Sortilin mediates the release and transfer of exosomes in concert with two tyrosine kinase receptors

Cancer cells influence their cellular microenvironment through intercellular communication. Growing evidence indicates that microenvironment control is supported by the release of extracellular microvesicles called exosomes. Exosomes are small membrane-bound vesicles, ranging in size from 40 to 100 nm that are derived from endosomes (Pan et al., 1985). They have the ability to transfer their cellular content to neighboring cells and to modify the microenvironment to promote tumor-induced immune suppression, angiogenesis and pre-metastatic niche formation (Simons and Raposo, 2009; Théry et al., 2009). The role of exosomes is likely to be dictated by their contents, which are composed of microRNAs (miRNAs), RNAs, lipids and more than 400 proteins. However, the function of some of these exosomal proteins is as yet unknown.

The multifaceted receptor sortilin (also known as neurotensin receptor 3) plays a multitude of roles in the cell, as a receptor or co-receptor, in the transport of proteins such as neurotensin and neurotrophins to the plasma membrane and to lysosomes, and in regulated secretion (Nykjaer and Willnow, 2012). A number of reports have shown that sortilin expression is elevated in several human cell lines (Dal Farra et al., 2001; Giorgi et al., 2008; Truzzi et al., 2008). Some of these reports have suggested that sortilin can play a dual role through its direct interaction with neurotensin or with mature neurotrophin receptors in cells that stimulates an autocrine and/or paracrine function, which could be a mechanism associated with cell survival and tumorigenesis (Dal Farra et al., 2001; Truzzi et al., 2008). A growing number of reports have highlighted that there can be metalloproteinase-mediated shedding of the extracellular domain of sortilin into the culture medium in several cell types (Hermey et al., 2006; Massa et al., 2013; Navarro et al., 2002). However, the mechanism of sortilin release from these cells and the consequence for the microenvironment is uncertain. In this study, we examined closely the secretion mechanism involved in the release of the extracellular domain of sortilin from human lung cancer cells (A549), together with the effect on endothelial cells. We show for the first time that sortilin is a key component of exosome-mediated communication between A549 and endothelial cells. Using in vitro models, we demonstrate that sortilin is essential for exosome release and is involved in the assembly of a new complex containing two tyrosine kinase receptors, which is increased in non-small cell lung cancer. This complex comprises tropomyosin-related kinase B (TrkB, also known as NTRK2), which is implicated in the activation and invasion of A549 lung cancer cells through the binding of the brain-derived neurotrophin factor (BDNF) (Zhang et al., 2010), and epidermal growth factor receptor (EGFR) (Mendelsohn and Baselga, 2003), both associated with sortilin. This newly reported TrkB–EGFR–sortilin (TES) complex in exosomes has a function in the activation and migration of endothelial cells and could play a role in lung cancer angiogenesis.

Sortilin, a member of the Vps10p domain family of receptors, is a type 1 membrane protein receptor (Fig. 1A). It is expressed in most tissues, notably the brain, central nervous system, muscle and other peripheral cell types. The expression of sortilin is upregulated in a number of cancers including breast, colon and prostate cancer. Sortilin exists in two forms, a full-length form (110 kDa) and a truncated form (95 kDa), corresponding to its large luminal domain (or ectodomain), which has been previously detected in the supernatant medium from sortilin-overexpressing cells (Navarro et al., 2002). To determine whether shedding of the ectodomain of sortilin occurs, we evaluated the expression of different sortilin constructs in two cell lines, human embryonic kidney cells (HEK293), to validate the constructs used in previous studies (Finan et al., 2011; Nielsen et al., 2001), and the cell line of interest, lung adenocarcinoma epithelial cells (A549). We observed that human kidney cells expressed a low endogenous level of full-length sortilin (110 kDa) (Fig. 1B), detected using an ectodomain-specific antibody against sortilin. In order to observe sortilin expression, we expressed various tagged and untagged versions of full-length and C-terminally truncated constructs of sortilin in HEK293 cells. It is interesting to note that the cytoplasmic tail of sortilin encodes two sorting motifs, a tyrosine-based YSVL motif and an acidic dileucine cluster motif, which facilitate both endocytosis and intracellular trafficking of sortilin (Nielsen et al., 2001). Full-length sortilin (110 kDa), when overexpressed and tagged either with GFP (130 kDa) or V5 (115 kDa), was detected in HEK293 cells (Fig. 1B), whereas immunopositive 95-kDa sortilin products were only detected in the medium (using the ectodomain-specific antibody) (Fig. 1B). Truncation of the C-terminus of sortilin results in the generation of smaller products at ∼95 kDa. We verified that these products still contained their appropriate tags using antibodies against GFP and V5 (Fig. 1B). We found truncated sortilin in the medium only with the ectodomain-specific antibody, with the loss of GFP- and V5-tagged proteins further indicating shedding of the C-terminus (Fig. 1B). Likewise, when we performed the same analysis in lung cancer cells (A549), we obtained similar results regarding the loss of tagged sortilin and the shedding of sortilin into the medium (Fig. 1C). To determine the subcellular localization of sortilin, we performed an immunocytochemistry analysis of both full-length and cytoplasmic-truncated sortilin in A549 cells. At 18 h post-transfection of GFP-tagged full-length and truncated sortilin, we found that full-length sortilin localized to the perinuclear region, such as in the trans-Golgi network (TGN), as confirmed by analysis of the colocalization with a TGN marker TGN46, which gave a high Manders' coefficient (Fig. 1D,E). Truncated sortilin distribution was more widespread, with notable peripheral and cell surface localization, and the analysis of its colocalization with TGN46 gave a lower Manders' coefficient (Fig. 1D,E). Hence, the shedding of sortilin and its subcellular localization in A549 cells (Fig. 1C,D) is in accordance with previous reports (Finan et al., 2011; Nielsen et al., 2001).

