Metformin counteracts stimulatory effects induced by insulin in primary breast cancer cells

Insulin and Metformin were purchased from Merck Life Science (Milan, Italy). The MEK inhibitor trametinib, the insulin receptor inhibitor OSI-906 and the PI3Kα inhibitor alpelisib were obtained from MedChemExpress (DBA, Milan, Italy). The CXCR4 antagonist AMD3100 was purchased from Santa Cruz Biotechnology (DBA, Milan, Italy). All compounds were dissolved in DMSO, except insulin, metformin and AMD3100 that were solubilized in water.

BCAHC-1 cells were isolated and characterized as previously described [27], and maintained in DMEM/F-12 with phenol red, supplemented with 5% FBS and 100 μg/ml penicillin/streptomycin. Cells were grown at 37 °C in a humidified 5% CO₂ and switched to a medium without serum and phenol red the day before treatments to be processed for experiments. CAFs were obtained as previously described [27, 28] from 10 invasive ductal breast carcinomas and pooled for the subsequent studies. Briefly, specimens were cut into 1–2 mm diameter pieces, placed in a digestion solution comprising 400 IU collagenase, 100 IU hyaluronidase, 10% serum, antibiotics, and antimycotics, and incubated overnight at 37 °C. After centrifugation at 90×g for 2 min, the supernatant containing fibroblasts was centrifuged at 485×g for 8 min; the pellet obtained was suspended in Medium 199 and Ham’s F12 mixed 1:1 (supplemented with 10% FBS and 100 μg/ml penicillin/streptomycin). CAFs were then expanded into 10-cm Petri dishes and stored as cells passaged for three population doublings within a total of 7 to tissue dissociation. Primary cell cultures of fibroblasts were characterized by immunofluorescence with human anti-vimentin (V9) and human anti-cytokeratin 14 (LL001) (Santa Cruz Biotechnology, DBA, Milan, Italy) (data not shown). FAPα antibody (H-56, Santa Cruz Biotechnology, DBA, Milan, Italy) was used to characterize activated fibroblasts (data not shown). We used CAFs passaged for up to 10 population doublings for the experiments to minimize clonal selection and culture stress, which could occur during extended tissue culture.

Total RNA was extracted, and cDNA was synthesized by reverse transcription as previously described [29]. The expression of selected genes was quantified by real-time PCR using platform Quant Studio7 Flex Real-Time PCR System (Thermo Fisher Scientific, Milan, Italy). Gene-specific primers were designed using Primer Express version 2.0 software (Applied Biosystems) and are as follows: 5′-CGAGCCCTTTGATGACTTCCT-3′ (c-Fos forward) and 5′-GGAGCGGGCTGTCTCAGA-3′ (c-Fos reverse); 5′-AGCTGTGCATCTACACCGAC-3′ (cyclin D1 forward) and 5′-GAAATCGTGCGGGGTCATTG-3′ (cyclin D1 reverse); 5′-AAGCCACCCCACTTCTCTCTAA-3′ (ACTB forward) and 5′-CACCTCCCCTGTGTGGACTT-3′ (ACTB reverse); 5′-CCTTGGAGCCAAATTTAAAACCT-3′ (CXCR4 forward) and 5′-GCTGGACCCTCTGCTCACA-3′ (CXCR4 reverse); 5′-CAGATGCCCATGCCGATTCT-3′ (CXCL12 forward) and 5′-TTCTTCAGCCGGGCTACAAT-3′ (CXCL12 reverse). Assays were performed in triplicate and the results were normalized for actin beta (ACTB) expression and then calculated as fold induction of RNA expression.

PCR arrays were performed using a TaqMan™ Human Tumor Metastasis Array (Thermo Fisher Scientific, Milan, Italy) according to the manufacturer’s instructions. The amplification reaction and the results analysis were carried out using platform Quant Studio7 Flex Real-Time PCR System (Thermo Fisher Scientific, Milan, Italy).

