Extracellular vesicles from young women’s breast cancer patients drive increased invasion of non-malignant cells via the Focal Adhesion Kinase pathway: a proteomic approach

Whole blood was collected in sodium citrate tubes under Colorado Multiple Institutional Review Board (COMIRB) approved protocol. YWBC patients were between the ages of 18 and 45 and had no known autoimmune condition, no other significant comorbid conditions (i.e., active infection, heart disease, diabetes), no other diagnosis of other concurrent disease, and no systemic drug treatment or surgery prior to blood draw (see Additional file 1 for clinical details). Proteomic analysis of EVs included 10 nulliparous patients and 10 parous patients (1–6 children, time range since last pregnancy 0.33–4 years). Invasion assay analysis of EVs included 8 nulliparous patients and 10 parous patients (1–6 children, time range since last pregnancy 0.33–4 years). Age-matched healthy female donors had never been diagnosed with cancer, an autoimmune disorder, or any of the comorbid conditions listed above and had reported never having been pregnant. Nulliparous healthy donors were chosen as controls for this study because prior pregnancy is a known risk factor for YWBC and the role of EVs during and after pregnancy on subsequent breast cancer risk is unknown [41, 42]. Study data were collected and managed using REDCap electronic data capture tools hosted at the University of Colorado Anschutz Medical Campus [43]. Plasma was separated by centrifugation at 2000×g for 15 min at room temperature. The supernatant was collected and centrifuged at 2000×g for an additional 10 min at room temperature and stored at − 80 °C.

Plasma samples were thawed on ice and spun at 15,000×g for 10 min at room temperature. One milliliter of supernatant was collected and layered over a 1.5 × 10 cm high Sepharose CL-2B size-exclusion column (GE Healthcare, UK). Thirty 1-ml serial fractions were eluted by gravity filtration with 0.32% sodium citrate in PBS as previously described for EV isolation [44]. Fractions were analyzed for the presence of EVs by nanoparticle tracking analysis. Fractions 5 through 10 were identified as enriched in EVs and combined and concentrated using 100-kDa molecular weight cutoff ultrafiltration tubes (Sartorius). These purified EVs were either stored at − 80 °C for subsequent electron microscopy and proteomics analyses or stored at 4 °C for less than 1 week for use in functional assays.

The human breast cancer cell line MDA-MB231 [45] was cultured in RPMI (Corning) containing 10% human AB serum (Corning), 2 mM l-glutamine (Corning), 100 IU penicillin, and 100 μg/ml streptromycin (Corning) in a 37 °C incubator with 5% CO₂. The MCF10DCIS.​com cell line was cultured as previously described [46, 47]. The cells were tested every 3 months to confirm mycoplasma negativity (MycoAlert™ Mycoplasma Detection Kit, Lonza), and validated for authenticity by fingerprinting performed by Dr. Christopher Korch (University of Colorado Cancer Center Sequencing Facility). To make conditioned media, cells were grown to 80% confluency, rinsed with Hanks Buffered Saline Solution, and incubated at 37 °C in serum-free media for 4 h to minimize serum protein and EV contamination. Cells were then transferred to fresh serum-free media and incubated for 48 h at 37 °C. Cell debris was removed by centrifugation at 500×g for 5 min and 2000×g for 10 min. Supernatant was filtered through a sterile 0.22-μm syringe filter and stored at 4 °C. To isolate EVs, approximately 180 ml of conditioned media was concentrated to 1 ml by centrifugation in a 50-kDa molecular weight cutoff ultrafiltration tube (Sartorius) and isolated over a size-exclusion column as described above.

EV concentration and size were analyzed using a Nanosight NS300 instrument with a 532-nm laser (Malvern). Images were captured using an sCMOS camera, with a gain of 1.0, and camera level of 13. EVs purified by size-exclusion chromatography (SEC) were diluted 200-fold in phosphate-buffered saline (PBS) and injected using a Nanosight autopump (Malvern) in script mode commanding a set temperature of 22 °C, an infusion rate of 25 μl/min, and video capture of five consecutive 30-s videos with a 5-s delay. Data were captured and analyzed using NTA Analytical Software suite version 3.1 (Malvern) with a detection threshold of 5.0. The instrument was calibrated using 100 nm silicone beads. Samples that were below 20 particles per frame or above 100 particles per frame were re-diluted to a concentration within this range.

