Research ArticleBiochemistry

SUMOylation Mediates the Nuclear Translocation and Signaling of the IGF-1 Receptor

See allHide authors and affiliations

Science Signaling  09 Feb 2010:
Vol. 3, Issue 108, pp. ra10
DOI: 10.1126/scisignal.2000628


The insulin-like growth factor 1 receptor (IGF-1R) plays crucial roles in developmental and cancer biology. Most of its biological effects have been ascribed to its tyrosine kinase activity, which propagates signaling through the phosphatidylinositol 3-kinase and mitogen-activated protein kinase pathways. Here, we report that IGF-1 promotes the modification of IGF-1R by small ubiquitin-like modifier protein–1 (SUMO-1) and its translocation to the nucleus. Nuclear IGF-1R associated with enhancer-like elements and increased transcription in reporter assays. The SUMOylation sites of IGF-1R were identified as three evolutionarily conserved lysine residues—Lys1025, Lys1100, and Lys1120—in the β subunit of the receptor. Mutation of these SUMO-1 sites abolished the ability of IGF-1R to translocate to the nucleus and activate transcription but did not alter its kinase-dependent signaling. Thus, we demonstrate a SUMOylation-mediated mechanism of IGF-1R signaling that has potential implications for gene regulation.


The insulin-like growth factor 1 receptor (IGF-1R) is a receptor tyrosine kinase (RTK), and its signaling has crucial roles in development, aging, and cancer biology by regulating various biological responses, including cell growth, proliferation, differentiation, apoptosis, and anchorage-independent growth (1, 2). The main signaling pathways for IGF-1R–mediated functions depend on the activation of the phosphatidylinositol 3-kinase (PI3K)–Akt and mitogen-activated protein kinase (MAPK) pathways. The mechanisms by which the IGF-1R is activated are generally understood.

IGF-1R is a heterotetramer that consists of two ligand-binding, extracellular α subunits and two β subunits that are composed of a transmembrane domain, an intracellular tyrosine kinase (TK) domain, and a C-terminal tail. Ligand binding induces tyrosines within the TK domain to be transphosphorylated by the dimeric subunit partner. The phosphorylated residues serve as docking sites for other signaling molecules, such as insulin receptor substrates 1 to 4 (IRS1 to IRS4) and the adaptor protein Shc, which leads to the activation of the PI3K and MAPK pathways; however, these signaling pathways are shared with most other RTKs. Therefore, it is likely that IGF-1R–specific effects are also mediated through other, still unknown, pathways. In addition to the well-established RTK signaling pathways from the cell membrane, direct communication between the membrane and nucleus through nuclear translocation of RTKs has also been reported.

Epidermal growth factor receptor (EGFR) family members (35), fibroblast growth factor receptor 1 (FGFR1) and FGFR3 (6, 7), insulin receptor (IR) (8), and the vascular endothelium growth factor (VEGF) receptor fetal liver kinase 1 [Flk1, also known as kinase insert domain receptor (KDR)] (9, 10) can accumulate in nuclei, either as intact receptors or as proteolytically cleaved fragments, with or without their corresponding ligands. These nuclear RTKs act as transcription factors for the genes that encode cyclin D1 (3), FGF2 (11), and cyclooxygenase 2 (COX-2) (4) and as modulators for inducing the expression of the genes that encode c-Jun and cyclin D1 (12). The underlying mechanism for these actions is, however, still unclear. The presence of nuclear IGF-1R (nIGF-1R) in hamster kidney cells has also been reported (13); however, the biological relevance and the molecular mechanism for the nuclear import of IGF-1R remain obscure.

In addition to tyrosine phosphorylation, the actions of RTKs are tuned by other posttranslational modifications. Of these, the role of ubiquitination is the most studied. We and others have shown that modification by ubiquitin plays important roles in the endocytic sorting, degradation, and MAPK-dependent signaling of the IGF-1R (1416). Modification of proteins by small ubiquitin-like modifier (SUMO) protein is another type of posttranslational modification that is important for certain signaling pathways (17, 18). SUMOylation is mainly observed in nuclear and perinuclear proteins but also occurs in a few types of cell-surface proteins, including ion channels (19, 20), glutamate and kainate receptors (21), and the transforming growth factor–β (TGF-β) type I receptor (18). Four SUMO isoforms are found in mammalian cells, of which SUMO-1 is the most extensively studied.

The functional consequences of the modification of proteins by SUMO-1 vary depending on the target and include the regulation of cytoplasmic-nuclear transport, DNA repair, protein stability, chromosome separation, and transcription (22, 23). The SUMOylation process begins with the activation of SUMO by a SUMO-activating enzyme, such as SAE1 or SAE2 (SUMO E1). Activated SUMOs are subsequently transferred to the specific SUMO-conjugating enzyme UBC9 (SUMO E2) (24) and finally transferred to lysine residues in targeted proteins. A hallmark of SUMOylation is that only a small fraction of the targeted substrate is modified at any given time (25). To our knowledge, SUMOylation of RTKs has not been reported. Here, we demonstrate that SUMO-1 is conjugated to three evolutionarily conserved lysine residues of IGF-1R in a ligand-dependent manner and that SUMOylation of IGF-1R is imperative for nuclear translocation of the receptor. We further demonstrate that nIGF-1R binds to genomic DNA and provide evidence suggesting that the IGF-1R acts as a transcriptional enhancer.


SUMOylation of IGF-1R is ligand-dependent

To determine whether IGF-1R is SUMOylated in cells, we used lysates of human melanoma cells (the DFB cell line), grown in the presence or absence of serum, in immunoprecipitation reactions with an antibody against IGF-1R and analyzed them by Western blotting with an antibody against SUMO-1. A 145-kD band was detected in cells grown in the presence of serum (Fig. 1A), which reflected the modification of IGF-1R by multiple SUMO-1 molecules. Furthermore, transient transfection with a plasmid containing HA-UBC9 [which encodes the hemagglutinin (HA)–tagged E2 ligase UBC9 for the SUMO-conjugation reaction] enhanced the SUMOylation of IGF-1R, as shown by the increased intensity of the 145-kD band, and also slightly increased the abundance of SUMOylated IGF-1R in serum-depleted cells (Fig. 1A). In addition to the unmodified β subunit (95 kD), the 145-kD band was also detectable with an antibody against the β subunit of IGF-1R, which was visible after long exposure (Fig. 1A). Control immunoprecipitation experiments were also performed (fig. S1A). As loading controls, the abundances of UBC9 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) in whole-cell extracts were compared (Fig. 1A). HA-UBC9 was recognized by its increased mass relative to that of its untagged, endogenous counterpart.

