Differential gene expression induced by insulin and insulin-like growth factor-II through the insulin receptor isoform A.

The human insulin receptor (IR) exists in two isoforms (IR-A and IR-B). IR-A is a short isoform, generated by the skipping of exon 11, a small exon encoding for 12 amino acid residues at the carboxyl terminus of the IR alpha-subunit. Recently, we found that IR-A is the predominant isoform in fetal tissues and malignant cells and binds with a high affinity not only insulin but also insulin-like growth factor-II (IGF-II). To investigate whether the activation of IR-A by the two ligands differentially activate post-receptor molecular mechanisms, we studied gene expression in response to IR-A activation by either insulin or IGF-II, using microarray technology. To avoid the interfering effect of the IGF-IR, IGF-II binding to the IR-A was studied in IGF-IR-deficient murine fibroblasts (R- cells) transfected with the human IR-A cDNA (R-/IR-A cells). Gene expression was studied at 0.5, 3, and 8 h. We found that 214 transcripts were similarly regulated by insulin and IGF-II, whereas 45 genes were differentially transcribed. Eighteen of these differentially regulated genes were responsive to only one of the two ligands (12 to insulin and 6 to IGF-II). Twenty-seven transcripts were regulated by both insulin and IGF-II, but a significant difference between the two ligands was present at least in one time point. Interestingly, IGF-II was a more potent and/or persistent regulator than insulin for these genes. Results were validated by measuring the expression of 12 genes by quantitative real-time reverse transcriptase-PCR. In conclusion, we show that insulin and IGF-II, acting via the same receptor, may differentially affect gene expression in cells. These studies provide a molecular basis for understanding some of the biological differences between the two ligands and may help to clarify the biological role of IR-A in embryonic/fetal growth and the selective biological advantage that malignant cells producing IGF-II may acquire via IR-A overexpression.

exon 11, a small exon encoding for 12 amino acid residues at the carboxyl terminus of the IR ␣-subunit. The relative abundance of the two IR isoforms is regulated by tissue-specific factors, stage of development, and cell differentiation (1)(2)(3). Genetic studies carried out in transgenic mice have shown that fetal growth in response to IGF-II is partially mediated by the IR (4 -6), and we have recently demonstrated that IR-A is the predominant isoform in fetal tissues and binds IGF-II with high affinity (7).
We also demonstrated that malignant transformation is associated with both IR overexpression and an increased relative abundance of IR-A, both in epithelial and in mesenchymal tumors (8 -13), and that IR-A relative abundance may further increase with cells dedifferentiation, as observed in thyroid cancer (12,14). Accumulating evidence also indicates that IR-A overexpression may play a significant role in growth promotion and apoptosis protection of malignant cells when tumors produce IGF-II (13,15). In contrast, IR-B is the predominant IR isoform in normal adult tissues that are major targets for the metabolic effects of insulin (adipose tissue, liver, and muscle) (1,2,16).
The binding characteristics of insulin and IGF-II to IR-A and the biological effects of IR-A stimulation by IGF-II have been studied previously in a variety of models (7). In particular, we studied IGF-II binding to the IR-A in IGF-IR-deficient murine fibroblasts (R Ϫ cells) transfected with the human IR-A cDNA (R Ϫ /IR-A cells). This study revealed that IGF-II displaces labeled insulin from IR-A with a lower affinity than insulin (ED 50 ϭ 2.5 versus 0.9, respectively). However, unexpectedly, IGF-II was a more efficacious mitogen than insulin in these cells. In contrast, insulin was more potent than IGF-II in stimulating glucose uptake (7). In accordance with our findings, it was independently shown that IGF-II is stronger than insulin in inducing growth in IR-transfected R Ϫ cells (17). These findings were confirmed and extended in SKUT-1 human rabdomyosarcoma cells, which lack functional IGF-IR and express almost only IR-A. In SKUT-1 cells, IGF-II was significantly more potent than insulin in stimulating the Shc/ERK pathway, whereas insulin was more potent than IGF-II in stimulating IR autophosphorylation and the IRS-1/phosphatidylinositol 3-kinase/Akt pathway. As a result, IGF-II was more potent than insulin in inducing cell chemoinvasion, whereas insulin was slightly more effective in apoptosis protection (13). Taken together, these studies indicate that insulin and IGF-II, by binding to the same receptor, may induce the preferential activation of different intracellular pathways. These differences may result in significant differences in the biological effects between the two ligands.