For reporter gene assays, cells (1 × 10⁵) were plated into 24-well dishes with 500 mL/well culture medium containing 5% FBS and transfected for 18 h using X-treme GENE 9 DNA Transfection Reagent, as recommended by the manufacture (Merck Life Science, Milan, Italy). A mixture containing 0.5 mg of gene reporter (Fos-luc, kindly obtained from Dr. K. Nose, Department of Microbiology, Showa University School of Pharmaceutical Sciences, Hatanodai, Shinagawa-ku, Tokyo, Japan; Cyclin D1-luc, kindly obtained from Dr. R. G. Pestell, Kimmel Cancer Center, Department of Cancer Biology and Jefferson Stem Cell and Regenerative Medicine Center, Thomas Jefferson University, Philadelphia, PA), and 5 ng of pRL-TK was then transfected, after 18 h cells were treated with insulin in the presence or absence of metformin, OSI-906, trametinib and alpelisib for additional 12 h. Luciferase activity was measured using the Dual Luciferase Kit (Promega, Milan, Italy) according to the manufacturer’s recommendations. Firefly luciferase activity was normalized to the internal transfection control provided by the Renilla luciferase activity. Normalized relative light unit values obtained from cells treated with vehicle (−) were set as onefold induction, upon which the activity induced by treatments was calculated. For gene silencing experiments, cells were plated onto 10-cm dishes and transfected by X-treme GENE 9 DNA Transfection Reagent for 24 h before treatments with a control vector and the plasmid DN/c-Fos (kindly obtained from Dr. C. Vinson, NIH, Bethesda, USA) encoding a c-Fos mutant that heterodimerizes with c-Fos dimerization partners but not allowing DNA binding.

Cells were grown in 10-cm dishes, exposed to treatments and then lysed in 500 μl RIPA buffer with protease inhibitors (1.7 mg/ml aprotinin, 1 mg/ml leupeptin, 200 mmol/l phenylmethylsulfonyl fluoride, 200 mmol/l sodium orthovanadate and 100 mmol/l sodium fluoride). Samples were then centrifuged at 13,000 rpm for 10 min and protein concentrations were determined using BCA protein assay according to the manufacturer’s instructions (Thermo Fisher Scientific, Milan, Italy). Equal amounts of whole-protein extract were resolved on 8–10% SDS polyacrylamide gels and transferred to a nitrocellulose membrane (Merck Life Science, Milan, Italy), which were probed with primary antibodies against: phosphorylated ERK1/2 (E4), ERK2 (C-14), AKT, c-Fos (E-8), CXCR4 (4G10), Insulin Rα (N-20) and β-actin (AC-15) (Santa Cruz Biotechnology, DBA, Milan, Italy), cyclin D1 (TA801655) (OriGene Technologies, DBA, Milan, Italy), p-IR (Y1146), p-AKT (D9E) and CXCL12 (3740) (Cell Signaling Technology, Euroclone, Milan, Italy). Proteins were detected by horseradish peroxidase-linked secondary antibodies (Bio-Rad, Milan, Italy) and then revealed using the chemiluminescent substrate for western blotting Clarity Western ECL Substrate (Bio-Rad, Milan, Italy).

Protein precipitation from conditioned medium derived from CAFs was carried out using the precipitation method with acetone [30]. Briefly, four volumes of ice-cold acetone (Merck Life Science, Milan, Italy) were added to one volume of sample. The mixture was vortexed and incubated at − 20 °C overnight. This was followed by centrifugation at 10,000×g for 15 min at 4 °C. Afterwards, the supernatant was discarded, the pellet was air dried, then it was dissolved in 2× Laemmli buffer and used in the appropriate experiments. In western blot analysis, the protein loading of conditioned medium samples was checked by Ponceau red staining [0.1% Ponceau S (w/v) in 5% acetic acid] of the blotted membranes.

