Metastatic Breast Carcinoma–Associated Fibroblasts Have Enhanced Protumorigenic Properties Related to Increased IGF2 Expression

Metastatic breast tumors are almost always incurable: They progressively become resistant to chemotherapy as well as targeted therapies even if there is an initial response. The usual explanation for this inexorable process is selection of treatment-resistant clones, generated by the inherent genomic instability of breast tumors. However, there is little clinical evidence for the selection of any “resistance gene” in metastatic breast cancers, even with next-generation sequencing (1). The classic resistance gene, ABCB1 coding for P-glycoprotein, has not been clinically validated (2). One hypothesis to explain the incurability of metastatic lesions is that the tumor microenvironment protects tumor cells in metastatic lesions, more so than in the primary tumor. Moreover, this protection may vary from metastatic site to site, which may explain the frequent clinical observation of differential response to therapy in different metastatic sites. Much attention has centered on the essential interplay of the cells of the tumor microenvironment (such as fibroblasts, lymphocytes, macrophages, etc.) in regulating the biology and progression of cancer cells. Extensive differences in tumor stroma compared with normal stroma have been widely observed and several studies have shown that carcinoma-associated fibroblasts (CAFs) may affect tumor cells' sensitivity to cancer therapy. For instance, Straussman and colleagues (3) showed that CAFs may cause drug resistance in melanomas via the secretion of HGF. They also found that coculture with CAFs rescued a HER2⁺breast cancer cell line from sensitivity to lapatinib. However, this effect was not observed when cancer cells were treated with cytotoxic therapy. Although it is well established that CAFs from primary breast tumors (pCAFs) show significant molecular and phenotypic changes compared to normal fibroblasts (NFs), which confer on them a broad protumorigenic ability, there have been almost no studies of CAFs obtained from metastatic sites, and thus, the contribution of metastatic CAFs (or mCAFs) to the incurability of metastasis is presently unknown. Indeed, there was previously no evidence that mCAFs are different from CAFs present within primary tumors. Through an ongoing metastatic tumor banking effort, we were able to isolate and grow eight mCAFs from different metastatic breast tumor sites. Here, we show that mCAFs show significant and marked phenotypic differences compared with both pCAFs and NFs. These differences are reflected in molecular differences especially regarding the expression of growth factors such as IGF2 and IFNβ, which have clear therapeutic significance.

Antibodies against total and phospho-tyrosine, ERK (T202/Y204), Akt (S473), and STAT3 (Y705) were purchased from Cell Signaling Technology. α-SMA, FAP, actin, cytokeratin, vimentin, E-cadherin, N-cadherin, and secondary antibodies were also from Cell Signaling Technology. The total and p-IGF1βR antibodies and corresponding total IGF2 protein (Cell Signaling Technology) were a kind gift from Dr. Michael Pollak (McGill University, Montreal, Canada). Recombinant Human IGF2 was obtained from Peprotech. Doxorubicin and paclitaxel were from Jewish General Hospital, McGill University. CFSE was from Stemcell Technologies. Phorbol myristate acetate (PMA) and ionomycin were from Sigma-Aldrich. LEAF-purified anti-human CD3, CD3-PE, and CD69-FITC were from BioLegend. Frozen PBMCs (human peripheral blood mononuclear cells) were from Stemcell Technologies with the health donor lot No. 1903040032. MDA-MB-436, BT-20, MDA-MB-231, and MCF-7 cell lines were purchased from ATCC. The cell lines were thawed for a maximum of 2-month experimental use, and then new vials were thawed. Periodically, the Mycoplasma was tested with MycoAlert PLUS Mycoplasma Detection Kit (Lonza).

Primary and metastatic breast cancer tissue was obtained from patients with breast cancer at the Jewish General Hospital (JGH) in Montreal. The collection and use of human samples was approved by the Institutional Review Board (IRB), JGH (No. 05-006), which is in accordance with the Declaration of Helsinki and the Belmont Report. Patients signed informed consent to provide samples for the JGH Biobank for studies approved by the IRB of the JGH. pCAFs and NFs were described in ref. 4 and mCAFs were generated with the same procedure. Normal, primary and metastatic CAF tissue culture was carried out according to the work of Hosein and colleagues (4). All CAF experiments were performed within two to 10 passages after collection from tumor samples. After the isolation of CAFs from biopsies, Mycoplasma testing was performed once the cell growth was stable, and then CAFs were frozen and stored in liquid nitrogen. For every new experiment, one frozen cell vial was used for 2-month in vitro culture. Periodically, the culturing CAFs were tested for Mycoplasma with a Mycoplasma Detection Kit (Lonza).

