The Nuclear Receptor Coactivator AIB1 Mediates Insulin-like Growth Factor I-induced Phenotypic Changes in Human Breast Cancer Cells

The nuclear receptor coactivator AIB1 belongs to the p160/SRC (steroid receptor coactivator) family consisting of SRC-1 (1), TIF-2 (GRIP1; ref. 2) and AIB1 (ref. 3; ACTR/RAC3/TRAM-1/SRC-3; refs. 4, 5, 6, 7). The AIB1 gene is amplified in several human cancers, such as breast, ovarian, pancreatic, and gastric cancer (3, 8, 9). Amplification of the AIB1 gene is detected in 5 to 10% of primary breast tumors and AIB1 is highly expressed in many breast tumor specimens (3, 10, 11, 12). AIB1 enhances in vitro the transcriptional activity of the estrogen receptor (ER; refs. 3, 4, 7) and binds directly to ER in vivo (13). Furthermore, AIB1 is rate-limiting for estrogen-mediated growth of MCF-7 human breast cancer cells (14). AIB1 gene expression is up-regulated by selective ER modulators, such as tamoxifen (15) and the estrogenic activity of selective ER modulators, can be increased by AIB1 and an AIB1 isoform (16). However, emerging data suggest that the role of AIB1 is not restricted to nuclear receptor signaling. Disruption of p/CIP, the mouse homologue of AIB1, results in a pleiotropic phenotype, including reduced female reproductive function and blunted mammary gland development in mice (17, 18). Interestingly, embryonic tissues from p/CIP-knockout mice show severe defects in the insulin-like growth factor I (IGF-I) and growth hormone-signaling pathways (18). Consistent with this role for AIB1 in growth factor signaling, we have found that an isoform of AIB1 (Δ3-AIB1) overexpressed in breast tumors strongly enhances epidermal growth factor-mediated transcription in squamous cell carcinoma cells (19). Also, AIB1 overexpression is positively correlated with the expression of p53 and HER2/neu in breast tumors (20). p/CIP also plays a role in CREB binding protein–dependent transcriptional activation induced by IFN-γ and 12-O-tetradecanoylphorbol-13-acetate (21), and a study of Taiman, the Drosophila homologue of AIB1, indicates that AIB1 is also involved in the control of cell motility as well as platelet-derived growth factor/vascular endothelial growth factor signaling (22, 23). Taken together, these data suggest that AIB1 may well be an important factor for growth factor-mediated signaling pathways. In this study, we report that selective reduction of endogenous AIB1 levels in MCF-7 cells reveals a significant role for this coactivator in IGF-I signaling in human breast cancer cells.

MCF-7 and MDA MB-231 human breast cancer cells were maintained in Improved Modified Eagle’s Medium (IMEM; Invitrogen, Carlsbad, CA) with 10% fetal bovine serum. Recombinant human IGF-I (R&D Systems, Minneapolis, MN) was resuspended in 10 mmol/L acetic acid +0.1% bovine serum albumin and used at 100 ng/mL (13 nmol/L). ICI 182,720 (Tocris Cookson, Ellisville, MO) was resuspended in etomidate and used at a concentration of 10 nmol/L.

Twenty-one–mer oligoribonucleotides of sense and antisense RNA strands were synthesized corresponding to nucleotides 564–582 of the AIB1 coding region. The AIB1 sense oligoribonucleotide sequence as follows: r(GGUGAAUCGAGACGGAAAC)dTT. The AIB1 antisense oligoribonucleotide sequence as follows: r(GUUUCCGUCUCGAUUCACC)dTT. The control siRNA is a scrambled sequence and does not target any known mammalian mRNA (Qiagen, Valencia, CA). The control sense oligoribonucleotide sequence as follows: r(UCCGUUUCGGUCCACAUUC)dTT. The control antisense oligoribonucleotide sequence as follows: r(GAAUGUGGACCGAAACGGA)dTT (Qiagen). The lyophilized double-stranded RNA was reconstituted in 1 mL of annealing buffer [100 mmol/L potassium acetate, 30 mmol/L HEPES-KOH, 2 mmol/L magnesium acetate (pH 7.4)], giving a final concentration of 20 μmol/L. The solution was then heated for 1 minute at 90°C and incubated at 37°C for 60 minutes.