Fig. 1

SUMOylation of IGF-1R. (A) DFB cells that were transfected with a plasmid encoding HA-UBC9 (HA-UBC9) or that were untransfected (Unt) were grown in basal (10% serum) or in serum-free medium (SFM), immunoprecipitated (IP) with an antibody against IGF-1Rβ, and analyzed by Western blotting [immunoblotting (IB)] with antibodies against SUMO-1 or IGF-1Rβ. The transfection efficiency of cells with the UBC9-encoding plasmid was determined by the analysis of whole-cell extracts. GAPDH was used as loading control. (B) DFB cells transfected with a plasmid encoding HA-SUMO-1 were grown in basal or SFM conditions, immunoprecipitated with an antibody against IGF-1Rβ, and analyzed with antibodies against HA or IGF-1Rβ. S1, S2, and S3 indicate IGF-1Rβ proteins that have been modified with one, two, or three SUMO-1 moieties, respectively. (C) The transfection efficiency was confirmed by analysis of whole-cell extracts with an antibody against HA. (D) DFB cells grown under basal or SFM conditions were immunoprecipitated with antibodies against either IGF-1Rβ or IGF-1Rα under denaturing conditions and analyzed by Western blotting with an antibody against SUMO-1. Equal loading of each lane was confirmed by incubating the blot with an antibody against IGF-1R. (E) Serum-starved DFB cells were stimulated with IGF-1 for the indicated times, immunoprecipitated with antibodies against IGF-1Rβ or UBC9, and analyzed by Western blotting with antibodies against SUMO-1 or IGF-1Rβ, after which the blots were stripped and then incubated with antibodies against IGF-1Rβ or UBC9 to confirm equal loading. The data shown [(A) to (E)] are representative of at least triplicate experiments.

Oligo-SUMOylation was further confirmed when we detected that HA-SUMO-1 in transfected DFB cells coimmunoprecipitated with IGF-1R to give three bands with molecular masses of 112, 125, and 145 kD (S1, S2, and S3), which corresponded to IGF-1R proteins that had undergone mono-, di-, and tri-SUMO-1 conjugation, respectively (Fig. 1B). Endogenous SUMOylated proteins are predicted to be rapidly deSUMOylated (17), which might explain why the endogenous S1 and S2 proteins were harder to detect than was S3. To investigate the subunit specificity for SUMOylation, we performed denaturing immunoprecipitation assays with antibodies against either the α or the β subunit, which showed that conjugation with SUMO-1 occurred only in the sample immunoprecipitated with the antibody against the β subunit; this finding suggested that the β subunit was the direct target for SUMOylation.

Considering that SUMOylation of IGF-1R is dependent on the presence of serum, we investigated the possibility of ligand-dependent SUMOylation of IGF-1R and what the kinetics of this modification might be. Serum-starved DFB cells were stimulated with IGF-1 over a number of time points (Fig. 1E). IGF-IR was immunoprecipitated from these samples and analyzed for modification by SUMO-1. The S3 IGF-1R band appeared 4 hours after stimulation and peaked in abundance at 6 to 8 hours after stimulation, after which its abundance declined (Fig. 1E). Because UBC9 associates with SUMO target proteins, the association of UBC9 with IGF-1R upon ligand stimulation was examined. UBC9 also exhibited an association with IGF-1R in a ligand-dependent manner, appearing 2 hours after stimulation, peaking in abundance at 6 hours, and declining thereafter (Fig. 1E).

SUMO–IGF-1R is localized in the nucleus

Previous studies have correlated modification by SUMO with the recruitment of target proteins to the nucleus (22, 24). To address whether modification by SUMO-1 might cause the nuclear recruitment of IGF-1R, we investigated the subcellular localization of IGF-1R in DFB and human embryonic kidney (HEK) 293 cells by fractionation and Western blotting analysis as described in the Supplementary Materials. As expected, IGF-1R was present in membrane fractions, but it was also found in nuclear extracts (Fig. 2A). The 145-kD band, which was detected with an antibody against IGF-1Rβ, appeared exclusively in the nuclear fraction (Fig. 2A). Addition of N-ethylmaleimide (NEM), an inhibitor of SUMO-1 hydrolases (the enzymes that remove deSUMOylated target proteins) (26), to the fractionation lysis buffers enabled the detection of the 112- and 125-kD bands in the nuclear fractions (Fig. 2A). This suggests that these slow-migrating bands represented SUMOylated IGF-1R and that the modified receptor was nuclear. Seemingly, the mono- and di-SUMOylated β subunits (112 and 125 kD, respectively) were readily deSUMOylated, whereas tri-SUMOylation of IGF-1R was relatively stable. To investigate whether the intact receptor (consisting of the α and β subunits) or its proteolytic fragments were localized to the nucleus, we also performed Western blotting with an antibody against IGF-1Rα. The α subunit (native size, 120 kD) together with the β subunit (native size, 95 kD) was present in the nuclear fraction (Fig. 2A), suggesting that nIGF-1R was an intact receptor. Consistent with our earlier results (Fig. 1D), no slow-migrating bands were detectable with an antibody against IGF-1Rα, which confirmed that the α subunit was not a target for modification by SUMO.

Fig. 2

Nuclear translocation of IGF-1R. (A) DFB and HEK 293 cells grown under basal conditions were fractionated in the presence or absence of NEM. The membrane (M) and nuclear (N) fractions of each cell type were analyzed by Western blotting with antibodies against IGF-1Rβ and IGF-1Rα. Histone H3 was used as a nuclear marker. Data are representative of four experiments. (B) The nonnuclear, soluble nuclear, and pelleted nuclear fractions prepared from HEK 293 cells were analyzed by Western blotting with an antibody against IGF-1Rβ. The abundance of c-Myc and histone H3 were analyzed to assess loading. Data are representative of four experiments. (C) Untreated DFB cells (none) or DFB cells treated with hypertonic medium (HT) or dansylcadaverine (DC) were fractionated and analyzed for the presence of IGF-1Rβ. Data are representative of three experiments. (D) DFB cells, grown under basal conditions, were treated with dimethyl sulfoxide (DMSO), as a control, or AG1024 (50 μM) for 24 hours. Cells were then either analyzed by Western blotting with an antibody against phosphorylated tyrosine residues (P-Tyr) to determine the abundance of phosphorylated IGF-1R or fractionated and analyzed to determine the subcellular distribution of IGF-1Rβ. Data are representative of three experiments.