Cells (1 × 10⁶) of each CAF were plated in a T75 tissue culture flask and cultured in 10-mL DMEM medium without FBS at 37°C, 5% CO₂ in a humidified atmosphere incubator for 3 days. At day 3, the supernatant was collected and centrifuged to remove the cellular debris (1,200 rpm/minute × 5 minutes). Cell-free conditioned media (CM) was then added to breast cancer cell line cultures for further experiments.

Transwell migration assay was performed in 6.5-mm-diameter Boyden chambers with a pore size of 8.0 μm (Corning). MDA-MB-436 cells (5 × 10⁴) were suspended in 0.5% DMEM medium and placed in the upper compartment of transwell chambers and 700 μL of CM from the different CAFs was placed in the lower compartment. After incubation in a 37°C, 5% CO₂ humidified atmosphere for 24 hours, cells in the lower surface of the top compartment were fixed in 4% formaldehyde for 20 minutes followed by staining with 0.05% crystal violet, and five random fields of each were counted at ×200 magnification. For invasion assays, cells were placed in 24-well Matrigel-coated invasion chambers (Corning). The lower chambers were filled with 700-μL CM from the distinct CAFs as a chemoattractant. A suspension of 5 × 10⁴MDA-MB-436 cells was plated in the upper compartment. The cells were incubated in a 37°C, 5% CO₂ humidified atmosphere for 24 hours, and then the membrane insert was fixed with 4% formaldehyde for 20 minutes and stained with 0.05% crystal violet, and stained cells counted.

Breast cancer cell lines were incubated with CM from different CAFs (indirect coculture), the CAFs themselves (direct coculture) or with IGF2 (20 ng/mL), then lysed in RIPA buffer (50 mmol/L HEPES, 150 mmol/L NaCl, 1% NP-40, 0.5% sodium deoxycholate, and 0.1% SDS) containing protease inhibitors. Cell lysates were centrifuged at 14,000 rpm, 4°C for 20 minutes to remove cellular debris, and then the supernatants were mixed with 6× SDS sample buffer and boiled for 5 minutes. Tumor samples from mice were homogenized in RIPA buffer containing protease inhibitors. For immunoprecipitation, tumor lysates containing 500 μg protein were incubated overnight with 2 μg of Ab and 35 μL of 50% protein G-Sepharose (GE Healthcare Bio-Science Inc). The immunoprecipitates were washed three times and boiled in 6× SDS sample buffer. Following separation by SDS-PAGE and immunoblotting, proteins were transferred to polyvinylidene difluoride membranes (Immobilon-p, Millipore) and immunoblotted with the indicated antibodies.

MDA-MB-436 cells were incubated with CM from the different CAFs for 3 days in 96-well plates. The supernatant was then replaced with 200-μL 10% AlamarBlue for a 3-hour incubation at 37°C, 5% CO₂ humidified atmosphere. The resulting fluorescence was read on a FLUOstar Optima, using 560-nm (Excitation) and 590-nm (Emission) filter settings and results were analyzed by plotting fluorescence intensity.

MDA-MB-436 cells were transfected with GFP plasmid for 48 hours; the cells were then trypsinized and 1 mL of complete DMEM medium was added. Plasmid containing clones were selected in 1 mg/mL G418 supplemented medium and individual colonies were transferred to 12-well plate using cloning rings (Sigma-Aldrich) for monoclonal expansion.

MDA-MB-436 cells (10⁶or 0.7 × 10⁶) were mixed with pCAFs (CAF53), mCAFs (BM113), and N-Fs (Norm2) at a ratio of 3:1 or 1:1, respectively; they were injected into the mammary fat pads of 6-week-old female athymic nude mice (Charles River Laboratories). Tumor growth was monitored twice a week and measurements taken with a digital vernier caliper. Mice with palpable (around 150–200 mm³) tumors were administrated via intraperitoneal injection with IgG isotype vehicle control antibody or IGF-2 blocking antibody BI836845 (100 mg/kg; Boehringer Ingelheim) twice per week for 2 weeks. Tumor volume (mm³) was determined with the formula (length × width²)/2. All the animal experiments were approved by the McGill University Animal Care Committee (Protocol Number: 2017-7931).