To gain further insight into the mechanism used for sortilin release into the medium, we first analyzed the transport routes that sortilin could use for secretion from the cell. One such possible route that sortilin could use is through exosomes, which are actively secreted through an exocytosis pathway normally used for receptor discharge and intercellular cross-talk (Raposo and Stoorvogel, 2013; Record et al., 2014). In order to confirm that A549 cells secrete exosomes into the culture medium, we first checked for the cell expression of the tetraspanin proteins CD63 and CD81, which are enriched in late endosomes and abundant in exosomes (Escola et al., 1998; Raposo et al., 1996). These were detected by fluorescence-activated cell sorting (FACS) after staining for CD63 and CD81 with fluorescent antibodies, or for EGFR, an exosome marker (supplementary material Fig. S1). To screen for exosomes secreted from A549 cells, we used two procedures. The first was the previously developed FACS-based assay (Ostrowski et al., 2010). In brief, exosomes present in the culture medium were captured onto beads coated with an antibody to the exosome marker CD63, and were detected by FACS after staining with for CD81 with fluorescent antibodies. Thus, we showed that A549 cells secrete exosomes into the culture medium (Fig. 2A). We used a number of compounds that are known to affect protein transport and stimulate secretion at different stages. Brefeldin A (BFA) is known to affect the classical protein transport pathway, preventing anterograde traffic from the endoplasmic reticulum (ER) to the Golgi and resulting in retrograde transport of proteins from the Golgi to the ER (Lippincott-Schwartz et al., 1989; Wilson et al., 2011). As shown in Fig. 2B, treatment with 1 µg/ml BFA had a major effect on sortilin release in exosomes into the medium under basal conditions. Likewise, treatment with the antibiotic tunicamycin, which inhibits synthesis of the lipid-linked oligosaccharide precursor important for glycoprotein traffic from the ER and Golgi compartments, prevented the release of sortilin into the medium (Fig. 2B). When cells were treated with dimethyl amiloride (DMA), which is known to perturb secretion (Savina et al., 2003), exosomes isolated from A549 cells showed a marked reduction in sortilin (Fig. 2B). We next tested the effect of a reagent called Monensin (Mon), previously shown to affect exosome secretion (Savina et al., 2003). We observed that Mon treatment significantly reduced intracellular levels of sortilin and potentiated the secretion of sortilin into the medium (Fig. 2B). We obtained similar results to those observed with sortilin for EGFR in whole-cell lysates and exosomes under the different treatments (Fig. 2B). The second procedure is a quantitative CD63 and CD81 enzyme-linked immunosorbent assay (ELISA)-based analysis, which involves an ultracentrifugation step to harvest exosomes from culture medium. As shown in Fig. 2C, the average number of exosome particles in control conditions for CD63 and CD81 that were obtained per ml of A549 culture medium was ∼4×10⁹. We observed a significant reduction in the exosome population for both CD63 and CD81 secreted from A549 cells when using the chemical inhibitors BFA (CD63, P = 0.001; CD81, P = 0.0009), tunicamycin (CD63, P = 0.0006; CD81, P = 0.0009) and DMA (CD63, P = 0.0057; CD81, P = 0.0084) (Fig. 2C). In contrast, Mon treatment resulted in a significant elevation of secreted exosomes (CD63, P = 0.005; CD81, P = 0.006) (Fig. 2C). To establish that sortilin is secreted through endocytic exosomes, we purified the exosome population secreted by A549 cells. Western blotting carried out on the exosomal extracts confirmed the presence of exosomal proteins known to be enriched in A549 exosomes, such as the multivesicular body (MVB) marker Tsg101, which is important for membrane protein transport (Babst et al., 2000) and indicates an endosomal origin of the vesicles, the molecular chaperone heat shock protein 90 (Hsp90) and the exosomal protein flotillin (Fig. 2D). We confirmed that sortilin, in particular its extracellular domain, is enriched in A549 exosomes (Fig. 2D). Next, we immunocaptured purified exosomes with anti-CD63 or non-related-serum-coated beads followed by western blotting with for the ectodomain of sortilin (Fig. 2E). The results further confirmed that sortilin is secreted through endocytic exosomes. To prove that intracellular sortilin localizes to late endosomal membranes, we performed electron microscopy analysis using immunogold labeling against the sortilin protein on ultrathin cryosections of A549 cells. Fig. 2F shows unlabeled images obtained from A549 cells, including whole cells and MVBs, facilitating identification of organelle structures. Fig. 2G also shows localization of sortilin (5-nm gold particles) to intraluminal vesicles (ILVs), with the average ILV size of 48 nm within MVBs. Higher magnification showed that sortilin protein was localized to ILVs and to the outer membrane of MVBs (Fig. 2G). When we performed similar EM immunogold labeling of CD63-purified exosomes, we found that sortilin was localized to CD63-positive exosomes (Fig. 2H). Taken together, our results confirm that sortilin is found in MVBs and is secreted from the cell in exosomes.