The fractionation procedure that we used isolates nuclear proteins that are free of Golgi- and endoplasmic reticulum (ER)–specific proteins (27). Through the use of markers for plasma membrane [Na+- and K+-dependent adenosine triphosphatase (Na+,K+-ATPase)], early endosomes (EEA1), cytoplasm (GAPDH), the ER (calnexin), and nuclei (histone H3), we confirmed that there was no contamination of the nuclear fractions by any of these components (fig. S1B). We also performed our analyses with three commonly used antibodies against IGF-1R to detect the receptor in the nucleus (fig. S1B).

To investigate the distribution of IGF-1R within the nucleus, we extracted nuclei by a standard, high-salt nuclear matrix procedure (3, 28, 29). This method enables separation of soluble nuclear proteins (in the high-salt fraction) from the nuclear pellet, which contains proteins tightly bound to DNA and nonsoluble nuclear proteins in a membrane environment. nIGF-1R was localized predominantly in the nuclear pellet; however, because the IGF-1R was extractable by a high concentration of salt (Fig. 2B), this indicated that nIGF-1R was also soluble outside the nuclear membrane bilayer. The S1 and S3 bands were detectable exclusively in the nuclear pellet (Fig. 2B).

Blocking receptor internalization with dansylcadaverine or hypertonic medium strongly suppressed the nuclear accumulation of IGF-1R, which suggested that IGF-1R was translocated to the nucleus from the cell membrane (Fig. 2C). Tyrphostin AG1024, an established inhibitor of the kinase activity of IGF-1R (30), was used to evaluate the role of receptor kinase activity in the nuclear translocation process. AG1024 strongly reduced the nuclear translocation of IGF-1R relative to that in control cells (Fig. 2D), which suggested that the kinase activity was required for nuclear translocation of the receptor. Inhibition of the phosphorylation of IGF-1R upon treatment of cells with AG1024 was also confirmed (Fig. 2D).

SUMOylation enhances the nuclear accumulation of IGF-1R

To address whether SUMOylation promoted the nuclear recruitment of IGF-1R, we analyzed the abundance of nIGF-1R in the context of the overabundance of UBC9 by transfection of cells with a plasmid encoding UBC9 or by knockdown of UBC9 with UBC9-specific interfering RNA (RNAi) (Fig. 3A). Overexpression of UBC9 increased the abundance of nIGF-1R relative to that in mock-transfected cells, whereas knockdown of UBC9 decreased the abundance of nIGF-1R, which suggested that increased SUMOylation of IGF-1R enhanced its nuclear translocation (Fig. 3A). The 145-kD band was detectable in the nuclei after a long exposure of the Western blot and its abundance was increased in UBC9-overexpressing cells relative to that in mock-transfected cells.

Fig. 3

Kinetics of the nuclear translocation of IGF-1R. (A) DFB cells were left untransfected (Unt) or were transfected with a plasmid encoding HA-UBC9 (+UBC9) or with UBC9-RNAi (−UBC9) and were then grown under basal conditions before being fractionated and analyzed by Western blotting to determine the presence of IGF-1Rβ. Triple-SUMOylated IGF-1R is indicated with S3. (B) The abundance of UBC9 in whole-cell extracts of samples from the experiments described in (A) was analyzed to confirm transfection efficiency in these experiments. (C) Untransfected DFB cells (Mock), DFB cells transfected with a HA-UBC–expressing plasmid (+UBC9), or DFB cells transfected with UBC9-specific RNAi (UBC9-RNAi) were starved of serum and then stimulated with IGF-1 for the indicated times, after which they were fractionated and analyzed by Western blotting for the presence of IGF-1Rβ. Mono- and tri-SUMOylated IGF-1R are labeled with S1 and S3, respectively. (D) Serum-starved DFB cells were stimulated with IGF-1 for the indicated times. IGF-1R–SUMO-1 and IGF-1R–UBC9 complexes were detected by in situ PLA (red dots) and cells were counterstained with Hoechst (blue) to visualize nuclei. The negative control was obtained by omitting one of the primary antibodies. (E) Cells were treated as described in (D) but were analyzed for the presence of IGF-1Rβ (red) and nuclei (blue) by confocal immunofluorescence microscopy. (F) Quantification of positive cells. On average, 100 cells per condition were analyzed in three independent experiments. Means and SDs are shown. P < 0.01. Data shown [(A) to (F)] are representative of at least triplicate experiments.

Considering that SUMOylation of IGF-1R was ligand-dependent, we next sought to determine whether nuclear translocation of the receptor was affected by IGF-1. Serum-starved cells, which were devoid of nIGF-1R, were stimulated with IGF-1 (Fig. 3C). Four to eight hours after stimulation, accumulation of nIGF-1R was detected. This response was enhanced when UBC9 was increased in abundance and was abrogated by UBC9-specific RNAi (Fig. 3C). After a long exposure of the Western blot, both the 145- and 112-kD bands were detectable in the cells that had an increased abundance of UBC9, whereas they were undetectable in cells transfected with UBC9-specific RNAi (Fig. 3C), which suggested that the lower-migrating bands were indeed SUMO-modified IGF-1R.

An in situ proximity ligation assay (PLA) (31), in which the signals detected are strictly dependent on the simultaneous recognition of the target by two antibodies, was applied to investigate the subcellular localization of endogenous IGF-1R–SUMO-1 and IGF-1R–UBC9 complexes and the kinetics of their formation upon stimulation of cells with IGF-1. SUMOylated IGF-1R was detectable 6 to 8 hours after stimulation and was predominantly localized perinuclearly (Fig. 3D), consistent with our earlier findings (Figs. 1E and 2B). The IGF-1R–UBC9 complex was detectable in the cytoplasm 2 hours after stimulation with IGF-1 and was mainly detected perinuclearly 6 hours after stimulation (Fig. 3D). Intranuclear IGF-1R–SUMO-1 and IGF-1R–UBC9 complex dots were also identified. The IGF-1–induced nuclear translocation of IGF-1R was confirmed by immunofluorescence confocal microscopy (Fig. 3E). Although the in situ PLA signal was recognized as a clearly punctuate fluorescent dot, this did not necessarily indicate a redistribution of IGF-1R into vesicular structures, because confocal microscopy displays a diffuse distribution. Indeed, the distinct punctuate dot was due to the fluorophore labeling of the rolling-circle amplification (RCA) oligonucleotide product (see Materials and Methods).

