Semaphorin-7A on Exosomes: A Promigratory Signal in the Glioma Microenvironment

Exosomes are one of the most important mediators of the cross talk occurring between glioma stem cells (GSCs) and the surrounding microenvironment. We have previously shown that exosomes released by patient-derived glioma-associated stem cells (GASC) are able to increase, in vitro, the aggressiveness of both GSC and glioblastoma cell lines. To understand which molecules are responsible for this tumour-supporting function, we performed a descriptive proteomic analysis of GASC-exosomes and identified, among the others, Semaphorin7A (SEMA7A). SEMA7A was described as a promigratory cue in physiological and pathological conditions, and we hypothesised that it could modulate GSC migratory properties. Here, we described that SEMA7A is exposed on GASC-exosomes’ surface and signals to GSC through Integrin β1. This interaction activates focal adhesion kinase into GSC and increases their motility, in our patient-based in vitro model. Our findings suggest SEMA7A-β1-integrin as a new target to disrupt the communication between GSCs and the supporting microenvironment.


Introduction
Despite a state-of-the art treatment, the median overall survival of glioblastoma multiforme (GBM)'s patients is 14 months [1]. The lack of therapeutic efficacy is due to the great intertumour and intratumour cellular and molecular heterogeneity, biological aggressiveness, the ability to develop drug resistance, as well as the infiltrative nature of tumour cells into the surrounding brain parenchyma [2].
Infiltration makes impossible a radical surgery thus favouring recurrences within 1 to 2 cm from the original tumour mass, appearing few months after the first diagnosis and treatment or being already present at the time of the initial presentation [3]. Moreover, single infiltrating cells are often spread throughout the entire brain parenchyma, escaping from surgery and treatment [4].
Several studies suggest that GBM's tumour bulk contains a subpopulation of self-renewing and highly tumorigenic stem cells named glioma stem cells (GSCs) [5,6], with tumour-initiating properties that contribute to tumour growth and malignancy [7] through their sustained proliferation, As assessed by Western blot, exosome preparations were enriched in CD63 and CD9, two transmembrane tetraspanins used as exosomal markers, and in TSG101 (tumour susceptibility gene 101)-a key component of the ESCRT-I complex (Endosomal Sorting Complex Required for Transport). Conversely, exosome preparations did not express Calnexin, an Endoplasmic Reticulum Protein, highly expressed in the cellular counterpart, thus excluding the presence of cellular contaminants ( Figure 1B). As assessed by Western blot, exosome preparations were enriched in CD63 and CD9, two transmembrane tetraspanins used as exosomal markers, and in TSG101 (tumour susceptibility gene 101)-a key component of the ESCRT-I complex (Endosomal Sorting Complex Required for Transport). Conversely, exosome preparations did not express Calnexin, an Endoplasmic Reticulum Protein, highly expressed in the cellular counterpart, thus excluding the presence of cellular contaminants ( Figure 1B).
Sequence database searches were performed using the MASCOT software (Matrixscience.com) giving a list of 316 elements (Tables S1 and S2). Four filtering steps have been applied to the list to fish out interesting candidates. (1) Significant MASCOT score: Since on average, individual ions scores > 44 indicate identity or extensive homology (p < 0.05), we excluded from Mascot lists proteins identified with a Mascot protein score < 44. After this step, a list of 215 digits remained (140 for S82 and 75 for S104). (2) Single digit for one protein: some proteins are present also in truncated forms and they are therefore found in different bands. All the multiple digits for the same protein were removed leaving a list of 139 elements (91 for S82 and 48 for S104). (3) Only human proteins: Mascot analyses were performed also using mammals as taxonomy. Bovine proteins corresponding to human ones and identified with the same scores were considered as putative contaminants, and therefore excluded. After these first steps a list of 85 elements, exclusively human, was left (62 in sample S82 and 23 in sample S104). (4) Proteins present in both S82-and S104-derived exosomes: comparing the two samples we found only 15 proteins present in both lists (Table 1).