Cells were grown on a cover slip, exposed to treatments and then fixed in 4% paraformaldehyde in PBS, permeabilized with 0.2% Triton X-100, washed 3 times with PBS and incubated at 4 °C overnight with primary antibodies anti-CXCL12 (3740) (Cell Signaling Technology, Euroclone, Milan, Italy). After incubation, the slides were extensively washed with PBS, probed with Alexa Fluor conjugated secondary antibodies (Thermo Fisher Scientific, Milan, Italy) for 1 h at room temperature. Finally, cells were washed three times with PBS, incubated with DAPI (4′,6-diamidino-2-phenylindole) (1:1000) for 3 min and, after washing, immunofluorescence images were obtained using the Cytation 3 Cell Imaging Multimode reader (BioTek, AHSI, Milan Italy) and analyzed by the Gen5 software (BioTek, AHSI, Milan Italy).

Cells were exposed to treatments, washed twice with PBS, fixed in 4% paraformaldehyde in PBS for 10 min, washed briefly with PBS and then incubated with Phalloidin-Fluorescent Conjugate (Santa Cruz Biotechnology, DBA, Milan, Italy). The images were obtained using the Cytation 3 Cell Imaging Multimode reader (BioTek, AHSI, Milan Italy) and analyzed by the Gen5 software (BioTek, AHSI, Milan Italy).

To analyze cell cycle distribution, BCAHC-1 cells (1 × 10⁵) were cultured in regular medium in 6 well plates and shifted in medium without serum. Next, BCAHC-1 cells were exposed to treatments, then pelleted, once washed with PBS and fixed in 50% methanol overnight at − 20 °C, before to stain with a solution containing 50 µg/ml propidium iodide (PI) in 1×PBS, 20 U/ml RNAse-A and 0.1% Triton (Merck Life Science, Milan, Italy). Cell phases were estimated as a percentage of a total of 10 000 events. Samples were then analyzed with CytoFLEX flow cytometry (Beckman-Coulter, Milan, Italy).

BCAHC-1 cells (1 × 10⁴) were seeded in 24-well plates in regular growth medium, washed once they had attached, incubated in medium containing 2.5% charcoal-stripped FBS and then exposed to treatments. Treatments were renewed every day. The proliferation rate was calculated counting the cells on day 5 using the Countess Automated Cell Counter, as recommended by the manufacturer’s protocol (Thermo Fisher Scientific, Milan, Italy).

BCAHC-1 cells were cultured in regular growth medium to 90% confluence. Cells were then trypsinized, counted and seeded (1 × 10³) in 6-well plates in medium containing 2.5% charcoal-stripped FBS and then exposed to treatments, as indicated. Treatments were renewed every 3 days. After 10 days, cells were washed with PBS, fixed in acetone:methanol (1:1) for 3 min at room temperature and then stained with Giemsa (Merck Life Science, Milan, Italy) for 10 min. A total of 10 pictures for each condition was detected using a digital camera and colony number was measured by ImageJ program.

For spheroid generation, 100 μl/well of BCAHC-1 cell suspensions (1 × 10⁴) were placed into 2% agar coated 24-well plates in medium containing 2.5% charcoal stripped FBS. Three days after seeding, tumor spheroids (a single spheroid per well) were exposed to treatments and a 50% medium and treatment renewal was performed every 2 days. Images were obtained on day 20 using a conventional inverted microscope, thereafter cell number per spheroid was determined by trypsinizing three different spheroids, mixing the cell suspension with trypan blue and counting the number of viable cells. The total number of cells obtained was divided by the number of trypsinized spheroids.

CAFs were placed in medium without serum and treated with vehicle (−) or 10 nM insulin for 4 h. Thereafter, CAFs were washed and fresh medium was added without serum for additional 8 h. The supernatants were then collected, centrifuged at 3500 rpm for 5 min to remove cell debris and used as conditioned medium in the appropriate experiments.

