IGF-I, IGF-II, and Insulin Stimulate Different Gene Expression Responses through Binding to the IGF-I Receptor

Insulin and the insulin-like growth factors (IGF)-I and -II are closely related peptides important for regulation of metabolism, growth, differentiation, and development. The IGFs exert their main effects through the IGF-I receptor. Although the insulin receptor is the main physiological receptor for insulin, this peptide hormone can also bind at higher concentrations to the IGF-I receptor and exert effects through it. We used microarray gene expression profiling to investigate the gene expression regulated by IGF-I, IGF-II, and insulin after stimulation of the IGF-I receptor. Fibroblasts from mice, knockout for IGF-II and the IGF-II/cation-independent mannose-6-phosphate receptor, and expressing functional IGF-I but no insulin receptors, were stimulated for 4 h with equipotent saturating concentrations of insulin, IGF-I, and IGF-II. Each ligand specifically regulated a group of transcripts that was not regulated by the other two ligands. Many of the functions and pathways these regulated genes were involved in, were consistent with the known biological effects of these ligands. The differences in gene expression might therefore account for some of the different biological effects of insulin, IGF-I, and IGF-II. This work adds to the evidence that not only the affinity of a ligand determines its biological response, but also its nature, even through the same receptor.


INTRODUCTION
Insulin and the closely related insulin-like growth factors (IGF)-I and -II are important for the regulation of metabolism and cell growth, survival, motility, differentiation, and development (1)(2)(3)(4)(5)(6). These ligands bind to closely related receptor tyrosine kinases. The main physiological receptor for insulin is the insulin receptor, while the IGFs mainly exert their effects through the IGF-I receptor (7,8). The insulin receptor exists under two isoforms, A and B, due to alternative splicing of exon 11 of the insulin receptor gene (9,10).
Insulin-like growth factor-II in mammals also binds to the IGF-II/cation-independent mannose-6-phosphate receptor, which is thought to act as a scavenger for IGF-II rather than a signaling receptor (11,12). Its presence on most cells however complicates the study of IGF-II binding and signaling mediated through the IGF-I receptor.
Binding of the ligands to the insulin or IGF-I receptor leads to autophosphorylation of the receptor on tyrosine residues. This creates binding sites for SH2 and PTB domain-containing docking proteins such as IRS-1-4 and Shc, and stimulates the tyrosine kinase activity of the receptor, enabling it to phosphorylate multiple cytoplasmic substrates, which activates signaling cascades, resulting in ligand-specific biological effects (4,13).
Both the ligands and the receptors are closely related (and therefore the ligands can bind to their non-cognate receptors) and the signaling pathways they activate are largely overlapping (14). Microarray profiling showed that the two receptors are capable of stimulating the same gene expression response (15). Nevertheless, insulin is mainly a metabolic regulator, while the IGFs exert mainly mitogenic effects (growth, proliferation . . .). The molecular basis of this signaling specificity is still not understood (6,16).
As mentioned, the three ligands can also bind to their noncognate receptors, though with lower affinity, and by doing so they can exert different effects in comparison to the cognate ligand. Frasca et al. and Morrione et al., e.g., showed independently that IGF-II is more potent in stimulating proliferation through the insulin receptor A isoform than insulin (17,18). Frasca et al. also showed that insulin is a more potent metabolic regulator through this isoform than IGF-II (17). Pandini et al. found that insulin and IGF-II induce different gene expression patterns after binding to the A isoform of the insulin receptor (19). Malaguarnera et al. found that proinsulin binds with high affinity the insulin receptor isoform A and predominantly activates the mitogenic pathway (20). Also, insulin analogs with different residence times on the insulin receptor have been shown to have different relative potencies for mitogenic versus metabolic signaling (21)(22)(23). Previous work from our laboratory has described an insulin mimetic peptide that despite binding to the insulin receptor with an affinity similar to insulin's is less potent in stimulating thymidine incorporation and induces a different gene expression response in comparison to insulin (24). All in all, it is becoming increasingly clear that various ligands acting through the same receptor may activate different patterns of end-point cellular effects ("differential signaling").
In this study we measured gene expression by microarray profiling after stimulating mouse fibroblasts expressing the IGF-I receptor, but devoid of insulin and IGF-II/cation-independent mannose-6-phosphate receptors (25) with equipotent concentrations of insulin, IGF-I, and IGF-II. During the analysis the focus was on finding differences, rather than similarities, in gene expression between the three ligands. The results show that insulin, IGF-I, and IGF-II indeed create different gene expression responses when stimulating the IGF-I receptor. We hope that these results and further studies will lead to a better understanding of the signaling specificity and different biological effects of these three ligands.