SEMA7A Is Released on Exosomes
We focused our attention on Semaphorin-L SEMAL, also known as SEMA7A, since it has been described as a promigratory protein in different physiological conditions [25,27]. Indeed, we hypothesised that SEMA7A, carried by exosomes, could provide a promoting signal for the invasive properties of GSCs.
To verify our hypothesis, we first validated the presence of SEMA7A on exosomal preparations from the producing cell lines used for the proteomic analysis ( Figure 2A). Western Blot on S82 and S104 exosomal lysates and on the paired secreting cells lysates confirmed the presence of SEMA7A in both samples. We also analysed exosomes derived from a human fibroblast cell line Wi38, a mesenchymal cell line derived from nontumour tissue, as control. We observed that SEMA7A is expressed in the cells, but not released in vesicles produced by the healthy cell type, suggesting that the release of SEMA7A in exosomes is a peculiarity of mesenchymal stromal cells resident in the glioblastoma microenvironment. Table 1. List of 15 human proteins shared by two GASC-derived exosome preparations analysed by proteomics (Exo_S82, Exo_S104). Seven proteins were already described in Exocarta database and shown in the right column is a list of sources of exosomes in which they were observed. In bold are highlighted the six proteins not previously referenced in Exocarta.  Membranes were hybridised with anti-TSG101 (Tumour Susceptibility Gene 1) and beta-Actin to show the reliability of exosomes and cell lysates respectively. (C) Western blot analysis for SEMA7A, CD9 and TSG101 in Exo#1 and Exo#2 preparations either untreated (−) or subjected to proteinase K treatment (P.K., +).

Expression of SEMA7A Receptors on GSC
Since SEMA7A is exposed on the surfaces of GASC-exosomes, we verified whether it could interact with receptors expressed on the GSC membrane.
SEMA7A has been shown to bind to PlexinC1 [25] and to β1-integrin receptors [31]. Plexin C1 is a member of a large family of transmembrane receptors with high affinity for semaphorins [32]. In neurons, Sema7A-PlexinC1 signalling regulates synapse development and neuroglial plasticity [33,34], while, in cancer, PlexinC1 is involved in cell migration and proliferation [35].
The importance of β1-integrin receptors in SEMA7A signalling has been demonstrated by Pasterkamp et al. by blocking the binding between Sema7A and its receptor, resulting in the inhibition of Sema7A-dependent neurite outgrowth in olfactory bulb neurons [31].
Therefore, we first verified, by flow cytometry, the presence of the SEMA7A receptors PlexinC1 and β1-integrin on the surface of GSC. As shown in Figure 3, 97.5 ± 2.12% of GSC expressed β1-integrin at high intensity, while the percentage of cells expressing PlexinC1 was almost undetectable (mean = 0.3 ± 0.14% positive cells).
In light of this observation, we hypothesised that SEMA7A exposed on GASC-exosomes could signal, through its interaction with β1-integrin, on GSC. The presence of SEMA7A was further confirmed in the 4 GASC cell lines and relative exosomes, subsequently used in functional experiments ( Figure 2B).
SEMA7A is a glycosylphosphatidylinositol-linked membrane protein that, once released by the cells, can be either found in the vesicular membrane or inside the membrane bilayer [30].
To elucidate where SEMA7A is accumulated in exosomes, we treated intact exosomes with Proteinase K, a broad substrate-specific endopeptidase. As showed in Figure 2C, Proteinase K completely removed SEMA7A, similarly to the multipass membrane tetraspanin CD9, while the inner TSG101 (Tumour Susceptibility Gene-101) was not affected by digestion, supporting the idea that SEMA7A is exposed on the external side of exosomes and therefore could signal directly on the target cells.

Expression of SEMA7A Receptors on GSC
Since SEMA7A is exposed on the surfaces of GASC-exosomes, we verified whether it could interact with receptors expressed on the GSC membrane.
SEMA7A has been shown to bind to PlexinC1 [25] and to β1-integrin receptors [31]. Plexin C1 is a member of a large family of transmembrane receptors with high affinity for semaphorins [32]. In neurons, Sema7A-PlexinC1 signalling regulates synapse development and neuroglial plasticity [33,34], while, in cancer, PlexinC1 is involved in cell migration and proliferation [35]. The importance of β1-integrin receptors in SEMA7A signalling has been demonstrated by Pasterkamp et al. by blocking the binding between Sema7A and its receptor, resulting in the inhibition of Sema7A-dependent neurite outgrowth in olfactory bulb neurons [31].
Therefore, we first verified, by flow cytometry, the presence of the SEMA7A receptors PlexinC1 and β1-integrin on the surface of GSC. As shown in Figure 3, 97.5 ± 2.12% of GSC expressed β1-integrin at high intensity, while the percentage of cells expressing PlexinC1 was almost undetectable (mean = 0.3 ± 0.14% positive cells). Histograms overlays show isotype control staining profile (green histograms) versus specific antibody staining profile (red histograms).

