 (32D IGF-IR/IRS1 cells) and the other does not (32D IGF-IR cells). Both cell lines grow exponentially in the first 48 h after shifting from IL-3 to IGF-I, although 32D IGF-IR/IRS-1 cells continue to grow (and form tumors in animals), whereas 32D IGF-IR cells eventually differentiate along the granulocytic lineage (38,42). The previous experiment showed that IRS-1 is present in the nuclei of 32D IGF-IR/IRS-1 cells (Fig. 7C). Fig. 9 shows that rRNA synthesis is lower in 32D IGF-IR cells than in 32D IGF-IR/IRS1 cells at both times. When the bands were counted, rRNA synthesis was increased in 32D IGF-IR/IRS1 cells from 2-to 3-fold relative to 32D IGF-IR cells. The labeling of nascent rRNA was essentially abolished when the cells where treated with 0.05 g/ml actinomycin, which specifically inhibits rRNA synthesis (Fig. 11A).

Co-precipitation of Nucleolin and IRS-1 Is Limited to the
Nuclei-We have already mentioned that IRS-1 and nucleolin interact with each other (10). We have asked whether subcellular localization of IRS-1 plays a role in their interaction, as it seems to do with the SV40 T antigen and UBF. For this purpose, we used LNCaP/IRS-1 cells (37). These are human prostate-cancer cells stably transfected with a plasmid expressing IRS-1 (the parental cells do not express IRS-1). We selected these cells because they express unusually high amounts of nucleolin (roughly10 -15 times the levels in MEF), thus making it easier to detect.
After subcellular fractionation, both the cytosol and the nuclear fractions were monitored by Western blot for the presence of nucleolin. Fig. 10 shows that nucleolin is abundant in these cells and is present in roughly equal amounts in the cytosolic and nuclear fractions (the purity of the fractions was monitored as in Fig. 1). When the two fractions were immunoprecipitated with an antibody to IRS-1, nucleolin was found only in the immunoprecipitate from the nuclei, although IRS-1 was found in both fractions. This experiment was repeated several times, both with this cell line and another LNCaP cell line in which, besides IRS-1, IGF-IR expression was increased (37). In all instances, and regardless of growth conditions, nucleolin was immunoprecipiated by an anti-IRS-1 antibody only in the nuclear fraction. It should be noted that LNCaP cells (parental or derived) were grow in serum-free medium, although IGF-I partially increases their growth (37).
Effect of Actinomycin D on rRNA synthesis and UBF in 32D IGF-IR IRS-1 Cells-Actinomycin D, at very low concentrations, inhibits almost exclusively rRNA synthesis (48), a finding confirmed in 32D IGF-IR IRS-1 cells (Fig. 11A). The effect is dramatic, as incorporation of 32 P decreases to a level less

IRS-1 Translocation and rRNA Synthesis
than 1% of untreated cells. Fig. 11B shows that at the same time, UBF is no longer detectable in the treated cells, whereas it is clearly visible in the untreated cells. Thus, IRS-1 has no effect on rRNA synthesis in cells, where UBF has been decreased by treatment with Actinomycin D.

DISCUSSION
Our data confirm and extend the results of Lassak et al. (27) and Prisco et al. (28) that IRS-1 can translocate to the nuclei of cells. New findings in this communication include: 1) IRS-1 translocates not only to the nuclei but especially to the nucleoli of Rϩ cells stimulated with IGF-I; 2) nuclear/nucleolar translocation is confirmed in different cell lines, like Rϩ and RϪ/vsrc cells; 3) more important, we find that nuclear IRS-1 immunoprecipitates at least two nucleolar proteins, UBF and nucleolin. The finding with UBF is especially significant, since this protein resides exclusively in the nucleolus, and an interaction with IRS-1 can only occur in the nucleolus, or, at the most, the nucleus; 4) nuclear/nucleolar translocation of IRS-1 correlates with an increase in rRNA synthesis. The two most important questions raised by our observations are the evidence for nuclear translocation and its biological significance.