To investigate whether sortilin plays a role in A549 exosomes, we generated stable knockdowns of sortilin and Rab27a, a small Rab GTPase involved in MVB docking at the plasma membrane, as a control for blocking the exosome release, using short hairpin RNA (shRNA) (Peinado et al., 2012). We confirmed the efficiency of sortilin and Rab27a knockdown compared to pLKO control by measuring both mRNA and protein levels (Fig. 3A,B). Exosomes were purified by differential ultracentrifugation and quantified by CD63 and CD81 ELISA from the culture medium of cells stably expressing one of two shRNAs against sortilin or an shRNA against Rab27a. We observed a significant decrease in the exosome population for both CD63 and CD81 secreted from A549 cells stably expressing shRNA for both sortilin and Rab27a (Fig. 3C).

We analyzed the subcellular distribution of CD63 by immunofluorescence in A549 cells (supplementary material Fig. S2A). In a comparison of the control and Rab27a-knockdown cells, CD63, a marker of MVBs, appeared to be homogenously distributed throughout the cell, but the intensity and clustering was more noticeable with Rab27a silencing (supplementary material Fig. S2A). Likewise, sortilin-knockdown cells showed a similar staining pattern to Rab27a, where the intensity of CD63 staining was different from control. In contrast, the knockdown of the Golgi-localized protein γ-adaptin ear-containing, ARF-binding 1 (GGA1) caused perinuclear CD63 staining with some clustering, compared to the control (supplementary material Fig. S2A). Furthermore, quantitative analysis of CD63–FITC fluorescence revealed that both sortilin- and Rab27a-knockdown cells had significantly higher CD63 staining compared to the pLKO control (supplementary material Fig. S2B). Pulse-chase analysis of CD63 over a 24-h period demonstrated that sortilin knockdown impaired exosome secretion from A549 cells, as observed by CD63 accumulation in the cells and a reduction in the amount of CD63-enriched exosomes released into the culture medium (Fig. 3D,E). Surprisingly, electron microscopy analysis revealed that the number of ILVs found in MVBs was significantly increased in sortilin-knockdown cells compared to those observed from pLKO control cells (Fig. 3F). In addition, the sortilin phenotype resulted in larger ILVs than those observed in pLKO control cells (Fig. 3F). Likewise, we observed an increased population of ILVs in MVBs after Rab27a knockdown (Fig. 3F). In contrast, analysis of ILVs in Rab27a-knockdown cells revealed no change in the surface area of the ILVs compared to pKLO control, whereas the surface area was significantly smaller than that observed after sortilin knockdown (Fig. 3F). Both the pulse-chase and immunofluorescence analysis indicate that CD63 secretion is perturbed in sortilin-depleted cells. To examine the CD63 exosome population more closely, we performed a sucrose density gradient analysis (supplementary material Fig. S2C). After exosome purification, the 100,000 g exosome pellet was fractionated through an overlaid sucrose gradient where CD63-positive exosomes have been shown to sediment at ∼1.10–1.19 g/ml (Bobrie et al., 2012). Similarly, we observed that CD63-positive exosomes fractionated at between 1.10 and 1.19 g/ml, with a peak in the 1.13 g/ml fraction, along with EGFR and flotillin (supplementary material Fig. S2C). In contrast, both sortilin and Rab27a shRNA impaired the secretion of CD63-positive exosomes compared to the pLKO control (supplementary material Fig. S2C). However, in the case of sortilin shRNA, we observed that a population of exosomes remained, as judged by the presence of flotillin within the exosome fractions (supplementary material Fig. S2C). These results show that the number of ILVs observed per MVB was significantly increased after both sortilin and Rab27a knockdown. In contrast, the size of ILVs in the MVBs was much larger with sortilin shRNA than those observed with Rab27a shRNA, highlighting the different role(s) played by sortilin and Rab27a in MVB maturation and exosome secretion. These results are the first documented evidence of the importance of sortilin in controlling exosome secretion from A549 cells.