The numbers of cells in which IGF-1R was localized to the nucleus were quantified (Fig. 3F). After 8 hours of stimulation with IGF-1, 78% of the cells exhibited nIGF-1R. The association of IGF-1R with UBC9 and SUMO-1 was also quantified (Fig. 3F), which demonstrated that the SUMOylation of IGF-1R occurred with kinetics that paralleled those of the nuclear translocation of the receptor. nIGF-1R was found not only in cultured cells but also in tissues, specifically by immunohistochemical analysis of melanoma tissues (fig. S1C).

SUMOylation at Lys1025, Lys1100, and Lys1120 is a prerequisite for the nuclear translocation of IGF-1R

Our results so far suggested that the SUMO modification was mainly related to nIGF-1R and that an overall alteration in SUMOylation affected nuclear accumulation of the receptor. To directly address whether SUMOylation of IGF-1R was essential for its nuclear translocation, we individually replaced each of the 24 lysine residues within IGF-1Rβ (Fig. 4A) by arginines. The effect of each mutation on the nuclear translocation of IGF-1R was tested by transfecting IGF-1R–deficient leiomyosarcoma SKUT-1 cells (32) with plasmids encoding each mutant and investigating the resultant subcellular distribution of each mutant IGF-1R by fractionation and Western blotting analysis. Although the abundance of membrane-bound IGF-1R was equal for all mutant proteins, three mutations resulted in substantially lower amounts of nIGF-1R than that observed in cells containing wild-type IGF-1R (Fig. 4, B and C). These three mutations—K1025R, K1100R, and K1120R—impaired the nuclear translocation of the corresponding mutant IGF-1Rs by 76, 43, and 29%, respectively, whereas the other mutations had no detectable effects. A tri-SUMO-site mutant (TSM) IGF-1R, which contained the mutations K1025R, K1100R, and K1120R, was as abundant at the plasma membrane as was the wild-type IGF-1R, whereas the nuclear fraction of cells transfected with this construct was nearly devoid of the mutant receptor (Fig. 4B). This suggested that these three lysine residues had a substantial effect on the nuclear transport of IGF-1R.

Fig. 4

Nuclear translocation of IGF-1R requires site-specific tri-SUMO-1 modification. (A) Schematic structure of IGF-1R with the lysine residues of the β subunit denoted. (B) IGF-1R–deficient SKUT-1 cells, transfected with plasmids encoding wild-type (WT) IGF-1R or the indicated mutant IGF-1Rs, were fractionated and analyzed by Western blotting with an antibody against IGF-1Rβ. Data are representative of five experiments. (C) Quantification of the abundance of nIGF-1R based on five independent experiments. Means and SDs are indicated. A representative experiment involving all 24 mutants is shown in fig. S2A. (D) IGF-1R–deficient SKUT-1 cells, cotransfected with a plasmid encoding HA-SUMO-1 and with plasmids encoding wild-type IGF-1R or the indicated mutant IGF-1Rs, were immunoprecipitated with an antibody against IGF-1Rβ and analyzed by Western blotting with antibodies against HA or IGF-1Rβ to detect SUMOylated IGF-1R. Mono-, di-, and tri-SUMOylated IGF-1R species are indicated by S1, S2, and S3, respectively. Data are representative of three experiments. (E) SUMOylated IGF-1R was visualized by in situ PLA (red) in SKUT-1 cells transfected with empty vector, plasmid encoding WT IGF-1R, or a plasmid encoding the indicated mutant IGF-1R. The lower panel shows the quantification of the positive cells. On average, 100 cells per condition were analyzed by automated image analysis in three independent experiments. Means and SDs are shown. (F) IGF-1R–deficient SKUT-1 cells, transfected with plasmids encoding WT IGF-1R or the indicated mutant IGF-1Rs, were grown under basal conditions (10% serum), immunoprecipitated with an antibody against phosphorylated tyrosine residues (P-Tyr), and analyzed by Western blotting for the presence of IGF-1Rβ. Data are representative of three experiments. (G) IGF-1R–deficient SKUT-1 cells, transfected with plasmids encoding WT IGF-1R or the indicated mutant IGF-1Rs, were serum-starved for 24 hours, stimulated with IGF-1 for the indicated times, immunoprecipitated with antibody against P-Tyr residues, and analyzed by Western blotting with an antibody against IGF-1Rβ. Data are representative of three experiments. (H) IGF-1R–deficient SKUT-1 cells, transfected with plasmids encoding WT IGF-1R or the indicated mutant IGF-1Rs, were stimulated with IGF-1 for the indicated time points. Cells were then fixed and analyzed to determine the amount of cell-surface IGF-1R by enzyme-linked immunosorbent assay. Means and SDs (n = 4 experiments) are shown.

To verify whether these lysine residues were also SUMOylation sites, we investigated lysates of SKUT-1 cells cotransfected with plasmids encoding HA-SUMO-1 and either wild-type IGF-1R or one of the various mutants of IGF-1R for their SUMOylation of IGF-1R. Because the K821R and K1203R mutations had no effect on the nuclear accumulation of IGF-1R (Fig. 4B), they were used to distinguish the effects of the K1025R, K1100R, and K1120R mutations from those of other lysine-to-arginine mutations. Whereas wild-type IGF-1R and the K821R and K1203R mutant IGF-1Rs showed the characteristic tri-SUMO-1 bands when analyzed by Western blotting, the K1025R, K1100R, and K1120R mutants were poorly SUMOylated, showing weak mono- and di-SUMO-1 bands. The TSM IGF-1R did not undergo any detectable SUMOylation (Fig. 4D). Investigation of the subcellular localization of SUMO-1–modified wild-type and mutant IGF-1Rs by in situ PLA showed consistent results (Fig. 4E). Because each signal detected by in situ PLA represented one interaction, we could quantify the number of PLA signals per single cell in 100 transfected cells for each condition by automated image analysis (Fig. 4E). Whereas cells containing wild-type IGF-1R exhibited on average 30 associations per cell, cells transfected with plasmids encoding K1025R, K1100R, or K1120R mutant IGF-1Rs showed 6, 11, and 15 signals per cell, respectively. Furthermore, the number of associations between the TSM IGF-1R and SUMO-1 was few, comparable to that of mock-transfected cells.