EVs purified by SEC were incubated on formar-coated grids and negatively stained using 5% uranyl acetate. The grids were rinsed, and the size and morphology of EVs analyzed using a Technai 10 Transmission Electron Microscope (Field Emissions Inc.). Images were captured at 25,000× using a First Light digital camera (Gatan) (CU AMC Electron Microscopy Center, Aurora, CO).

EV samples purified by SEC were analyzed via mass spectrometry (CU AMC Mass Spectrometry and Proteomics Shared Resource, Aurora, CO). The samples were digested according to the FASP protocol using a 30-kDa molecular weight cutoff filter [48]. In brief, samples were mixed in the filter unit with 8 M urea in 0.1 M ammonium bicarbonate (ABC), pH 8.5 and centrifuged at 14,000×g for 15 min. The proteins were reduced by addition of 100 μl of 10 mM DTT in 8 M urea and 0.1 M ABC, pH 8.5; incubated for 30 min at room temperature; and centrifuged. Subsequently, 100 μl of 55 mM iodoacetamide in 8 M urea and 0.1 M ABC, pH 8.5 was added to the samples, incubated for 30 min at room temperature in the dark, and centrifuged. The pellets were washed three times with 100 μl 8 M urea in 0.1 M ABC, pH 8.5, then three times in 100 μl of 0.1 M ABC buffer. The pellets were digested overnight at 37 °C with 0.02% Protease Max (Promega). Peptides were recovered by transferring the filter unit to a new collection tube and spinning at 14,000×g for 10 min. To complete peptide recovery, the filters were rinsed twice with 50 μl 0.2% FA and 10 mM ABC and collected by centrifugation. The peptide mixture was desalted and concentrated on a C18 Tip (Thermo Scientific Pierce).

Samples were analyzed on a Q Exactive quadrupole orbitrap mass spectrometer (Thermo Fisher Scientific) coupled to an Easy-nLC 1000 UHPLC (Thermo Fisher Scientific) through a nanoelectrospray ion source. Peptides were separated on a self-made 15-cm C18 analytical column (100 μm × 10 cm) packed with 2.7 μm Phenomenex Cortecs C18 resin [49]. After equilibrations with 3 μl 5% acetonitrile and 0.1% formic acid, the peptides were separated by a 180-min linear gradient from 2 to 32% acetonitrile with 0.1% formic acid at 350 nl/min. LC mobile phase solvents and sample dilutions used 0.1% formic acid in water (buffer A) and 0.1% formic acid in acetonitrile (buffer B) (Optima™ LC/MS, Fisher Scientific). Data acquisition was performed using the instrument supplied Xcaliber™ (version 3.0) software. The mass spectrometer was operated in the positive ion mode and in the data-dependent acquisition mode. In one scan cycle, peptide ions were first scanned by full MS at resolution 60,000 (FWHM at m/z 200), and then, the top 12 intensive ions (2 m/z isolation window) were sequentially subjected to HCD fragmentation and detected at resolution 15,000. Dynamic exclusion was set to 20 s. Spray voltage was set to 2.5 kV, S-lends RF level at 55, and heated capillary at 275 °C.

MS/MS spectra data were extracted from raw data files and exported as mascot generic format files (mgf) using MassMatrix. The mgf files were then searched against the SwissProt database using an in-house Mascot™ server (version 2.2.06, Matrix Science). Mass tolerances were ± 10 ppm for MS peaks and ± 0.1 Da for MS/MS fragment ions. Trypsin specificity was used, allowing for one missed cleavage. Methionine oxidation, proline hydroxylation, protein N-terminal acetylation, and peptide N-terminal pyroglutamic acid formation were allowed for variable modifications while carbamidomethyl of Cys was set as a fixed modification. All raw or processed data files are available upon request.

Scaffold (version 4.4, Proteome Software) was used to filter tandem MS-based peptide and protein identifications. Peptide and protein identifications were accepted if they could be established at greater than 95% and 99% probability, respectively, as specified by the Peptide Prophet algorithm. Protein identifications also required at least two identified unique peptides.