MATERIALS
Fibroblasts from mice knockout for IGF-II and the IGF-II/cationindependent mannose-6-phosphate receptor were a gift from Dr. Kurt von Figura (25). Insulin was from Novo Nordisk A/S, Denmark, and IGF-I and IGF-II from Novozymes GroPep, Thebarton, SA, Australia. 125 I-IGF-I was prepared by Novo Nordisk A/S. Unless otherwise specified all chemicals were from Sigma-Aldrich, Denmark.
The mouse fibroblasts, devoid of IGF-II and the IGF-II/cationindependent mannose-6-phophate receptor, did not bind 125 Iinsulin, indicating the absence of biologically active insulin receptors (results not shown), but did bind 125 I-IGF-I. From the below mentioned homologous competition assay data, we found that approximately 75,000 IGF-I receptor sites/cell are present on this cell line. AFFINITIES OF IGF-I, IGF-II, AND INSULIN FOR THE  IGF-I RECEPTOR   To determine the apparent affinities of the ligands for the IGF-I receptor on the mouse fibroblast cell line, homologous and heterologous radioligand competition assays were performed in quadruplets. Cells were detached with 10 mM EDTA (Gibco, Invitrogen, Denmark). Three million cells per milliliter were incubated with a constant concentration of 125 I-IGF-I (20,000 cpm/ml) and increasing concentrations of cold IGF-I, IGF-II, or insulin for 2.5 h (time needed to reach steady-state binding) at 15°C in Hepes Binding Buffer (100 mM Hepes, 120 mM NaCl, 5 mM KCl, 1.2 mM MgSO 4 , 1 mM EDTA, 10 mM Glucose, 15 mM Na Acetate, and 1% BSA). After centrifugation unbound 125 I-IGF-I was removed and cell-bound 125 I-IGF-I was counted in a Wallac WIZARD gamma counter (PerkinElmer). K d values were calculated after fitting the data to a one-site model using a program developed in our laboratory by Ronald M. Shymko and Andreas V. Groth.

PREPARATION OF THE CELLS FOR THE MICROARRAY EXPERIMENTS
Mouse fibroblasts were seeded out into 145 cm 2 TC dishes (Nunc, Denmark) at two million cells per dish and subsequently allowed to recover for 24 h. In quadruplets, but at the same cell passage and after washing the cells twice with D-PBS (w/o Calcium and Magnesium; Gibco, Invitrogen, Denmark), the cells were serum starved for 24 h and afterward either left unstimulated or stimulated for 4 h with 20 nM IGF-I, 177 nM IGF-II, or 5168 nM insulin. These concentrations compensate for the relative affinities of the ligands for the receptor, measured as described above.

ISOLATION AND PURIFICATION OF TOTAL RNA
Total RNA was isolated by using the TRI® reagent method (Molecular Research Gene, USA) and cleaned up using the RNeasy™ Mini Kit (Qiagen) according to the manufacturers' protocol. RNA quality was verified by 1% agarose gel electrophoresis. Concentration and purity were determined by measuring absorbance at A 260 and A 280 in a spectrophotometer (Brinkmann Eppendorf BioPhotometer, Germany).

cRNA GENERATION AND HYBRIDIZATION TO GENE CHIP MICROARRAYS
cRNA was produced using the One-cycle Target Labeling Kit (Affymetrix, Santa Clara, CA, USA). One-cycle Target Labeling Kit and procedures followed protocols in the GeneChip Expression Analysis Technical Manual (Affymetrix, Santa Clara, CA, USA). Fragmented biotin-labeled cRNA was hybridized to Affymetrix GeneChip® Mouse Genome 430 2.0 Arrays according to manufacturer's protocol. The arrays were incubated at 45°C for 16 h under rotation (60 rpm), washed in the GeneChip® Fluidics Station (Affymetrix) and scanned using the GeneChip® Scanner 3000.