To characterize basic receptor functions, we examined the phosphorylation, internalization, and signaling of the SUMO site mutants of IGF-1R (K1025R, K1100R, K1120R, and TSM) and compared them to those of wild-type IGF-1R. Phosphorylation of the IGF-1R mutants was intact relative to that of wild-type IGF-1R both under basal conditions (Fig. 4F) and upon stimulation with IGF-1 (Fig. 4G). The mutant receptor proteins also retained their ability to be endocytosed upon ligand stimulation (Fig. 4H). To study the signaling capabilities of the receptors, we transiently transfected HEK 293 cells with plasmids encoding wild-type IGF-1R or the SUMO site mutants. Upon stimulation with IGF-1, cells transfected with plasmids encoding either wild-type IGF-1R or one of the SUMO site mutants exhibited an equally increased abundance of phosphorylated Akt (pAkt) and phosphorylated extracellular signal–regulated kinase (pERK) relative to that in mock transfected cells (Fig. 5A). This suggested that downstream signaling was not impaired by the introduction of the SUMO site mutations in IGF-1R. All three SUMOylation sites (Lys1025, Lys1100, and Lys1120) are localized within the kinase domain of IGF-1R. The crystal structure of the kinase domain of phosphorylated IGF-1R, obtained from the Research Collaboratory for Structural Bioinformatics (RCSB) Protein Data Bank (PDB code 1K3A) (33), shows that Lys1025, Lys1100, and Lys1120 are exposed at the surface of the receptor (Fig. 5C) and are thereby accessible to SUMOylation enzymes. Analysis of sequence alignments demonstrated that Lys1025, Lys1100, and Lys1120 are conserved in the closely related insulin receptor (IR) and the insulin receptor–related receptor (IRR) (Fig. 5D). These residues are also conserved between species (Fig. 5D).

Fig. 5

Further characterization of the SUMO-site mutant IGF-1R proteins and of nIGF-1R. (A) Lysates from serum-starved HEK 293 cells, transfected with empty vector or plasmids encoding WT IGF-1R or the indicated mutant IGF-1Rs, were stimulated with IGF-1 for 10 min and analyzed by Western blotting for the presence of phosphorylated ERK (pERK) and phosphorylated Akt (pAkt). The blots were then stripped and incubated with antibodies against Akt or ERK to demonstrate equal loading of lanes. The experiments were repeated twice with similar results. (B) The transfection efficiency of the experiments presented in (A) was verified by analyzing the Western blots described in (A) with antibodies against IGF-1Rβ and GAPDH, as a loading control. Three experiments were performed with similar results. (C) Crystal structure of the phosphorylated kinase domain of IGF-1R from the RCSB PDB (PDB code 1K3A). The residues Lys1025 (K1025), Lys1100 (K1100), and Lys1120 (K1120) are indicated. N and C indicate the N and C termini, respectively. (D) The upper panel shows the amino acid alignment of the IGF-1R sequence that contains K1025, K1100, and K1120 and the corresponding regions of the closely related insulin receptor (IR) and insulin receptor–related receptor (IRR) obtained with Kalign (61). The bottom panel shows the alignment of the same sequence in humans and in the indicated species homologs. Conserved SUMO acceptor lysine residues are boxed. Abbreviations for the amino acids are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; R, Arg; S, Ser; T, Thr; V, Val; and Y, Tyr.

nIGF-1R binds to putative enhancer regions and enhances transcription

To explore the functional relevance of nIGF-1R, we investigated the DNA-binding capacity of the receptor in vitro by performing electrophoretic mobility shift assays (EMSAs) with randomly synthesized, biotin-labeled, double-stranded, 40–base pair (bp) oligonucleotides (table S1). After nuclear extracts prepared from DFB cells were mixed with DNA probe number 2, an electrophoretic mobility shift was obtained (Fig. 6A), which represented the binding of proteins (for example, transcription factors) to this probe. The protein-DNA complex was supershifted by the addition to the reaction mixture of a specific antibody against the IGF-1R (Fig. 6A), which indicated that IGF-1R was associated with this DNA sequence either alone or in complex with other proteins. A similar supershift band was observed for probe number 5. The specificity of these bands was controlled for by competitive binding with a 200-fold molar excess of unlabeled probes, which eliminated the supershift bands (Fig. 6A). From the 17 probes tested, four (probe numbers 2, 5, 6, and 14) were supershifted by the antibody against IGF-1R (fig. S2B), and no shift was observed when the antibody against IGF-1R was used alone. Collectively, these data indicate that the IGF-1R interacted physically with double-stranded DNA, and given that it did not bind to 13 of the 17 probes, this binding is likely to be specific.

Fig. 6

IGF-1R binds to DNA and enhances transcription. (A) Randomly synthesized, biotinylated probes 2 and 5 (table S1) were subjected to EMSA. In lanes 5 and 7, unlabeled probes were added at 200 times the molar excess for competition. Bands were detected with horseradish peroxidase–conjugated streptavidin. Experiments were performed four times with similar results. NE, nuclear extract; ab, IGF-1Rβ antibody. (B) Left panel: seven ChIP-isolated DNA fragments to which IGF-1R bound (table S2) were inserted upstream of the GL4.23 luciferase reporter construct. pGL4.23 indicates no insertion. HEK 293 cells were cotransfected with reporter constructs containing the different fragments and with empty vector, plasmid encoding WT IGF-1R, or plasmid encoding the indicated mutant IGF-1R proteins, and luciferase assays were performed seven times. Data were normalized to the activity of Renilla luciferase. Statistical analysis (by Student’s t test) of IGF-1R compared to mock: 1.15, P = 0.008; 10.5, P = 0.00002; and 10.6, P = 0.0006. Right panel: HEK 293 cells were cotransfected with the reporter constructs containing the 1.15, 10.5, and 10.6 inserts and with empty vector, plasmid encoding WT IGF-1R, or plasmid encoding the indicated mutant IGF-1R proteins. The assays were performed five times independently with similar results. The inhibitory effects of the K1025R and TSM mutant IGF-1Rs on WT IGF-1R–induced transcription were calculated. Means and SDs are shown. P < 0.008. (C) HEK 293 cells transfected with empty vector, plasmid encoding WT IGF-1R, or plasmid encoding the TSM mutant IGF-1R were fractionated and the membrane and nIGF-1Rβ content in each sample were compared by Western blotting analysis with an antibody against IGF-1Rβ. Lamin was used as a marker of the nuclear fraction. Data are representative of four experiments. (D) The IGF-1R–enriched regions obtained from the ChIP-seq were classified according to their position relative to known genes and their distribution is illustrated graphically. Upstream regions were defined as 20 kb upstream of an annotated transcript start site (5′UTR + 20 kb upstream). Downstream regions were defined as 20 kb downstream of an annotated transcript end site (3′UTR + 20 kb downstream).