Transwell 8 μm polycarbonate membranes (Costar, Merck Life Science, Milan, Italy) were used to evaluate in vitro cell migration. In 300 μl serum free medium, BCAHC-1 cells (5 × 10⁴) previously exposed to treatments were seeded in the upper chamber. Conditioned medium of CAFs was added to the bottom chambers. 8 h after seeding, cells on the upper surface of the membrane were then removed by wiping with Q-tip, and migrated cells were fixed with 100% methanol, stained with Giemsa (Merck Life Science, Milan, Italy), photographed using a digital camera and counted using the WCIF ImageJ software.

The Matrigel drop assay was performed as previously described [31, 32] and modified by suspending 3 × 10⁴/drop of BCAHC-1 cells, previously exposed to treatments, in 15 μl of complete medium and gently mixed with 15 μl of Corning® Matrigel® Growth Factor Reduced (GFR) Basement Membrane Matrix (Biogenerica, Catania, Italy). The cell/matrigel suspension was layered onto the surface of 12-well plate to form a well-defined drop and placed at 37 °C to solidify. 2 ml of conditioned media collected from CAFs previously treated with vehicle (−) or 10 nM insulin were placed over the drop. Cells were observed at specified time points and drops were photographed using the ImageFocus Plus V2 software of the inverted Oxion Inverso microscope (Euromex Microscopen bv, The Netherlands). The number of cells migrated out of the drop was measured.

Bioinformatics studies were performed using the Invasive Breast Cancer Cohort data of the publicly available dataset The Cancer Genome Atlas (TCGA). The patient clinical information, gene expression data (RNA Seq V2 RSEM) and the pathway activity (z-scores) were retrieved from UCSC Xena (https://​xenabrowser.​net/​) on the 7th March 2022. Samples (n. 1247) were filtered for missing values.

The survival analyses were performed using the IR-mediated signaling pathway activity and the IR and CXCR4 gene expression data of the TCGA patients along with the overall survival (OS), and disease-free interval (DFI) information. The survivALL package was employed to examine Cox proportional hazards for all possible points of separation (low–high cut-points). The cut-point with the lowest p-value was selected [33], therefore dividing the patients with high (n = 488) and low (n = 87) IR-mediated signaling pathway activity, high (n = 166) and low (n = 90) IR expression or high (n = 145) and low (n = 107) CXCR4 expression. The Kaplan Meier survival curves were generated using the R survival and survminer packages.

The statistical analysis was performed using ANOVA followed by NewmanKeuls’ test to determine differences in means. Densitometric analysis was performed using the freeware software ImageJ that allowed to quantify the band intensity of the protein of interest with respect to the band intensity of the loading control. All bioinformatics analyses were carried out using R Studio. Box plots were performed with the tidyverse package and the related statistical analysis was carried out by using the t-test. Heatmaps were drawn on the log₂ fold changes of gene expression using the pheatmap package. p‐values < 0.05 were considered statistically significant.

