Background
Triple-negative breast cancer (TNBC), a breast cancer subtype characterized by lack of estrogen and progesterone receptor expression and absence of EGFR-2 (HER2/erbB2) overexpression, accounts for 15–20% of all breast cancers and most commonly affects young women under the age of 45. Unfortunately, approximately 30–40% of TNBC patients develop brain metastases, which have a particularly poor prognosis with a median survival of less than 5 months [
1,
2]. These brain metastases not only lead to death, but also cause severe cognitive complications that negatively affect both the physical and psychological well-being of patients, significantly reducing their overall quality of life (loss of visual acuity, behavioral disorders, etc.) [
3]. To date, no targeted therapy exists, highlighting the critical need to identify key molecular players that promote brain tropism and secondary tumor formation in the brain microenvironment. Differential transcriptomic studies of primary and/or secondary tumor samples have established predictive metastatic signatures [
4‐
7]. The development of brain metastasis involves, among many other steps in the metastatic cascade, the interaction and transmigration of cancer cells across the blood–brain barrier (BBB) and the establishment of a supportive environment for tumor growth in the brain parenchyma [
8]. These steps require dynamic interactions between cancer cells and the brain microenvironment [
9].
In this context, we have shown that the tyrosine kinase receptor TrkA, was overexpressed in up to 20% [
10] of TNBC cases and was involved in both tumorigenesis and the metastasis in vitro and in vivo [
11‐
14]. TrkA is the high-affinity receptor of nerve growth factor (NGF) [
15,
16] and in TNBC, both NGF and its precursor the proNGF have been found to increase invasion and migration through TrkA-mediated mechanisms [
12,
17]. These two growth factors act through TrkA phosphorylation and underlying signaling pathway [
14,
17]. However, for the first time, we have shown that NGF and proNGF can also activate TrkA-independent signaling through [
18] the formation of TrkA/CD44v3 [
11,
19] and TrkA/EphA2 [
20] receptor complexes respectively. Notably, proNGF is the common form of NGF in the brain [
21]. Individually, brain metastasis has been linked to TrkA, EphA2 and its downstream signaling partner Src [
14,
22,
23]. The aim of this study was to investigate the involvement of proNGF and the TrkA downstream activation pathways in the development of the TNBC cell metastatic process. By mimicking the last key steps of brain metastasis using a human BBB in vitro model, in vitro organotypic matrix, ex vivo mouse organ section cultures and mouse xenograft experiments, we show that proNGF promotes brain metastasis of breast cancer cells through the formation of the TrkA/EphA2 complex.
Methods
Cell culture and transfections
MDA-MB-231 and BT549 breast cancer cells were obtained from the American Type Culture Collection (ATCC, Rockville, MD, USA) and were incubated in a humidified incubator (37 °C, CO2 5%) in Minimum Essential Medium Eagle (MEM, Gibco™) supplemented with 10% (v/v) inactivated fetal bovine serum (FBS, Life Technologies), penicillin (40 IU/mL, Sciencell), streptomycin (40 IU/mL, Sciencell), gentamicin (50 mg/mL, Gibco™), Zell shield® (1X, Minerva-Biolabs®), and MEM supplement nonessential amino acid (NEAA 1X, Gibco™). ProNGF (cleavage resistant) from Alomone Labs was dissolved in filtered distilled water and used at a concentration of 0.5 nM on TNBC cells.
Endothelial cells were differentiated from CD34 + hematopoietic stem cells isolated from umbilical cord blood according to the method reported by Pedroso et al. [
24]. The protocol was approved by the French Ministry of Higher Education and Research (reference: CODECOH DC2011-1321) and by the local investigational review board (Béthune Maternity Hospital, Beuvry, France). Briefly, CD34 + cells were seeded at a density of 100 000 cells/cm2 in 24-well plates (Corning Inc.) and were incubated (37 °C, 5% CO2) in Endothelial Cell Medium (ECM, Sciencell) with 20% (v/v) heat inactivated FBS and Vascular Endothelial Growth Factor (VEGF, 50 ng/mL, PreproTech Inc.) for 20 days. Then, endothelial cells were cultured in gelatin-coated Petri dishes (100 mm diameter, Corning) in ECM-5 medium containing ECM medium supplemented with 5% (v/v) FBS, 1% (v/v) Endothelial Cell Growth Supplement (ECGS, Sciencell) and gentamicin (50 μg/mL, Biowest).