To gain further insights on the molecular mechanisms differentially activated by either IGF-II or insulin in R Ϫ /IR-A cells, we investigated gene expression in response to either ligand using microarray technology. Microarray techniques have emerged as a new potent approach for the global analysis of gene transcription. We used Affymetrix MG-U74A Gene-Chips to measure changes in mRNA levels for ϳ6,000 functionally characterized murine genes and ϳ6,000 expressed sequence tags (ESTs). We found that 45 genes are differentially transcribed in response to either insulin or IGF-II in R Ϫ /IR-A cells. We also validated these results by evaluating the expression profile of 12 genes by quantitative real-time reverse transcriptase-PCR. These findings provide a molecular basis for understanding the biological differences between insulin and IGF-II after binding to the same receptor.

EXPERIMENTAL PROCEDURES
Materials-The pNTK2 expression vector containing the cDNA for the A (Ex11Ϫ) isoform of the human IR was kindly provided by Dr. Axel Ullrich (Martinsried, Germany). Fetal calf serum, glutamine, Lipo-fectAMINE, DNAase I were from Invitrogen; RPMI 1640 medium, Dulbecco's modified Eagle's medium, bovine serum albumin (BSA, ra-dioimmunoassay grade), bacitracin, phenylmethylsulfonyl fluoride, puromycin, porcine insulin were from Sigma; IGF-II was obtained from Calbiochem Laboratories. TRIzol reagent and Superscript Choice system were purchased from Invitrogen; Oligotex mRNA kit and RNeasy Mini kit were obtained from Qiagen; BioArray HighYield RNA transcript labeling kit (ENZO Bioarray kit) was obtained from Affymetrix .
Cells-R Ϫ mouse fibroblasts (mouse 3T3-like cells derived from animals with a targeted disruption of the IGF-IR gene, expressing ϳ5 ϫ 10 3 native insulin receptors/cell) were kindly provided by Dr. R. Baserga (Philadelphia, PA) and were routinely grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. R Ϫ cells grown in 35-mm plates until 60 -70% confluent were co-transfected with 2 g of pNTK2 expression vector containing the cDNA encoding for the A (Ex11Ϫ) isoform of the human IR (18) and with the pPDV6ϩ plasmid encoding for the puromicyn resistance gene. Cells were subsequently subjected to antibiotic selection in medium supplemented with 2.4 g/ml puromicin for 3 weeks. Stably transfected cells were then cloned, and a cell clone with ϳ5 ϫ 10 5 receptors/cell was obtained, as described previously (7). Receptor content was evaluated in selected cell clones by enzyme-linked immunosorbent assay (10).  cRNA Preparation-R Ϫ /IR-A cells were grown until 80% confluent and serum-starved for 24 h. Cells were then stimulated with 10 nM of either insulin or IGF-II for 30 min or 3 or 8 h. Total RNA was isolated by TRIzol reagent, and mRNA was purified from total RNA using Oligotex mRNA kit, according to the protocol recommended by Affymetrix. mRNA (2 g) was then used to synthesize double-stranded cDNA by Superscript Choice system with T7-(dT)24 as a primer. Biotinlabeled cRNAs were in vitro transcribed using the ENZO BioArray kit and fragmented to produce a distribution of RNA fragments with size ranging from ϳ35 to 200 bases. Samples of fragmented cRNA (15 g) were hybridized for 16 h at 45°C to MG-U74A mouse arrays (Affymetrix). Analysis of the scanned chips was carried out using Affymetrix Microarray Suite version 5.0 (MAS5).

Differential Transcriptional Effects of Insulin and IGF-II
Data Treatment-Raw data from GeneChip microarrays were converted with the MAS5 software into a single, tab-delimited text file reporting, for each probe set, the "signal" and "detection" values from all experimental points. This file was subsequently processed with Microsoft Excel as follows. For each probe set, average signal was calculated across all experimental points. The average signal column was used to sort rows by increasing signal and to normalize individual microarray columns using a moving average (window of 200 probe sets) of increasing signal.