Human samples were harvested and fixed in 10% neutral-buffered formalin (Thermo Fisher Scientific), processed, and embedded in paraffin. After deparaffinization and rehydration, 5-μm sections were prepared and antigenic epitopes were retrieved in Tris-EDTA buffer (pH 9.0) in a steamer. Sections were then incubated with primary anti-IGF2 antibody (Santa Cruz Biotechnology, gift from Dr. Pollak, McGill University) overnight at 4°C. The slides were then washed in PBS three times for 5 minutes and incubated with an anti-rabbit secondary antibody conjugated to horseradish peroxidase for 1 hour at room temperature. After addition of 3,3′-diaminobenzidine (DAB) according to the manufacturer's instructions (Sigma), the slides were sealed using VectaMount AQ Aqueous Mounting Medium (Cedarlane).

Total RNA from CAFs was extracted using the AllPrep DNA/RNA/miRNA Universal Kit (Qiagen) following the manufacturer's instructions; the quality and quantity of the total RNA were measured using an Agilent Model 2100 Bioanalyzer, and samples showing a RIN >9 were selected for further study. Samples were stored in RNase-free water at −80°C.

RNAseq analysis was performed at the McGill University and Genome Quebec Innovation Center. Libraries were generated from 500 ng of total RNA using the TruSeq stranded Total RNA Sample Preparation Kit (Illumina TS-122-2301, San Diego, California, United States) with Ribo-Zero Gold rRNA removal kit (Illumina, San Diego, California, United States), as per the manufacturer's recommendations. Libraries were quantified using the Quant-iT PicoGreen dsDNA Assay Kit (Life Technologies) and the Kapa Illumina GA with Revised Primers-SYBR Fast Universal kit (Kapa Biosystems). Average size fragment was determined using a LabChip GX (PerkinElmer) instrument. The libraries were sequenced using the Illumina HiSeq 2500 PE100 sequencer, 100 nucleotide paired-end reads, generating approximately 60 million reads sample. The sequencing reads were trimmed using CutAdapt and mapped to the human reference genome (hg19) using STAR aligner (version 2.4.0e), with default parameters, and annotated using the Gencode V19 (version 19, December 2013) annotation. FeatureCounts was used for expression quantification. Differential expression analyses was performed using the Bioconductor package DESeq2. Specifically, “SummarizedExperiment” objects containing count matrices were generated from BAM files using the “summarizeOverlaps” function of the “GenomicAlignments” package.

We plotted the first and second principal components obtained from the DESeq2 package using log-transformed data. RNAseq data are accessible at the ENA accession number PRJEB34465. The ENA accession number is PRJEB34465.

For RT-PCR analysis, total RNA was isolated from 10⁶cells, using a Qiagen Kit. Total RNA (1 μg) was used for the first complementary strand generation using QuantiTect Reverse Transcription Kit (Qiagen). The primers for gene expression analysis are as following: Human FAP (NM_004460): F- GATACCACTTACCCTGCGTATG, R- ACAATCCCATGTCTGCCAG; Human NANOG (NM_024865): F- GAAATACCTCAGCCTCCAGC, R- GCGTCACACCATTGCTATTC; Human SNAI1 (NM_178310): F- GCACTCAGACGCCAAGAA, R- GGACCATCCCTCCTACCTG; Human GAPDH (NM_002046): F- AATCCCATCACCATCTTCCAG, R- TTCACACCCATGACGAACAT. All primers displayed 90% efficiency with a single melting curve in CFX-96 Real-Time System (Bio-Rad). Expression levels of the housekeeping gene gapdh were used as controls.

Validated siRNA targeting IGF1R (SI00017521 and SI02624552) and scramble control were purchased from Qiagen and used according to the manufacturer's instructions.

