Tumor-derived exosomes promote the in vitro osteotropism of melanoma cells by activating the SDF-1/CXCR4/CXCR7 axis

Melanoma (SK-Mel28, WM266, LCP and LCM) and breast cancer (MDA-MB231) cell lines (ATCC, Rockville, MD, USA) were cultured in Exo-free complete medium. Overnight starvation with FBS-depleted medium was used to synchronize the cell cycle before each experiment. Scraps of cancellous bones were recovered from 5 healthy subjects suffering of post-traumatic orthopaedic surgery and used to generate small bone fragments of 3–5 mm³, or cultured for 24 h with free-culture medium to achieve bone conditioned medium (BCM) [8].

Exosomes were purified by ultracentrifugation of supernatants from 48-h cultured melanoma cells [15]. Briefly, dead cells, debris, protein aggregates and microvesicles were removed by both centrifugation and mechanical filtration using Millipore filter of 0.22 µm diameter. Supernatants were then twice centrifuged at 100,000×g for 70 min at 4 °C to obtain Exos that were stored at − 80 °C in PBS aliquots of 100 µl. A limited number of samples were randomly selected to verify the size distribution and concentration of vesicles by using the NanoSight NS300 instrument (Malvern Instruments, Malvern, UK), while the transmission electron microscopy (TEM) defined the morphology of vesicles. After the measurement of protein amount using the Bradford protein assay (Bio-Rad), Exo preparations from each sample were verified by measuring the expression of CD63, CD81 (eBioscence) and CD9 (BD Pharmigen) by flow-cytometry [16] with dedicated mouse anti-human monoclonal antibodies (MoAbs). For this purpose, 30 µg of Exos were previously conjugated with 4 µm diameter aldehyde/sulfate latex beads (Invitrogen, Carlsbad, CA) [17], while mouse IgG1 was the isotypic control. Moreover, to further validate the purity of Exo preparations, western blots (WB) were performed to measure the levels of CD81, TSG101, calnexin (CANX) and bovin serum albumin (BSA) in accordance to Minimal Information for Studies of Extracellular Vesicles (MISEV) guidelines [18].

The ability of melanoma cells to incorporate Exos was also investigated by confocal microscopy (Nikon Instr., Lewisville, TX). Briefly, 1 × 10⁴ melanoma cells were cultured for 4 h with 50 µg/ml of Exos previously bound to a red lipophilic fluorescent dye (PKH26; Sigma-Aldrich, St Louis, MO, USA) [14]. Then, cells were stained with FITC-conjugated phalloidin (Invitrogen), while nuclei counterstained with DAPI (4′,6-diamidino-2-phenylindole; Sigma Aldrich).

Trans-well plates of 8 µm diameter (Corning Incorporated, NY) were used to investigate the migratory behaviour of melanoma cells, while invasiveness was assessed by the BioCoat Matrigel cell culture chambers (Becton–Dickinson Bioscience, MA). MDA-MB231 cells were the positive control in relation to their metastatic bone tropism [19]. For both migration and invasion assays, 1 × 10⁴ cells were seeded onto the upper chamber in presence of RPMI supplemented with 1% FBS. The lower chamber was filled with 10% FBS or bone fragment as chemoattractant, while 1% FBS was the negative control. Then, cells adherent on the upper surface of the membrane were removed at different time points (24 and 48 h), while those trapped in the underside of the insert were fixed with 4% paraformaldehyde, stained with DAPI and visualised under a UV microscope (Leica, Heidelberg, Germany). DAPI⁺ cells were counted in ten random fields of 0.2 mm² at 40× magnification.

Further experiments explored the Exo role in influencing the osteotropic attitude of melanoma cells. To this, both migration and invasion assays were completed as previously described although the lower chamber was filled with Exos (50 µg/ml) derived from autologous (a-Exos) or heterologous melanoma cells (h-Exos). Each experiment was completed in triplicate.