Human brain pericytes were provided by Professor Takashi Kanda (Department of Neurology and Clinical Neuroscience, Yamaguchi University Graduate School of Medicine, Ube, Japan). Human brain pericytes (Cosmo Bio Co., Ltd, Japan) are seeded in Petri dishes (100 mm diameter, Corning) coated with collagen I (100 μg/mL, Corning) and maintained in Dulbecco’s Modified Eagle Medium (DMEM, Life technologies) containing D-glucose (4.5 g/L) supplemented with 10% (v/v) FBS, penicillin (40 IU/mL), streptomycin (40 IU/mL) and L-glutamine (2 mM, Merck).
TrkA and TrkA KD overexpressing breast cancer cells were generating by electroporation using Amaxa Nucleofector technology (Lonza, Switzerland) according to the manufacturer’s instructions. For the establishment of stable lines, cells are maintained under selection pressure with geniticin (50 mg/mL, Gibco™).
For Src FRET biosensor experiments [
25], cancer cells were seeded in 100 mm diameter dishes (24 h, 37 °C, CO
2 5%). The plasmid (4 µg) is diluted in a buffer solution (500 µL, JetOptimus
® Buffer) and a transfection agent was added (5 µL, 10 min, TA, JetOptimus
® reagent). The mixture was incorporated into the cell medium (6 h, 37 °C, CO
2, 5%). In the case of RNA interference (siRNA), a solution containing medium (MEM; 10% FBS; 500 µL), Interferin (20 µL) and siRNA (40 µM; 1 µL) was incorporated into the cell medium after incubation (20 min; RT). The cells were incubated for 36 h at 37 °C.
Establishment of the human BBB in vitro model
According to the protocol of Vandenhaute et al. [
26], after a treatment with Ethylene-diamine-tetra-acetic acid solution/trypsin (EDTA 0.025% (w/v), 1X Trypsin, Biowest), Human endothelial cells are seeded at 70 000 cells/cm2 rate on Matrigel® (Growth Factor Reduced, Corning) coated insert filters (3 μm pores, Corning) and cultivated in ECM-5 medium (37 °C, 5% CO2). Pericytes were treated with the EDTA/trypsin solution and then seeded at a 13 000 cells/cm2 rate on collagen I (100 μg/mL)-coated 1-well plates (Corning). After 6 days of culture alone, the endothelial cells were transferred on top of the pericyte culture and incubated and cultivated for an additional 6 days (37 °C, 5% CO2). After 6 days of coculture, the required time for the induction of BBB properties by pericytes, endothelial cells were then called brain-like endothelial cells (BLECs) and had the properties of brain endothelial cells. The human BBB model was then used for experiments.
Endothelium permeability assay to Lucifer Yellow
The integrity of the BLEC monolayer was evaluated by measuring the diffusion of a BBB integrity marker that faintly crosses the BBB. To do so, the inserts containing the BLEC monolayer or only the coated insert (Matrigel®) were transferred into 12-well plates containing Ringer-HEPES (RH) buffer (NaCl 150 mM, KCl 5.2 mM, CaCl2 2.2 mM, MgCl2 0.2 mM, NaHCO3 60 mM, HEPES 5 mM, glucose 2.8 mM; pH 7.4). The integrity marker Lucifer yellow (LY), diluted in RH buffer (50 μM, Sigma‒Aldrich), was then added to the upper compartment. Every 20 min up to one hour, the insert filter was transferred into another well filled with RH buffer. At the end of the kinetic evaluations, the fluorescence intensity of aliquots from the initial solution, the lower compartments at each time point of the kinetics, and the upper compartment was quantified using a multiplate reader (425/538 nm, Synergy H1, BioTek Instruments). To obtain a concentration-independent parameter, the clearance principle was used. The slopes of the clearance curves for the insert with endothelial cells and the coated filters were PSt and PSf, respectively, where PS = permeability x insert surface area. The endothelial permeability (PSe) was calculated according to the formula 1/PSe = 1/PSt-1/PSf. The endothelial permeability coefficient (Pe, expressed in cm/min) was obtained by dividing the value of PSe by the insert filter surface area (1.12 cm2).