To measure IGF2 production by different cells, 10,000 cells per well were seeded in a 96-well plate cultured alone or with the addition of 3,000 MDA-MB-436 or MCF-10 cells per well for 48 hours at 37°C, 5% CO₂ humidified atmosphere with 10% DMEM medium. Then, the cells were cultured in 200-μL DMEM without FBS for another 3 days and IGF2 levels measured using an ELISA Kit (AnshLabs) according to the manufacturer's instructions. IFNβ and CXCL-12 levels were measured on the supernatants from the different cultured CAFs using the corresponding ELISA kit (R&D Systems).

MDA-MB-436 parent cells or incubated with the CM from the distinct CAFs were washed twice with PBS and then harvested. Detached cells were resuspended in PBS supplemented with 0.5% FBS (1 × 10⁶cells/80 μL). Fluorochrome-conjugated mAbs against human CD44-FITC and CD24-AF647 (BD Biosciences) were added to the cell suspension as recommended by the manufacturer, and incubated at 4°C in the dark for 30 minutes with swirling tubes every 15 minutes. The labeled cells were analyzed on a BD FACS Aria II.

Whole blood was collected from healthy volunteer donors. Blood was then mixed with Ficoll-Paque PLUS (GE Healthcare) and PBS in a ratio of 1:1:1. The mixture was centrifuged at 400 × g for 35 minutes at room temperature. The PBMCs at the interface were washed with PBS at 300 × g at 4°C twice and labeled with CFSE. We then stimulated them with PMA (50 ng/mL)/ionomycin (500 ng/mL) for 2 or 3 days. PBMCs were also incubated in a 96-well plate, which was coated with ultra-LEAF purified anti-human CD3 antibody (10 μg/mL) overnight for 2 days. At the end of the incubation, PBMCs were stained with CD3-PE and CD69-FITC for T-cell activation assay on BD FACS Aria II.

All statistical analyses were performed using the Student t test with a 95% confidence interval with P < 0.05 interpreted as statistically significant.

Stromal fibroblasts were isolated from 16 patients from normal breast tissue or from primary or metastatic breast tumors, and designated NFs, pCAFs, and mCAFs, respectively (Supplementary Table S1). We obtained two mCAFs from the same patient, one from a bone metastasis (BM69) and one from a liver metastasis (BM69L), the latter collected 1 year after the bone metastasis were collected. Each CAF sample showed strong vimentin and absent cytokeratin 18 staining (Supplementary Fig. S1A) as well as a mesenchymal cell shape. We observed that mCAFs displayed higher α-smooth muscle actin (SMA) protein but similar fibroblast activation protein (FAP) protein expression compared with pCAFs and NFs (Supplementary Fig. S1B and S1C).

To examine how CAFs affect breast cancer cells, we chose triple-negative breast cancer (TNBC) cell lines, MDA-MB-436 and BT-20, which represent the most aggressive form of breast cancer (5). We found that CM from mCAFs stimulated MDA-MB-436 cell proliferation significantly more than CM from pCAFs and NFs (Fig. 1A). Similarly, mCAFs induced stronger breast cancer cells MDA-MB-231 proliferation compared with pCAFs.

Although the difference was less evident in MCF-7 cells (Supplementary Fig. S2A). CM from mCAFs also significantly increased the capacity of MDA-MB-436 and BT-20 cells to form colonies compared with CM from NFs and pCAFs (Fig. 1B). Moreover, compared with CM from NFs and pCAFs, CM from mCAFs induced significantly more migration and invasion in MDA-MB-436 cells tested in Boyden chambers (Fig. 1C and D).

To investigate the effects of direct coculture of CAFs with TNBC cells, we grew CAFs as a feeder layer over which GFP-tagged TNBC cells were plated. The breast cancer cells formed spheroids when cocultured over CAFs. We observed that breast cancer cell spheroids on mCAFs were significantly larger compared with those cultured over NFs and pCAFs (Fig. 1E). Taken together, these results suggest that mCAFs have greater protumorigenic effects in vitro than pCAFs and normal fibroblasts.

