The authors have declared that no competing interests exist.
Conceived and designed the experiments: MAB YHI ASO DY. Performed the experiments: MAB YHI ASO DHF SAB. Analyzed the data: MAB YHI ALS AVL CF CMP DY. Contributed reagents/materials/analysis tools: ASO DHF SAB LMS. Wrote the paper: MAB YHI ASO DHF ALS LMS CF CMP DY.
Current address: Mayo Clinic Cancer Center, Rochester, Minnesota, United States of America
Therapies targeting the type I insulin-like growth factor receptor (IGF-1R) have not been developed with predictive biomarkers to identify tumors with receptor activation. We have previously shown that the insulin receptor substrate (IRS) adaptor proteins are necessary for linking IGF1R to downstream signaling pathways and the malignant phenotype in breast cancer cells. The purpose of this study was to identify gene expression profiles downstream of IGF1R and its two adaptor proteins. IRS-null breast cancer cells (T47D-YA) were engineered to express IRS-1 or IRS-2 alone and their ability to mediate IGF ligand-induced proliferation, motility, and gene expression determined. Global gene expression signatures reflecting IRS adaptor specific and primary vs. secondary ligand response were derived (Early IRS-1, Late IRS-1, Early IRS-2 and Late IRS-2) and functional pathway analysis examined. IRS isoforms mediated distinct gene expression profiles, functional pathways, and breast cancer subtype association. For example, IRS-1/2-induced TGFb2 expression and blockade of TGFb2 abrogated IGF-induced cell migration. In addition, the prognostic value of IRS proteins was significant in the luminal B breast tumor subtype. Univariate and multivariate analyses confirmed that IRS adaptor signatures correlated with poor outcome as measured by recurrence-free and overall survival. Thus, IRS adaptor protein expression is required for IGF ligand responses in breast cancer cells. IRS-specific gene signatures represent accurate surrogates of IGF activity and could predict response to anti-IGF therapy in breast cancer.
The insulin-like growth factor (IGF) pathway mediates cancer cell proliferation, survival, and metastasis. These ligands interact with the type 1 IGF receptor (IGF-1R) and a number of monoclonal antibodies and tyrosine kinase inhibitors have been developed and tested in clinical trials. Although clinical benefit has been demonstrated in some cancers [
Insulin receptor substrate (IRS) proteins play a critical and differential role in mediating receptor tyrosine kinase activity in breast cancer cells [
Herein, we demonstrate that IRS adaptor proteins are required for IGF-ligand induced biology and gene transcription. Target gene validation confirmed that both distinct and overlapping patterns of IRS-regulated gene expression are evident in response to IGF pathway activation. The Late IRS-1 gene signature reported the highest significance in terms of functional pathway analysis and gene set enrichment in molecular breast tumor subtypes. A high correlation to the Late IRS-1 gene signature was a marker of poor prognosis independent of nodal and/or hormone receptor status. IRS gene enrichment in luminal B breast tumors was an independent predictor of both recurrence-free and overall survival. As a result, IRS adaptor signatures may distinguish patients that would benefit from anti-IGF targeted therapeutics.
T47D-YA, and T47D-YA/IRS-1/2 cells were generated and described previously [
Immunoblotting was performed as previously described [
Cells were plated in 24-well plates at a density of 10,000 cells per well, allowed to equilibrate overnight and starved in SFM media for 24 hours prior to treatment with IGF-I. After 3 days of treatment, growth was assessed via the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay as described previously (176). 60 μL of 5 mg/mL 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide solution in SFM was added to each well. After incubation for 4 h at 37°C, wells were aspirated and formazan crystals were lysed with 500 μL of solubilization solution (95% DMSO + 5% IMEM). Absorbance was measured with a plate reader at 570 nm using a 650 nm differential filter to assess growth.
Cells were plated in an 8-well scratch-wound plate at a density of 1x104 cells, allowed to equilibrate overnight, starved in SFM overnight and a scratch induced manually employing a P10 pipette tip. The media was supplemented with or without IGF-I and monitored for 24 by light microscopy and concurrent image acquisition. Values represent mean area cleared in IGF treated groups vs. control SFM groups.
Cells were examined by Boyden chamber assay as previously described [
Cells were plated at a density of 3x106 in 150 mm dishes, allowed to equilibrate overnight, and the media replaced by SFM alone for 24 h prior to stimulation. At time = 0 cells were treated with SFM+FN alone or with IGF-I. Total RNA was collected using RNeasy mini kit (Qiagen) or PerfectPure RNA tissue kit (5Prime) at 4 h and 24 h. RNA quantity was determined by 260:280 assay and quality using the Agilent Bioanalyzer 2100 to ensure banding conservation. Isolated RNA samples were then submitted to the Biomedical Genomics Center—Microarray Facility University of Minnesota for biotin labeling, synthesis and hybridization to the Affymetrix U1330 Plus 2.0 arrays. Signatures are available through the GEO database (GSE78916).