Such normalization corrected signal non-linearity and allowed comparison of any experimental point with any other. Stimulation points were compared with the controls or between each other through pairwise log 2 ratio calculation and averaging. Standard deviations (S.D.s) of these average log 2 ratios were also calculated. To obtain a more reliable estimate of variability, for each probe set, we also calculated the root mean square standard deviation (r.m.s.s.d.), encompassing all S.D.s of the average log 2 ratios. In fact, although the S.D. of a single duplicate comparison can easily be aberrantly high or low by chance, the r.m.s.s.d. from many duplicate comparisons is a more stable and reliable parameter.
An additional test was performed on these data, based on the "detection" call (present, absent, marginal). In synthesis, if a gene is induced in a certain experimental point, it must be called expressed in that point (not necessarily in the control). Otherwise, if it is suppressed, it must be called expressed in the control. At the end of this process, the following data were obtained for each gene: 1) normalized expression levels for all individual control and stimulated points; 2) average log 2 ratio for each experimental condition with respect to the control or to another experimental condition of choice; 3) S.D. for each average log 2 ratio, and r.m.s.s.d.; 4) call compatibility for each comparison. The first filter was the call compatibility, after which the other parameters were included and "tuned" in a statistical test aimed at identifying significantly regulated genes. The test requires that after subtraction of m*S.D. or of m*r.m.s.s.d., the average log 2 ratio is still higher than a threshold value of T. The tunable values in this test are m, the S.D./r.m.s.s.d. multiplier, and T, the threshold -fold change. To optimize test tuning, we systematically evaluated the false discovery rate, that is, the percentage of the sequences that could have passed the test by chance. False discovery can be estimated by generation, through data permutation, of mixed couples of microarray data that are not expected to display significant gene regulation. Existing microarray analysis tools such as significance analysis of microarrays (19) support data permutation. Differently from significance analysis of microarrays, our modified test weights overall variability of each probe set across all duplicates, which allows more reliable detection of tiny differences in gene expression. 2 We also implemented a permutation strategy and estimated the false discovery rate of our analysis based on 1,260 permutations. The test tuning parameters showing the best performance with the present data were T ϭ 0.4 and m ϭ 1.5, with which we could detect 259 regulated genes with an false discovery rate below 10%. Test tuning for identifying genes differentially regulated by insulin and IGF-II was slightly different, with T ϭ 0.4 and m ϭ 1. The false discovery rate above 10% indicated the necessity for real-time PCR validation of these data.
As a control of data robustness, we also used dChip (20) to normalize the data. We saw a lower coefficient of variation in dChip-normalized triplicates and could confirm Ͼ90% of the genes originally identified on MAS5-normalized data as regulated by insulin and/or IGF-II. Interestingly, dChip normalization rendered non-significant the regulation of a gene we had already validated by real-time PCR, which indicates that different normalization procedures may also yield non-overlapping false negatives. We therefore decided to make available for download the two spreadsheets containing, respectively, MAS5-normalized and dChip-normalized data.
The analysis spreadsheets and the raw CEL files can be downloaded (www.ircc.it/ϳemedico/FOG/data). The original data will also be submitted to the NCBI's Gene Expression Omnibus public data base (www.ncbi.nlm.nih.gov/geo) at a later date. Further information is available from the authors on request.
Hierarchical Clustering-Hierarchical clustering of the selected genes was performed using the computer program Cluster (21) and visualized using the program TreeView (available at rana. stanford.edu/software).
Real-time Polymerase Chain Reaction-Primer Express software (PE Applied Biosystems, Foster City, CA) was used to design appropriate primer pairs and fluorescent probes. Primer pairs and probes with 5Ј-FAM reporter dye and 3Ј-TAMRA quencher dye were synthesized by MWG-Biotech (Ebersberg, Germany). Probe and primers for endogenous control (glyceraldehyde-3-phosphate dehydrogenase) were from predeveloped TaqMan assay reagents (Applied Biosystems). Quantitative realtime PCR was performed on Abi Prism 7700 (PE Applied Biosystems) using Sybr Green PCR Master Mix and Taqman Universal PCR Master Mix (PE Applied Biosystems) following manufacturer's instructions. To normalize gene expression, a parallel amplification (six replicates) of endogenous and target genes was performed with Sybr Green reagents. For Taqman analysis, all reactions (six replicates) were performed by co-amplifying in the same tube endogenous and target genes. To check reaction sensitivity, in preliminary experiments, serial dilutions of each cDNA (1, 1:10, 1:100, 1:1,000) were amplified for endogenous and target genes. The reaction efficiency resulted similar in simplex and duplex reactions (i.e. slope ϭ 3.6 Յ x Ն3; correlation coefficient Ն0.99). Relative quantitative evaluation (PE Applied Biosystems user bulletin number 2) of target gene levels was performed by comparing ⌬Ct, as described previously (22).