For siRNA experiments, 24 hours before transfection, MCF-7 cells were plated in a 10-cm dish at 50% confluency in IMEM +10% FBS. For each transfection, 60 μL of 20 μmol/L AIB1 siRNA were diluted with 1 mL of IMEM plus 60 μL of Oligofectamine (Invitrogen). The siRNA and Oligofectamine solutions were allowed to complex at room temperature for 15 minutes before treatment of the cells. After washing, the siRNA-Oligofectamine complex was added to the cells in 5 mL of IMEM and incubated for 4 hours at 37°C. A total of 1.5 mL of IMEM +30% FBS was added to the transfected cells and incubated for 16 to 18 hours at 37°C. For proliferation assays, the transfected cells were trypsinized and placed into 96-well plates and treated with IMEM +1% charcoal-stripped calf serum (CCS) containing 100 ng/mL IGF-I or 10% FBS (serum). Cell number was determined by a WST-1 colorimetric assay (Roche Diagnostics, Indianapolis, MN). For cell cycle or apoptosis analysis, the transfected cells were trypsinized and placed into 60-mm dishes and treated with IMEM +1% CCS with 100 ng/mL IGF-I or 10% FBS for 24 hours. Cell cycle analysis was done with the Vindelov method of nuclei preparation for flow cytometry DNA analysis. The percentage of cells in early and late apoptosis (percent cell death) was analyzed by staining with Annexin V-FITC (Trevigen, Inc., Gaithersburg, MD).

MCF-7 cells were transfected with either AIB1 or control siRNA as described from the siRNA proliferation assays. After 16 to 18 hours, 7000 cells were resuspended in 0.35% soft agar and layered on top of 1 mL of 0.6% solidified agar in a 35-mm dish with 100 ng/mL IGF-I or 10% FBS (serum). IMEM +1% CCS were included in both layers. The soft agar colonies were allowed to grow at 37°C for 10 to 15 days. Cell colonies with a diameter of ≥80 μm were counted with an image analyzer (Omnicon TCA, Biologics, Gainesville, VA). Experiments were carried out in triplicate.

To prevent cell attachment, 60-mm dishes were coated with 10 mg/mL poly(2-hydroxyethyl methacrylate; poly-HEMA, Sigma-Aldrich, St. Louis, MO) diluted in etomidate and allowed to dry completely. Estrogen-stripped MCF-7 cells were transfected with AIB1 or control siRNA as described above for the proliferation assays. After 48 hours, the transfected cells were plated onto the poly-HEMA coated dishes in IMEM +1% CCS with 100 ng/mL IGF-I or 10% FBS (serum). After 24 hours, the cells were harvested either for cell cycle analysis or apoptosis analysis as described above. Each experiment was carried out in duplicate.

Attached cells were transfected with siRNA and treated with growth factors for 48 hours. Unattached cells (poly-HEMA), a pool of MCF-7 or MDA MB-231 cells, was transfected with siRNA under normal adherent culture conditions. After 24 hours, the transfected cells were trypsinized, washed in 5% CCS + IMEM, and plated in poly-HEMA coated dishes for 24 hours with appropriate treatment media. Cells were then washed with cold PBS, lysed in 50 mmol/L Tris-HCl (pH 8.0), 150 mmol/L NaCl, 40 mmol/L β-glycerophosphate-Na, 0.25% Na-deoxycholate, 1% NP40, 50 mmol/L NaF, 20 mmol/L Na PP_I, 1 mmol/L EGTA, 1 mmol/L Na₃VO₄, and 1× complete protease inhibitor (Roche Diagnostics) and incubated on ice for 20 minutes. The lysate was centrifuged at 10,000 × g at 4°C for 15 minutes. The lysate was boiled in SDS-PAGE buffer with reducing agents, proteins were resolved by electrophoresis on a 4 to 20% Tris-glycine gel, transferred to a polyvinylidene difluoride membrane, and the membrane was incubated for 1 hour at room temperature with 5% milk in PBST (PBS, 0.2% Tween 20). Primary and secondary antibodies were diluted in 5% milk in PBST, and incubations were done at room temperature for 1 hour. The antibodies used were raised against: AIB1 (BD Transduction Laboratories, San Jose, CA); β-actin (Chemicon International, Inc., Temecula, CA); IRS-1 (Upstate, Lake Placid, NY); IRS-2 (Upstate); ER-α clone 1D5 (DakoCytomation, Carpinteria, CA); cyclin D1 (Neomarkers/Lab Vision, Fremont, CA); Bcl-2 (Santa Cruz Biotechnology, Santa Cruz, CA); and IGF-I receptor (IGF-IR) α, HER2/erbB2, phospho-AKT (Ser⁴⁷³), AKT, phospho-p44/42 MAPK (Thr²⁰²/Tyr²⁰⁴), and p44/42 MAPK (Cell Signaling Technologies, Beverly, MA). Relative band intensities were assessed with densitometry and corrected for β-actin loading.