DATA ANALYSIS
The quality of the arrays was verified by quality control in the R package 1 from Bioconductor 2 . The probe level data (CEL files) were transformed into expression values using R and the GC-RMA package from Bioconductor (see text footnote 2) (26). Briefly, the background was subtracted, the data were normalized by the quantile normalization method and the expression values of a probe set were summarized into one expression value.
For data analysis, the expression values were imported into the software package DNA-Chip Analyzer (dChip) (version 2008), freeware developed by Li and Wong (27) 3 . When generating original lists of transcripts, a fold change and p-value cut-off of respectively 1.2 and 0.05 were chosen. The lower confidence bound of fold changes was used for filtering and the threshold for absolute difference between two group means was set to 35. Using these cut-offs gave empirical median false discovery rates (FDR) of maximum 2% after running 100 permutations in dChip (FDRs were 0% for all but the lists of genes regulated by insulin, IGF-I, or IGF-II in comparison to the control). dChip recommends a median FDR of <5 or 10%. Composing a list of transcripts regulated by insulin, IGF-I, and IGF-II together or separately was done by selecting transcripts that fulfilled the above-mentioned criteria for the ligands in comparison to the control. In order to generate lists containing transcripts only regulated by one of the ligands, transcripts were selected that fulfilled the criteria for one of the ligands in comparison to the control and in comparison to the two other ligands. Transcripts that also fulfilled the criteria for one of the other ligands in comparison to the control were excluded. The resulting transcripts, uniquely regulated by one of the ligands, were afterward filtered for a fold change of 1.5 in comparison to the control, in order to focus the below mentioned functional analysis on the transcripts with the highest biological relevance. In order to study differences between one ligand and the two other ligands as a group, transcripts were selected that fulfilled the criteria for the two ligands in comparison to the control and to the other ligand. The resulting transcripts were afterward filtered for a fold change of 1.5 in comparison to the control, in order to focus the below mentioned functional analysis on the transcripts with the highest biological relevance.
Identification of gene function themes and canonical pathways was done using the web-based software Ingenuity Pathways Analysis (IPA) 4 . IPA takes the gene IDs in the dataset file and maps them to genes in the Ingenuity Pathways Knowledge Base (IPKB). The functional and canonical pathway analyses identified the molecular and cellular functions and canonical pathways that were most significant to the data set. This significance value is a measure for how likely it is that genes from the dataset file under investigation participate in that function. In this method, the p-value is calculated by comparing the number of user-specified genes of interest that participate in a given function or pathway, relative to the total number of occurrences of these genes in all functional/pathway annotations stored in the IPKB. Ingenuity uses a right-tailed Fisher's Exact Test in order to calculate a pvalue. In the right-tailed Fisher's Exact Test, only over-represented functional/pathway annotations, annotations which have more Functions/Canonical Pathways Analysis Genes than expected by chance ("right-tailed" annotations), are used.

PREPARATION OF TOTAL RNA FOR qRT-PCR
To validate the microarray data two-step RT-PCR was performed on a subset of genes. To perform the validation on biological replicates, new (in comparison to the RNA used for the arrays) total RNA samples were prepared at three different cell passages.

qRT-PCR
The total RNA was reverse transcribed into single-stranded cDNA using the Transcriptor First Strand cDNA Synthesis Kit (Roche Applied Science) according to the manufacturer's protocol. The cDNA was transcribed using FastStart TaqMan Probe Master (Rox) (Roche Applied Science). Probes were purchased from Universal ProbeLibrary (Roche). Probes were selected and primer sequences designed using the ProbeFinder software (Universal ProbeLibrary, Roche). The primers were purchased from DNAtechnologies, Denmark. Primers and probes used are listed in Table 1. Per qRT-PCR assay the cDNA samples were run in  (28). To compare the qRT-PCR data with the microarray results, negative microarray fold changes were converted into values between 0 and 1. When multiple probe sets for one gene were regulated on the microarrays, the average fold change was calculated. Significant differences in the qRT-PCR data were calculated by a two-tailed t -test.