Resultant proteomic data from cell line-derived EVs were compared by overall enrichment scores using DAVID Bioinformatics Resource [50, 51]. Proteins with 6 or more spectral matches were included in the analysis, and enrichment scores greater than 1.5 were reported. Proteomic data from patient EVs, in which more than two groups were compared, were analyzed using online statistical software MetaboAnalyst 3.0 [52]. Data was normalized by sum, auto-scaled (mean-centered and divided by the standard deviation of each variable), and multivariate and statistical analyses such as t tests, volcano plots, and partial least squares discriminant analysis (PLS-DA) were performed. Normalized data was exported from MetaboAnalyst 3.0, and further statistical analyses were performed using GraphPad Prism 7.

ImageLock 96-well plates (Essen Bioscience) were coated with 0.2 mg/ml Matrigel diluted in 1× PBS (Corning Life Sciences) for 2 h at room temperature and rinsed twice with 1× PBS. MCF10DCIS.​com cells were seeded at 4000 cells per well and incubated overnight to 100% confluency at 37 °C in low-serum culture media containing 1% horse serum, ± 5 × 10⁸ EVs, and ± 3 μM FAK inhibitor (PF-573.228, Sigma). For the migration assays, uniform scratch wounds were created in the center of each well using the IncuCyte Wound Maker (Essen BioScience). Cells were washed, incubated in low-serum media, and bright-field images were taken every 2 h using an IncuCyte ZOOM® live cell imaging instrument (Essen BioScience). After 24 h, images were analyzed using IncuCyte ZOOM® analysis software and the density of cells in each wound area was calculated. For the invasion assays, a 2-mg/ml Matrigel pad was layered over the cells after wounding and images were captured for 48 h [53]. Flow cytometry was performed on a BD Fortessa X-20 after staining cells treated as described above with antibodies specific for total FAK (Biolegend, clone W16060A) and phosphorylated FAK (Fisher Scientifics, clone 31H5L17).

MCF10DCIS.​com cells were plated at 40,000 cells per well in 96-well plates and cultured overnight at 37 °C. EVs from YWBC patients, healthy donors, or MDA-231 cells were then added and incubated for 18 h. Cells were then trypsinized, washed, and RNA isolated using a NucleoSpin RNA isolation kit according to the manufacturer’s instructions (Macherey-Nagel). RNA expression of genes related to cancer pathways (PanCancer Cancer Pathways Panel, #XT-CSO-PATH1) and cancer progression (PanCancer Progression Panel, #XT-CSO-PROG1) were measured using NanoString technology. Data were normalized and analysis performed using the NanoString nSolver 3.0 software to conduct nCounter advanced analysis (NanoString Technologies, Seattle, WA, USA).

Two groups were compared using an unpaired two-tailed Student’s t test, three or more groups were compared using one-way ANOVA, and three or more groups with multiple measures were compared by two-way ANOVA using GraphPad Prism software version 6.0. Where appropriate, p values are adjusted for multiple comparisons and multiple measurements.

To determine if the proteomic content of EVs correlates with function and could potentially serve as a biomarker of a breast cancer patient’s disease state, we first compared the proteomic content and function of EVs from two representative breast cancer cell lines differing in metastatic attributes. We characterized EVs secreted by the MDA-MB231 breast cancer line, aggressive triple negative breast cancer cells that form invasive carcinomas in xenograft models and which have been demonstrated to produce EVs that increase the motility of less invasive breast cancer lines [45, 54]. The second cell line used was MCF10DCIS.​com cell line, which readily forms DCIS-like tumors and can become invasive in some conditions [47, 55]. EV characteristics of this cell line have not been previously described. We found that the physical properties of EVs from these breast cancer cell lines were similar to each other and to previously reported studies of EVs. Specifically, both cell lines produced approximately 100 nm particles that eluted between fractions 5 and 11 of the size-exclusion column (column volumes 5 ml through 11 ml, Fig. 1a) and had similar average particle sizes and protein concentrations (Fig. 1b) consistent with previous reports for MDA-MB231 EVs (Fig. 1b, [56]). These EVs also expressed the putative EV proteins Hsp70, CD63, and CD9 by western blot (Fig. 1c). These proteins are detected at multiple molecular weights, likely due to their known isoforms and differential glycosylation patterns [57‐60]. In fact, treatment of EV lysate with N-glycosidase F reduced CD63 to a single detectable band at the reported molecular weight (Fig. 1c). Consistent with previous studies, EVs produced by invasive MDA-MB231 cells significantly increased the migration and invasion of MCF10DCIS.​com cells across a scratch wound (Fig. 1d, e). In contrast, EVs produced by the less aggressive MCF10DCIS.​com cells did not increase invasion of MCF10DCIS.​com cells but did increase migration in the scratch wound assay. There was no significant effect of EVs from either cell line on the motility of MDA-MB231 cells in these assays (data not shown), suggesting that these cells are already maximally invasive.