Total RNA was extracted and DNase treated with the RNeasy Mini kit (Qiagen). Real-time reverse transcription-PCR was done with the SuperScript One-Step Reverse Transcription-PCR with Platinum Taq system (Invitrogen). Samples were reverse transcribed for 30 minutes at 58°C, followed by a denaturing step at 95°C for 5 minutes and 40 cycles of 15 s at 95°C and 1 minute at 58°C. Fluorescence data were collected during the 58°C step with the Cycler iQ Detection System (Bio-Rad Laboratories, Hercules, CA). The primers and probes for real-time reverse transcription-PCR measurement were as follows: AIB1 forward primer, 5′-CAGTGATTCACGAAAACGCA-3′; AIB1 reverse primer, 5′-CAGCTCAGCCAATTCTTCAAT-3′; AIB1 probe, 6FAM-TGCCATGTGATACTCCAG AAG-Black Hole Quencher 1 (BHQ1); glyceraldehyde-3-phosphate dehydrogenase forward primer, 5′-CCCACATGGCCTCCAAGGAGTA-3′; glyceraldehyde-3-phosphate dehydrogenase reverse primer, 5′-GTGTACATGGCAACTGTGAGGAGG-3′; and glyceraldehyde-3-phosphate dehydrogenase probe, 6FAM-ACCCCTGGACCAGCCCCAGC-TAMRA.

The method used to prepare samples for cDNA microarray analysis is outlined in the Affymetrix Gene Chip Expression Analysis Technical Manual, Section 2: Eukaryotic Sample and Array Processing. Total RNA was harvested from estrogen-stripped MCF-7 cells transfected with AIB1 or control siRNA for 48 hours (RNeasy Mini kit, Qiagen). Twenty micrograms of total RNA were used to synthesize double-stranded cDNA. T7 oligo(dT) primers [5′-GGCCAGTGAATTGTAATACGACTCACTATAGGGAGGCGG-(dT)₂₄-3′] were used to prime the first-strand cDNA synthesis. Synthesis of biotin-labeled cRNA from double-stranded cDNA was done with the Enzo BioArray High Yield RNA Transcript Labeling kit (Enzo Life Sciences, Farmingdale, NY). Twenty micrograms of fragmented biotin-labeled cRNA were hybridized to the Affymetrix Human Genome U133A GeneChip (Affymetrix, Santa Clara, CA). MCF-7 cells either transfected with the control or AIB1 siRNA were analyzed using four U133A chips.

For each of the eight arrays, measurements from 22,283 genes were obtained. For each gene, an indicator of its expression level is given as either present, absent, or marginal call because it is assigned by Affymetrix Microarray suite software. Gene expression measurements that were <10 were given a threshold value of 10. Log base 2 transformations were applied to the gene expression values to reduce variation and to make the data more normally distributed. Each gene that had an absent call from five or more of eight measurements was eliminated. This reduced the data set to 12,057 genes. The following analysis was done on the 12,057 genes with BRB-Array Tools (V.3.0.1). The intensity values were first normalized such that the median log intensity of each of the eight arrays is equal to the median log intensity for the reference array (BRB array tools arbitrarily choose the second array to be the reference array). Two sample t tests were carried out to compare the expression intensity between AIB1 high (control siRNA transfected) and AIB1 low (AIB1 siRNA transfected) arrays for each of the 12,057 genes with four arrays per group using the randomized variance model. A total of 124 genes showed a statistically significant difference between AIB1 high and AIB1 low at P ≤ 0.0025 from the univariate test. The 124 genes include genes that were up or down-regulated by AIB1. The genes were also ranked according to P value.