SEMA7A and Exosomes Activates FAK Signalling through β1-Integrin on GSC
The most relevant pathway activated by Integrin receptors is the nonreceptor protein tyrosine kinase focal adhesion kinase (FAK), which is rapidly phosphorylated after integrin receptor binding, leading to the activation of cytoskeleton proteins responsible for integrin signal propagation [36].
To verify whether SEMA7A activates FAK pathway, we treated GSC with exogenous recombinant SEMA7A-Fc, for 2, 5 and 10 min at the concentration of 10 ng/mL and 100 ng/mL, and assessed the level of tyrosine phosphorylation (Tyr397) of FAK (p-FAK) by Western blot ( Figure 4A). As shown in Figure 4B, SEMA7A-Fc treatment increased the level of FAK phosphorylation and the peak of p-FAK was observed at 5 min with a concentration of 100 ng/mL (p < 0.0001 vs. untreated cells, Ctrl). Likewise, the treatment of GSC with exosomes, at a concentration of 10 μg/mL, increased the level of p-FAK after 5 and 10 min (p = 0.0002 vs. Ctrl cells). Conversely, maintaining exosomes for 30 min, diminished the p-FAK at levels of control ( Figure 4C,D). We therefore choose the concentration of 100 ng/mL for SEMA7A-Fc and 10 μg/mL for GASC-exosome to exploit their effect on GSC motility. In light of this observation, we hypothesised that SEMA7A exposed on GASC-exosomes could signal, through its interaction with β1-integrin, on GSC.

SEMA7A and Exosomes Activates FAK Signalling through β1-Integrin on GSC
The most relevant pathway activated by Integrin receptors is the nonreceptor protein tyrosine kinase focal adhesion kinase (FAK), which is rapidly phosphorylated after integrin receptor binding, leading to the activation of cytoskeleton proteins responsible for integrin signal propagation [36].
To verify whether SEMA7A activates FAK pathway, we treated GSC with exogenous recombinant SEMA7A-Fc, for 2, 5 and 10 min at the concentration of 10 ng/mL and 100 ng/mL, and assessed the level of tyrosine phosphorylation (Tyr397) of FAK (p-FAK) by Western blot ( Figure 4A). As shown in Figure 4B, SEMA7A-Fc treatment increased the level of FAK phosphorylation and the peak of p-FAK was observed at 5 min with a concentration of 100 ng/mL (p < 0.0001 vs. untreated cells, Ctrl). Likewise, the treatment of GSC with exosomes, at a concentration of 10 µg/mL, increased the level of p-FAK after 5 and 10 min (p = 0.0002 vs. Ctrl cells). Conversely, maintaining exosomes for 30 min, diminished the p-FAK at levels of control ( Figure 4C,D). We therefore choose the concentration of 100 ng/mL for SEMA7A-Fc and 10 µg/mL for GASC-exosome to exploit their effect on GSC motility.

SEMA7A and Exosomes Increases GSC Motility
We performed a motility scratch assay to investigate whether SEMA7A-Fc treatment affected the migratory properties of GSC. A gap was obtained in confluent GSC by scratching the cell monolayer with a tip and the distance covered after treatment with recombinant SEMA7A-Fc (100 ng/mL) and GASC-exosomes (10 μg/mL) were measured, after 8 and 24 h. As shown in Figure 5, cells treated with SEMA7A-Fc were significantly faster in repairing the gap compared to control (covered distance in 8 h = 169 ± 21.9 μm vs. 69.62 ± 25.50 μm in SEMA7A-Fc treated and Ctrl cells, respectively). Likewise, a significant improvement in the motility of GSC was observed after treatments with exosomes (covered distance in 8 h = 198.8 ± 63. 9 μm). Thus, SEMA7A represents a promigratory signal for GSCs, and this effect is presumably mediated by FAK pathway activation.