To investigate the capacity of cells to take up exosomes, we utilized human umbilical vein endothelial cells (HUVECs) as target cells. First, we assessed whether exosomes could be incorporated into the target cell. We collected A549 conditioned cell culture medium after 48 h and purified the exosome fraction. The exosomes were then labeled with a low-density-lipoprotein dye (Dil), and HUVECs that had been transfected with turbo-green were incubated with the labeled exosomes for 24 h. We performed a z-stack analysis and found that exosomes were incorporated into HUVECs (supplementary material Fig. S3A). In a more detailed study, we examined the fate of exosomes. Purified exosomes were labeled with the dye PKH67 and added to HUVECs. After incubating HUVECs with exosomes prepared from the pLKO control for 3 h at 37°C, cells were fixed and analyzed by confocal microscopy. We observed efficient internalization of pLKO exosomes by HUVECs (Fig. 4A,B), which could be blocked by pre-treating HUVECs with the dynamin inhibitor dynasore or by keeping the cells at 4°C during the incubation (Fig. 4A,B). When HUVECs were incubated with exosomes derived from sortilin-knockdown cells, the level of exosome uptake was significantly reduced. Likewise, Rab27a knockdown affected exosome uptake into HUVECs (Fig. 4A,B). We next analyzed whether we could observe the transfer of sortilin-containing exosomes to HUVECs. To facilitate this analysis and to improve the sensitivity of the assay, we chose to perform a 2-day pulse-chase experiment (Fig. 4C). In brief, sortilin was overexpressed in A549 cells by transient transfection and, at 18 h post-transfection, a pulse-labeling was performed for 5 h followed by a 24-h chase. After the 24-h chase, we harvested the labeled A549 exosomes from the culture medium by differential centrifugation and incubated HUVECs with the labeled exosomes for 24 h. Remarkably, after immunoprecipitation with the ectodomain-specific sortilin antibody, we pulled-down two labeled SDS-resistant products in A549 cells expressing sortilin, which migrated at between ∼350–400 kDa (Fig. 4D, A549 WCL lane). Interestingly, the higher product (∼400 kDa) was secreted from A549 in exosomes (Fig. 4D, A549 Exo lane) and transferred to HUVECs (Fig. 4D, HUVEC WCL lane). We performed a control experiment to rule out the possibility that the higher molecular mass product was formed artificially during immunoprecipitation. We confirmed that this product was unable to form in a mixed immunoprecipitate of both WCL and exosomes (Fig. 4E). Given that the size of these products was much larger than the predicted size of sortilin (95 kDa) secreted from cells, these results suggested that sortilin was part of a high-molecular-mass complex. Sortilin is a multi-functional receptor that interacts with a number of ligands including the apoptotic death receptor p75 neurotrophin (p75^NTR, also known as NGFR) and the cell survival receptor, tropomyosin-related kinase (Trk) receptor. We screened a number of possible interacting partners by subjecting metabolic-labeled A549 cells overexpressing sortilin to co-immunoprecipitation using antibodies against sortilin, EGFR, TrkB and p75^NTR. We found that sortilin physically interacts with both EGFR and TrkB but not with p75^NTR in A549 exosomes (Fig. 4F). These results were confirmed when we performed two successive immunoprecipitations. After the first precipitation with anti-TrkB antibody under native conditions, the precipitates were released and exposed to re-immunoprecipitation with anti-EGFR antiserum followed by western blotting with anti-sortilin antibody, indicating the presence of the complex (Fig. 4G). When the first precipitation was carried out with anti-TrkB antiserum followed by precipitation with p75^NTR or a non-related serum, no products were detected (Fig. 4G). Taken together, these data point to a trimeric complex comprising sortilin, TrkB and EGFR, which has a predicted molecular mass of ∼360–410 kDa depending on the presence of truncated (95 kDa) or full-length TrkB (145 kDa), respectively. Given that cross-talk between EGFR and TrkB has been reported previously (Qiu et al., 2006), their detection in exosomes reinforces their potential role in the cellular communication of cancer cells with the microenvironment. Hence, these results, demonstrating the presence of a new trimeric sortilin–TrkB–EGFR complex in exosomes that can be transferred to target cells, represent an important finding.