To study the role of AIB1 in IGF-I signaling, we used siRNA directed at nucleotides 564–582 of AIB1 to selectively reduce AIB1 gene expression in the MCF-7 breast cancer cell line. This region bears no significant homology to other coactivators or sequences in the human genome database. The AIB1 siRNA reduced the AIB1 mRNA (Fig. 1,A) and AIB1 protein in a concentration-dependent fashion (Fig. 1,B). The siRNA concentration used in our experiments, 2 × 10^-7 mol/L, produced a >80% reduction in cellular AIB1 protein levels (Fig. 1,B). In addition, after a single transfection of siRNA, we found that the AIB1 protein levels were still repressed after 6 days (Fig. 1,C). A control-scrambled sequence siRNA had no effect on AIB1 gene expression (Fig. 1 A–C). The expression of a number of unrelated proteins such as actin or HER2 was unaltered by either siRNA. Thus, application of AIB1-directed siRNA was an effective and selective method of long-term suppression of endogenous AIB1 levels, enabling experiments to determine the impact of reducing endogenous AIB1 on MCF-7 cells.

An important growth factor induced phenotypic change, and a hallmark of malignant transformation is the ability of cells to form colonies in soft agar. This anchorage-independent growth involves cell cycle regulatory genes, as well as genes that prevent apoptosis under anchorage-independent conditions (i.e., anoikis). IGF-I has been reported as a powerful inducer of anchorage-independent growth in MCF-7 breast cancer cells, and we first determined if AIB1 played a role in this IGF-I effect. IGF-I induced a 3-fold increase in the number of MCF-7 colonies formed, and this increase was completely negated by treatment with AIB1 siRNA (Fig. 2,A). In contrast, a similar 3-fold increase in colony formation observed with serum treatment was unaffected by the reduction of AIB1 (Fig. 2 A), suggesting a selective role of AIB1. In 1% CCS, the colony count is indistinguishable from the baseline of this assay, and this is unaffected by AIB1 siRNA treatment. The effects of AIB1 siRNA on IGF-I–induced colony formation were not due to nonspecific activation of RNA degradation because parallel experiments in which endogenous AIB1 were reduced by the expression of tetracycline-regulated ribozymes (14) showed the same selective inhibition of IGF-I–induced colony formation (data not shown).

Increases in colony formation in anchorage-independent growth due to growth factor stimulation result from a balance between increased numbers of cells entering the cell cycle and reduction in anoikis. Therefore, it was of interest to determine which of these processes was affected by the reduction of AIB1. To examine this, we grew MCF-7 cells on poly-HEMA–coated dishes that prevent attachment and forces the cells to grow anchorage independently (24). Fluorescence-activated cell sorting analysis showed that reduction of AIB1 levels had no significant effect on progression of the cells into G₂ + S either under basal, IGF-I, or serum-induced growth conditions (Table 1,A). Annexin V staining to quantitate apoptosis, however, showed that IGF-I was unable to rescue the cells with reduced AIB1 levels from anoikis, whereas the rescue by serum was unaffected (Fig. 2 B). The increase in anoikis in 1% CCS in the presence of AIB1 siRNA did not register in the soft agar colony count because this was already at background levels of this latter assay in the absence of AIB1 siRNA treatment and is therefore not sensitive to additional reductions. Thus, there are no factors in 1% CCS that stimulate colony formation, but AIB1 siRNA can still cause apoptosis in these cells because there are no factors in 1% CCS to rescue them. Overall, the results suggest that IGF-I–induced survival of cells growing in suspension depends on an AIB1-dependent signaling pathway, whereas serum does not.

To determine whether the effect of AIB1 siRNA were unique to anchorage-independent growth conditions, we also determined its effects under anchorage-dependent growth conditions. Attached cells show a reduced growth rate (∼30%) after reduction of AIB1 and the stimulation by IGF-I (1.7-fold) or by serum (2.8-fold) was not affected by reduced AIB1 levels. These data imply that cell attachment can circumvent the rate-limiting signal pathways required for IGF-I–dependent survival of cells in suspension (Fig. 3,A). Interestingly, AIB1 siRNA treatment reduced by 34% the number of MCF-7 cells in G₂ + S phases of the cell cycle (Table 1,B), which explains the reduction of the overall growth rate independent of exogenously added growth stimulus. Under attached growth conditions, there was a trend to increased apoptosis with reduction in AIB1 with siRNA, but the differences were not statistically significant (Fig. 3 B). In the MCF-7 Rz29 cell lines (14), similar results were also obtained (data not shown).