Genes Regulated by Insulin and/or IGF-II in R Ϫ /IR-A Cells
To analyze gene expression profiles following IR-A activation by either insulin or IGF-II, R Ϫ /IR-A cells were stimulated with either ligand (10 nM) for various time intervals (0.5, 3, and 8 h). Biotinylated cRNA probes were generated from the RNA extracted from control and ligand-stimulated cells and hybridized to microarray membranes containing the entire mouse genes, according to Affymetrix procedure. Using the analysis strategy  Interferon activated gene 202 Zinc finger protein 216 Growth factor-inducible immediate early gene Ϫ2.9

Genes Similarly Regulated by Both Insulin and IGF-II
Two hundred and fourteen genes and ESTs were regulated with a similar pattern by both insulin and IGF-II (Fig. 1A). Sixty genes (Table I) and 52 ESTs were similarly up-regulated by the two hormones, whereas 48 genes (Table II) (Tables I and II).
In some genes, up-or down-regulation was transient and only detected at one time point, whereas it was persistent in other cases, as indicated by cluster analysis (Fig. 1A). The

Differential Transcriptional Effects of Insulin and IGF-II
number of genes and ESTs similarly up-regulated or downregulated by the two ligands at the different time points is indicated in Fig. 2.

Genes Differentially Regulated by Insulin and IGF-II
Comparative analysis of the ϳ6,000 genes and ϳ6,000 ESTs on the cDNA microarrays revealed that 45 transcripts (27 genes and 18 ESTs) were differentially regulated by insulin and IGF-II.
Transcripts Regulated by Only One Ligand-Eighteen of these differentially regulated genes (10 genes and 8 ESTs) were responsive to only one of the two ligands. Twelve transcripts (7 genes and 5 ESTs) responded only to insulin (3 up-regulated and 9 down-regulated, Table III), whereas 6 transcripts (3 genes and 3 ESTs) responded only to IGF-II (5 up-regulated and 1 down-regulated; see Table III).
Three genes selectively up-regulated by insulin are genes involved in angiogenesis regulation and differentiation: mrp/

FIG. 3. Expression patterns in genes differentially regulated by insulin and IGF-II.
The expression profile of 16 differentially regulated genes fell into one of the three patterns shown. In A, genes were persistently up-regulated by both ligands with IGF-II being more potent, especially at 8 h. In B, genes were up-regulated by both ligands; gene expression, however, remained above basal levels after IGF-II, whereas it decreased significantly after insulin. In C, genes were persistently down-regulated by both ligands with IGF-II being more potent than insulin at 8 h.
plf, proliferin, and Hox-3.1. Mrp/plf and proliferin are highly homolog proteins that belong to the superfamily of prolactin/ growth hormone and act as endogenous regulators of angiogenesis. These factors play a role in blood vessel formation and remodeling by affecting endothelial cell proliferation, apoptosis, and migration. However, these two factors have opposite functions as mrp/plf is an inhibitor of angiogenesis, whereas proliferin is a promoter of angiogenesis (23)(24)(25)(26). Hox-3.1 is a gene of the homeobox family that is involved in development regulation. Genes down-regulated by insulin include transcripts encoding for TGF␤1, a negative growth regulator, and for TAK1 (TGF-␤-activated kinase), a caspase-independent antiapoptotic factor (27)(28)(29)(30). Mesenchyme fork head-1, a transcription factor that affects adipocyte metabolism and that is increased by high fat diet, seems to counteract most of the symptoms associated with obesity, including hypertriglyceridemia and diet-induced insulin resistance, a protection against type 2 diabetes (31,32).
Transcripts that are responsive and up-regulated only by IGF-II include GNB-1-like EST and the Bat-4 gene. GNB-1 encodes for a protein that releases cGMP-phosphodiesterase from inhibition (33,34). Bat-4 encodes for a protein of the family of the major histocompatibility complex. Only one gene was responsive only to IGF-II and down-regulated, protein kinase C , which plays a role in insulin receptor signaling, differentiation, and survival of T-cells and in multiple processes essential for angiogenesis, regulation of cell cycle progression, and formation and maintenance of actin in the cytoskeleton (35)(36)(37).