The differential effect of EVs from invasive and non-invasive breast cancer cells on the motility and invasive capabilities of the DCIS.com cells suggests there may be proteomic differences between EVs isolated from these cell lines. To address this question, we compared the total protein content of EVs isolated from two independent preparations of conditioned media from MDA-MB231 and MCF10DCIS.​com cells by mass spectrometry. Shared proteins identified in EVs from both cell lines were enriched for vesicle proteins (enrichment score of 7.4), membrane components (2.97), regulators of cell death (2.86), contractile fibers (2.56), adhesion molecules (1.89), and proteins that promote cell motility (1.84), and included putative EV proteins such as Hsp70, CD63, and CD9. EVs produced by the invasive MDA-MB231 cells were significantly enriched for proteins involved in vesicle formation (enrichment score of 6.16), protein synthesis (4.9), proteolysis (3.56), and glycolysis (1.54). In contrast, MCF10DCIS.​com EVs were significantly enriched for membrane proteins (12.65 enrichment score), adhesion molecules (10.33), proteins involved in cellular migration (4.21), and components of the extracellular matrix (3.65). Reflecting these differences, the most abundant proteins uniquely identified in MDA-MB231 EVs were those involved in transcriptional regulation (splicesome, transcription factors, ribosomal proteins, tRNA ligases), proteolysis (proteasome units, pyrophosphatase), EV formation (annexin and vesicle markers LAMP-1 and EEA1), cell cycle (NUMA1), and cell motility and adherence to extracellular matrices (vitronectin, collagen, filamin proteins, and EDIL3) (Table 1). In contrast, the most abundant proteins uniquely identified in EVs from the MCF10DCIS.​com cells were cellular adhesion proteins (cadherin family members, laminin proteins, proteoglycans, syndecan-1, EPCAM, b-catenin, and collagen), regulators of cellular proliferation (CD109, RARRES1, PTGFRN, FAT1, S100A14, and amphiregulin), and metabolic proteins (calcium-binding proteins, serine proteases, and cholesterol- and lipoprotein-binding proteins) [61, 62]. These results suggest that the protein content of EVs from MDA-MB231 and MCF10DCIS.​com cells reflects the biologic differences between these invasive and non-invasive breast cancer cells and thus may contribute to their altered functional activity.

Since the largest functional differences were observed in the 2D scratch wound invasion assay through the Matrigel pad (Fig. 1e), we next determined whether EVs isolated from the peripheral blood of human YWBC patients increased invasion in breast cancer cells compared to EVs from healthy donors. EVs from 18 YWBC patients and 10 healthy donors (Additional file 1) were isolated using size-exclusion chromatography. Similar to EVs isolated from the breast cancer cell lines, the majority of small particles from both YWBC patients and healthy donors eluted in fractions 5 through 10 (Additional file 2). The average particle size and yield were similar between YWBC patients and healthy donors (Fig. 2a), and isolated EVs had a classically spherical cup-shaped appearance by electron microscopy (Fig. 2b). Incubation with EVs from YWBC patients significantly increased the density of MCF10DCIS.​com cells inside scratch wounds compared to untreated controls or EVs from healthy donors (Fig. 2c). To compare across multiple patients and assays, the average percent invasion of experimental conditions was compared at the time point in which the untreated controls reached 50% invasion (Fig. 2d). Compared to the corresponding untreated controls, incubation with EVs from 13 of the 18 YWBC patients and only 1 of the 10 healthy donors significantly increased MCF10DCIS.​com invasion over untreated control cells (Fig. 2e). On average, EVs from YWBC patients significantly increased MCF10DCIS.​com cell invasion (74.5% wound closure) over that of untreated controls (50.2% wound closure) or cells treated with EVs from healthy donors (55.6% wound closure) (Fig. 2f). Functional EVs were identified in patients across risk factor subsets, including subtype, stage, body mass index, and parity (Additional file 3).