AFFINITIES OF IGF-I, IGF-II, AND INSULIN FOR THE IGF-I RECEPTOR
In order to stimulate the IGF-I receptor on mouse fibroblasts with concentrations that are adjusted for the relative affinities of IGF-I, IGF-II, and insulin for the receptor, the apparent affinities of the three ligands were measured by allowing the cold ligands to compete with 125 I-IGF-I for binding to the IGF-I receptor (Figure 1). IGF-I had a K d value of 1.49 ± 0.14 nM, IGF-II a K d value of 13.11 ± 0.69 nM, and insulin of 383 ± 27 nM. These results are in accordance with the known relative affinities of the ligands for the IGF-I receptor (29). Taking these relative affinities into account, it was decided to stimulate the cells for 4 h with 20 nM IGF-I, 177 nM IGF-II, or 5168 nM insulin, concentrations then are near saturation of the receptor with either ligand.

GLOBAL GENE REGULATION PATTERNS
A total of 698 transcripts were regulated by both insulin and the IGFs (fold changes and p-values for these transcripts are in Table  S1 in Supplementary Material). Table 2 shows the number of transcripts regulated by each ligand in comparison to the control and the number of transcripts commonly regulated between ligands. Fold changes and p-values for these transcripts can be found in Table S2 in Supplementary Material (IGF-I), Table S3 in Supplementary Material (IGF-II), and Table S4 in Supplementary Material (insulin). All the transcripts regulated in common between ligands were either up-regulated by all regulating ligands or down-regulated by all regulating ligands. Even though the three ligands stimulate similar responses, the overlap is partial and we identified transcripts selectively regulated by each ligand.

Transcripts selectively regulated by IGF-I
A total of 75 transcripts were only regulated by IGF-I (Table 3; fold change cut-off 1.5). Fold changes and p-values for insulin and IGF-II can be found in Table S5 in Supplementary Material. According to IPA the top five molecular and cellular functions these transcripts are involved in are molecular transport, protein trafficking, post-translational modification, protein folding, and cell morphology.

Transcripts selectively regulated by IGF-II
Eight transcripts were only regulated by IGF-II (see Table 4; fold change cut-off 1.5; for fold changes and p-values for insulin and IGF-I: see Table S6 in Supplementary Material). Two of these transcripts were TNF receptor-associated factor 1 (Traf1) and TRAF and TNF receptor-associated protein (Ttrap), which are functionally related proteins.

FIGURE 1 | Affinities of insulin, IGF-I, and IGF-II for the IGF-I receptor.
To determine the apparent affinities of the ligands for the IGF-I receptor on the mouse fibroblasts, homologous and heterologous radioligand competition assays were performed in quadruplets. Three million cells/ml were incubated with a constant concentration of 125 I-IGF-I (20,000 cpm/ml) and increasing concentrations of cold IGF-I, IGF-II, or insulin for 2.5 h (time needed to reach steady-state binding) at 15°C. After centrifugation unbound 125 I-IGF-I was removed and bound 125 I-IGF-I was counted in a gamma counter. Specifically bound 125 I-IGF-I/specifically bound 125 I-IGF-I at 0 nM cold ligand was plotted versus the concentration of cold ligand. K d values were calculated after fitting the data to a one-site model using a program developed by Ron M. Shymko and Andreas V. Groth. Results are averages ± standard deviations.

Transcripts selectively regulated by insulin
Four transcripts were only regulated by insulin (see Table 5; fold change cut-off 1.5; for fold changes and p-values for IGF-I and IGF-II: see Table S7 in Supplementary Material).

Transcripts selectively or more potently regulated by the IGFs than by insulin
Sixty five transcripts fulfilled the set criteria for IGF-I and IGF-II in comparison to the control and to insulin. The IGFs regulated  46 transcripts that were not regulated by insulin in comparison to the control ( Table 6). Interestingly, the 19 transcripts that were also regulated by insulin were always more regulated by the IGFs than by insulin. The top five molecular and cellular functions in IPA for these genes were cellular development, cellular growth and proliferation, cell cycle, gene expression, and cell death and survival. Two of the top five canonical pathways represented by these genes were ErbB signaling and neuregulin signaling. The regulated transcripts in these pathways were amphiregulin, epiregulin, heparin-binding EGF-like growth factor, FBJ osteosarcoma oncogene, and Jun oncogene for ErbB signaling and amphiregulin, epiregulin, heparin-binding EGF-like growth factor, ERBB receptor feedback inhibitor 1, and myelocytomatosis oncogene for neuregulin signaling.