Most important for us is to determine the biological significance of IRS-1 nuclear translocation. In other words, what does nuclear IRS-1 do different from cytoplasmic IRS-1? A clue to a biological function of nuclear IRS-1 is its preferential localization to the nucleoli and the accompanying dramatic increase in nucleolar size. The increase in nucleolar size does not occur in RϪ cells stimulated by IGF-I. The nucleolus is the site of ribosomal RNA synthesis, which, in turn, depends on the activity of RNA polymerase I (29). The activity of the rDNA promoter and its polymerase is regulated by a few proteins, among which is prominent is the role of UBF (29,30). It is therefore intriguing that IRS-1 in the nucleus co-precipitates UBF; the appropriate controls indicate that this interaction does occur. Interestingly, IRS-1 immunoprecipitates largely the active form of UBF (29), UBF1. IRS-1 also immunoprecipitates nucleolin, and only in nuclear extract. However, the interaction with UBF1 is the most significant. It confirms the after fractionation into a nuclear and a cytosolic fraction. The first two lanes are Western blots of the fractions (C for cytosol, N for nuclear). The amounts of nucleolin are roughly similar in the two fractions. All other lanes are blots of immunoprecipitates, using an antibody to IRS-1 to co-precipitate nucleolin. Nucleolin is immunoprecipitated only in the nuclear fraction. Different growth conditions (fetal bovine serum, serum-free medium, or IGF-I) gave the same results. The lower panel shows that IRS-1 was immunoprecipitated in both fractions, although in two conditions, IRS-1 is higher in the nuclear than in the cytosolic fraction. For the lysates, we used 40 g of protein (5 g for IRS-1), and for the immunoprecipitates, 400 g of protein.
nucleolar localization of IRS-1 (there is no UBF in the cytosol), which is further supported by the finding that a mutant IRS-1 that does not translocate to the nucleus (28) also fails to coprecipitate UBF. This result suggests a role of nuclear IRS-1 in rRNA synthesis. In agreement with this suggestion, we show a correlation between nuclear/nucleolar localization of IRS-1 and increased rRNA synthesis compared with cells with cytosolic IRS-1 or without IRS-1. Thus, there is a substantial difference in rRNA synthesis between 32D IRS-1 (cytoplasmic IRS-1) and 32D IGF-IR/IRS-1 cells (nuclear IRS-1). Second, in rapamycintreated 32D IGF-IR/IRS-1 cells, the cytosolic localization of IRS-1 is accompanied by a marked reduction in rRNA synthesis with respect to untreated cells. Rapamycin inhibits p70 S6K and causes the differentiation of 32D IGF-IR/IRS-1 cells (42). TOR proteins, which are inhibited by rapamycin, are known to stimulate rRNA synthesis and rRNA processing (reviewed in Ref. 43). Our findings are in agreement with the fact that rRNA synthesis decreases in differentiated cells (44) and that the nucleolus actually disappears in terminally differentiated cells. The chick erythrocyte nucleus is the best example of nucleolar involution in differentiated cells (45). Finally, rRNA synthesis is much lower in 32D IGF-IR cells that do not express IRS-1 than in 32D IGF-IR/IRS-1 cells, where IRS-1 is in the nucleus. Interference with UBF inhibits the effect of IRS-1 on rRNA synthesis (Fig. 11). We therefore suggest that nuclear IRS-1 plays an important role in increasing rRNA synthesis and that it does so by interacting with the regulator of RNA polymerase I activity, UBF1.
Increased rRNA synthesis results in increased cell size. Cell size is usually determined by the amount of protein/cell (46,47), which in turn requires an increase in the amount of rRNA (47,48). An increased synthesis of rRNA implies a larger number of ribosomes, hence more protein synthesis and an enlargement of cells (48). Indeed, in murine hemopoietic 32D cells, we have shown that cells expressing IRS-1 are larger than 32D cells not expressing IRS-1, even when both cell types are growing exponentially (42). Interestingly, the size of 32D IGF-IR/ IRS1 cells decreases when the cells are treated with rapamycin. The importance of IRS-1 and its downstream signaling on cell size in vivo are also supported by several reports in the literature. Mice with deleted IRS-1 (49) or p70 S6K (50) genes are smaller than their wild-type littermates. But the importance of IRS-1 and p70 S6K in cell-size regulation was rigorously demonstrated by the observations that homologues of either IRS-1 (51) or the S6 kinase (52) regulate cell size in Drosophila.
Clearly, IRS-1 cannot be an absolute requirement for rRNA synthesis and cell size; otherwise 32D cells would not grow in IL-3. Indeed, deletion of IRS-1 or chico in Drosophila or mice results in animals that are smaller, but viable. Our studies offer for the first time a molecular explanation. Even in cells without IRS-1, there is rRNA synthesis, but a nuclear IRS-1 increases it considerably, presumably by interacting with UBF1. In other words, in animals with compromised IRS-1 signaling, body size is reduced, indicating that IRS-1 does indeed contribute an additional (and not redundant) stimulus to growth in size. On the basis of our results, we suggest that nuclear/nucleolar IRS-1 is the important contributor to cell growth.