SEMA7A and Exosomes Increases GSC Motility
We performed a motility scratch assay to investigate whether SEMA7A-Fc treatment affected the migratory properties of GSC. A gap was obtained in confluent GSC by scratching the cell monolayer with a tip and the distance covered after treatment with recombinant SEMA7A-Fc (100 ng/mL) and GASC-exosomes (10 µg/mL) were measured, after 8 and 24 h. As shown in Figure 5, cells treated with SEMA7A-Fc were significantly faster in repairing the gap compared to control (covered distance in 8 h = 169 ± 21.9 µm vs. 69.62 ± 25.50 µm in SEMA7A-Fc treated and Ctrl cells, respectively). Likewise, a significant improvement in the motility of GSC was observed after treatments with exosomes (covered distance in 8 h = 198.8 ± 63. 9 µm). Thus, SEMA7A represents a promigratory signal for GSCs, and this effect is presumably mediated by FAK pathway activation.

Impact of SEMA7A-β1-Integrin Receptor Pathway on GSC Motility
To show that β1-integrin is the intermediate receptor between SEMA7A-exosomes and FAK in GSC, we first tested a functional blocking antibody to β1-integrin on GSC and evaluated changes of p-FAK, as described above (Figure 4). Figure 6A,B shows that (1) FAK was quickly phosphorylated after exposure to SEMA7A-Fc and GASC-exosomes; (2) FAK phosphorylation was abrogated when cells were treated by anti-β1-integrin antibody; and (3) FAK phosphorylation was not recovered despite the addition of SEMA7A-Fc or GASC-exosomes.
Therefore, the inactivation of β1-integrin receptor completely neutralised the activation of the FAK pathway stimulated by SEMA7A or GASC-exosomes. This effect was specific to β1 blocking antibody, since treating GSC with an unrelated antibody (anti-mouse IgG), had no effect on SEMA7A-mediated FAK phosphorylation.
To test whether the molecular interference on SEMA7A-β1-integrin-FAK might impact on GSC motility, we performed a Scratch assay to evaluate the GSC's migratory performances, following exposure to recombinant SEMA7A-Fc alone, in combination with anti-β1-integrin blocking antibody or with anti-mouse IgG as control. Since our hypothesis is that SEMA7A transported by exosomes in the glioma microenvironment could be responsible, at least in part, of the promigratory effect on GSC, we performed a Scratch assay on GSC exposed to exosomes alone or in combination with β1-Integrin-blocking antibody. Histograms in Figure 6C clearly show that the distance covered by SEMA7A-Fc and exosomes-treated cells measured after 8 h (133.3 ± 42.26μm and 174 ± 51.23 μm, respectively), was significantly longer when compared with cells treated with anti-mouse IgG antibody (79.30 ± 20.66 μm). When the binding of SEMA7A-exosomes with β1-integrin receptor was