Previous studies have shown that a crosstalk between estrogen and insulin transduction pathways promotes the progression of BC [27, 34, 35]. Therefore, we began our study analyzing the expression and activity of IR in the TCGA cohort of ER-positive and negative BC patients. First, we ascertained that the IR mRNA expression as well as the IR-mediated signaling pathway are significantly higher in ER-positive than ER-negative BCs (Fig. 1A, B). In addition, stratifying the BC patients on the basis of the PAM50 gene signature, both IR expression and signaling were found higher in the BC luminal A and luminal B patients respect to the other BC subtypes (Additional file 1: Fig. S1A, B). In accordance with these data, was found enriched and associated with a worse overall survival (OS) (Fig. 1C) in ER-positive respect to ER-negative BC patients. Pre‑clinical and currently ongoing clinical trials indicate that the insulin-sensitizer metformin elicits anti-tumor effects in ER-positive BCs and is associated with a lower risk of developing BC in patients with T2D respect to control patients [11, 20, 36‐40]. Hence, we aimed to ascertain whether metformin may inhibit insulin-dependent signaling pathways like phosphoinositide 3-kinase (PI3K)/AKT and mitogen-activated protein kinase (MAPK). To this end, we used as a model system a BC cell line namely BCAHC-1, which is characterized by the peculiar expression of IR and the 46 kDa ERα splice variant (ERα46) [27, 41, 42]. First, the insulin-induced IR phosphorylation was prevented using the IR inhibitor OSI-906, but not in the presence of the PI3K inhibitor alpelisib and the MEK inhibitor trametinib (Fig. 1D). The activation of AKT upon insulin exposure was totally abrogated by OSI-906 and alpelisib, but only partially dampened by trametinib (Fig. 1E) as previously reported [43], whereas the insulin-promoted activation of the ERK1/2 pathway was inhibited by OSI-906 and trametinib but not using alpelisib (Fig. 1F). Next, we assessed that metformin reduces the IR phosphorylation prompted by insulin, without any change in total IR protein expression (Fig. 1G). Moreover, we determined that metformin prevents the activation of both AKT and ERK1/2 upon insulin exposure in BCAHC-1 cells (Fig. 1H, I). On the basis of these findings and in line with previous studies showing that PI3K/AKT and ERK1/2 transduction pathways mediate the expression of insulin target genes, such as c‐Fos and Cyclin D1 [27, 44‐46], we found that insulin induces c-Fos expression in BCAHC-1 cells at both mRNA (data not shown) and protein levels (Fig. 2A–C). Worthy, OSI-906, alpelisib, trametinib and metformin abrogated either the protein increase (Fig. 2A–C) or the activation of a promoter construct of c-Fos transiently transfected in BCAHC-1 cells (Fig. 2B–D). Similarly, we established that OSI-906, alpelisib and trametinib prevent the Cyclin D1 protein increase and the transactivation of a Cyclin D1 promoter construct upon insulin exposure (Fig. 2E, F). Reminiscing our previous findings showing that c-Fos is involved in the regulation of Cyclin D1 [27], we demonstrated that the protein induction of Cyclin D1 by insulin is prevented transfecting BCAHC-1 cells with a DN/c-Fos expression vector (Fig. 2G). Furthermore, we showed that metformin abolishes the Cyclin D1 protein increase (Fig. 2H) and the activation of a Cyclin D1 promoter construct (Fig. 2I) triggered by insulin. Taken together, these findings indicate that in BCAHC-1 cells metformin inhibits the insulin-induced IR/PI3K/MAPK/c-Fos transduction pathway, thus preventing the up-regulation of Cyclin D1. In line with the acknowledged role of Cyclin D1 as a regulator of G1/S transition in the cell cycle [47], we found that metformin leads to the arrest within the G0/G1 phase of the cell cycle in BCAHC-1 cells (Fig. 3A, B). Interestingly, the results of flow cytometric analysis of propidium iodide (PI)-staining showed that insulin induces S-phase entry in BCAHC-1 cells, however this effect is no longer evident in the presence of metformin (Fig. 3A, B) and using OSI-906, alpelisib or trametinib (data not shown). In accordance with these observations, the proliferation and colony-forming ability of BCAHC-1 cells upon insulin treatment were abolished by metformin (Fig. 3C–E) as well as in the presence of OSI-906, alpelisib, trametinib (Additional file 2: Fig. S2A–C). In addition, the spheroid expansion of BCAHC-1 cells prompted by insulin was inhibited by metformin (Fig. 3F, G) as well as OSI-906, alpelisib and trametinib (Additional file 2: Fig. S2D, E). Overall, these data indicate that metformin impairs the growth responses induced by insulin in BCAHC-1 cells.