Transcripts Regulated by Both Ligands but with a Different Time Pattern-Twenty-seven transcripts (17 genes and 10 ESTs) were regulated by both insulin and IGF-II but showed, at least in one time point, a significant difference between the two ligands ( Table IV ). The expression profile of 16 of these transcripts followed three major patterns: pattern a transcription was persistently up-regulated by both ligands with IGF-II being more potent; pattern b transcription was transiently up-regulated by both ligands; it persisted, however, above basal levels after IGF-II, whereas levels after insulin were significantly lower; pattern c transcription was persistently down-regulated by both ligands with IGF-II being more potent than insulin.
Five genes fell in pattern a (Fig. 3). These genes encode for: acidic nuclear phosphoprotein 32, a protein that is expressed at high levels during embryogenesis and decreases as tissues terminally differentiate (38 -40); RhoGDI-1, a GTP-binding protein; surfeit 6, a nucleolar matrix protein ubiquitously expressed with nucleic acid binding properties (41)(42)(43); transcription-associated zinc ribbon protein, a protein that binds to transcription factors and to DNA; helicase-lymphoid-specific, a protein involved in DNA repair.
Eight genes followed pattern b (Fig. 3). Up-regulation of these genes was more persistent following IGF-II than following insulin. Two of these genes play an important role either in cell-substrate adhesion (integrin-␣ 5 ) or in cell-to-cell interaction (ICAM). One gene is a regulator of G-protein signaling (retinally abundant regulator G-protein). TDAG51 is a gene involved in apoptosis regulation of T-cells that is required for Fas expression and is also involved in apoptosis resistance and growth dysregulation in cancer, as shown in the melanoma model (44). Three genes in this group are involved in various aspects of the metabolism: cholesterol metabolism (7-dehydrocholesterol reductase), phosphate transport (Glvr-1), selenium supply, and prevention of oxidative stress (selenoprotein P).
Three genes followed pattern c: these included topoisomerase inhibitor-suppressed and two negative regulators of cell cycle, cyclin G2 and Gas1. These genes were more potently downregulated by IGF-II with a significant difference at 8 h (Fig. 3).

Validation of Microarray Data
To validate the microarray transcription data, 12 genes belonging to different functional categories were selected for realtime PCR confirmation. To minimize data variation, the same cDNA samples used for microarray analysis were used also for real-time PCR. Genes selected and primer pairs used are reported in Table V. At least two independent experiments were carried out for each gene. In these 12 genes, the magnitude of the regulation obtained by real-time PCR was similar to that obtained by microarray analysis. The expression profile of 6 representative genes by the two techniques is shown in Fig. 4. Data obtained in the 12 genes of the validation group at the different time points by either microarrays or real-time PCR are highly correlated (r 2 ϭ 0.7096, p Ͻ 0.0001, Spearman correlation) (Fig. 5).

DISCUSSION
Both IR isoforms (IR-A and IR-B) are expressed by most human tissues (1, 2). The relative abundance of these two FIG. 4. Validation of microarray data by quantitative real-time PCR. Among the 45 genes differentially regulated by insulin and IGF-II, 12 genes were analyzed also by real-time PCR (A and B). The -fold induction was computed using a standard curve analysis and normalized to the level of glyceraldehyde-3-phosphate dehydrogenase as described under "Experimental Procedures." All PCR reactions were performed at least in duplicate. The expression profile after either insulin or IGF-II stimulation (-fold change, mean Ϯ S.D.) is shown for 6 representative genes and compared with data obtained by microarray analysis. isoforms is strictly regulated by tissue-specific factors (1,2). In general, IR-B is more expressed than IR-A by classical insulin target tissues (1,2). Recently, we have tried to clarify the biological role of IR-A. We found that IR-A binds not only insulin but also IGF-II with high affinity and behaves as a second physiologic receptor for IGF-II in fetal cells (7). Moreover, IR-A was found overexpressed both in epithelial and mesenchymal malignancies, including cancer of the breast, colon, lung, thyroid, and myosarcomas (11)(12)(13). In these malignancies, IR-A overexpression activates an autocrine/paracrine loop involving IGF-II that appears to promote growth, protection from apoptosis, and cancer progression (12). Recently, an independent study has extended these findings to ovary cancer (15). Furthermore, we have shown that IR-A overexpression profoundly affects the binding and signaling specificities of IR/IGF-IR hybrids (hybrid-R), heterodimeric receptors containing an IR, and an IGF-I-R hemireceptor, which are abundant in all tissues that express both IR and IGF-IR (10,45). Taken together, these data suggest that IR isoform switching is a major regulator of the IGF system (45).