Impact of SEMA7A-β1-Integrin Receptor Pathway on GSC Motility
To show that β1-integrin is the intermediate receptor between SEMA7A-exosomes and FAK in GSC, we first tested a functional blocking antibody to β1-integrin on GSC and evaluated changes of p-FAK, as described above (Figure 4). Figure 6A,B shows that (1) FAK was quickly phosphorylated after exposure to SEMA7A-Fc and GASC-exosomes; (2) FAK phosphorylation was abrogated when cells were treated by anti-β1-integrin antibody; and (3) FAK phosphorylation was not recovered despite the addition of SEMA7A-Fc or GASC-exosomes.
inhibited, cells were significantly slower (37.32 ± 22.95 μm) and their migration was not restored after SEMA7A-Fc or exosomes introduction (62.07 ± 31.46 μm and 55.20 ± 30.43 μm, respectively). Therefore, our results strongly suggest that SEMA7A exposed on GASC-released exosomes, upon binding to β1-integrin on the surface of GSCs, accelerates their migration, which is correlated with tumour aggressiveness. Representative Western blotting for FAK and pFAK of GSC treated with anti-Immunoglobulin G (Anti-IgG) antibody, 5 μg/mL for 65 min, anti-Immunoglobulin G (anti-IgG)antibody, 5 μg/mL for 60 min and SEMA7A 100 ng/mL for 5 min, anti-IgG antibody 5 μg/mL for 60 min and GASC-derived exosomes 10 μg/mL for 5 min, SEMA7A 100 ng/mL for 5 min, GASC-derived exosomes 10 μg/mL for 5 min, anti-Integrin beta 1 blocking antibody 5 μg/mL for 65 min, anti-Integrin blocking antibody 5 μg/mL for 60 min and SEMA7A 100 ng/mL for 5 min and anti-Integrin blocking antibody 5 μg/mL for 60 min and GASC-derived exosomes 10 μg/mL for 5 min. (B) Densitometric analysis was performed using the software available in the Gel-doc instrument (Alliance Uvitec, Ltd. Cambridge, UK) to quantify the level of FAK phosphorylation. The histogram represents results reported as the Fold Change of p-FAK of different treatments vs. cells treated with the anti-IgG antibody. Values were calculated as the ratio of IOD (Integrated Optical Density) pFAK/FAK after normalisation on beta actin level. (C) A motility assay was performed in cells treated as described in B. Histogram represents the distance covered after 8 h, calculated measuring the width of the gap between the two margins of scratched monolayer. Data are presented as means ± standard deviation of 4 replicates. * p < 0.05 vs. Anti-IgG, ** p < 0.05 vs. Anti-IgG+SEMA-7A, *** p < 0.05 vs. Anti-IgG+ Exo, § p < 0.05 vs. with anti-Immunoglobulin G (Anti-IgG) antibody, 5 µg/mL for 65 min, anti-Immunoglobulin G (anti-IgG)antibody, 5 µg/mL for 60 min and SEMA7A 100 ng/mL for 5 min, anti-IgG antibody 5 µg/mL for 60 min and GASC-derived exosomes 10 µg/mL for 5 min, SEMA7A 100 ng/mL for 5 min, GASC-derived exosomes 10 µg/mL for 5 min, anti-Integrin beta 1 blocking antibody 5 µg/mL for 65 min, anti-Integrin blocking antibody 5 µg/mL for 60 min and SEMA7A 100 ng/mL for 5 min and anti-Integrin blocking antibody 5 µg/mL for 60 min and GASC-derived exosomes 10 µg/mL for 5 min. (B) Densitometric analysis was performed using the software available in the Gel-doc instrument (Alliance Uvitec, Ltd. Cambridge, UK) to quantify the level of FAK phosphorylation. The histogram represents results reported as the Fold Change of p-FAK of different treatments vs. cells treated with the anti-IgG antibody. Values were calculated as the ratio of IOD (Integrated Optical Density) pFAK/FAK after normalisation on beta actin level. (C) A motility assay was performed in cells treated as described in B. Histogram represents the distance covered after 8 h, calculated measuring the width of the gap between the two margins of scratched monolayer. Data are presented as means ± standard deviation of 4 replicates. * p < 0.05 vs. Anti-IgG, ** p < 0.05 vs. Anti-IgG+SEMA-7A, *** p < 0.05 vs. Anti-IgG+ Exo, § p < 0.05 vs. SEMA-7A, • p < 0.05 vs. Exo. Statistical analysis was performed by repeated measurements one-way ANOVA followed by Bonferroni Multiple Comparison post-test (GraphPad Prism 5, San Diego, California). Therefore, the inactivation of β1-integrin receptor completely neutralised the activation of the FAK pathway stimulated by SEMA7A or GASC-exosomes. This effect was specific to β1 blocking antibody, since treating GSC with an unrelated antibody (anti-mouse IgG), had no effect on SEMA7A-mediated FAK phosphorylation.
To test whether the molecular interference on SEMA7A-β1-integrin-FAK might impact on GSC motility, we performed a Scratch assay to evaluate the GSC's migratory performances, following exposure to recombinant SEMA7A-Fc alone, in combination with anti-β1-integrin blocking antibody or with anti-mouse IgG as control. Since our hypothesis is that SEMA7A transported by exosomes in the glioma microenvironment could be responsible, at least in part, of the promigratory effect on GSC, we performed a Scratch assay on GSC exposed to exosomes alone or in combination with β1-Integrin-blocking antibody. Histograms in Figure 6C clearly show that the distance covered by SEMA7A-Fc and exosomes-treated cells measured after 8 h (133.3 ± 42.26µm and 174 ± 51.23 µm, respectively), was significantly longer when compared with cells treated with anti-mouse IgG antibody (79.30 ± 20.66 µm). When the binding of SEMA7A-exosomes with β1-integrin receptor was inhibited, cells were significantly slower (37.32 ± 22.95 µm) and their migration was not restored after SEMA7A-Fc or exosomes introduction (62.07 ± 31.46 µm and 55.20 ± 30.43 µm, respectively). Therefore, our results strongly suggest that SEMA7A exposed on GASC-released exosomes, upon binding to β1-integrin on the surface of GSCs, accelerates their migration, which is correlated with tumour aggressiveness.