The functional effects of EVs from YWBC patients on breast cancer cell invasion suggest that their protein content might be distinct from that of healthy donors. To elucidate the specific differences between healthy and breast cancer-associated human EVs, we compared the proteomic content of EVs isolated from 20 YWBC patients to that of EVs isolated from 10 healthy donors (Additional file 1) using a simple top-down proteomic approach [63].

Of the 571 proteins identified, YWBC EVs contain 85 unique proteins, 76 proteins that overlap with MDA-MB231 EVs (Fig. 3a), and 70 proteins that overlap with MCF10DCIS.​com EVs. We consistently identified typical EV markers CD9, CD81, CD63, and HSP70, and proteins vital for EV formation such as Rab proteins, tetraspanins, clatherin components, and myosin proteins in EVs from both healthy donors and YWBC patients [1]. Proteins unique to YWBC EVs and EVs from the cell lines that were not identified in HD EVs include serpin B3, tripeptidyl-peptidase 2, prolactin-inducible protein, tetraspanin-15, and proteasome subunits (Additional file 4).

To compare EVs from YWBC and healthy donors, multivariate analysis using the partial least squares discriminant analysis method was performed and variables were sorted into principal components based on their ability to discriminate between YWBC and healthy donor EV content (Fig. 3b). Volcano plot analysis identified 46 proteins that significantly differ between YWBC and healthy donor EVs with a fold change threshold of 0.2 and a p value threshold of 0.05 (Fig. 3c, Additional file 4). Of interest in breast cancer, EVs from YWBC patients had increased levels of Mucin 1, TIMP-1, Myc Target Protein, and Latent TGFB binding protein1. Further, proteins such as Mucin 5b, Mucin 1, TIMP1, and Laminin B1 were significantly enriched not only in breast cancer EVs as compared to healthy donor EVs, but also specifically in the EVs that increased invasion compared to those that had no effect on invasion (Fig. 3d).

We next sought to determine whether cancer cell signaling pathways are altered after treatment with EVs. We compared gene expression in MCF10DCIS.​com cells treated with EVs isolated from YWBC patients, healthy donors, invasive MDA-MB231 cells, and untreated cells using NanoString analysis of the Cancer Pathway and Cancer Progression gene sets. nSolver and nCounter analysis identified significantly altered expression of genes related to cell motility and EMT, cell adhesion, angiogenesis, and proliferation (Table 2). KEGG pathway analyses revealed EV-induced alterations of the Focal Adhesion Kinase (FAK) pathway. This was supported by our proteomics above, which identified EV-associated proteins that have demonstrated involvement in FAK-mediated cell motility (tetraspanin-15, proteasome subunits) [64‐66]. These proteins were identified exclusively in YWBC and MDA-MB231 EVs, but not in healthy donor EVs, suggesting a potential role of the FAK pathway in the functionality of these EVs (Additional file 4). Additionally, a number of proteins known to upregulate the FAK pathway were significantly enriched in EVs that increased cell invasion in our functional assays (TIMP1, Mucins, and Laminin B1). This led us to further investigate the role of the FAK pathway in driving EV-induced invasion. As shown in Fig. 4, several genes in the FAK pathway were significantly altered after treatment with breast cancer EVs.

Since proteomic and genomic analysis implicated the FAK pathway as a potential driver of the functional effects of breast cancer EVs, we hypothesized that inhibition of the FAK signaling pathway using a validated inhibitor (PF-573.228) may reduce EV-induced breast cancer cell invasion. PF-573.228 is an established small molecule inhibitor that has a 50- to 250-fold selectivity for FAK over other protein kinases [67]. To determine whether inhibition of FAK affects breast cancer cell invasion, we performed the invasion assay with or without EV treatment and in the presence or absence of the FAK inhibitor. Incubation with the FAK inhibitor decreased FAK phosphorylation, detected by flow cytometry (Fig. 5a). As reported above, treatment with EVs from MDA-MB231 cells increased the invasion of MCF10DCIS.​com cells through a Matrigel pad (Fig. 5b, c). Although FAK inhibition in the absence of EVs did not affect invasion, the FAK inhibitor significantly abrogated the increase in EV-stimulated DCIS.com cell invasion to levels matching the untreated control (Fig. 5b, c).