Differential Transcriptional Effects of Insulin and IGF-II 42187
A puzzling finding was the observation that, after IR-A binding, IGF-II is a stronger mitogen than insulin, whereas insulin is a stronger activator of glucose uptake (7). For these studies, to avoid the interference of IGF-IR that may also bind IGF-II with high affinity, we used R Ϫ mouse embryo fibroblasts, cells with disrupted igf-1r by homologous recombination, which express low endogenous IR levels (17,46). When transfected with human IR-A cDNA, these cells (R Ϫ /IR-A cells) expressed the human IR-A at high level (ϳ5 ϫ 10 5 /cell). Such a model avoids cross-reactivity of ligands with receptors other than IR-A.
By using the same model, we now show that upon binding to IR-A, insulin and IGF-II cause a differential gene expression. To analyze the physiological profile of gene expression by microarrays, we exposed these cells to a low dose (10 nM) of each of the two hormones and carried out a three-point time course (30 min and 3 and 8 h) as it is known that the expression of genes may vary with the time elapsed after stimulation. We addressed in a careful way the problem of data variability, a limitation of microarray analysis. To deal with variability, we obtained data from two separate experiments, each containing three time points, and used stringent selection criteria that allowed us to define as significant a 1.3-fold difference in the transcript relative abundance with a very small probability that a gene classified as regulated could be a false positive. In accordance with these selection criteria, it was recently reported that a Ͻ2-fold variation in gene expression can be biologically relevant (47).
Out of ϳ6,000 genes and ϳ6,000 ESTs, a total of 259 genes and ESTs were regulated by one or both ligands (insulin and IGF-II). The majority of these genes (n ϭ 214) were similarly regulated by insulin and IGF-II. The magnitude of gene expression variation ranged from ϩ1.3 to ϩ4.2 and from Ϫ1.3 to Ϫ7.9, values similar to those reported by other authors (47,48).
Two recent studies have reported microarray analysis of gene expression in IR-transfected cells (47,48). However, these authors have used a different cell model and different stimulation and analysis modalities. Dupont et al. (48) used NIH-3T3 cells transfected with IR and stimulated cells for 90 min. Mulligan et al. (47) used 3T3-L1 cells transfected with a TrkC/IR chimeric receptor and stimulated cells for 4 h. In addition, both studies addressed the differences in gene expression between insulin and IGF-I on their cognate receptor, and none evaluated the IR-A isoform. Many genes stimulated by insulin in our model had not been reported previously (47,48). Interestingly, 6 genes (integrin ␣ 5 , early growth response 1, Jun oncogene, IGF binding protein 10, T-cell death-associated gene 51, immediate early protein Gly96) were reported previously to be regulated only by IGF-I in cells overexpressing the IGF-IR but not in cells overexpressing the IR (48). The reason for this discrepancy is unclear. Possible explanations may include the different experimental system and methodology. Moreover, one intriguing possibility is that IR-B and IR-A isoforms have a different capability to regulate gene transcription.
Remarkably, 45 genes and ESTs were differentially expressed in response to either insulin or IGF-II. These genes were mainly involved in cell cycle and apoptosis regulation, angiogenesis, cell-to-matrix or cell-to-cell interaction, and signal transduction. In 12 of them (validation group), the relative abundance of differentially regulated genes was also confirmed by quantitative real-time PCR with expression profiles very similar to those obtained by the microarray analysis.
Eighteen of these transcripts (40%) were responsive to only one ligand; 12 of them were responsive only to insulin, and 6 were responsive only to IGF-II. The remaining 27 differentially regulated genes (60%) responded to both ligands but with a different magnitude or a different time course; in all of them, IGF-II was more potent than insulin either in inducing upregulation or in inducing down-regulation.