The basal expression of 27 genes mainly implicated in cell migration and bone tropism were investigated by real-time (RT)-PCR in unstimulated melanoma cells using a 96-well custom plate (BioRad). GAPDH was the housekeeping gene to calculate the 2^−Δct. In addition, the potential effect on gene expression levels after 6-h of melanoma cell stimulation by BCM, Exos or both, was explored by QX200 droplet digital (dd)-PCR. Data were analyzed by the QuantaSoft software (BioRad). Genes and relative primers used for RNA amplification are listed in Additional file 1: Table S1. Experiments were completed in biological triplicate and results were expressed in terms of fold change using GAPDH as the housekeeping gene.

Based on the results of dd-PCR, membrane and intracellular expression of CXCR4 and CXCR7 by melanoma cells were investigated by flow cytometry using dedicated mouse anti-human MoAbs (Abcam), while the mean fluorescence intensity (MFI) ratio was calculated with respect to IgG1 isotypic control. Both membrane and cytoplasmic protein fractions from melanoma cells were obtained using the Mem-PERM Plus Kit (Thermo Scientific) and the relative CXCR4 and CXCR7 levels measured by WB. Either pan-cadherin or alpha-tubulin (Abcam) were used as intra-assay control for membrane and intracellular measurements, respectively, as described [20]. Protein expression was calculated in terms of optical density (O.D.) by ImageQuantTL (GE Healthcare, UK), while differences between stimulated (BCM or Exos) and unstimulated cells were expressed in terms of ratio [21]. Additionally, CXCR4 and CXCR7 levels on melanoma-derived Exos were measured by flow-cytometry as previously described.

Further experiments assessed the contribute of either CXCR4 or CXCR7 in conditioning the osteotropism of melanoma cells. Therefore, their expression was restrained by small interfering RNAs (siRNAs) using the following primers: 5′-GGCAGUCCAUGUCAUCUACTT-3′ and 5′-GUAGAUGACAUGGACUGCCTT-3′ for CXCR4; 5′-GGAUGACACUAAUUGUUAGTT-3′ and 5′-CUAACAAUUAGUGUCAUCCTT-3′ for CXCR7 (Life Technologies, CA, USA). Briefly, 1 × 10⁶ LCP and SK-Mel28 cells were treated for 48 h with 3.75 µl/ml of Lipofectamine 3000 Reagent (Invitrogen, CA, USA) to induce transient transfection in presence of anti-CXCR4 (30 nmol/l) or anti-CXCR7 (60 nmol/l) siRNAs. Cells treated with 3.75 µl/ml of Lipofectamine 3000 reagent or scramble probes (Ambion) were the controls.

Since both CXCR4 and CXCR7 share the SDF-1 ligand, silenced melanoma cells were explored in their propensity to migrate and invade toward human-recombinant SDF-1 (R&D Systems, MN, USA) either in presence or absence of Exos. The optimal concentration of SDF-1 used in these experiments, namely 100 ng/ml, was similar to that detected in BCM by ELISA (96.4 ± 17.2 ng/ml).

Data were analysed by the Mann–Whitney test for not-parametric distributions, while One-way ANOVA was used to analyse the differences between basal levels of CXCR4 and CXCR7 by melanoma cells. Data were was considered statistically significant with a p < 0.05.