To determine genes that might be critical targets of AIB1 during proliferative responses in MCF-7 cells, we compared gene array analysis of MCF-7 cells after treatment with AIB1 or control siRNA for 2 days. The Affymetrix U133A human gene chip that represents 33,568 transcripts was used to evaluate gene expression profiles under these different conditions, and we found 124 genes that showed a statistically significant difference between AIB1 high (control siRNA transfected) and AIB1 low (AIB1 siRNA transfected) at P ≤ 0.0025 (Table 2). Interestingly, several genes that are known to be critical for cell cycle regulation, anoikis, and apoptosis, notably cyclin D1, Bcl-2, and MAPK [extracellular signal-regulated kinase (ERK) 1/2] were highly dependent on AIB1 levels for their sustained expression (highlighted in Table 2). In our array analysis, we also found that the expression of a number of genes was up-regulated by reducing AIB1 with siRNA (Table 2), indicating that high AIB1 levels normally suppress expression of these genes.

In a separate set of array analyses, we also examined the effect of a short-term 1 hour treatment by IGF-I on gene expression after treatment of MCF-7 cells with AIB1 siRNA and control siRNA. This was to determine whether there were critical immediate early genes that IGF-I–regulated dependent on the AIB1 levels. However, at P ≤ 0.0025, this analysis did not differ significantly from the AIB1 siRNA versus control siRNA analysis. Thus, 1 hour of IGF-I treatment after pretreatment with AIB1 siRNA did not alter gene expression in comparison to cells pretreated with AIB1 siRNA alone (this comparison is not shown).

To assess whether altered mRNA expression due to AIB1 reduction was also reflected at the protein level, we did a series of Western blot analysis for those proteins known to be crucial for cell growth and survival. The protein levels of cyclin D1, Bcl-2, and ERK2 (but not ERK1) were reduced (90, 40, and 40%, respectively) by lowering the AIB1 levels by >90% in attached MCF-7 cells (Fig. 4). AIB1 siRNA reductions in expression of these genes were also observed in the presence of IGF-I (Fig. 4). IGF-I clearly induced the expression of cyclin D1 (4-fold) and caused a smaller 1.7-fold increase in the protein expression of ERK2 (Fig. 4). The basal levels of both cyclin D1 and Bcl-2 were increased under anchorage-independent conditions (Fig. 4), and IGF-I did not increase these levels any further (Fig. 4). Overall, it is clear that in basal or IGF-I–treated conditions or in cells attached or in suspension, targeting of AIB1 was effective in producing decreases in cyclin D1, Bcl-2, and ERK2 protein levels. Maintenance of the expression of these genes by AIB1 could be involved in both basal and IGF-I–induced growth responses in MCF-7 cells.

A somewhat surprising aspect of our cDNA array analyses was that after 48 hours of exposure to AIB1 siRNA, we did not see significant changes in the mRNA levels of molecules that specifically transmit IGF-I signaling such as the IGF-IR, IRS-1, IRS-2, or IGF-binding protein family members. We had conjectured that changes in the expression of these molecules after reduction of AIB1 might explain the selective effect on IGF-I but not serum signaling, and thus, we next determined the impact of the reduction of AIB1 on the protein levels of IGF-I–signaling molecules to pick up potential posttranscriptional effects. We did Western blot analysis of IGF-I–signaling molecules in the presence of control or AIB1 siRNA in both attached and suspension growth (Fig. 5, left and right panels). Under the anchorage-independent conditions, which had revealed the dependence of IGF-I survival signals on AIB1 (see Fig. 2), we found that IGF-I receptor protein levels were reduced by 50% upon reduction of AIB1. Furthermore, we also observed a smaller 10 to 20% reduction in IRS-1 protein levels in the presence of IGF-I (Fig. 5,A, right panel). IRS-2 protein was not altered. IGF-IR levels in attached cells showed a similar reduction of the IGF-I receptor protein after AIB1 siRNA treatment and no detectable alteration in IRS-1 or IRS-2 proteins (Fig. 5, left panel). Thus, we conclude that the IGF-I receptor is dependent on AIB1 for its protein expression levels in MCF-7 cells. This regulation appears to be mediated at the posttranscriptional level because we did not detect significant changes in the IGF-IR mRNA levels on the cDNA array analysis.