Cancer cell trans-BBB migration experiments
The trans-BBB migration experiment is performed according the protocols of Drolez et al. and Vandenhaute et al
. [
26,
27]. Briefly, cancer cells were loaded with Green-CellTracker™ (5-chloro-methylfluorescein-diacetate (CMFDA), 2.5 μg/mL, Invitrogen) dye according to the manufacturer’s instructions. Then, the cells were rinsed with phosphate-buffered saline (PBS, NaCl 8 g/L, KCl 0.2 g/L, KH
2PO
4 0.2 g/L, NaHPO
4 2.86 g/L, pH 7.4) and treated with EDTA solution (5 mM, 10 min, 37 °C) before being mechanically detached and resuspended in high glucose DMEM supplemented with 1% (v/v) FBS. Cancer cells were placed in contact with endothelial cells (80 000 cells/filter, 16 h, 37 °C). ProNGF (0.5 nM, Alomone-Labs) was added to the abluminal compartment. At the end of the incubation time, the filters were rinsed (DMEM high glucose) to remove nonadherent cells and fixed with paraformaldehyde (PAF 4% (w/v), 10 min, in darkness, 20 °C). The nuclei were counterstained with Hoechst 33258 (1 mM, 10 min, in darkness), and filters were separated before being mounted on slides and coverslipped with fluorescent mounting medium (Dako
®). Images were acquired using a Plan Fluor 20x/0.45 air objective on an inverted microscope (Eclipse TiU, Nikon) with the accompanying Nikon software (NIS element AS 4.60). Transmigrating cells were counted manually on the entire filter surface (100 fields per filter).
Brain & liver slices assay
Brains and livers from 6- to 10-week-old female C57BL/6 mice (Jackson Laboratories) were extracted in ice-cold dissection medium (MEM 75% (v/v), Hanks’ balanced salt solution (HBSS) 25% (v/v), NEAA 1X, penicillin 40 IU/mL, streptomycin 40 IU/mL). Brains were fixed and stabilized by water-resistant glue (SuperGlue, Loctite®) on a vibratome stage. The brain tissue was sectioned horizontally at a thickness of 300 μm by using a vibratome (Leica, VT1200S). Brain sections were transferred to filtered inserts (SARSTEDT) in 6-well plates (Greiner Bio-one) with incubation medium (MEM 50% (v/v), HBSS 25%, NEAA 1X, penicillin 40 IU/mL, streptomycin 40 IU/mL, ZellShield 1X) in the bottom well (1 mL, 37 °C, CO2 5%, ovn). Cancer cells were loaded with Green-CellTracker™ (CMFDA) dye and then incubated in Matrigel® (600 000 cell, 37 °C, CO2 5%) and placed on a sterile plastic spacer (4 mm diameter) for 1 h. Then, the spacer was removed, and the tumor cells were incubated in contact with the organ Sect. (72 h, 37 °C, CO2 5%). The interface between the tumor cells and the tissue section was observed using a Plan Fluor 10x/0.30 objective on a laser scanning confocal microscope (LSM 880, Zeiss). The number of cancer cells invading the area were analyzed manually with ImageJ software (v2.3.0/1.53f).
Breast cancer cells were seeded in brain or liver organotypic extracellular matrix (BIOMIMESYS® hydroscaffold, HCS Pharma) and incubated in a humidified chamber (37 °C, CO2 5%, 3 weeks). Then, all cells were fixed (PFA 4% (w/v), 10 min, 20 °C) and labeled with Hoechst 33258 (10 min) and Alexa Fluor™ 488 Phalloidin (2 h, 20 °C, ThermoFisher). The microscopy images were acquired using an automated ImageXpress® Micro Confocal microscope (Molecular Devices) in confocal mode with a Plan Apo Lambda 20x/0,45 objective. The cell number and colony shape were segmented and quantified with both ImageJ and Imaris 9.8 (Bitplane) software.