The first set of experiments evaluated the migration tendency of melanoma cell lines. As shown in Fig. 1a, c, melanoma cells stimulated for 24-h with 10% FBS increased their basic migratory propensity with respect to unstimulated cells (p < 0.01 in all instances) and this occurred in a fashion almost similar to MDA-MB231. Both SK-Mel28 and WM-266 cells showed the highest migratory potential (39.6 ± 9.0 and 31 ± 7.5 cells/0.2 mm²) as compared to LCP and LCM (14.8 ± 1.9 and 17 ± 6.8 cells/0.2 mm², respectively). However, the migratory capacity of LCP cells also increased once exposed to the stimuli of bone fragments (24.3 ± 4.4 cells/0.2 mm²; p < 0.01), as for MDA-MB231 cells (46.2 ± 3.6 cells/0.2 mm²; p < 0.01). By contrast, bone fragments exerted only a modest effect on LCM, WM-266 and SK-Mel28. Similar results were obtained by measuring the invasiveness in the same melanoma populations (Fig. 1b, c), as well as in cells stimulated for 48-h (data not shown). In relation to these preliminary findings and for the next experiments, LCP cells were arbitrarily defined as ‘osteotropic’, whereas WM-266 and SK-Mel28 as ‘not-osteotropic’. Moreover, 24-h stimulation was arbitrarily defined as the optimal time point for the next trans-well assays.

Next, we investigated the potential influence of Exos on both migration and invasiveness of melanoma cells. Exosomes were purified from supernatants of WM-266, SK-Mel28 and LCP melanoma cells at a mean concentration of 1.6 × 10¹⁰ vesicles/ml, 1.1 × 10¹¹ vesicles/ml and 1.9 × 10¹¹ vesicles/ml, respectively. Figure 2 illustrates specific characteristics of Exos, including the 30–150 nm diameter revealed in more than 80% of vesicles (panel a), the typical cup-shaped morphology by TEM (panel b), and the presence of CD81, CD63 or CD9 tetraspanins (panel c). Western blot analyses (Fig. 2d) demonstrated that Exo preparations were positive for typical markers of EVs (CD81 and TSG101), while excluded the possible contamination by large-EVs (> 200 nm) or non-EV structures, such as protein aggregates, since all the samples resulted negative for both CANX and BSA as compared to complete control medium. Finally, Fig. 2e evidences the uptake of red-fluorescent labeled Exos by SK-Mel28 cells whose cytoplasm was engulfed of clusters formed by h-Exo from LCP.

We then explored the effect of stimulation with both Exos and bone fragments on the migration and invasiveness of not-osteotropic cells, namely SK-Mel28 and WM-266. As shown in Fig. 3a, both a-Exos and h-Exos from non-osteotropic cells poorly improved the migration of SK-Mel28 (19.5 ± 7.8 and 23.2 ± 4.1 cells/0.2 mm², respectively) and WM-266 cells (10.8 ± 3.2 and 16.2 ± 4.8 cells/0.2 mm²) as compared to controls (SK-Mel28: 18.2 ± 4.8 cells/0.2 mm²; WM-266: 13.3 ± 2.4 cells/0.2 mm²; p = ns in both instances). By contrast, the migration was significantly increased in the presence of h-Exos from osteotropic LCP (SK-Mel28: 38.4 ± 5.3 cells/0.2 mm²; WM-266: 27.1 ± 5.2 cells/0.2 mm²; p < 0.05), although this effect was abrogated by the removal of bone fragment (SK-Mel28: 20.9 ± 6.1 cells/0.2 mm²; WM-266: 15.5 ± 2.6 cells/0.2 mm²). Similar results occurred by exploring the invasive behaviour of SK-Mel28 and WM-266 (Fig. 3b). Taken together, these data suggested a potential role of Exos from melanoma cells in conferring osteotropic function in heterologous cells.

To identify possible deregulated genes reflecting the osteotropic propensity of melanoma cells, we compared the transcriptional profiles of osteotropic LCP to not-osteotropic SK-Mel28 and WM-266 cells. Therefore, gene expression analyses investigated the expression of several genes potentially implicated in cell migration and chemotaxis.

As showed in Fig. 4, genes regulating the EMT-machinery, including N-Cadherin (N-CAD), SNAIL, ZEB and TWIST1, were found variably up-regulated in both osteotropic and not-osteotropic cells with the exclusion of E-CAD. The mRNA basal levels of N-CAD (p < 0.05), ZEB1 (p < 0.001), MMP1 (p < 0.05) and MMP2 (p < 0.001) were significantly increased in not-osteotropic cells with respect to osteotropic LCP. However, the other genes resulted poorly expressed in explored cell lines.

