Changes in insulin receptor signaling underlie neoadjuvant metformin administration in breast cancer: a prospective window of opportunity neoadjuvant study

A single-arm, “window of opportunity” neoadjuvant metformin study (ClinicalTrials.gov identifier: NCT00897884) was conducted at Mount Sinai and Princess Margaret hospitals in Toronto, ON, Canada, from March 2009 to June 2011 (Additional file 1) [24]. Women with newly diagnosed, treatment-naive, early-stage breast cancer were recruited for participation in the study, regardless of tumor subtype. Patients <70 years of age with normal hepatic, cardiac and renal function were eligible, whereas those receiving neoadjuvant therapy, with known diabetes, blood glucose of ≥7 mmol/L or exposure to metformin in the preceding 4 weeks were ineligible. Eligible patients were administered 500 mg of metformin three times daily for ≥2 weeks after diagnostic core biopsy up to and including the night before surgery. Diagnostic core biopsy and surgical tumor specimens were collected, fixed in formalin and embedded in paraffin (FFPE) pre- and post-metformin treatment, respectively. Core biopsy samples were placed immediately in formalin and formalin-fixed for 8 to 24 hours. Excisional samples were fixed in formalin for a minimum of 24 hours and a maximum of 96 hours, with most fixed for under 72 hours. Tumor characteristics were obtained from standardized pathology reports. All patients included in the study provided informed consent for participation in the trial and for the publication of associated results. The study was conducted in compliance with the research ethics requirements of Princess Margaret and Mount Sinai hospitals and was approved by the Ontario Cancer Research Ethics Board.

To ensure optimal performance and to confirm specificity, antibodies against phosphorylated AMPK (p-AMPK) T172, phosphorylated acetyl coenzyme A carboxylase (p-ACC) S79, phosphorylated protein kinase B (p-PKB)/Akt S473 and phosphorylated extracellular signal-regulated kinase (p-ERK) T202/Y204 were optimized for immunohistochemistry (IHC) using FFPE MCF-7 breast cancer cells (American Type Culture Collection, Manassas, VA, USA). For p-PKB/Akt S473 antibody optimization, MCF-7 cells were starved of serum overnight and then treated with either 200 nM of the dual phosphatidylinositol 3-kinase/mammalian target of rapamycin (PI3K/mTOR) inhibitor BEZ235 (Selleck Chemicals, Houston, TX, USA) for 1 hour or media containing 10% fetal bovine serum (Gibco/Life Technologies, Burlington, ON, Canada) and 100 ng/ml heregulin (R&D Systems, Minneapolis, MN, USA) for 20 minutes. To optimize the antibody against p-ERK1/2 T202/Y204, MCF-7 cells were starved of serum overnight and then treated with either 10 μM of the Mitogen-activated protein kinase kinase 1/2 inhibitor U0126 (Sigma-Aldrich, St Louis, MO, USA) for 1 hour or media containing 100 ng/ml epidermal growth factor (EGF; PeproTech, Rocky Hill, NJ, USA) for 10 minutes. For p-AMPK T172 and p-ACC S79 antibodies, MCF-7 cells were either left alone or treated with 2 mM 5-aminoimidazole-4-carboxamide-1-β-d-ribofuranoside (AICAR; Toronto Research Chemicals, Toronto, ON, Canada) for 3 hours. After each respective treatment, cells were collected on ice and divided into two aliquots: one for lysis and Western blot analysis and the other for fixing, embedding and IHC analysis. Immunohistochemical staining (performed as described below) for p-PKB/Akt S473 was observed in serum- and heregulin-stimulated cells as both nuclear and cytoplasmic, whereas treatment with BEZ235 significantly reduced S473 staining (Additional file 2). Staining for p-ERK1/2 T202/Y204 was extremely low in U0126-treated cells, but it was increased in cells treated with EGF (Additional file 2). Likewise, staining for p-AMPK T172 and p-ACC S79 was evident in AICAR-treated cell pellets, and signaling for these proteins was diminished in untreated cells (Additional file 3). The signaling changes observed by IHC were validated for each antibody by Western blot analysis. Additional antibody optimization for IHC was also performed using tissue sections from FFPE human breast tumors to ensure an adequate dynamic range of staining in standard pathology samples.

The antibody for organic cation transporter 1 (OCT1) was optimized and validated using tissue sections from normal human liver and breast, as well as breast tumor tissue (Additional file 4). The IR antibody has been described previously [17].

Staining for p-AMPK T172 (1:50 dilution; Cell Signaling Technology, Danvers, MA, USA), p-ACC S79 (1:100 dilution; EMD Millipore, Billerica, MA, USA), p-ERK1/2 T202/Y204 (1:50 dilution; Cell Signaling Technology) and OCT1 (1:75 dilution; LifeSpan BioSciences, Seattle, WA, USA) was performed on a Ventana BenchMark XT automated slide preparation system using the iVIEW 3,3′- diaminobenzidine (DAB) detection kit (both from Ventana Medical Systems, Tucson, AZ, USA) and the IR (1:50 dilution; EMD Millipore) was assessed using the ultraView Universal DAB detection kit (Ventana Medical Systems). For p-PKB/Akt S473, staining was performed manually. Sections were incubated overnight in primary antibody (1:50 dilution; Cell Signaling Technology), and biotin-streptavidin-horseradish peroxidase secondary antibodies (Vector Laboratories, Burlingame, CA, USA; and ID Labs, London, ON, Canada) and DAB were used for detection. Sections from embedded MCF-7 cell pellets (described above) were included as controls for each batch of tumor sections.

All sections were scored manually by a pathologist (MCC) using the Allred system [26], which incorporates the proportion of cells staining positive (scored from 0 to 5) and intensity of staining (scored from 0 to 3) to yield values of 0 or 2 to 8. Scoring was performed in a blinded fashion; the pathologist was unaware of which biopsy and surgical specimens were from the same patient. For p-PKB/Akt, scores were generated for both nuclear and cytoplasmic staining, and an overall score was calculated as the mean of both values.

Because many of the variables had markedly skewed distributions, nonparametric rank-based methods were used to analyze the data, and the medians (upper limit of first and third quartiles (interquartile range, IQR)) were used to summarize the distributions. Change in biomarker scores was calculated as postmetformin minus premetformin, and the Wilcoxon signed-rank test was used to test whether the change was distributed symmetrically around zero. To investigate the extent to which changes in serum insulin, tumor IR expression and p-Akt within the same patient correlated with the patient’s change in Ki67, we scaled the insulin, IR and p-Akt variables such that each one’s largest observed reduction became −1. The three variables were then summed, resulting in a summary variable with the properties that a person with large reductions in all three variables would score near −3 and one with little change in any of the three would score near zero. Spearman’s rank correlations were used to quantify the correlations between variables.