The greatest difference in favor of IGF-II was observed for acidic nuclear phosphoprotein 32 (ANP32) gene (40). ANP32 is a protein involved in embryogenesis that is switched off in differentiated cells. Although ANP32 is not an oncogene and can actually inhibit transformation, it increases resistance to apoptosis (38 -40). In humans, several ANP32 isoforms have been described; one of these isoforms is increased in prostate cancer and has been suggested to play a role in this malignancy (38 -40).
Two genes, ␣ 5 ␤ 3 integrin and TDAG51, reported previously to be responsive to IGF-I and not to insulin, in our system demonstrated themselves to be more strongly regulated by IGF-II than by insulin. ␣ 5 ␤ 3 integrin has been recently described to be an important determinant of the IGF-IR activation (49) and to play a role in DNA synthesis, cell proliferation, and migration. Previously, we have shown that IR-A preferential expression, in contrast to IR-B predominant expression, plays a role in the activation of the IGF-I system by multiple mechanisms, including direct binding of IGF-II and formation of IR/IGF-IR hybrid receptor (hybrid-R A ), that bind both IGFs with high affinity and insulin with low affinity (45). Up-regulation of ␣ 5 ␤ 3 integrin by IR-A phosphorylation (more potently stimulated by IGF-II) may be an additional mechanism contributing to the activation of the IGF system (50). TDAG51 is a positive regulator of apoptosis in T-cells (44), and transcriptional up-regulation of this molecule by two antiapoptotic factors such as insulin and IGF-II may be an unexpected finding. However, although considered proapoptotic, TDAG51 is also up-regulated by IGF-I (48) and platelet-derived growth factor (51) through the activation of the Akt pathway (51), which is strongly involved in survival. TDAG51 regulation by IR-A may reflect a mechanism common to different growth factors and may underscore the fact that apoptosis is also regulated by the IR-A isoform, by both proapoptotic and antiapoptotic factors, as shown previously for IGF-IR (44,52,53).
The gene encoding for 7-dehydrocholesterol reductase was also significantly more responsive to IGF-II than to insulin. The protein encoded by this gene is a key enzyme in the metabolism of cholesterol but is also an important factor in morphogenesis. Mutations of this gene lead to the Smith-Lemli-Opitz syndrome, characterized by multiple congenital anomalies such as microcephaly, cleft palate, visceral malfor- mations, postnatal failure to thrive, and mental retardation (54 -56). This observation is consistent with the important role of IGF-II during embryogenesis and fetal development (4,5). Among genes responsive to IGF-II more than to insulin are also selenoprotein P, a selenium supply protein involved in the prevention of oxidative stress (57)(58)(59), and the regulator of G-protein signaling. The preferential stimulation of these genes by IGF-II may reflect the protective role of IGF-II on cell survival and G-protein receptor differential regulation by the two ligands. Finally, two genes involved in the negative control of cell cycle (cyclin G2-and growth arrest-specific 1) were more strongly down-regulated by IGF-II than by insulin.
In contrast, Mrp/plf and proliferin, both involved in angiogenesis regulation, were significantly up-regulated only by insulin and not by IGF-II. These results have not been reported before and may indicate a possible, previously unrecognized, role of IR-A in the control of angiogenesis.
In conclusion, we show for the first time that insulin and IGF-II may differentially affect gene expression in cells expressing the fetal IR isoform and lacking the IGF-IR. Although some genes appear to be regulated only by insulin, 6 genes are selectively regulated by IGF-II, and other genes show a more potent and/or persistent regulation after IGF-II than after insulin. These findings may appear surprising when considering that IR-A binds IGF-II with a lower affinity than insulin (7). However, these data are in agreement with our previous data showing that IGF-II is a more potent mitogen than insulin in R Ϫ cells transfected with the IR-A (7). Similar data were also reported by Morrione et al. (17). Moreover, we also previously found that IGF-II is more potent than insulin in stimulating chemoinvasion in SKUT-1 leiomyosarcoma cells that overexpress IR-A and lack IGF-IR (13). In cells that express both IR and IGF-IR, the relative abundance of the two IR isoforms affects the relative potency of insulin and IGF-II signaling to hybrid-Rs, as hybrid-R A binds IGF-II much better than insulin (45), and according to the present data, we may expect that the two ligands elicit partially different transcriptional effects also via hybrid-R A . These findings may help to clarify the biological role of IR-A in embryonic/fetal growth and the biological advantage that malignant cells acquire by overexpressing IR-A.