To elucidate whether sortilin-containing exosomes could be transferred to endothelial cells and were functionally active in the recipient cells, we investigated the effect of exosomes derived from A549 cells expressing sortilin on HUVECs. There is growing evidence that exosomes mediate cell-to-cell communication in a variety of biological processes. Hence, in addition to direct cell-to-cell interaction or secretion of active molecules, they are now considered another class of signal mediators. Exosomes contain over 400 proteins, some of which are involved in metabolism or linked to the cytoskeleton, but, most importantly, they also contain proteins involved in cell signaling pathways. The influence of these mediators on target cell signaling pathways largely remains to be elucidated (Mathivanan and Simpson, 2009). The EGFR signaling cascade modulates many pathways involved in cell proliferation, survival, adhesion, migration and differentiation (Citri and Yarden, 2006). The EGFR signaling pathway is connected to three major mitogen-activated protein kinase (MAPK)-pathways and the phosphoinositide 3-kinase (PI3K)/AKT survival pathway. The MAPK pathway consists of the ERK1/2 module, the p38 MAPK module, and the JNK module. To gain insight into the activities of these MAPK modules in HUVECs exposed to sortilin exosomes derived from A549 cells, we used arrays to analyze for the presence of phosphorylated MAPKs to explore the effect of A549 exosomes containing sortilin on signaling pathways in the target cell (Fig. 5A,B). We observed that exposure of HUVECs to A549 sortilin-containing exosomes resulted in a 3–50-fold induction (i.e. phosphorylation elevation) of several proteins involved in cell growth and differentiation such as ERK1/2 (also known as MAPK3 and MAPK1, respectively), p38δ (also known as MAPK13), AKT1/2, Hsp27, RSK1/2, GSK3α/β and JNK2 (Fig. 5A,B). Conversely, we observed a reduction in phosphorylation of all MAPK proteins analyzed when HUVECs were exposed to A549 sortilin-depleted exosomes (Fig. 5A,B). To explore further the effect of exosomes from the phenotypic cells used in this study, we chose to study a time course of several phosphorylated MAPK proteins analyzed in the array of HUVECs. We observed the phosphorylation of several kinases including ERK1/2, AKT, GAB1 and EGFR in HUVECs exposed to pLKO exosomes for 90 min (supplementary material Fig. S4A,B). We showed above that HUVECs pre-treated with dynasore or incubated at 4°C were unable to internalize exosomes (Fig. 4A,B). However, although it was possible that the signaling pathways could be initiated from the cell surface by the binding of exosomes to the surface of the target cell, we showed that cell surface binding of exosomes to the target cell was not sufficient to transmit a signal to the target cell, as judged by reduced phosphorylation of ERK1/2, AKT, GAB1 and EGFR (supplementary material Fig. S4A,B). Likewise, phosphorylation of the MAPKs was reduced in HUVECs that were incubated with both sortilin shRNA exosomes and Rab27a shRNA exosomes (supplementary material Fig. S4A,B).