These findings demonstrate that both osteotropic and not-osteotropic cells showed at baseline a similar EMT signature, poorly correlated with the bone tropism.

Since the baseline EMT gene expression profile was not correlated with the osteotropic behaviour of LCP, SK-Mel28 and WM-266, we hypothesized that a major effect may occur within the bone microenvironment. To verify this, we measured the expression of the same genes following the stimulation of melanoma cells with BCM. As shown in Fig. 5a, osteotropic LCP underwent a significant up-regulation from baseline of CXCR7 (2.9 ± 0.6-fold increase; p < 0.001), CXCR4 (2.3 ± 0.5-fold increase; p < 0.01) and PTH-rP (2.2 ± 0.7-fold increase; p < 0.05), while MMP1 and TGF-β1 were slightly increased (≈ 1.5-fold increase; p < 0.05). By contrast, the transcriptional profile of not-osteotropic SK-Mel28 and WM-266 cells remained almost unchanged following the BCM stimulation, except for a minimal increment of CXCR7 (≈ 1.5-fold increase; p = ns) for SK-Mel28. Similar experiments explored the effects induced by h-Exos from osteotropic LCP on SK-Mel28 and WM-266 cells (Fig. 5b). As shown, the stimulation with h-Exos dramatically increased the expression of CXCR7, as compared to unstimulated cells, both in SK-Mel28 (3.3 ± 0.8-fold increase; p < 0.05) and WM-266 cells (3.1 ± 1.3-fold increase; p < 0.01). On the contrary, no transcriptional modifications were observed in not-osteotropic WM-266 cells stimulated with a-Exos, thus suggesting that CXCR7 up-regulation was putatively responsible of the osteotropic behavioral changes induced by LCP-derived Exos. To validate these results, we further investigated the effects of BCM or h-Exos stimulation on protein levels of both CXCR4 and CXCR7 in melanoma cells.

To this purpose, we analyzed by flow-cytometry both membrane and intracellular levels of CXCR4 and CXCR7 in either stimulated or unstimulated cells. As shown in Table 1, LCP and both SK-Mel28 and WM-266 similarly expressed membrane (6.5%, 10.1% and 8.4%, respectively) and intracellular (88%, 84.6% and 87.1%) levels of CXCR4. By contrast, the levels of both membrane and intracellular CXCR7 by LCP (17.5% and 94.7%, respectively) were higher with respect to both SK-Mel28 (10.4% and 82.5%) and WM-266 cells (9.4% and 79.5%). As shown in Fig. 6a, following the stimulation with BCM, the percentage of LCP and SK-Mel28 expressing either membrane or intracellular CXCR4 remained unchanged with respect to unstimulated cells in a fashion almost similar to CXCR7. The stimulation of WM266 produced similar results. However, the MFI ratio relative to membrane CXCR7 by LCP cells significantly increased (13.6 ± 1.2 vs 8.3 ± 0.9; p < 0.05) after the stimulation with BCM, while remained unchanged in SK-Mel28 (7.6 ± 0.9 vs 7.5 ± 1.3) and WM-266 (data not shown). By contrast, the stimulation of SK-Mel28 cells with h-Exos from osteotropic LCP up-regulated the membrane levels of CXCR7 as compared to unstimulated cells (30.5% vs 10.4%; p < 0.01), while the intracellular expression was unaffected (87.1% vs 82.5%). In addition, the MFI ratio relative to membrane expression of CXCR7 was also enhanced in SK-Mel28 cells by the stimulation with h-Exos with respect to baseline (12.3 ± 1.0 vs 7.7 ± 1.8; p < 0.05). Similar results were observed for WM-266 cells stimulated with h-Exos from LCP (data not shown). Consistent with gene expression analyses, levels of CXCR4 by not-osteotropic cells were not influenced by LCP-derived Exos in a fashion similar to those of both CXCR4 and CXCR7 by LCP stimulated with h-Exos from SK-Mel28.