Western blotting
Cells were lysed in buffer (HEPES 40 mM, pH 7.5; EDTA 1 mM pH 8.0; NaCl 120 mM; 10 mM; NaPPi 50 mM NaF 50 mM; Na3VO4 1.5 mM; Triton-X100 1% (v/v); sodium lauryl sulfate (SDS) 0.1% (v/v); PMSF 1 mM; protease cocktail inhibitor 1% (v/v); glycerol 10% (v/v)) and then frozen (12 h, − 80 °C). The lysates were recovered by scraping and centrifuged (13 800 g; 10 min; 4 °C). Protein extracts in Laemmli buffer (Tris HCl 63 mM; glycerol 10% (v/v); SDS 2% (w/v); β-mercaptoethanol 5% (v/v); bromophenol blue 0,025% (v/v); pH 6,8) were loaded (40 µg/well) and separated on polyacrylamide gel (agarose 10%, 180 V constant, 1h15). Proteins were transferred (Tris 25 mmol/L, glycine 192 mmol/L, methanol 15% (v/v), H2O) into PVDF membrane (100 V, 1.5 h). The membranes were saturated (1 h, 20 °C, shaking) in TBS-0.1% (v/v) Tween-20 (TBST) with 5% (w/v) BSA. Then they were incubated with the primary antibodies diluted in the saturation buffer (BSA 5% (w/v), 16 h, 4 °C, shaking). The following primary antibodies were used: anti-beta-Actin (Sigma-Aldrich, A2066, rabbit), anti-pTrkA (Tyr674/675, Cell Signaling, 9141, rabbit), anti-TrkA (Cell Signaling, 2510, rabbit), anti-Src (Cell Signaling, 2123, rabbit), anti-pSrc (Tyr416, Cell Signaling, 2021, rabbit). After washing (TBST 0.1% (v/v), 5 × 5 min, 20 °C), the membranes were incubated with the secondary antibody (HRP-linked Anti-rabbit, Goat, 1/5000, 7074, Cell signaling) diluted in TBST (0.1% (v/v)). After the washing step, the chemiluminescence reaction was carried out (West Pico, ThermoScientific). Photons were detected in the darkroom using a camera (FUJIFILM LAS-4000), and the results were processed using ImageJ software.
Immunocytochemistry
Cells were seeded on type I collagen coated (100 µg/mL) compartmentalized slides (Thermo Fisher Nunc™ Lab-Tek™). At the end of the experiment, the cells were fixed with paraformaldehyde solution (PFA 4% (w/v), pH 7.4; 10 min, 4 °C). The cells are permeabilized with a solution of Triton-X100 (0.3% (v/v), 2 × 5 min, 20 °C, sigma). Subsequently, wells were treated with saturation buffer (NDS 5% (v/v), BSA 1% (w/v), 1 h, 20 °C) before the primary antibody incubation (4 °C, ovn). The following primary antibodies were used: anti-Claudin-5 (Invitrogen, 34–1600, rabbit), anti-EphA2 (R&D systems, AF3035, goat), anti-TrkA (Alomone, ANT-018, rabbit), anti-Src (Cell signaling, 2123, rabbit), anti-pSrc (Tyr416, Cell signaling, 2021, rabbit) and anti-VE-cadherin (Invitrogen, 36–1900, rabbit). After washing step (PBS, 4 × 5 min, 20 °C), appropriated AlexaFluor conjugated-secondary antibodies were incubated (1.5 h, 20 °C). The compartmentalized slides were coverslipped with the fluorescent mounting medium (Dako
®). Labeling was visualized using Plan-Apochromat 63x/1.4 Oil objective with laser scanning confocal microscope (LSM 880, Zeiss). For colocalization experiments, the image resolution was calculated according to the Rayleigh criterion (
r = 0.61 (λ/NA)). Then, the images were quantified with the EzColocalization plugin [
28] on ImageJ software to estimate the threshold overlap score (TOS) of the different labeling.
Proximity ligation assay (PLA)
Cells were seeded on type I collagen coated (100 µg/mL) compartmentalized slides (Thermo Fisher Nunc™ Lab-Tek™). At the end of the experiment, the cells were fixed with paraformaldehyde solution (PFA 4% (w/v), pH 7.4; 10 min, 4 °C). The cells were treated with saturation buffer (NDS 5% (v/v), BSA 1% (v/v), 1 h, 20 °C) before the primary antibody incubation (4 °C, ovn). After washing step (PBS, 4 × 5 min, 20 °C), PLA was performed using a Duolink in Situ-Red kit rabbit/goat (Sigma-Aldrich) according to the manufacturer’s protocol. Labelling was visualized using Plan Fluo 100x/1.3 Oil objective with laser scanning confocal microscope (LSM 880, Zeiss). In-house automatic script on ImageJ was computed to estimate the total number of TrkA/EphA2 red dots per cell. First, the channels were split and an appropriate background subtraction was found for each channel to enable accurate quantification. To do this, Gaussian blur, sharpen function and threshold were optimized together to enhance the noise/signal ratio. Based on fluorescence intensity, size and circularity of the particles, the PLA signals and the nuclei were quantified.