Exosomes have been associated with tumor-promoting activities. Tumor exosomes display angiogenic properties, and hypoxic tumor cells enhance angiogenesis by exosome secretion (Park et al., 2010). Exosomes have also been implicated in the activation of relevant oncogenic pathways by secretion of EGFR signaling ligands, such as amphiregulin, which is displayed on the surface of exosomes in a signaling-competent orientation (Higginbotham et al., 2011). The presence on the surface of exosomes of active ligands for receptors, such as EGFR, which is overexpressed in many cancers, also suggests the possible involvement of exosomes in providing a favorable environment to circulating tumor cells to allow the development of metastasis (Demory Beckler et al., 2013). Cell migration is essential for many physiological processes including angiogenesis. As a functional readout of exosome activity, we assessed the effect of exosomes produced from A549 cells on recipient cancer cell invasion, a hallmark of malignancy. Using a Boyden chamber assay, we observed that pLKO-A549 (control) exosomes efficiently attracted HUVECs though the transwell, as judged by a high level of fluorescence (Fig. 6A,B). Next, we investigated the effect of sortilin-depleted A549 exosomes upon HUVEC invasion in the transwell assay. Sortilin depletion resulted in nearly a 5-fold reduction of HUVECs migrating through the chamber compared to pLKO control. Similar results were obtained for both sortilin shRNAs (shRNA no. 1, P = 0.001; shRNA no. 2, P = 0.003) (Fig. 6A,B). In the same way, we found that Rab27a-knockdown exosomes blocked HUVEC invasion in the transwell at a similar level (shRNA no. 1, P = 0.0009; shRNA no. 2, P = 0.005) to that observed for sortilin knockdown (Fig. 6A,B). We found we could significantly recover the block of HUVEC invasion imposed by sortilin-knockdown exosomes after adding back purified sortilin-containing A549 exosomes at 8 h (P = 0.01) (Fig. 6A,B). Exosomal induction of proteins promoting angiogenesis in target cells was corroborated by the exposure of HUVECs to sortilin-containing A549 exosomes. Several relevant proteins [endothelin-1, IL-8, thrombospondin-2, urokinase-type plasminogen activator (uPA) and VEGF] were upregulated in HUVECs upon exposure to sortilin-containing A549 exosomes (Fig. 6C). In contrast, these same proteins were not elevated in HUVECs upon exposure to sortilin-depleted A549 exosomes (Fig. 6C). Furthermore, we observed upregulation of serpin E1 [also known as plasminogen activator inhibitor (PAI)-1] in HUVECs treated with sortilin-depleted A549 exosomes (Fig. 6C). Similar results were obtained when Rab27a-depleted A549 exosomes were used (Fig. 6C).

A role for the multifaceted receptor sortilin in exosome release and transfer is supported here by three major findings that advance our understanding of exosome release and communication with the vascular microenvironment. First, we have shown that sortilin uses exosomes for secretion and that sortilin is a key component required for exosome secretion and function (Fig. 7). Second, we identified a new heteromeric receptor complex composed of TrkB, EGFR and sortilin (herein named the TES complex) in exosomes, which mediates communication and signaling events between A549 and endothelial cells. Third, our data suggest a paracrine function for sortilin and its partners in exosome transfer and the control of the microenvironment.