We have determined previously that AIB1 plays a rate-limiting role in the estrogen-induced proliferation of MCF-7 cells (14). There is much evidence for cross-talk between the IGF-I and ER-signaling pathways (25, 26, 27) and the expression of IGF-1R, IRS-1, cyclin D1, Bcl-2, and MAPK are all regulated by estrogen (28, 29, 30, 31, 32, 33). We therefore hypothesized that some if not all of the effects that we saw on gene expression by reducing cellular levels of AIB1 might be mediated through a reduction in ER signaling. Interestingly, we found that the expression of ER-α itself was not reduced under basal conditions after AIB1 siRNA treatment under attached conditions (Fig. 6,A). However, in suspension conditions, we observed a decrease in ER levels in the presence of AIB1 siRNA. Thus, it was possible that the ER reduction alone could explain some of the effects of AIB1 siRNA in suspension. To examine the role of estrogen in the gene expression changes that we observed in the presence of AIB1 siRNA, we repeated the analysis of protein expression changes in the presence of the antiestrogen ICI 182,720 (ICI, Faslodex), which blocks estrogen binding to the receptor, as well as down-regulates the ER levels (Fig. 6,A, right panel; ref. 34). We argued that if the effects of AIB1 were dependent on ER, then in the presence of antiestrogen, we would no longer see the AIB1 siRNA effect. Interestingly, AIB1 siRNA-dependent reductions in cyclin D1, Bcl-2, and ERK2 that we had observed in Fig. 4 were all preserved in the presence of ICI under attached conditions (Fig. 6,B). However, the basal reduction in cyclin D1 was only 30% compared with the >80% reduction that we saw in the absence of ICI. This suggests that a large portion of the basal expression of cyclin D1 is maintained through estrogen signaling. In contrast, the magnitude of the AIB1 siRNA reduction in Bcl-2 and ERK2 were comparable in the presence or absence of antiestrogen. Although the IGF-IR is known to be an estrogen-induced gene, AIB1 siRNA treatment was still able to decrease IGF-IR expression irrespective of ICI (compare Fig. 5 with Fig. 6 B). Overall, this data indicates that a large portion of the basal and IGF-I–induced changes in gene expression that we observed that were dependent on AIB1 were occurring independent of ER signaling.

To determine whether the dependence of IGF-I signaling was observed in other breast cancer cell lines, we examined the effect of AIB1 siRNA on the human breast cancer cell line, MDA MB-231. These cells are ER negative and proliferate in response to IGF-I (28). To determine whether similar genes were involved in the AIB1 effect in these cells as MCF-7 cells, we did immunoblot analysis of cell cycle and apoptosis genes, as well as for molecules involved specifically in IGF-I signaling. Similar to MCF-7 cells, cyclin D1 and ERK2 gene expression were dependent on maintaining high AIB1 levels (Fig. 7,A). However, Bcl-2 levels were unaffected by AIB1 siRNA treatment. Unlike MCF-7 cells, none of these genes were induced by IGF-I in the MDA MB-231 cells (Fig. 7,A). Interestingly, in MCF-7 cells, we observed that IGF-IR and to a certain extent IRS-1 expression were dependent on AIB1, but in MDA MB-231 cells, no changes in the expression of any of these proteins was observed (Fig. 7 B).

The data thus far indicated that neither the estrogenic effects nor the regulation of IGF-I–signaling molecules was the principal mechanism whereby AIB1 specifically effected IGF-I signaling. We therefore sought to determine whether down-regulation of AIB1 directly affected activation of IGF-I–signaling pathways. For these experiments, we pretreated MCF-7 or MDA MB-231 cells (attached or suspended) for 48 hours with AIB1 siRNA and then determined the responsiveness of the cells to short-term treatment with IGF-I (0 to 30 minutes; Fig. 8). We examined the levels of AKT and its phosphorylation status on Ser⁴⁷³ as a readout of the sensitivity of the PI3k pathway and measured ERK 1/2 levels and their phosphorylation as a read out of MAPK activation. In attached conditions, in both MCF-7 and MDA MB-231 cells, we observed a reduction in the ratio of phospho-AKT to AKT, indicating a decreased sensitivity of the PI3k pathway to IGF-I (Fig. 8, A and B). However, the reduction in MCF-7 cells was less than that observed in MDA MB-231 cells, and the reduction in the phospho-AKT to AKT ratio was primarily due to an increase in AKT levels rather than a reduction in phosphorylation of AKT (Fig. 8,A). Under suspension conditions (equivalent to the soft agar colony formation conditions) the IGF-I–induced changes in phospho-AKT to AKT ratio are unaltered in MCF-7 cells by AIB1 siRNA treatment, but the ratio was still reduced in MDA MB-231 cells (Fig. 8, C and D).