FRET imaging in living cells
Two hours before image acquisition, the cells expressing the Src biosensor were starved in fresh MEM without phenol red containing FBS 0.1% (v/v). During image acquisition, the cells were maintained in an incubation chamber (37 °C, CO2 5%) installed on a laser scanning confocal microscope (LSM 880, ZEISS). Images were acquired using a plan-apochromat 63 × /1.4 oil objective. The lasers were tuned to emit 458 nm and 514 nm wavelengths laser lines through a 470 to 500 nm bandpass emission filter (BP470–500) for ECFP detection and a 530 nm longpass emission filter (LP530) for YPET detection. For the sensitized emission method, the emission of the ECFP, FRET and YPET channels was recorded. The fluorescence emission ratios were computed using ImageJ software to address a Src activity channel. For the acceptor photobleaching method, the images in the ECFP and YPET channels were collected prior to and following acceptor fluorescent photobleaching.
Tumor xenograft growth in Severe combined immunodeficient mice (SCID) mice
MDA-MB-231 TrkA- or TrkA KD-overexpressing cells were subcutaneously injected (3 000 000 cells/mouse) into the flanks of six-week-old female SCID mice. Three weeks after injection, the mice were randomized into 4 groups according to the different treatments. Five treatments of entrectinib (per os, 30 mg/mouse) and/or siEphA2 (subcutaneous, 7.5 µg/mouse) were given to mice in a 2- to 3-day interval. Tumor volume was quantified throughout the experiment by measuring the length (l) and width (w) and was calculated as π/6 × l × w × (1 + w)/2. The mice were sacrificed under isoflurane anaesthesia when the tumors reached a critical size. (around 4 cm3), and the organs were immediately removed and stored in liquid nitrogen or fixed by immersion (PFA 4% (v/v), 4 °C, ovn) and then stored for tissue clearing (PBS, 4 °C).
Whole-mount immunostaining and tissue clearing
Perfusion-fixed whole mouse brains were processed using an adapted version of the iDISCO + protocol described previously [
29,
30]. The samples were first dehydrated in a series of methanol gradients then bleached overnight with 5% H
2O
2 at 4 °C, and delipidated overnight in a solution of 66% dichloromethane/33% methanol at 4 °C. Two rinses in 100% methanol were performed after the bleaching and delipidation steps. The samples were then gradually rehydrated from methanol to PBS, and permeabilized and blocked-in incubation solution (PBS, 0.2% gelatin, 1% Triton X100, 0.05% sodium azide) for 4 days. The brains were subsequently incubated with primary antibodies (anti-hHLA, Goat, 1/1000, Sigma) in incubation solution for 10 days, rinsed several times in incubation solution, and further incubated with secondary antibodies (Alexa Fluor 647 anti-Goat, Donkey, 1/1000, Sigma) for 5 days. Unbound antibodies were washed out with several rinses in PBS + 1% Triton X-100. After immunostaining, the brains were dehydrated in a series of methanol gradients and incubated overnight in 66% dichloromethane/33% methanol. On the next day, the methanol was washed out by a final 1-h incubation in 100% dichloromethane. Finally, the samples were cleared by immersion in dibenzyl ether for at least 2 h in rotation. When transparency was achieved, a fresh solution of dibenzyl ether was used for storage, and the samples were kept protected from light at room temperature until imaging.
Lightsheet imaging and analysis
Imaging of cleared tissues was performed in dibenzyl ether on an Ultramicroscope instrument (LaVision BioTec, BioImaging Center of Lille) and using either a 1.1x/0.1NA MI PLAN objective or a 4x/0.3NA LVMI-Fluor objective (LaVision Biotec). The following parameters were used: z-step was set to 2–4 µm, laser width and numerical aperture were kept at maximum, and for mosaic acquisitions a 10% overlap between tiles was configured. Acquisitions were saved as tiff sequences and converted to the Imaris file format using Imaris Converter (Bitplane).