The neurotensin-3 receptor sortilin plays a multitude of roles, as a receptor or co-receptor (Nykjaer et al., 2004), in protein trafficking (from the TGN to the plasma membrane and endosomes) (Chen et al., 2005) and in protein trafficking to the lysosome for protein degradation (Canuel et al., 2008). Our observation that the knockdown of sortilin affects both the number and size of ILVs found in MVBs, and their secretion into the culture medium, indicates that sortilin has a role in this process, which was previously unknown. Furthermore, we showed that the size of ILVs in the MVBs observed with sortilin shRNA is significantly larger than those observed with Rab27a shRNA. This might indicate that sortilin controls the size of exosomes and could be involved in a scission event when ILVs are formed in the MVB. There are three major steps in the formation of ILVs in the MVB where sortilin could participate in their biogenesis. The first is cargo recognition and sorting, the second is membrane budding and the third is vesicle scission from the endosome membrane. The MVB pathway is well characterized in yeast and has been shown to require the vacuolar protein sorting (vps) class E genes that assemble into five endosomal sorting complex required for transport (ESCRT) complexes (ESCRT-0, ESCRT-I, ESCRT-II, ESCRT-III and Vps4) (Henne et al., 2011). The ESCRT pathway is necessary for the final membrane scission step generating the MVBs. It has been previously reported that the HRS gene of the ESCRT-0 complex might be involved in exosome formation and secretion (Tamai et al., 2010). A recent study has shown that silencing of genes encoding components of the ESCRT-0 complex (HRS, STAM1 or TSG101) reduces exosome secretion and alters the size and/or protein composition of the ILVs (Colombo et al., 2013). However, it still remains to be determined whether this class of proteins is important for exosome function and microenvironment control. By contrast, Rab27a plays a different role in exosome secretion; it is well documented that it is important for the docking of the MVB at the plasma membrane (Ostrowski et al., 2010). We observed that exosomes derived from sortilin- or Rab27a-depleted cells are not readily internalized by HUVECs (Fig. 4A,B). These results could be explained by the differences observed in the exosome content from exosomes derived after transfection with sortilin shRNA and Rab27a shRNA (supplementary material Fig. S2C). The sucrose density gradient analysis revealed that there were reduced protein levels of the exosome marker CD63 when sortilin or Rab27a was depleted. At the same time, the marker flotillin was maintained in sortilin-depleted exosomes, whereas it was lost upon Rab27a knockdown. Interestingly, CD63 is part of a large tetraspanin superfamily that, at the plasma membrane, can interact with cell adhesion molecules such as integrins, which are involved in cell adhesion, motility and survival (Berditchevski, 2001; Pols and Klumperman, 2009). The ability of CD63 exosomes to interact with cell adhesion molecules at the plasma membrane of a target cell is not well characterized. However, it is clear that sortilin is important for exosome secretion and that the knockdown of sortilin modifies the exosome population (CD63), which might affect exosome internalization.

The role of sortilin in human neural cells was initially described as an intracellular transport protein for neurotrophins and proneurotrophins (Chen et al., 2005) and, subsequently, sortilin was shown to associate with Trk receptors (including TrkB) to enhance anterograde transport, neuronal survival and neurotrophin signaling (Vaegter et al., 2011). TrkB can bind BDNF and neurotensin-4 and -5, and is believed to be a potential target in several cancers. Although the actual mechanism of action of TrkB remains elusive, a number of studies have shown that TrkB activates PI3K/AKT and MAPK (MEK to ERK) signaling (Ho et al., 2002; Kim et al., 2004). Similarly, we observed an induction of PI3K/AKT signaling pathways in HUVECs exposed to exosomes containing sortilin and its partners EGFR and TrkB. In contrast, these same signaling pathways could be blocked when we used sortilin-depleted exosomes. These results suggest that exosomes containing the TES complex might allow communication with the microenvironment to promote cell survival and the angiogenesis process.

We observed, using a cell invasion assay, that HUVECs invading a transwell are significantly blocked in the presence of sortilin-depleted exosomes and this block is restored when supplemented with sortilin-containing exosomes. These results are not surprising, as sortilin has previously been shown to play a role in the migration (Martin et al., 2003) and release of chemokines in microglia (Dicou et al., 2004), and in dendritic spine maturation in the cerebral cortex (Gandou et al., 2010). At the same time, we demonstrate that angiogenic factors such as endothelin-1, IL-8, thrombospondin-2 and VEGF are significantly affected when we deplete sortilin from exosomes. Aberrant expression of endothelin-1, or overexpression of endothelin receptors is now recognized as a common mechanism (both autocrine and paracrine) that contributes to tumor initiation and the progression of various solid tumors, including ovarian, prostate, colon, breast, bladder and lung cancers (Rosanò et al., 2013). Interestingly, we observed a switch in the angiogenic factors with a significant elevation of serpin E1 and downregulation of uPA when HUVECs were exposed to sortilin- or Rab27a-depleted exosomes. PAI-1 has multiple roles, including cell de-adhesion, proliferation and apoptosis and cell signaling, indicating that PAI-1 expression in the tumor enhances cancer progression. Elevated levels of both uPA and PAI-1 are associated with a poor prognosis in many cancers. However, many studies have suggested that PAI-1 at a high level prevents tumor growth through inhibiting angiogenesis (McMahon et al., 2001). Taken together, our data demonstrate that the downregulation of sortilin expression in exosomes from A549 cells impairs the angiogenesis process and results in elevated levels of PAI-1 and a reduction in uPA, inhibiting the angiogenesis process.