Nuclear translocation of IRS-1 is now rigorously demonstrated. The evidence from this and previous papers (27,28) can be summarized as follows. 1) The IRS-1 antibodies show nuclear localization in RϪ cells expressing T antigen or v-src that are known to interact with IRS-1. Parental RϪ cells are negative for nuclear IRS-1. 2) Nuclear localization is increased in Rϩ cells after stimulation with IGF-I. IRS-1 is not detectable in the nuclei of Rϩ cells stimulated with epidermal growth factor (data not shown) or in RϪ cells, regardless of growth conditions. 3) The anti-IRS-1 antibodies failed to give a positive stain with two cell lines (LNCaP and parental 32D cells) in which IRS-1 is not expressed (36,37). 4) Translocation into the nuclei is also supported by the use of an IRS-1 with a FLAG tag (Ref. 28, and this paper). The FLAG antibody fails to stain cells that do not express the IRS-1/FLAG construct. In RϪ cells expressing this construct, FLAG is found essentially in the cytoplasmic fraction. Only in Rϩ cells stimulated with IGF-I is FLAG-tagged material detectable in the nuclei. 5) Controls with different antibodies and competing peptides confirm that the protein detected in Western blots, in confocal microscopy and in immunohistochemistry is bona fide IRS-1. Rϩ cells show some IRS-1 staining even in SFM (see Fig. 1). Rϩ cells, however, are known to secrete small amounts of IGF-I, and their IRS-1 is tyrosyl-phosphorylated (albeit weakly) even when the cells are in SFM (34).
There is a discrepancy between subcellular fractionation and confocal microscopy in terms of the fraction of IRS-1 translocated to the nuclei. For instance, in RϪ/T cells, it would seem, from confocal microscopy, that all or most of IRS-1 is in the nucleus. There are specks of material stained with the IRS-1 antibody outside the nuclei, but much too little in comparison to the amount revealed by subcellular fractionation. Our expla- FIG. 11. Effect of Actinomycin D on rRNA synthesis and UBF in 32D IGF-IR IRS-1 cells. The concentration of actinomycin D was 0.05 g/ml. Ribosomal RNA synthesis was determined as in the previous figures in cells growing exponentially or in cells treated for 24 h with actinomycin D (panel A). The bands from autoradiography were counted in a liquid scintillation counter. In panel B, aliquots of the same cells treated in similar manner were stained for UBF before or after treatment with actinomycin D. UBF is detectable in the exponentially growing cells, but not in the cells treated with actinomycin D. nation is leakage from the nuclei during fractionation. It is true that the T antigen is recovered only in the nuclear fraction, but PCNA is not. PCNA is a nuclear protein, and yet it is recovered in equal amounts in both fractions. Perhaps some proteins leak out of the nuclei more than others, and IRS-1 could be one of them. A second possibility is that IRS-1 could be present in the nucleus in the form of cleavage products, as in the case of ErbB-4 (24). However, in our subcellular fractionation studies, the size of IRS-1 was full length (see Figs. 1 and 2), and no other bands were detected, even though some smaller bands could be detected in the cytosol. Although we cannot exclude the presence of small amounts of nuclear IRS-1 even in RϪ cells, the general evidence is that under certain circumstances, IRS-1 can translocate to the nuclei. We suggest that, in this case, confocal microscopy may be more reliable than subcellular fractionation.
It has recently been reported that IRS-3 (23) translocates to the nucleus. In this paper, the authors reported that IRS-1 did not translocate to the nucleus. The discrepancy between the data of Kabuta et al. (23) and those of Lassak et al. (27), Prisco et al. (28), and ours is probably due to the different cells used.
Finally, the co-localization of IRS-1 and nucleolin is in agreement with the reported interaction of IRS-1 with nucleolin (10). Nucleolin plays a role in rRNA processing (53). We have shown in this paper that IRS-1 co-precipitates nucleolin in the nuclei, but not in the cytosol, despite the fact that in LNCaP cells, nucleolin is abundant in both fractions. The biological significance of the IRS-1/nucleolin interaction is less obvious than the UBF/IRS-1 interaction. However, we find it interesting that their close interaction may be limited to the nuclear environment.
In conclusion, our data give rigorous evidence that IRS-1 can be translocated to the nuclei of mouse embryo fibroblasts by two oncogenes and by stimulation of the wild-type IGF-IR with IGF-I. From these experiments, it is reasonable to hypothesize a role of nuclear IRS-1 in rRNA synthesis as indicated by its localization in the nucleoli, its interaction with UBF1 in the nuclear fractions, and its correlation to increased rRNA synthesis. Finally, it is unlikely that nuclear translocation of IRS-1 is only an artifact of tissue cultures, as nuclear IRS-1 has been reported in tissue sections from human breast cancers (54) and human medulloblastomas (27).