Scanning electron microscopy
Three-dimensional extracellular matrix samples were fixed with 1% glutaraldehyde in 0.1 M sodium cacodylate buffer overnight. After washing, the samples were treated with 1% osmium tetroxide in water in the dark for 1 h. The samples were dehydrated with a series of increasing ethanol concentration. After two pure ethanol incubations, the sample was washed with hexamethyldisilazane and then dried (20 °C, overnight). Finally, extracellular matrix was mounted on stubs and observed with a secondary electron detector in a Zeiss SEM Merlin Compact VP operating at 1 kV.
Tissues were collected in RNAse-free Precellys tubes (Ceramic beads CK14) with QIAzol lysis reagent (QIAGEN). RNA extractions were performed according to manufacturer’s instructions using the RNeasy Mini Kit (QIAGEN).
Reverse transcription PCR (RT-PCR)
RNA was reverse transcribed using SensiFAST cDNA Kit (Ozyme) according to the manufacturer’s recommendations. PCR was performed with reversely transcribed RNA using ONE Green FAST qPCR Premix (Ozyme). The primer sequences used in this study were for: human beta-2-microglobulin (Forward 5′- CCAGCAGAGAATGGAAAGTC -3′; Reverse 5′- GATGCTGCTTACATGTCTCG -3′) and mouse beta-2-microglobulin (Forward 5′- CTGCTACGTAACACAGTTCCACCC -3′; Reverse 5′-CATGATGCTTGATCACATGTCTCG -3′). The PCR products (for human beta-2-microglobulin) were analyzed by agarose gel electrophoresis (1%) followed by ethidium bromide staining.
Discussion
In this study, we confirmed that proNGF induces the formation of the TrkA/EphA2 complex in TNBCs independently of TrkA phosphorylation. Furthermore, we showed that proNGF-induced Src phosphorylation is dependent on the TrkA/EphA2 complex. From a mechanistic point of view, we showed that proNGF is involved in brain metastasis through this complex. In addition, proNGF promotes transmigration across the BBB, invasion into the brain parenchyma and growth of TNBCs cells in extracellular matrices that mimic the brain. These findings correlate with Src activation, and brain tropism in vivo can be blocked by co-inhibition of TrkA and EphA2. For the past few decades, the incidence and death of breast cancer kept increasing worldwide and the development of metastasis is particularly harmful in breast cancer, even more so, in TNBC [
34,
35]. The level of TrkA found in these cancer types is comparable to levels of other receptor tyrosine kinases, in particular Met, which is known to induce metastasis. Met amplification is 1–3% in lung adenocarcinoma [
36] and reaches 20% in the case of concomitant EGFR mutation [
37]. Interestingly, the oncogenic activity of growth factor receptors is related to their kinase activity [
10,
38]. Thus, many therapies for brain metastasis are kinase inhibitors that target receptor tyrosine kinases (RTKs) or downstream signaling molecules [
39]. In the case of TrkA, lestaurtinib, larotrectinib and entrectinib have been developed. However, the efficiencies of all three inhibitors are moderate, and only a limited benefit is shown in the context of TrkA oncogenic fusions, which reflects the extent of oncogene dependence on kinase activity. In breast cancer, these oncogenic fusions represent less than 1% of cases [
40]. In our study, TrkA overexpression likely acted through its kinase activity to a certain extent. TrkA kinase activity is known to enhance the aggressiveness and metastasis of TNBC [
38] and other breast cancer subtypes, such as HER2-positive breast cancer [
41]. Interestingly, our results demonstrated that TrkA tyrosine kinase-independent pathways are also involved. Indeed, we demonstrated the recruitment of the EphA2 receptor under the effect of proNGF, as previously shown [
20]. Interestingly, EphA2 is also overexpressed and associated with a decrease in disease-free survival and overall survival in TNBC patients [
42]. In cancer, EphA2 undergoes an oncogenic switch and a loss of dependence on its ligand Ephrin A1. Moreover, the expression levels of EphA2 and Ephrin A1 are inversely correlated. EphA2 is thus, by its transactivation, involved in many oncogenic processes, such as tumor cell proliferation, survival and metastasis [
43‐
45], and is associated with a poor prognosis in TNBC patients [
42]. The present work showed that Src activation through the TrkA/EphA2 axis is critical for metastasis of TNBC cells to the brain, hence the relevance of conducting studies in this field to effectively block this pathway in TNBC cells and reduce the incidence of brain metastases. Src signaling pathways are mediated by membrane proteins, including integrins, growth factor receptors, and EPH family receptors. With these multiple key players as targets, Src is known to not only promote the proliferation but also the migratory and invasive capacities of cancer cells, which leads to increased metastatic spread [
46]. Our results demonstrated that Src is activated in TNBCs during the different stages of brain metastasis, particularly during BBB transmigration and brain parenchyma invasion. This work is in accordance with existing studies in the literature that found that Src inhibition significantly suppressed transmigration of breast cancer cells through the BBB [
23]. In this work, we observed that Src activation may be linked to the formation of the TrkA/EphA2 complex induced by proNGF. Moreover, we observed that this activation depends on TrkA phosphorylation and on the recruitment of EphA2, which supports this activation in the absence of TrkA phosphorylation. These results are in line with the fact that EphA2 induces Src in cancers to support invasive migration [
47].