To verify these data, WB experiments were completed on the membrane and cytosolic fractions of SK-Mel28. As shown in Fig. 6b, c, the stimulation with h-Exos from LCP induced a significant up-regulation of membrane CXCR7 expression (1.53 ± 0.22-fold increase; p < 0.05), while the cytoplasmic fraction was apparently not influenced (1.06 ± 0.1-fold increase; p = ns). Additionally, no modification of CXCR4 expression was revealed in either membrane or cytoplasmic fractions of stimulated SK-Mel28 as compared to unstimulated cells (1.1 ± 0.07 and 0.88 ± 0.18-fold increase, respectively).

Levels of CXCR4 and CXCR7 were then measured by flow cytometry in Exos from osteotropic LCP. As shown in Fig. 6d, both chemokine receptors were not revealed in these Exos, thus excluding a direct Exo-mediated transfer of these molecules onto the plasma membrane of SK-Mel28 as a putative mechanism involved with CXCR7 up-regulation.

To investigate the functional relevance of chemokine receptors in the osteotropism of melanoma cells, silencing of either CXCR4 or CXCR7 were carried out by siRNAs. Efficient silencing of both mRNA and protein levels in LCP and SK-Mel28 cells was verified by both dd-PCR and WB. As shown in Fig. 7a, more than 50% of mRNA levels of CXCR4 and CXCR7 were restrained in both cell lines as compared with untreated cells. A similar decrease in terms of protein expression was also demonstrated. To avoid the possible contamination from other soluble factors released by bone fragments, human-recombinant SDF-1 was used as chemoattractant for migration and invasion assays (Fig. 7b, c). The effect of SDF-1 alone or in combination with h-Exos on CXCR4 and CXCR7 expression by LCP and SK-Mel28 cells was investigated by dd-PCR (Additional file 2: Figure S1) obtaining similar results to those observed by BCM stimulation. Similarly to previous experiments using bone fragments, migration of both untreated and scramble-treated LCP was significantly increased by SDF-1 as compared to unstimulated cells (13.7 ± 2.4 vs 6.1 ± 1.7 cells/0.2 mm² and 14.9 ± 4.5 vs 5.2 ± 1.1 cells/0.2 mm², respectively; p < 0.05) (Fig. 7b). However, the migration of cells stimulated with SDF-1 was dramatically impaired by siRNA of either CXCR4 (9.5 ± 1.9 cells/0.2 mm²) or CXCR7 (7.8 ± 2.4 cells/0.2 mm²) as compared to both untreated and scramble-treated cells (p < 0.05), while h-Exos were ineffective. Furthermore, both untreated and scramble-treated SK-Mel28 were not influenced by SDF-1 in their migratory property with respect to unstimulated cells (16.1 ± 3.5 vs 15.9 ± 2.8 cells/0.2 mm² and 17 ± 3.7 vs 16.7 ± 3.2 cells/0.2 mm², respectively; p = ns), and no effect was induced by SDF-1 in silenced cells (Fig. 7c). However, the concomitant stimulation of SK-Mel28 cells with SDF-1 and h-Exos from LCP significantly increased the migration of both untreated (33.6 ± 3.4 cells/0.2 mm²; p < 0.001) and scramble-treated cells (34.2 ± 4.1 cells/0.2 mm²; p < 0.001), while this effect was dampened in CXCR4 (25.4 ± 3.7 cells/0.2 mm²) and CXCR7 silenced cells (22.5 ± 2.8 cells/0.2 mm²). Similar results were obtained when LCP and SK-Mel28 cells were investigated for their invasive capacity under the same conditions.

These data proved that SDF-1 drives the chemotaxis of osteotropic melanoma cells through the CXCR4/CXCR7 axis that can be activated in not-osteotropic cells by Exos through the up-regulation of membrane CXCR7 molecules.