Discussion
It is well known that glioblastoma cells subvert their microenvironment from a tumour-suppressive to a tumour-supporting condition, which promotes proliferation, angiogenesis and invasion [7]. This phenomenon requires a continuous cross-talk between tumour cells and nontumour components, such as stromal cells, extracellular matrix and immune cells [37].
In our previously established in vitro model of GBM microenvironment, we found that exosomes, released by GASC, have a tumour-supporting role towards GSC [19].
Therefore, we performed a descriptive proteomic analysis of exosomes released by GASC, to understand which components could be involved in the GASC tumour-supporting function.
Results from mass spectrometry revealed 15 proteins present in both exosomes preparations analysed. Overall, all the proteins found in this study could contribute to the tumour-supporting function of GASC' exosomes. Fibronectin1 FBN1, TIMP metalloprotease inhibitor 1)(TIMP1) and Phospholipid transfer protein (PLTP) were described as promoter of metastasis, resistance to apoptosis and migratory properties in ovarian cancer, melanoma and glioma, respectively [38][39][40]; Collagen Alpha-1 (VI) chain (COL6A1) and Collagen Alpha-2 (VI) chain (COL6A2) are involved in the guidance of neural crest cells, during CNS development [41] and in the motoneuron axon growth [42]; plasminogen activator inhibitor-1 (PAI-1), is a well-known marker overexpressed in several forms of cancer [43]; Galectin-3-binding protein promotes integrin-mediated cell adhesion [44] and has been found elevated in the serum of patients with cancer [45], although not present in Exocarta, it has been described in the exosomes of ovarian carcinoma cells [46] and it has been suggested to be an important component of tumour microenvironment [47]; Serotransferrin (transferrin) is an abundant blood plasma glycoprotein whose main function is to bind and transport iron throughout the body. Interestingly, Carlsson et al., 2013 [48] showed that~5% of human serotransferrin glycoforms bind galectin-3 and are targeted to a different endocytic pathway and that the galectin-3-bound glycoform is increased in cancer. Although eight out of 15 proteins have been previously described in exosomes of some sources, five of them are completely new. Among these, we focused our attention on SEMA7A, because of its role in physiological and pathological conditions. One of the first study showed that SEMA7A has a function on the immune system, being a potent activator of monocytes, stimulating their chemotaxis and production of inflammatory cytokines [49]. Pasterkamp et al. were the first to describe SEMA7A as a guidance signal responsible to stimulate axon outgrowth, during the formation of lateral olfactory tracts [31]. The authors showed that SEMA7A binds to integrin beta1 receptors and activates the mitogen activated protein kinase pathway (MAPK). Moreover, the use of a beta-1 integrin inhibitory antibody blocked the neurite outgrowth. Moreover, SEMA7A is expressed in the secretome of U87 glioblastoma cells at higher level than in less aggressive and invasive cell lines (T98, U118) [50].
In summary SEMA7A can be regarded as a promigratory stimulus and since motility of glioma cells is one of the properties responsible for the infiltrative nature of GBM, we hypothesised that SEMA7A, released by GASC through exosomes, can act as one of the messages stimulating GSC motility/migration.