As several studies have reported that CAFs are implicated in drug resistance (6–8), we studied the effects of the different CAFs on sensitivity to the chemotherapeutic drug doxorubicin in MDA-MB-436 breast cancer cells. We observed that CM from mCAFs did not significantly alter the sensitivity of MDA-MB-436 cells to doxorubicin, compared with NFs and pCAFs, although there was a trend toward a decreased sensitivity with CM from mCAFs compared with the NFs (P = 0.08; Supplementary Fig. S2). We then tested drug sensitivity in a 3D model of coculture spheroid growth using GFP-tagged MDA-MB-436 cells. GFP-MDA-MB-436 cells were mixed with stromal fibroblasts (in a 1:3 ratio) or grown alone on Matrigel and then treated with drug for 10 days. Doxorubicin treatment induced a significant decrease in 3D growth of breast cancer cells in direct coculture with NFs and pCAFs but not with mCAFs (Fig. 1F). These results suggest that mCAFs in direct contact with breast cancer cells offer greater protection from the cytotoxic effects of doxorubicin than pCAFs and NFs.

We next examined whether the distinct CAFs differentially support tumor formation in vivo using orthotopic mouse xenografts of MDA-MB-436 cells. Cells were mixed with human NFs (Norm2), pCAFs (CAF53), or mCAFs (BM-113) at a 3:1 tumor cell:fibroblast ratio and then injected into the mammary fat pads of nude mice. We observed significantly larger tumors in mice coinjected with MDA-MB-436 and mCAFs compared with pCAFs and NFs (Fig. 2A and B). We then increased the relative number of fibroblasts to a ratio of 1:1 with tumor cells, maintaining the same number of total cells injected (1.3 × 10⁶). As expected, owing to the decreased number of breast cancer cells, we observed that the emergence of the tumor was delayed (Fig. 2C). However, although the primary tumor size was comparable between the pCAFs and mCAF group but still significantly greater than the NF group, all mice injected with mCAFs displayed signs of metastatic disease, including ascites and intraperitoneal tumor deposits (Fig. 2D and E; Supplementary Fig. S3A–S3C). Only one of three mice with a pCAF coinjection developed metastatic disease. Taken together, these results indicate that mCAFs are more potent in enabling tumor growth as well as metastasis than pCAFs and NFs.

Epithelial–mesenchymal transition (EMT) is associated with the development of drug resistance, metastatic disease as well as stem cell–like properties (9). Breast cancer stem cells have been associated with a CD44^highCD24^lowpopulation. Approximately 10% of our MDA-MB-436 cells express these markers, suggesting that a small subpopulation of MDA-MB-436 cells may have properties of breast cancer stem cells. We found that incubation with CM from pCAFs modestly increased the CD44^highCD24^lowpopulation in MDA-MB-436 cells compared with NFs (Fig. 3A). However, CM from mCAFs induced a much greater CD44^highCD24^lowpopulation, so much so that this subpopulation of cells constituted about 50% of MDA-MB-436 cells.

Cadherin switching from E-cadherin to N-cadherin expression is a critical cellular event in the process of EMT (10). We found that N-cadherin protein levels were increased, whereas E-cadherin protein levels decreased in MDA-MB-436 cells directly cocultured with mCAF, more so than with pCAF or NF cocultures, when compared with levels in MDA-MB-436 cells alone (Fig. 3B). We also found that mCAF direct coculture was more potent than direct coculture with pCAFs or NFs in increasing RNA levels of SNAI1 and NANOG, two master regulators of EMT and stemness (11), in MDA-MB-436 cells (Fig. 3C and D). Taken together, these results indicate that mCAFs are more potent EMT and cancer stem cell inducers in MDA-MB-436 cells than pCAFs and NFs.