Cells were plated at a density of 1x106 in 100mm diameter dishes, allowed to equilibrate and incubated overnight in SFM. Cellular RNA was isolated using the 5 Prime PerfectPure RNA tissue kit according to the manufacturer (Fisher Scientific). For quality control and to determine nucleic acid concentration, a 260/280 assay was performed on a spectrophotometer. Forward and reverse primers were designed to target the following transcripts:
All arrays were normalized using GC-RMA process embedded in GeneData refiner and further normalized to corresponding untreated states to isolate IGF response independently of basal differences between each of the cell lines. Student’s t tests were performed between groups using GeneData expressionist with P-values < 0.05 (Bonferroni Correction in select cases) and a minimum average fold change of 1.5 was employed. Hierarchical clustering was carried out on log2-transformed data generated using Gene Cluster 3.0. Data was visualized and images generated using Java TreeView. Molecular subtype classification and Kaplan-Meier analysis was performed as previously described (15, 16). IRS expression profiles are submitted in the Gene Expression Omnibus.
The T47D-YA breast cancer variant cell line does not express IRS adaptor proteins or respond to IGF ligands, yet they retain functional IGF-1R [
(A) Monolayer growth and motility of T47D-YA (YA), T47D-YA-IRS-1 (#10 and #20) and T47D-YA-IRS-2 (#1 and #6) were measured by MTT assay and (B) scratch-wound healing assay in response to IGF-I treatment. The graphs are presented as fold-change response vs. non-treated control and error bars represent standard deviation. (C) IGF-induced gene expression is IRS-dependent. cDNA microarray analysis was performed on IRS-null YA, IRS-1, and IRS-2 clones. The graph represents IGF-regulated probes in comparison to untreated samples that met both fold (1.5) and p-value (0.05) cutoff values. Hierarchical clustering was carried out on log2-transformed using Gene Cluster 3.0 and visualized in Java TreeView.
To assess the contribution of IRS adaptor proteins in the regulation of gene expression, IRS-null, IRS-1, and IRS-2 cells were stimulated with IGF-I for a period of 4 or 24 hours and cDNA microarray analysis performed (GSE78916). Unsupervised hierarchical clustering revealed significant gene induction by both IRS-1 and IRS-2 at both early and late time points (
To assess the value of IRS adaptor proteins in breast cancer outcome, distinct IRS isoform gene signatures were derived from the global gene expression patterns observed in response to IGF stimulation (
(A) Venn diagrams depicting four distinct IRS isoform gene signatures were derived from overlapping and differential global gene expression patterns in response to IGF-I. (B) Target gene validation confirms both distinct and overlapping patterns of IRS-regulated gene expression. Gene expression was normalized to RPLP0 and is presented as fold-change of treatment (black bars) vs. serum-free (white bars) conditions. Error bars represent standard deviation and all results are representative of at least three independent replicates. (C) IRS gene signature enrichment in breast tumor subtypes in the UNC337 cohort. Median expression values are represented here in graphical format with p-values included for each of the IRS gene signatures.
P-value indicates modified Fisher’s exact Probability Value and a high E-Scores (Enrichment Scores) indicates significant gene enrichment in the annotation cluster.
Signature | Pathway (KEGG) | P-value | Function (Annotation Cluster) | P-value | E-Score |
---|---|---|---|---|---|
P53 signaling pathway | 1.9E-02 | Transcription | 4.2E-02 | 1.65 | |
Cell cycle | 1.8E-03 | Mitosis | 2.1E-16 | 10.67 | |
Focal adhesion | 1.2E-02 | Regulation of protein kinase activity | 2.8E-04 | 3.25 | |
P53 signaling pathway | 1.1E-02 | Microtubule cytoskeleton | 3.1E-05 | 2.89 |
Using median expression values, IRS gene signatures were significantly enriched according to molecular breast tumor subtype (basal-like, claudin-low, HER2-enriched, luminal A, luminal B, and normal-like) in the UNC337 (GSE18229) cohort (
To examine the biological significance of IRS-induced gene expression, we evaluated the induction of TGFβ mRNA by both IRS-1 and IRS-2 in breast cancer cells. TGFβ2 was selected for evaluation as both IRS-1 and -2 T47D transfected cells regulated this gene but not the related growth factor TGFβ1 (
(A) Expression of TGFβ1 and TGFβ2 by qPCR in T47D-YA-IRS-1 (#10 and #20) and T47D-YA-IRS-2 (#1 and #6). (B) IGF-induced TGFβ2 expression in MCF10A, MCF-7L, MCF-7 ATCC, MDA-231 and F11 cells. For A & B, all cells were exposed to 5nm IGF-I for 4 hours prior to harvesting mRNA. Gene expression was normalized to RPLP0 and is presented as fold-change of treatment (black bars) vs. serum-free (white bars) conditions. (C) TGFβ2 expression was assessed by qPCR in an IRS-gene deletion mouse models (left) and IRS-overexpressing SH-EP neuroblastoma cells (right). (D) IRS-1, IRS-2 and TGFβ2 expression in a panel of patient breast tumors. Arrows indicate invasive breast carcinoma. Yellow bars signify high gene expression, blue bars signify low gene expression. E) pSMAD2 was examined by immunoblot at the indicated time points in MCF-7 cells. (F) Cell motility was examined by modified Boyden chamber assay. MCF-7 cells were incubated in the presence of neutralizing antibodies to either TGFβ1 or TGFβ2 and IGF-induced motility assessed. Error bars represent standard deviation and all results are representative of at least three independent replicates.