We confirmed both the presence of SEMA7A on four exosome preparations and, most importantly, its exposure on the external surface of exosomes. On the other hand, cytofluorimetric analysis revealed the presence of integrin beta 1, but not Plexin C1-two major SEMA7A receptors-on glioma stem cells (GSCs). This suggested that SEMA7A could directly signal to GSC through integrin beta1.
Integrins represent the major cellular receptors for extracellular matrix involved in the regulation of cell migration through their coupling with cytoskeletal and signalling molecules, clustering in focal adhesion in adherent cells and Focal adhesion kinase (FAK) has been established as a key component of signal transduction triggered by integrins [51]. Autophosphorylation and activation of FAK lead to modulation of cytoskeletal proteins, cytoskeletal reorganisation and force generation [52,53].
Regulation of cell migration by integrin signalling through FAK leading to cancer pathogenesis and aggressiveness has been assessed in many cell types [54]. β1 integrins have been implicated in brain invasion of glioma cells in animal models using antisense RNA to reduce integrin expression [55].
In our in vitro model, relying on GASC and GSC obtained from the same tumour, the treatment of GSC, both with recombinant SEMA7A-Fc and exosomes produced by GASC, stimulated a rapid FAK phosphorylation and significantly increased the motility of GSC. Using an antibody blocking β1-integrin receptors, we observed a reduction of FAK activation at the level of the control and a decreased speed of GSC, in covering the gap generated in the motility assay. Concomitant addition of SEMA7AFc or GASC-exosomes to anti-integrin beta1 antibody was not able to rescue FAK engagement or motility of GSC.
The involvement of SEMA7A-integrin β1 in cell migration and tumour invasiveness is not completely new: Black et al. identified a high expression of SEMA7A in ductal in situ breast cancer characterised by poor prognosis and distant metastases, and showed the involvement of SEMA7A in promoting tumour cell invasion and lymphangiogenesis, via activation of β1-integrin [29].
Moreover, other members of the semaphorin family were identified in exosomes of the glioblastoma microenvironment. SEMA3A released in GSC' exosomes disrupt the endothelial barrier, thus promoting the vascular permeability and invasion in the surrounding brain parenchyma [56].
The novelty of our work is to indicate, for the first time, the interaction SEMA7A-β1-integrin as a new mediator in the cross-talk occurring in the glioma microenvironment between exosomes produced by glioma stromal cells and GSC, becoming an interesting new possible therapeutic target In fact, integrin ligand binding and regulatory sites are externally exposed and made them good accessible drug candidates [57]. Indeed, many inhibitors have been developed. Of the number of clinical trials started, some reached late stage and some inhibitors have even launched for treatment. Of the Intβ1 inhibitors evaluated in the clinical trials, a pan β1 monoclonal antibody P5 (claimed to predominantly act on α5β1) is reported to enhance cisplatin efficacy in lung adenocarcinoma cells and is in use in phase 3 trial for non-small cell lung cancer [58].
In the light of our result it would be interesting to further investigate the possibility to reduce tumour infiltration by the use of β1 inhibitors like the P5 antibody.
Altogether our results give new insights on how stromal cells in the glioblastoma microenvironment could contribute to the increased aggressiveness of the tumour.