Selective gene regulation by insulin and IGF-II
Twenty transcripts fulfilled the criteria for insulin and IGF-II in comparison to the control and to IGF-I ( Table 7). Fourteen of these were not influenced by IGF-I in comparison to the control, while they were either down-regulated or up-regulated by insulin and IGF-II.
The top five molecular and cellular functions in IPA for the 14 genes specifically regulated by insulin and IGF-II were cell cycle, cellular assembly and organization, DNA replication, recombination and repair, cellular function and maintenance, and cell morphology.

Gene regulation by insulin and IGF-I
Eleven transcripts fulfilled the criteria for insulin and IGF-I in comparison to the control and to IGF-II ( Table 8). In contrast to the selective gene regulation by the IGFs and by insulin and IGF-II, 10 of these 11 transcripts were also, and more strongly, influenced by IGF-II.

VALIDATION OF THE MICROARRAY DATA BY qRT-PCR
To validate the microarray data qRT-PCR was performed for six transcripts on the total RNA of three independent biological replicates. These RNA samples are independent of the RNA used to generate the microarray data. Fold changes were calculated in comparison to the control and plotted in Figure 2. For the IGFs, the regulation trends from the microarray experiments ( Table 6) are confirmed by qRT-PCR for all six genes: the IGFs regulate these genes more potently than insulin. For insulin, gene regulation ( Table 6) was confirmed for four out of six genes (Areg, Egr2, HB-EGF, and Jun-B). In addition, for Ccnd1 the fold change was 1.51 on the array and 1.46 by qRT-PCR, two values that lay very close and are only just separated by the 1.5 fold change cut-off. In conclusion, the qRT-PCR data validate very well the microarray results.

DISCUSSION
We compared the gene expression responses stimulated by insulin, IGF-I, and IGF-II through the IGF-I receptor using Affymetrix gene expression profiling. In order to eliminate the influence of the affinity of the ligands stimulating the receptor, we stimulated the IGF-I receptor on a mouse fibroblast cell line with concentrations of insulin, IGF-I, and IGF-II that compensated for the relative affinities of the ligands for the receptor on this cell line. Our analyses revealed that these three ligands stimulate overlapping but specific gene expression responses.
Some of the regulated transcripts that appeared in our analyses were also found by Mulligan  Boucher et al. recently showed that IGF-I and insulin, at equal concentrations, regulate the expression of the same genes through the IGF-I receptor (15). Insulin does that with a smaller magnitude of response than IGF-I. We show here that when compensating for the different affinities of the ligands, each ligand does specifically influence the expression of certain genes through the IGF-I receptor.
Each ligand specifically regulated a group of transcripts that was not regulated by the other two ligands. When stimulating the IGF-I receptor with IGF-II for example, two of the eight specifically regulated genes were Traf1 and Ttrap. Traf1 was up-regulated by IGF-II and is an inhibitor of apoptosis, which may be due to increased activation of nuclear factorkappa B (NF-κB), an anti-apoptotic transcription factor (32)(33)(34). Ttrap was down-regulated by IGF-II and inhibits the transcriptional activation of NF-κB (35). These results are consistent with www.frontiersin.org  the known anti-apoptotic activity of IGF-II through the IGF-I receptor.
In order to identify common gene regulation patterns between ligands, we studied the gene expression induced by two ligands in comparison to the control and to the third ligand. Interestingly, a group of 65 transcripts was identified to be selectively or more potently regulated by the IGFs than by insulin. ErbB signaling and neuregulin signaling were significant canonical pathways over-represented in the data set; regulated transcripts in common between the two pathways were amphiregulin, epiregulin, and heparin-binding EGF-like growth factor (HB-EGF). These were up-regulated more potently by the IGFs than by insulin. Pandini et al. showed that amphiregulin, HB-EGF, and epiregulin were similarly up-regulated by insulin and IGF-II through the insulin receptor isoform A in mouse fibroblasts (19). Mulligan et al. showed that HB-EGF transcript expression was up-regulated more potently after signaling through the IGF-I receptor than through the insulin receptor in fibroblasts (30). Amphiregulin, HB-EGF, www.frontiersin.org  and epiregulin are all EGF receptor (also named ErbB-1 or HER1) ligands (36). HB-EGF acts both as a regulated autocrine/paracrine and a juxtacrine growth factor (36,37). Amphiregulin has been suggested to have both growth inhibitory and stimulatory effects (38). Epiregulin is a growth promoter in primary rat hepatocytes (39,40) and an autocrine growth factor in human keratinocytes (41). HB-EGF and amphiregulin also bind and activate ErbB-3 and HB-EGF binds and activates ErbB-4 (42), just like the neuregulins, which bind ErbB-3 and ErbB-4. HB-EGF induces chemotaxis after stimulation of ErbB-4 (43).
As for the IGFs, we identified 14 transcripts selectively regulated by insulin and IGF-II. Using the same analysis criteria, this was however not the case when looking at insulin and IGF-I as a group. Ten of the 11 transcripts that were regulated by insulin and IGF-I in comparison to the control and IGF-II were also regulated by IGF-II. So the IGFs on one hand and insulin and IGF-II on the other hand seem to provoke more similar gene expression patterns than insulin and IGF-I. This is in accordance with the numbers presented in Table 2. Of all the transcripts regulated by insulin in comparison to the control, a larger fraction was also regulated by IGF-II than by IGF-I, even though IGF-I overall regulated more transcripts than IGF-II.
Although some of the transcripts identified in this study were involved in metabolic functions, the overall biological patterns were of a non-metabolic nature. This is not surprising, considering the tissue origin of the cell line used. From this study, no general conclusions could thus be drawn on whether certain ligands created a more metabolic or mitogenic response in comparison to the other ligands.
Many of the functions, pathways, and genes mentioned above are consistent with the known effects of insulin, IGF-I, and IGF-II. One could thus speculate that these differences in gene expression might account for some of the different biological effects of these three ligands. It should be mentioned that these gene expression patterns were measured after stimulating the receptor with supraphysiological concentrations of ligands. Therefore studying the concentration dependence of these gene expression profiles, together with performing time series of gene expression, could provide a more subtle picture.
Since the influences of affinity of the three ligands were largely accounted for in this study, it is likely that the differences in gene expression are due to intrinsic properties of each ligand. Different suggestions have been made to explain the mechanism responsible for this signaling specificity. Both differences in ligand binding kinetics and internalization properties have been correlated with different responses after stimulating the insulin receptor with different ligands (21)(22)(23)(44)(45)(46). More studies are needed in order to clarify at which level the cellular signal of different ligands stimulating the same receptor diverges.