We next performed RNA sequencing (RNA-seq) analysis of all 16 fibroblasts grown in DMEM medium without FBS. Principal component analysis (PCA) revealed that pCAFs segregated with the NFs and away from the mCAFs (Fig. 4A). Moreover, the mCAFs segregated into two groups, one (cluster 1) with four mCAFs from skin, lung, and bone metastases and the second (cluster 2) containing the three mCAFs from liver metastases and one from a bone metastatic site. Interestingly, the two mCAFs from the same patient segregated separately, with the liver mCAF together with the other liver mCAFs. Differential expression analysis was performed and we found that 331 genes were significantly upregulated in mCAFs compared with pCAFs (>2-fold, P < 0.05, >10 reads in all metCAFs) and 256 genes were downregulated (<2-fold, P < 0.05, >10 reads in all pCAFs). The top 10 genes overexpressed in mCAFs are shown in Supplementary Table S2. The most differentially expressed gene is IGF2 (insulin-like growth factor 2). A DAVID Gene Ontology analysis of upregulated genes in mCAFs relative to pCAFs revealed that the biological processes most significantly enriched in this gene set included IFN signaling, extracellular matrix organization, response to hypoxia, and positive regulation of MAPK cascade (Fig. 4B). Ingenuity pathway analysis (IPA) analysis was performed on up- and downregulated genes. The top five most significant canonical pathways include hepatic fibrosis/hepatic stellate cell activation and IFN signaling. We then separately compared each of the two clusters of metCAFs to pCAFs and performed IPA analysis on significantly up- and downregulated genes. We found that the most significant canonical pathway altered in non-liver mCAFs (cluster 1) is “IFN signaling,” whereas that in cluster 2 was “hepatic fibrosis/hepatic cell stellate activation,” consistent with the predominance of liver metCAFs in this cluster. DAVID GO analysis of upregulated genes in cluster 1 also identified IFN signaling as the most significantly enriched biological process (Supplementary Fig. S4A). We then measured IFNß in the CM of these mCAFs using an ELISA assay and found that mean IFNβ levels were much higher in mCAFs compared with pCAFs and also 1.7 times higher in cluster 1 than in cluster 2 (P < 0.01; Fig. 4C). DAVID GO analysis of upregulated genes in cluster 2 indicated an enrichment in genes related to inflammatory response, response to hypoxia and ERK signaling (Supplementary Fig. S4B). Levels of mRNA of PDGFA growth factor, one of the genes in the ERK signaling group, is expressed >2-fold higher in cluster 2 compared with the mean levels in cluster 1 mCAFs by both RNA-seq and confirmed by qRT-PCR (Supplementary Fig. S5). In fact, RNA-seq shows that PDGFA mRNA is expressed at a mean of >2-fold higher levels in all mCAFs compared with all pCAFs (P < 0.05). Differentially expressed genes between the two clusters were consistently differentially expressed in the same direction (liver vs. bone) in the two samples from the same patient (BM69 liver and BM69 bone; Supplementary Table S3).

As noted, the most upregulated gene in mCAFs compared with pCAFs was IGF2. Using an ELISA assay, we confirmed that the transcriptional data were correlated with protein levels: IGF2 levels were significantly higher in the CM of mCAFs compared with pCAFs or NFs (Fig. 4D). Interestingly, secreted IGF2 levels increased markedly when there was a physical interaction between mCAFs and MDA-MB-436 cells, an increase that was not observed in direct coculture with the normal breast MCF10A cells (Fig. 4D) nor when the MDA-MB-436 cells were incubated with the CM from the mCAFs without direct contact between the cells (Supplementary Fig. S6). We then determined that CM from mCAFs increased phospho-IGF1R, p-STAT3, p-AKT, and p-ERK levels in MDA-MB-436 cells to levels comparable if not greater than that observed with direct IGF2 stimulation (Fig. 4E), confirming that CM from mCAFs can trigger signalling downstream of the Igf-1R. We found that phospho-IGF1R and phospho-AKT levels in MDA-MB-436 cells were increased to a greater extent by CM from mCAFs compared with CM from pCAFs and NFs (Fig. 4F). Similar results were observed with CM incubated with other TNBC cells (BT-20 and MDA-MB-231 cells; Supplementary Fig. S7A and S7B). These findings suggest that mCAFs more potently induce IGF signaling, consistent with their higher level of IGF2 secretion. Finally, high levels of IGF2 protein were observed in the stroma and the tumor cells of the clinical metastatic lesions from which the mCAFs were derived (Fig. 4G). Furthermore, IGF2 protein expression was also very high in breast cancer metastatic lesions in that it developed when MDA-MB-436 breast cancer cells were mixed in a 1:1 ratio with mCAFs (1:1) in mouse xenografts (Supplementary Fig. S3D–S3F).

To further investigate the role of IGF2 signaling in the effects of mCAFs on breast cancer cells, we silenced IGF-1R in MDA-MB-436 cells using siRNA (Supplementary Fig. S8). IGF-1R silencing in MDA-MB-436 cells significantly decreased breast cancer cell migration and invasion compared with a scrambled siRNA control (Fig. 5A). Silencing IGF-1R in MDA-MB-436 and in BT-20 cells markedly decreased clonogenic growth stimulated by CM from the different mCAFs compared with scrambled siRNA controls (Fig. 5B). Taken together, these findings support a role for IGF2 signaling in mediating mCAF effects on breast cancer cell lines.