The knowledge of key steps that occur in the brain metastasis process would have major implications for the design of improved therapies, which are still ineffective to date. Many factors, such as the mechanism of extravasation [
48], angiogenesis [
49] and tumoral persistence [
50], are still a challenge. Indeed, our knowledge of tumoral persistence within TNBC remains obscure, in part due to the lack of a relevant in vitro study model. For these reasons, in this study, we used multiple models to evaluate the distinct aspects of the tumor microenvironment. These aspects include the cellular component, where brain endothelial cells and glial cells are known to be key players in intercellular communication between tumor cells and promote metastatic progression [
51]. Beyond this cellular component, the three-dimensional structure of the extracellular matrix and its composition modulate the phenotype of tumor cells as well as their migratory and invasive properties [
52‐
54]. Both the composition and elasticity of the organotypic extracellular matrix (BIOMIMESYS®) are similar to those found in different organs, such as the brain and liver, which are preferential metastatic sites of TNBC cells [
55]. Through this work, we showed a differential response to TrkA/EphA2 axis inhibitors depending on the nature of the extracellular matrix in which the cancer cells were cultured. Nevertheless, we still do not know the underlying mechanisms responsible for these differences. Modulating the physical/chemical properties to determine whether the composition and/or the structure of the extracellular matrix have a specific effect could be an interesting approach to better understanding these responses [
56]. Indeed, by modulating the elasticity of a hydrogel, Kondapaneni and Rao demonstrated that brain-derived MDA-MB-231 (MDA-MB-231 BR) cells cultured in matrices with low elasticity maintained a dormant phenotype [
57]. Here, we showed that in MDA-MB-231 TrkA cells, inhibition of TrkA/EphA2 signaling also generated this phenotype. Thus, TrkA/EphA2 signaling would promote cell proliferation versus dormancy in very low elasticity matrices.
Breast cancer cells that metastasize to the brain have to develop new properties to cross the BBB. Recently, different molecular pathways that promote cancer cell transmigration across the BBB have been identified. For instance, metalloproteases such as MMP-9 [
58] or MMP-2 [
59] were found to be overexpressed in brain metastases of lung adenocarcinoma cells and breast cancer cells, respectively. These MMPs secreted by metastasizing tumor cells damage the integrity of the BBB by disrupting tight junctions shaped by Claudin-5, creating space for the invasion of cancer cells. Furthermore, some growth factors, such as VEGF, enhance the transendothelial migration process by promoting the adhesion of breast cancer cells to the endothelium and by disrupting VE-cadherin complexes that form tight junctions [
60]. Our results indicated, for the first time, that TrkA and proNGF improve the transmigration of cancer cells across the BBB by increasing their ability to form membrane protrusions with activated Src. Although the mechanisms by which cancer cells transmigrate across the BBB have been described as being either paracellular [
61,
62] or transcellular [
62], the key players in TNBC extravasation are not yet fully understood. The process of BBB transmigration requires dynamic changes in cancer cell shape, as well as the formation of specific protrusive structures that facilitate invasion through the blood vessel [
63]. The lack of understanding of the mechanisms and scarcity of accurate models to study metastasis of TNBC are major challenges for the development of effective TNBC treatments. To date, systemic treatments or targeted therapies have shown limited efficacy on metastatic cancers, with a limited increase in survival (5 months) for Troveldy, specifically [
64]. Therefore, our study models and the results obtained from using them are of major importance in the development of effective anti-metastatic therapies.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.