Through a chromatin immunoprecipitation (ChIP)–based cloning strategy (34), we sought to identify the genomic DNA sequences to which nIGF-1R bound. This strategy was previously used to identify previously uncharacterized target genes regulated by transcription factors such as E2F (34). In brief, immunoprecipitated DNA fragments from IGF-1R chromatin extracts were blunt-ended by treatment with polymerase and ligated into a pGEM-T Easy vector. To minimize nonspecific pull-down of DNA, we performed two sequential immunoprecipitations (34). To more efficiently determine the specific sequences to which nIGF-1R bound, we performed two negative control pull-down assays, one with an antibody against GAPDH and another with control immunoglobulin G (IgG). After the second immunoprecipitation reaction, unlike in the case of the pull-down reaction with the antibody against IGF-1R, there was no detectable DNA in either of the negative control reactions, which was suggestive of the successful isolation of genomic fragments specifically bound by nIGF-1R. The transformed clones were screened with vector-specific primers. We limited further studies to seven positive clones, because consideration of all of the potential sequences was outside the scope of this study. We sequenced the DNA fragments (table S2) and noted some of their characteristics (table S3). As shown, the sizes of the DNA fragments ranged between 126 and 427 bp, they were located on different chromosomes, and four of the sequences were intronic. The distances from these sequences to their respective closest genes were large (from 8 to 475 kb) (table S3), which suggested that the ChIP-purified sequences may represent enhancer regions (35).

To investigate whether the DNA fragments represented transcriptional regulatory elements, we subcloned them into the pGL4-minimal promoter vector, which encoded firefly luciferase harboring a minimal promoter with low basal activity; cloning of enhancer regions upstream of luciferase in this construct results in strong transcriptional activation of the gene. HEK 293 cells were transiently cotransfected with the pGL4 reporter construct, with or without the inserted ChIP sequences, together with either an empty vector or a plasmid encoding wild-type IGF-1R. HEK 293 cells were used in these experiments because of the efficiency with which they are transfected. The transcriptional activities of the reporter genes, each containing fragments of potential enhancer regions (ChIP inserts), were then determined (Fig. 6B). Cells transfected with the plasmid encoding wild-type IGF-1R showed a substantially elevated transcriptional activity for sequences 1.15, 10.5, and 10.6 relative to that in cells transfected with the empty vector. In addition to that of wild-type IGF-1R, the effects of TSM and the K1025R mutant IGF-1R on the transcriptional activity of these three sequences were also investigated. Transcriptional activity was much lower in cells transfected with the K1025R-expressing plasmid relative to that in cells transfected with the wild-type IGF-1R–expressing plasmid and was completely abrogated in cells transfected with a plasmid that encoded the TSM IGF-1R (Fig. 6B). The abundance of nIGF-1R was greater in cells cotransfected with a plasmid encoding wild-type IGF-1R than in cells transfected with either the empty vector or a plasmid encoding the TSM IGF-1R (Fig. 6C).

We used ChIP-seq (36) to further examine the interaction of IGF-1R with DNA on a genome-wide scale. Approximately 6 million reads were produced, of which about 3.9 million survived several rounds of filtering (see Materials and Methods for details) and were successfully aligned to the human reference genome. Analysis of the data set with MACS software (37) resulted in 568 candidate peaks, that is, statistically significant IGF-1R–enriched regions. The IGF-1R–enriched regions were divided into five classes on the basis of their location relative to known genes. Most of the IGF-1R–interacting sites (80%) were located distal from any annotated gene (intergenic), 6.3% were located in introns, 6.3% in exons, 3.4% were ≤20 kb upstream of an annotated transcript start site [5′ untranslated region (5′UTR) + 20 kb upstream], and 3.6% were ≤20 kb downstream of an annotated transcript end site (3′UTR + 20 kb downstream (Fig. 6D).


In this study, we present three findings, which together may represent a previously uncharacterized signaling mechanism of IGF-1R with functional relevance. First, internalized IGF-1R underwent SUMOylation in a ligand-dependent manner. Second, SUMOylated IGF-1R was predominantly localized perinuclearly and this modification was essential for the nuclear translocation of the receptor. Third, nIGF-1R bound to putative enhancer regions and enhanced transcription.

The functions of SUMOylated transmembrane proteins are poorly understood. SUMOylation of the ion channel proteins K2P1 and Kv1.5 causes their inactivation (19, 20). SUMOylation of the kainate receptor subunit GluR6 is kainate-dependent and enhances endocytosis, which inhibits synaptic transmission (21). SUMOylation of TGF-β receptor I affects its signaling by regulating the activation of Smad proteins and downstream responses, thus enhancing tumorigenesis (18). Indeed, loss of the control of SUMOylation contributes to cancer development as a result of the SUMOylation of several oncogenic proteins and tumor suppressors, such as p53, Mdm2, and c-Myb (3841). In addition, SUMOylation of the reptin chromatin complex regulates expression of the metastasis suppressor gene KAI1 and the invasive activity of cancer cells (42). IGF-1R mediates several important functions in development and cancer (1, 43). Our present findings, showing that SUMOylated IGF-1R is translocated to the nucleus in which it binds to DNA and increases gene transcription, may represent a mechanism that contributes to the diverse biological roles of IGF-1R and SUMO in cancer.

RTKs are found in the nucleus in two forms, either as intact molecules or as cytoplasmic fragments that are produced by the sequential action of two distinct membrane-localized proteases. Whereas the membrane-associated fragment is ubiquitinated and degraded by the proteasome, the cytoplasmic fragment is translocated to the nucleus (44). On the other hand, other RTKs such as EGFR and FGFR-I are translocated to the nucleus in intact form. Our results show that this is also the case for nIGF-1R (Fig. 1F). EGFR and FGFR-1 are not found in the nuclear envelope, but are present in the nucleoplasm in a nonmembranous environment (3, 29). We demonstrated that nIGF-1R was extractable from the cell nuclei by high concentrations of salt, indicating its presence in the nonmembranous environment. The mechanism by which the intact receptor is removed from a membrane bilayer remains to be elucidated. Sec61 is a part of the ER-associated degradation (ERAD) system, which is mainly known for retrotranslocating misfolded proteins in the ER and delivering such molecules intact to the cytoplasm for proteasomal degradation (45). Liao and Carpenter have demonstrated that the Sec61 translocon provides a mechanism for extracting EGFR from the membrane lipid bilayer, thus enabling its nuclear localization (29). The importance of Sec61 in the nuclear translocation of IGF-1R remains to be determined.