CONCLUSION
We studied the gene expression patterns after stimulating the IGF-I receptor with equipotent concentrations of IGF-I, IGF-II, and insulin by microarray gene expression profiling and found significant differences in responses between the three ligands. Each ligand specifically regulated a group of transcripts that was not regulated FIGURE 2 | Validation of microarray data by qRT-PCR. Two-step RT-PCR was performed on a subset of transcripts. 18S was used as an internal control. The results are expressed as fold change in comparison to the control (unstimulated samples). Full bars represent the microarray data ( Table 6). Bars with patterns represent the average qRT-PCR results ± standard deviations. Black: insulin, gray: IGF-I, light gray: IGF-II. Significant differences in the qRT-PCR data were calculated by a two-tailed t -test. ∧ Significantly up-regulated by insulin in comparison to the control at the 1.5 fold change and 0.05 p-value level. *Significantly more up-regulated by this IGF than by insulin at the 0.05 p-value level.
by the other two ligands. Also, insulin and IGF-I seemed to stimulate the least overlapping response. The different gene expression profiles for the three ligands might explain some of their different biological effects. These results also add to the accumulating evidence that different ligands can bind to the same receptor and stimulate different cellular responses and that the nature of a ligand bound to a receptor, and not just its concentration and affinity, is determinant for the downstream cellular response. Further studies should help bringing a mechanistic understanding to the different functional consequences of different ligands activating the same receptor.

ACKNOWLEDGMENTS
We thank Susanne Smed and Elisabeth Schiefloe for help with scanning of the microarrays. The Hagedorn Research Institute and the Receptor Systems Biology Laboratory were independent basic research components of Novo Nordisk A/S. Soetkin Versteyhe, Jane Palsgaard, and Maja Jensen were the recipient of an Industrial Ph.D. scholarship from the Danish Ministry of Science, Technology and Innovation. Steven Gray was the recipient of a BIO + IT postdoctoral fellowship from the Oresund IT Academy.