In contrast to the changes in AKT activation under all conditions and in both cell lines, treatment with AIB1 siRNA did not change the time course or the magnitude of ERK 1/2 phosphorylation after IGF-I stimulation (Fig. 8 A–D). Overall, these data indicate that AIB1 levels determine the sensitivity of the PI3k pathway to IGF-I signaling in these breast cancer cell lines under certain plating conditions, but changes in AIB1 levels appear to have little effect on MAPK signaling.

The role of AIB1 and other coactivators in nuclear receptor signaling is well established. In this study, we now show that AIB1 can also be rate-limiting for IGF-I–stimulated growth of MCF-7 human breast tumor cells. This is consistent with the finding in p/CIP (AIB1 mouse homologue)-knockout mice that IGF-I stimulation of proliferation of mouse embryonic fibroblast cells was significantly reduced by deletion of p/CIP (18). However, in contrast to Wang et al. (18), where only a few changes in gene expression were observed upon array analysis of pCIP^−/− versus wild-type mouse embryonic fibroblast cells, in MCF-7 cells, a number of critical signaling molecules were dependent on AIB1 for their expression and regulation. Most significantly altered was cyclin D1, whose expression was highly dependent on AIB1 expression under anchorage-dependent and -independent conditions. Cyclin D1 is an important cell cycle-regulating protein that in response to mitogenic stimulus associates with CDK4/6. The cyclin D1-CDK4/6 complex hyperphosphorylates the retinoblastoma protein, leading to the release of E2F. Free from its association with the retinoblastoma protein, E2F activates genes necessary for cell proliferation (35). Cyclin D1 is a critical downstream target of PI3k-signaling pathway for proliferation in MCF-7 cells (36). It is likely that the reduction in cyclin D1 under low AIB1 conditions would explain some of the cell cycle effects of AIB1 siRNA treatment of MCF-7 cells under anchorage-dependent conditions. However, it appears that the cells are less dependent on cyclin D1 for cell cycle control under anchorage-independent conditions. Our data confirm and extend previous observations that cyclin D1 gene regulation by estrogen is AIB1 dependent (37). In addition, we show that a portion of the basal expression of cyclin D1, as well as its IGF-I induction, occurs independent of estrogen signaling but is nevertheless AIB1 dependent (Fig. 6 B).

The antiapoptotic protein Bcl-2 was also found in our cDNA array study to be significantly down-regulated in the absence of AIB1. Bcl-2 is a proto-oncogene that is a part of a family of related proteins that functions to either promote or inhibit apoptosis and anoikis (38, 39). In the presence of AIB1 siRNA, Bcl-2 levels were decreased in both basal and IGF-I treatment conditions, regardless of attached or poly-HEMA growth conditions. Bcl-2 is thought to play a central role in the interplay of integrin and IGF-I signaling in anoikis (40, 41), and reductions in Bcl-2 may explain some of the abrogation of the anti-anoikis effect of IGF-I in suspension. Interestingly, Bcl-2 and cyclin D1 expression are interconnected. When Bcl-2 was overexpressed in a human breast epithelial cell line, MCF10A, it caused an increase in the expression of cyclin D1. Bcl-2 was also found to be able to increase the activity of cyclin D1 promoter in both MCF10A and MCF-7 cells independent of anchorage conditions (42). The effects of both the loss of Bcl-2 and AIB1 could therefore contribute to greater reduction in cyclin D1 levels.