Considering that a negative correlation between metformin and the metastatic spread of BC has been reported [48‐50], we then assessed whether metformin may abolish the capability of insulin to induce a metastatic gene signature in BCAHC-1 cells. In this vein, we performed a TaqMan Gene Expression Assay consisting of a Human Tumor Metastasis Array in BCAHC-1 cells exposed to insulin in the presence or absence of metformin. Among the genes that showed at least a 0.25 log2 fold change upon insulin treatment respect to vehicle and the reduction of this increase in the presence of metformin, we focused on the C-X-C Motif Chemokine Receptor 4 (CXCR4) due to its acknowledged role as prognostic marker and main driver of BC metastasis [51‐53]. Evaluating the expression of CXCR4 in ER-positive BC samples of the TCGA dataset, we assessed by pairwise comparison that CXCR4 levels are significantly higher in ER-positive BCs respect to matched normal tissues (Fig. 4A). Thereafter, we ascertained whether the mRNA expression of both CXCR4 and IR would be predictive of the outcome of ER-positive BC patients. In this regard, survival analyses revealed that high IR levels are associated with a short disease-free interval (DFI) in BC patients showing increased CXCR4 expression (Fig. 4B). Similarly, a correlation between high CXCR4 levels and poor prognosis characterizes BCs displaying enhanced IR levels (Fig. 4C). Considering that insulin modulates pro-metastatic mediators involved in the regulation of CXCR4 in BC cells [54, 55], we ascertained that insulin increases both mRNA (Fig. 4D) and protein levels (Fig. 4E) of CXCR4 in BCAHC-1 cells. Next, immunoblot experiments showed that either the IR inhibitor OSI-906 (Fig. 4F) or metformin (Fig. 4G) prevent the up-regulation of CXCR4 triggered by insulin, suggesting that metformin prevents the expression of CXCR4 mediated by insulin/IR axis in BCAHC-1 cells.

A crosstalk between tumor cells and cancer‐associated fibroblasts (CAFs) may fuel the growth and metastasis of BC [56‐59]. In particular, the CAF-secreted CXCR4 ligand, namely CXCL12 (also known as SDF-1), is an important factor linking stromal and BC cells toward aggressive malignant features [59‐64]. On the basis of the aforementioned findings and considering that insulin has been implicated in the release of pro-tumorigenic and inflammatory mediators by stromal cells [65, 66], we ascertained whether insulin may promote the expression and secretion of CXCL12 in breast CAFs. By real-time PCR, immunoblotting and immunofluorescence experiments we demonstrated that the expression of CXCL12 increases at both mRNA (Fig. 5A) and protein levels (Fig. 5B, C) in CAFs exposed to insulin. Moreover, CXCL12 was found up-regulated in conditioned medium (CM) of CAFs treated with insulin (Fig. 5D), as ascertained by acetone precipitation assays. Altogether, these data suggest that insulin induces the increase of the CXCR4 ligand CXCL12 and stimulates its release by CAFs.

Considering the pivotal role played by the CXCL12/CXCR4 axis in BC development and metastasis [51, 64, 67, 68], we evaluated whether insulin may engage the CXCL12/CXCR4 signaling, leading to aggressive features in BCAHC-1 cells. Hence, in order to visualize the F-actin pattern CM from CAFs exposed to vehicle and insulin was collected and used as culture medium in BCAHC-1 cells, which were previously treated with vehicle and insulin. These experimental conditions allowed us to observe a marked increase in F-actin assembly in BCAHC-1 cells treated with insulin and cultured with CM from insulin-stimulated CAFs (Fig. 6A, B). These effects were prevented using the specific CXCR4 antagonist AMD3100 and metformin (Fig. 6A, B). The aforementioned findings were corroborated by transwell migration assays performed in BCAHC-1 cells that were previously treated with insulin and cultured as described above. Of note, the migration of BCAHC-1 cells prompted by insulin was further potentiated culturing cells with CM obtained from insulin-treated CAFs (Fig. 6C, D), however this response was no longer evident treating BCAHC-1 cells with AMD3100 and metformin (Fig. 6C, D). Similarly, three-dimensional motility assays showed that AMD3100 and metformin prevent the outflow from the matrigel drop of insulin-treated BCAHC-1 cells cultured with CM from insulin-stimulated CAFs (Fig. 7A, B). Taken together, these results suggest that the insulin-activated CXCL12/CXCR4 axis sustains a paracrine feedforward loop coupling cancer and stromal cells, thus fostering a motile phenotype in BCAHC-1 cells.