Isolation and Culture of GASC and GSC
Cells were isolated from patients affected by a de novo supratentorial glioblastoma (GBM). All patients, not previously treated, underwent surgical resection of the tumour at the Neurosurgery Department of the Udine Hospital. The independent ethic committee of the Azienda Ospedaliero-Universitaria of Udine has approved the research (Consent 102/2011/Sper, 02 August, 2011 and Consent 196/2014/Em, 03 December, 2014). Written informed consents were obtained from all patients and clinical investigations have been conducted according to the Declaration of Helsinki. GASC were isolated from 6 tumour samples, as previously described [19]. In 4 cases, GSCs from the same tumour samples were cultured [59] in order to have 4 GASC-GSC pairs. See Appendix A.

Isolation of Exosomes
Cells were cultured in expansion medium for three passages and then seeded at 6000 cells/cm 2 in 100 mm Petri dishes. After 24 h, expansion medium was replaced with a serum-linoleic acid bovine serum albumin-depleted medium. Cells were maintained until 70-80% confluence (48 h). Wi38 cells were cultured in Dulbecco's Modified Eagle Medium D-MEM+10% exosome-depleted foetal bovine serum. Exosomes were isolated from the collected supernatants using ExoQuick-TC Exosome precipitation solution (System Biosciences, Palo Alto, California), according to manufacturer's protocol. The exosomal pellets were resuspended in phosphate-buffered saline (PBS) or radio immunoprecipitation assay (RIPA) buffer; see Appendix A.

Nanoparticle Tracking Analysis
Concentration and particle size of purified exosomes were measured by Nanosight (LM10, Malvern system Ltd., U.K.), equipped with a 405 nm laser. Briefly, each sample, once properly diluted, was recorded for 60 s with a detection threshold set at maximum. Temperature was monitored throughout the measurements. Vesicle size distribution and an estimated concentration of NTA (Nanoparticle Tracking Analysis) profiles were obtained from the given raw data files.

Proteomic Analyses
For proteomic analysis, 30 µg of total exosomal proteins obtained from two different GASC cultures (S82 and S104) were used. Protein lysates were separated on SDS-PAGE (Gel Tris-Glycine 10%) and the gel was cut to divide samples into six macrobands and treated essentially as previously described [60]. Fragments obtained by in-gel trypsin digestion were analysed by LC-MS/MS with ionic trap, on an Agilent 1200 series nanoHPLC interfaced to an HCTultra IT (Bruker Daltonics, Billerica, Massachussetts). Peptide masses and MS/MS spectra were exported as 'mgf' files and database search was performed with the MASCOT MS/MS (Mascot-Peptide mass finger-print) Ion Search option). See Appendix A.

Western Blot
Whole-cell and exosomes extract proteins were obtained by lysis in RIPA buffer. Thirty micrograms of proteins was resolved on SDS-PAGE, transferred and immobilised on a 0.45 µm nitrocellulose membrane (Amersham, London, UK).

Proteinase K Treatment of GASC-Exosomes
30 µg of GASC-derived exosomes resuspended in PBS were incubated with 100 µg/mL Proteinase K (Sigma-Aldrich, Italy) or the same volume of water as control, for 1 h, at 37 • C, with regular shaking.
Proteinase K was inhibited with phenylmethylsulfonyl fluoride PMSF 5 mM treatment for 10 min at 72 • C. Exosomes were then lysed and analysed by western blot.
For treatments with GASC-exosomes, GSCs were exposed to 10 µg/mL of exosomes for 5, 10 and 30 min. For treatments with anti-integrin beta1 blocking antibody, GSC were treated as detailed in Appendix A.
Cells were washed in cold PBS, immediately harvested in RIPA buffer, at 4 • C, and proteins were extracted as previously described. Twenty micrograms of proteins was resolved on 10% SDS/PAGE, and then transferred and immobilised on a 0.45 µm nitrocellulose membrane, (Amersham).
The membranes were incubated with rabbit-anti human to FAK antibody, rabbit-anti human to phospho-FAK (Tyr-397) antibody (Cell Signalling Technology, Danvers, Massachusetts) and with rabbit polyclonal to beta-Actin. Levels of FAK and p-FAK were evaluated by densiometric analysis, using ImageJ. IOD (Integrated Optical Density) was analysed for each condition and Fold Change of p-FAK was calculated vs. untreated (Ctrl) cells, after normalisation on β -actin expression.

Scratch Assay
GSC were seeded on 48-well plates at 3 × 10 4 cell/cm 2 . At confluence, the cell monolayer was straight-scraped with a p10 pipette tip. Cellular debris was removed and culture medium was replaced with fresh one containing the treatments described in Appendix A. Each experimental condition was performed four-fold. Images were acquired at 8 h and 24 h after treatments, by a Leica DMI 6000B microscope connected to a Leica DFC350FX camera (10× objective, numerical aperture 0.25, Wetzlar, Germany). Images were then compared and quantified by Image J.

Appendix A.2.3. Western Blot
Whole-cell extract proteins were obtained by cells' lysis in RIPA Buffer for 40 min on ice and centrifugation at 10.000 g for 15 min at 4 • C, and then the protein-enriched supernatants were collected.