In addition, we evaluated alternate IRS models to further confirm IRS-dependent TGFβ2 expression (
To test for functionality of IGF-induced TGFβ in MCF7 cells, we examined activation of the TGFβ signaling pathway. IGF transiently induced the phosphorylation of SMAD2, an effector of TGFβ signaling (
To evaluate the clinical impact of the IRS signatures, we analyzed their expression in the UNC337 and NK1295 cohorts. The Late IRS-1 signature was most significantly over-represented across the subtypes and as a result, its prognostic value was determined in breast cancer tumors. To this end, tumors were divided into three groups by Late IRS-1 expression values: “Strong IRS-1 Corr.” (upper 20% or most positive correlation values), “Inverse IRS-1 Corr.” (lower 20% or most negative correlation values) or “Weak IRS-1 Corr.” (remaining 60% or mid-range correlation values). In both the UNC337 and NKI295 cohorts, poor outcome was significantly associated with increased Late IRS-1 gene expression for both recurrence-free survival (RFS) (
Univariate Kaplan-Meier analysis of A) RFS and B) OS was assessed in the combined UNC3337 and NKI295 cohorts (n = 534). Tumors were subdivided and classified as one of the following: Strong IRS-1 Corr. (top 20% of all tumors), Inverse IRS-1 Corr. (bottom 20% of all tumors), and Weak IRS-1 Corr. (all remaining tumors). Corresponding p-values are depicted.
Separating tumors by molecular subtype revealed that Luminal B tumors with increased Late IRS-1 gene expression had poor RFS and OS (
RFS | OS | |||
---|---|---|---|---|
HR (95% CI) | P | HR (95% CI) | P | |
Age | 0.642 (0.378–0.987) | 0.0426 | 0.526 (0.248–1.009) | 0.0537 |
Size | ||||
2 | 1.464 (0.759–2.874) | 0.2559 | 1.635 (0.690–3.979) | 0.2640 |
3 | 2.736 (0.607–8.928) | 0.1687 | 8.229 (1.096–41.60) | 0.0419 |
Grade | ||||
2 | 0.588 (0.217–1.863) | 0.3412 | 1.30 (0.2.55–6.906) | 0.9699 |
3 | 0.871 (0.323–2.778) | 0.7994 | 1.495 (03.60–10.27) | 0.6092 |
Node | ||||
1 | 0.985 (0.503–1.938) | 0.9658 | 1.055 (0434–2.563) | 0.9045 |
2 | 0.774 (0.294–1.828) | 0.5713 | 1.069 (0.326–3.047) | 0.9054 |
IRS-1 Corr. | 2.463 (1.185–4.900) | 0.0167 | 2.831 (1.147–6.637) | 0.0250 |
At 5 years, a strong IRS-1 Late correlation resulted in significantly shorter recurrence (
Identifying IGF-dependent breast cancer tumors remains a challenge. While levels of total and IGF-1R have been identified as poor prognostic factors in breast cancer [
Using an
In addition to providing predictive biomarkers for anti-IGF-1R therapies, identification of key genes regulated by activation of this pathway may be useful. One of the genes we identified as regulated by IRS-1 (SLC7A11 or xCT) has a functional role in mediating response to reactive oxygen species [
We studied IGF signaling in this model system, it is also clear that the IRS adaptor proteins regulate insulin receptor signaling [
We conclude that IRS adaptor proteins represent potential predictive clinical biomarkers of breast cancer outcome and should be considered in conjunction with receptor expression. Furthermore, both receptor and adaptor protein targeting might result in enhanced suppression of growth factor signaling and inhibition of tumor growth.
Comparative analysis of T47D-YA/IRS-1 and MCF-7 gene arrays was performed and IGF-induced gene overlap determined. Arrays represent both temporal and directional overlap. Results were confirmed by Fisher’s exact test.
(TIFF)
Kaplan-Meier analysis stratified (n = 534) according to nodal and/or ERα status. Strong Late IRS-1 gene expression is associated with poor prognosis in all groups
(TIFF)
(DOCX)
This work was supported by a BCRF-AACR Grant for Translational Breast Cancer Research 07-60-26 (DY), Department of Defense Predoctoral Traineeship Award BC073039 (MAB), and the National Cancer Institute Cancer Center Support Grant P30 077598. We extend our appreciation to Eva Feldman for the use of the IRS overexpressing SH-EP human neuroblastoma cells.