Our data indicate that a large portion of the transcriptional control of genes, such as cyclin D1 and Bcl-2, occurs independent of the ER. AIB1 is known to coactivate a number of factors that are not nuclear receptors such as NF-κB (43, 44) and activator protein-1 (21). It is likely that some of these sites are in involved in AIB1 control of basal and IGF-I control of gene transcription in MCF-7 cells. Of note is that both cyclin D1 and Bcl-2 genes are estrogen-inducible genes that contain nonclassical ERE promoters because they do not contain half or full consensus ERE sites in their promoters. Both genes contain in common cAMP-response element, activator protein-1, and Sp-1 sites, which are necessary for estrogen induction of the promoter (29, 33, 45). Furthermore, IGF-I induction of the Bcl-2 promoter is dependent on a cAMP-response element site (46). Thus, it is possible that AIB1 may act at the cAMP-response element, activator protein-1, or Sp-1 sites to coactivate expression in the presence of IGF-I, which may not necessitate the presence of the ER.

Although the changes in cyclin D1, Bcl-2, and MAPK could explain the effects of AIB1 siRNA on basal proliferation, it is not clear that these changes would specifically affect the IGF-I versus serum pathway. In the analysis of the mouse embryonic fibroblast p/CIP^−/− cells, no real changes were observed in IGF-I–signaling molecule pathway molecules (18). Our study of MCF-7 cells indicates that AIB1 plays a role in controlling levels of IGF-IR, and this is especially pronounced under anchorage-independent conditions. In MDA MB-231 cells, however, the IGF-IR levels were unaltered by changes in AIB1 levels and thus do not appear to be involved in AIB1 control of IGF-I proliferation in these cells. As stated above, IGF-IR interplay with the integrin-signaling pathways plays a major role in the control of anoikis in many cell types (47, 48, 49), and it appears that AIB1 may be able to potentiate IGF-I inhibition of anoikis in MCF-7 cells by maintaining high IGF-IR levels. Importantly, the maintenance of the high IGF-IR levels occurred even when ER signaling was reduced, suggesting that AIB1 could maintain this phenotype even when breast cancer cells are growing hormone independently. Changes in IRS-1 levels have been observed to be AIB1 dependent in H-ras–induced tumors (50), but the decreases in IRS-1 we observed with AIB1 siRNA were small in MCF-7 cells and not observed in the MDA MB-231 cells.

A recent article by Zhou et al. (51) found that overexpressing AIB1 caused an increase in AKT levels and an increase in cell size in prostate cancer cells in a steroid-independent manner. In MCF-7 or MDA MB-231 cells, we did observe some changes in unphosphorylated AKT levels but did not observe changes in cell size when AIB1 was down-regulated. Differences in basal AKT regulation between cell types might be explained by differences in PTEN gene status in the cell lines used in these studies. The prostate cancer cell lines, LNCaP and PC-3, used in the Zhou et al. (51) study have a defective PTEN gene (52), whereas MCF-7 and MDA MB-231 breast cancer cells have a wild-type PTEN gene (53). It does seem clear from our data that the ability of IGF-I to activate AKT is dependent on AIB1 levels in these breast cancer cell lines at least under certain plating conditions. However, the most profound AIB1-dependent phenotypic change induced by IGF-I is observed when MCF-7 cells are growing anchorage independently. Recall that under these conditions we see little change in the IGF-I activation of AKT, thus it seems likely that AIB1 is also involved in other signaling pathways that predominate when the cells are growing in suspension.

In conclusion, our results suggest that IGF-I induced survival of cells in suspension depends on AIB1 signaling whereas the serum does not. Also, some IGF-I–induced changes in gene expression are dependent on AIB1. The observation that AIB1 is rate-limiting for aspects of IGF-I signaling is important because it implies that AIB1 could effect malignant phenotypic changes induced by growth factors, even if estrogen signaling is not intact. This may become relevant in advanced breast cancer, which is frequently estrogen receptor negative, but has amplified growth factor signaling from the IGF-I–signaling pathways (54). In addition, overexpression of AIB1 may play a role in potentiating growth factor signaling when estrogen signaling is blocked with antiestrogens, thus contributing to the antiestrogen-resistant phenotype in breast cancer. Our studies suggest a broader role of AIB1 in coactivating growth signaling that is not limited to hormonally responsive breast tissue but may be extended to other tissues that are responsive to IGF-I such as the pancreas (55) and prostate (56).