To further explore the role of IGF2 in mediating the effects of CAFs on breast cancer cells, we used an antibody that binds IGF1 and IGF2, BI836845 or xentuzumab (12), currently under investigation in clinical trials in breast cancer (clinicaltrials.gov NCT03659136). MDA-MB-436 cells were stimulated with IGF2 with or without preincubation with BI836845 (100 μg/mL) for 4 hours or grown in CM from the different mCAFs that had been treated or not with BI836845 (100 μg/mL) for 8 hours. We found that BI836845 significantly decreased both IGF2- or CM-induced IGF1R signaling in MDA-MB-436 cells (Fig. 5C). Furthermore, BI836845 significantly reduced CAF-CM–stimulated clonogenic growth of MDA-MB-436 cells (Fig. 5D). Consistent with these in vitro results, we observed that the growth of mouse xenografts of MDA-MB-436 cells coinjected with the BM113 mCAFs was inhibited by BI836845 treatment. On the other hand, BI836845 treatment had no significant effect on the growth of xenografts of MDA-MB-436 cells coinjected with pCAFs (Fig. 5E–H). Examination of tumor nodules showed markedly reduced levels of phospho-IGF1R in BI836845-treated MDA-MB-436 cells mixed with mCAFs tumors compared with those treated with vehicle (Fig. 5I).

There is early evidence that IGF2 may have immunosuppressive effects on T cells, by enhancing regulatory T-cell functions (13). There is also clear evidence from different model systems that CAFs regulate both the innate and adaptive immune response to tumor through the secretion of multiple cytokines, proteases, and growth factors (14, 15). Most of this evidence supports an immunosuppressive effect on T cells and dendritic cells (16). However, all of these studies were performed on pCAFs and not on mCAFs. Compared with pCAFs, our mCAFs showed a marked increase in RNA levels of several genes involved in the immune response such as CCL2 (>3-fold) and CXCL12 (>4-fold, confirmed by ELISA; Fig. 6A) as well as IFN-related genes (Fig. 4B and C). Given these results, we assessed the effects of pCAFs and mCAFs on T-cell proliferation. T cells labeled with CFSE from three healthy donors were cocultured with CAFs for 3 days and stimulated with PMA and ionomycin. We found that coculture with CAFs significantly decreased the proliferation rate of CD4⁺T lymphocytes, and that this decrease was greatest in direct coculture with the mCAF compared with the NF and the pCAF (Fig. 6B–D). Incubation of T cells with IGF2 by itself also decreased T-cell proliferation (Fig. 6B). We also examined the role of the CAFs on longer T-cell stimulation; PBMCs were stimulated with plate-bound anti-human CD3 antibody for 2 days, with different CAFs coculture or without. We found that mCAFs (BM87 and BM113) could markedly suppress CD3-induced T-cell activation, displaying lower CD69 expression, one of the earliest antigens expressed on T cells upon activation, compared with PBMCs alone or with NFs (Norm2) or pCAFs (CAF53) in coculture (Fig. 6G). These results expand and confirm our knowledge that CAFs from metastatic breast cancer have a stronger ability to suppress T-cell activation in response to PMA/IONO or anti-CD3 antibody stimulation than pCAFs or normal fibroblasts.

These results suggest that mCAFs may have more potent immunosuppressive effects on T effector cells than pCAFs and NFs. We then found that inhibiting IGF2 activity with BI836845 partially reversed mCAF suppression of T-cell activation (Fig. 6E and F), suggesting that both direct protumorigenic and immunosuppressive functions upregulated by mCAFs may involve IGF-2, and can be targeted by BI836845.