IGF-1R lacks a nuclear localization sequence (NLS) and does not associate with β-importin, a nuclear transport receptor. Even though a protein may lack an NLS, it can still be translocated to the nucleus in a process mediated by the NLS of an associated protein (46). Insulin receptor substrate 1 (IRS-1), which is important for the activation of IGF-1R–mediated signaling pathways, has a putative NLS and is translocated to the nucleus in response to IGF-1 (47). Whether the association between IGF-1R and IRS-1 augments the nuclear translocation of IGF-1R remains to be determined. According to our results, the critical event for nuclear translocation of IGF-1R is its conjugation with SUMO-1 at three conserved lysine residues, Lys1025, Lys1100, and Lys1120, all of which contribute to the SUMOylation of IGF-1R, because mutation of a single site alone reduced the extent of SUMOylation markedly but not completely. In this context, Lys1025 seems to be the most important SUMO target site, because the K1025R mutation had the strongest inhibitory effect on both the SUMOylation of IGF-1R and its nuclear translocation. Although SUMOylation enables the nuclear translocation of many proteins (48, 49), the mechanisms responsible for SUMO-mediated nuclear import of proteins are poorly understood. The presence of a SUMO E3 ligase and an isopeptidase, an enzyme that catalyzes deSUMOylation, at nuclear pore complexes has led to the suggestion of a model in which SUMOylation and deSUMOylation are coupled to targets shuttling into and out of the nucleus (50).

The residues Lys1025, Lys1100, and Lys1120 are not in consensus with the canonical SUMOylation motif ψKx(D/E) (51), where ψ is a large hydrophobic residue such as valine, isoleucine, leucine, methionine, or phenylalanine, and x is any residue. Nonconsensus SUMO sites have, however, been described (18, 5254), and it has been speculated that phosphorylation in close proximity to a SUMOylation site leads to a secondary structure that helps to determine the lysine residue(s) targeted by SUMO (55). Lys1025, Lys1100, and Lys1120 are in close proximity to tyrosine residues in the TK domain of IGF-1R. Additionally, inhibition of the kinase activity of IGF-1R attenuated trafficking of the receptor to the nucleus (Fig. 2D).

Our observation that the association between UBC9 and IGF-1R became detectable in the cytoplasm 2 hours after stimulation with ligand (Fig. 3D) suggests that this interaction occurs in endocytic vesicles subsequent to endocytosis of the receptor. Most of the UBC9–IGF-1R complexes, however, appeared at 6 hours and were localized perinuclearly and in the peripheral parts of nuclei (Fig. 3D). Spatiotemporally, the UBC9–IGF-1R association overlapped with that between SUMO-1 and IGF-1R, suggesting that SUMOylation of IGF-1R may take place at or in close proximity to the nuclear membrane (Fig. 3D). Nuclear-localized IGF-1R, as assessed by immunofluorescence, exhibited an intranuclear distribution different from that of UBC9-associated IGF-1R and most of it was seen in the nuclei (Fig. 3E).

Together, the findings from the microscopy study indicated that SUMO-modified IGF-1R was rapidly deSUMOylated after passage across the nuclear membrane. This notion is corroborated by the finding that most of nIGF-1R was unmodified, as assayed by Western blotting, and that SUMO–IGF-1R was detectable only in the nuclear pellet (Fig. 2B), indicating that SUMOylated IGF-1R was mainly localized at the nuclear membrane. Collectively, these data, together with the finding that SUMOylation was a prerequisite for the nuclear translocation of IGF-1R, strongly suggest that the detectable, unmodified receptors were deSUMOylated after import into the nuclei.

For the most part, different RTKs stimulate a similar repertoire of signaling proteins, yet, activation of a particular RTK leads to a particular biological response. One of the key open questions in cell signaling is how specificity in a biological response is generated after stimulation of a common set of signaling pathways by a given RTK. Here, we demonstrated the capacity of IGF-1R to bind, either directly or indirectly as part of a protein complex, to DNA. Further, the genome-wide interaction of IGF-1R with DNA was determined by ChIP-seq (36). The distribution of IGF-1R–enriched genomic regions relative to annotated genes (Fig. 6D) demonstrated that they were predominantly intergenic, although they were also found in introns and exons. These data emphasize that IGF-1R binds to genomic DNA and that the set of potential binding sites has the same signature as that of transcriptional regulatory enhancers. Furthermore, we observed substantially elevated transcriptional activity of reporter genes downstream of five out of the seven tested ChIP-enriched IGF-1R–binding sequences. These data strengthen the notion that nIGF-1R, directly or associated with other proteins, binds to enhancers and functions as a transcriptional cofactor. This might also be a general mechanism by which nuclear RTKs regulate the expression of their specific target genes.

Our findings add SUMOylation to the list of posttranslational molecular mechanisms that regulate the trafficking and signaling of IGF-1R. A future challenge is to investigate the interplay between these mechanisms and to unravel the physiological context in which they control receptor function in health and disease. With respect to the importance of both SUMO and IGF-1R in cancer development and progression, one could speculate about the consequences of losing control over the SUMOylation of IGF-1R, particularly with regard to its role in altered gene transcription. Furthermore, our discovery of the SUMOylation of IGF-1R and its potential effects on gene transcription may provide a better understanding of the unique actions of this receptor and may be valuable for the development of new therapeutic strategies.

Materials and Methods

In situ PLA

PLA has previously been used to detect protein-protein interactions and protein modifications by microscopy (31, 56, 57). We performed PLA to image the interactions between IGF-1R and SUMO-1 or UBC9 with the Duolink kit (Olink) according to the manufacturer’s recommendations. In brief, transfected SKUT-1 cells or nontransfected DFB cells on cover slides were fixed, permeabilized, and blocked as described in the Supplementary Materials. The cells were incubated overnight with mouse antibody against IGF-1R and rabbit antibody against SUMO-1 or with rabbit antibody against IGF-1R and mouse antibody against UBC9. Subsequently, oligonucleotide-conjugated secondary antibodies (PLA probes) were used. PLA probes hybridize to each other when in close proximity (40 nm) during the hybridization step. By adding a ligase, the two hybridized oligonucleotides are joined to form a closed circle. By taking advantage of the RCA method, which uses the ligated circle as a template, a concatemeric (repeated sequence) product is generated. Fluorescently labeled oligonucleotides were added, which hybridized to the RCA product and enabled its detection. The fluorescence signal from each detected pair of PLA probes was visualized as a distinct individual dot (8, 31, 58). The nuclei were counterstained with Hoechst and analysis of the images was carried out as described in the Supplementary Materials. The image analysis software BlobFinder ( was used to quantify the PLA signal.

Identification of ChIP-based sequences

DFB cells (30 × 106) at 80% confluency were serum-starved for 24 hours and stimulated with IGF-1 (50 ng/ml) for 8 hours. For cross-linking of proteins to DNA, formaldehyde was added directly to the culture medium to a final concentration of 1%, and the cells were incubated for 10 min at 37°C. Cross-linking was stopped with 125 mM glycine, and nuclear extracts were prepared with a ChIP assay kit (Upstate Biotechnology) according to the manufacturer’s protocol. Briefly, cells were lysed in SDS lysis buffer (Upstate Biotechnology) and sonicated in a BioRupter to yield an average DNA fragment size of 300 bp (range, 100 to 800 bp) as determined by analysis on an agarose gel. The lysate was precleared for 30 min with salmon sperm DNA and protein G agarose (Upstate Biotechnology). The resulting supernatant was incubated with antibody against IGF-1Rβ (Cell Signaling Technology), antibody against GAPDH, or rabbit IgG overnight. After a 1-hour incubation at 4°C with protein G agarose and serial washes, the protein-DNA complexes were eluted. The eluted samples from the ChIP were immunoprecipitated, washed, and eluted for a second time to minimize background, as described by Weinmann et al. (59). The cross-links were reversed by incubating the elution mix with 5 M NaCl at 65°C overnight. The DNA was purified with a polymerase chain reaction (PCR) purification kit (Qiagen). Blunt ends with an overhang were generated by incubating eluted DNA from the ChIP with 1 μl of Pfu DNA polymerase, 0.2 mM deoxynucleotide triphosphates (dNTPs), Pfu DNA Polymerase 10X Reaction Buffer (Promega) with MgCl2 at 70°C for 30 min. The blunt-ended pool of eluted DNA was purified again with the Qiagen PCR purification kit and ligated into the pGEM-T Easy Vector (Promega) according to the manufacturer’s protocol. Miniprep DNA was isolated from a few of the many colonies that were formed after a standard bacterial transformation protocol was performed. The positive colonies were identified and sequenced with T7 (5′-TAATACGACTCACTATAGGG-3′) and SP6 (5′-ATTTAGGTGACACTATAG-3′) primers.

Preparation of library and high-throughput sequencing

ChIP assays were performed as described above and a sequencing library was prepared from 10 ng of ChIP-DNA according to the manufacturer’s manual (“Preparing samples for ChIP sequencing of DNA,” part#11257047 revA, Illumina). Briefly, the ChIP-DNA was end-repaired with T4 DNA polymerase, Klenow DNA polymerase, and T4 polynucleotide kinase (PNK), followed by purification on a QIAquick PCR purification column (Qiagen). An A-base was ligated to the blunt ends of the DNA fragments with Klenow DNA polymerase and the sample was purified on a MinElute PCR purification column (Qiagen). Adapters for sequencing were ligated to the DNA fragments and the library was size-selected on an agarose gel. A 150- to 300-bp fragment was excised from the gel, purified on a Qiagen gel-extraction column, and amplified for 18 cycles of PCR, followed by purification with a MinElute PCR purification kit. The quality and quantity of the library was evaluated by the Agilent Technologies 2100 Bioanalyzer and a DNA 1000 kit. A 3 pM solution of the purified library was used in the cluster generation on the Cluster Station (Illumina Inc.). Sequencing was performed for 36 cycles with the Genome Analyzer I according to the manufacturer’s protocols. Image analysis and base calling was performed with the analysis pipeline that was supplied with the Genome Analyzer instrument.

Processing of the ChIP-seq data

A total of 10,733,610 unfiltered reads were obtained from the Genome Analyzer, of which 6,093,256 passed the initial quality filtering (chastity filtering with default parameters) and were aligned to the human reference genome (GRCh37/hg19) with MAQ software (60). With MAQ, 4,990,851 reads were aligned to the genome and these were subjected to further rounds of filtering as follows. Each of the 36 bases in a read was associated with a quality score from the Genome Analyzer and the sums of the quality scores of mismatched bases in the alignment were used to further refine the data set. We kept only those reads that had a unique best mapping position in the reference genome (according to the mapping quality score from MAQ) with a maximum sum of mismatch quality scores of 56 and with at most five mismatched bases. Thus, we obtained a final set of 3,945,265 reads. MACS software (37) was then used to find regions (peaks) with statistically significant enrichment with the default P value cutoff of 10−5 (a large number of different parameter sets for MACS were tried and yielded similar results). Only peaks containing at least six unambiguous unique reads were retained for further analysis, leading to a data set of 586 peaks.

Luciferase reporter assays

Luciferase reporter assays were performed by cotransfecting 4 × 104 HEK 293 cells in triplicate in 24-well plates by LT1 (Mirus) with 200 ng of reporter constructs [pGL4.23 (luc2/minP), Promega] with inserted ChIP sequences or the control vector without any modifications and with 100 ng of Renilla luciferase vector [pGL4.70 (hRluc), Promega]. Cotransfections were performed with 100 ng of empty backbone vector (mock), an expression vector encoding wild-type IGF-1R, or expression vector encoding a mutated IGF-1R. Twenty-four hours after transfection, the luciferase activity in the cell extracts was determined with a Dual Luciferase Assay System (Promega) according to the manufacturer’s instructions. Both firefly and Renilla luciferase activities were monitored with a Lumat LB9507 luminometer (Berthold Technology). Reporter activity was normalized by calculating the ratio of firefly to Renilla values. Normalized data were expressed as the relative luciferase activity in a given sample compared with that of cells transfected with the control vector (pGL4.23) alone.


We thank O. Emanuelsson for helpful discussions on the analysis of the ChIP-seq data. This study was supported by grants from the Swedish Cancer Society, the Cancer Society in Stockholm, the Swedish Research Council, the Swedish Children Cancer Society, Ingabritt and Arne Lundberg’s research foundation, the King Gustaf V research foundation, the Stockholm County Research Council, and the Karolinska Institute. Sequencing was performed by the SNP technology platform in Uppsala (

Supplementary Materials

Materials and Methods

Fig. S1. Further characterization of the SUMOylation and nuclear accumulation of IGF-1R.

Fig. S2. SUMO-1 binding site requirement and DNA binding of nIGF-1R.

Table S1. Sequences of the randomly synthesized double-stranded DNA fragments that were used for EMSAs.

Table S2. Sequences of the DNA fragments that were bound to IGF-1R.

Table S3. Features of the ChIP sequences.

Table S4. Sequences of the primers used to construct reporter plasmids.

Table S5. Sequences of the primers used to construct IGF-1R mutants.


References and Notes

View Abstract

Stay Connected to Science Signaling

Navigate This Article