The notion that the tumor microenvironment plays an essential role in breast cancer biology and therapeutic response is well established. The key protumorigenic function of breast CAFs is well supported by multiple studies including our own (17, 18). However, there is a gap in knowledge concerning the role of CAFs from metastatic lesions. One very recent study compared the estrogen-induced miRNA profile from one cutaneous metastatic breast tumor to primary breast tumor (pCAF) sites (19). Using our breast tumor biobank, we collected and analyzed eight mCAFs from different sites and performed the first study comparing mCAFs with CAFs from primary breast tumor (pCAF) sites. In this unique set, we found marked differences in in gene expression and ability to stimulate neoplastic growth between mCAFs and pCAFs and NFs. mCAFs conferred stronger growth stimulation and drug resistance on breast cancer cells, and expressed higher RNA levels of growth factors such as IGF2 and PDGFA than other CAFs. We then found that IGF2 protein levels were very high in mCAF-CM and that coculture of mCAFs with MDA-MB-436 cells increased IGF2 levels even more (two- to fourfold), together with IGF signaling activation. We also found evidence that the CM from mCAFs has immunosuppressive activity, in part due to IGF2 paracrine secretion. Finally, we observed that the use of an IGF1/2–neutralizing antibody, BI836845, blocked the in vitro and in vivo mCAF effects on cancer cell growth, IGF-1R signaling, tumor growth, and immunosuppressive effects. We also observed that an IFN gene signature was present within the transcriptome of mCAFs, confirmed by higher IFNß levels in mCAFs. Indeed, we recently reported that the IFN type I response was activated in a subset of pCAFs (5/23 pCAFs or 22%), with promitogenic effects (17).

Our findings are consistent with our recently published work, in which we reported that treatment of a TNBC xenograft model (MDA-MB-231) with an IGF-1R inhibitor resulted in a significant inhibition of lung metastases and not primary tumor growth in vivo (20). Singer and colleagues found that IGF2 and not IGF1 mRNA was increased in primary CAFs and that coculture with breast cancer cell lines markedly increased IGF2 expression (21). Moreover, high IGF2 expression has been found in CAFs from lung metastatic lesions in syngeneic mammary mouse models (22). Our findings are also consistent with the hypothesis that CAFs have immunosuppressive effects on T cells. CAFs secrete many cytokines that have direct effects on adaptive immunity and immunosuppressive effects have been reported in in vitro (15), genetically engineered (23), and syngeneic tumor mouse models such as the 4T1 mouse model (24, 25). Furthermore, CAFs from lung cancers with high Tregs in the stroma were able to induce FOXP3 expression in naïve CD4⁺T cells (26). pCAFs have been found to contribute to resistance to anti–PD-L1 treatment in a mouse model of colorectal cancer (24). One can speculate that combining BI835845 with anti–PD-L1 drugs may reveal a new potent therapeutic avenue for tumors expressing both IGF2 and PD-L1.

In summary, we have for the first time analyzed a set of CAFs from metastatic sites in breast cancer and have found significant molecular and phenotypic differences between mCAFs and CAFs from primary breast tumors. These differences may contribute to the intrinsic resistance of metastatic breast cancer to chemotherapy and their characterization provides novel therapeutic avenues in the treatment of metastatic breast cancer.

No potential conflicts of interest were disclosed.

Conception and design: Y. Gui, A. Aguilar-Mahecha, M. Basik

Development of methodology: Y. Gui, A. Hosein, M. Pollak, M. Basik

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Gui, A. Aguilar-Mahecha, U. Krzemien, A. Hosein, M. Buchanan, J. Lafleur, C. Ferrario, M. Basik

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Gui, A. Aguilar-Mahecha, M. Pollak, M. Basik

Writing, review, and/or revision of the manuscript: Y. Gui, A. Aguilar-Mahecha, M. Pollak, M. Basik

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Gui, A. Aguilar-Mahecha, M. Basik

Study supervision: A. Aguilar-Mahecha, M. Basik

We would like to thank Boehringer Ingelheim Pharmaceuticals, Inc for providing the BI-836845 Drug. This work was supported by an Innovation grant No. 702159 (INNOV13-2) from the Canadian Cancer Society to Dr. Basik and by the Quebec Breast Cancer Foundation and the FRQS Reseau de Cancer Axe cancer sein/ovaire funding for biobanking, also to Dr. Basik. Y. Gui was a recipient of a postdoctoral fellowship from Fonds de la recherche du Québec-Santé (FRQS) and supported by a CIHR/FRQS training grant in cancer research FRN 53888 of the McGill Integrated Cancer Research Training Program (MICRTP).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.