There are unprecedented reports suggesting that disruption of the Vps10 domain proteins, including sortilin, contribute to human diseases from neurodegeneration to cancer (Wilson et al., 2014). Furthermore, the expression levels of neurotrophic factors, including their receptors, are clearly elevated in cancer, and this elevation is potentially an important factor in the angiogenesis and metastasis process. We show for the first time, a new role for sortilin in the release and the assembly of a tyrosine kinase receptor complex (TES) in lung cancer cells. Taken together, these studies and our data suggest a paracrine function for sortilin and its partners in exosome transfer and the control of the microenvironment. This new complex containing sortilin could act as a molecular switch in cancer progression by promoting angiogenesis.

Cell culture reagents were from Invitrogen or Lonza. The Platinum® Quantitative PCR SuperMix-UDG was from Invitrogen, and the EasyTag Express Protein Labeling L-[³⁵S] methionine/cysteine mix were from PerkinElmer. All other chemicals were from Sigma, Ozyme and BD Bioscience. Rabbit polyclonal antisera recognizing sortilin was kindly provided by Claus Munck Petersen (Aahrus University, Denmark). Antisera used in FACS analysis were: mouse anti-human-CD63 (BD Bioscience) antibodies, for coupling to beads, and FITC-conjugated anti-CD63 and anti-CD81 (Biolegend) and rat anti-EGFR antibodies (Serotec). Antisera used for immunoblotting were: mouse anti-ECD-Sortilin (BD Bioscience); rabbit anti-p75^NTR and goat anti-ICD-sortilin (Santa Cruz Biotechnology); mouse anti-CD63 (Serotec); rabbit anti-Rab27a (Sigma); rabbit anti-EEA1 (Sigma); rabbit anti-flotillin and anti-TrkB (Cell Signaling); mouse anti-actin (Sigma); mouse anti-V5 and anti-GFP (Life Technologies); and horseradish peroxidase (HRP)-conjugated secondary antibodies (Dako).

Full-length human sortilin cDNA, V5- and GFP-tagged fusion proteins cloned in pEF6/V5-His TOPO vector or the pcDNA3.1/CTGFP TOPO vector were generously provided by Tae-Wan Kim (Columbia University Medical Center, NY). Lentiviral vector pCT-CD63-GFP expressing a CD63–GFP fusion protein was obtained from BioCat.

HUVECs (purchased from Lonza) were cultured in endothelial cell growth medium containing 2% fetal bovine serum (FBS) (EGM-2; Lonza) and growth factors. Human lung carcinoma cell line (A549) and human embryonic kidney 293 cells (HEK293T) were purchased from ATCC and maintained in DMEM GlutaMAX™ (Invitrogen) supplemented with 10% FBS and 1% non-essential amino acids. All cells were cultured in a humidified incubator set at 5% CO₂ and 37°C. For chemical compound treatments, in brief, A549 cells (2.5×10⁵) were transfected or not in six-well plates and treated with 2.5 µg/ml tunicamycin, 1 µg/ml brefeldin A (BFA), 15 nM dimethyl amiloride (DMA) or 2 µM Monensin (Mon) for 24 h.

For both transient and stable transfection, cells were transfected using JetPei transfection reagent (Polyplus transfection, Ozyme). Lentivirus-mediated RNA interference was used to generate sortilin-knockdown and Rab27a-knockdown stable cell lines using the previously described procedure (Magnaudeix et al., 2013). The shRNA sequences were as follows: for sortilin, TRCN0000005295 (5′-CCGGCCAGTGTACTTTACCAATATACTCGAGTATATTGGTAAAGTACACTGGTTTTT-3′) and TRCN0000005296 (5′-CCGGCGGATCAGTTAAGTGAAGAAACTCGAGTTTCTTCACTTAACTGATCCGTTTTT-3′); and for Rab27a, TRCN0000005295 (5′-CCGGCCAGTGTACTTTACCAATATACTCGAGTATATTGGTAAAGTACACTGGTTTTT-3′) and TRCN0000005296 (5′-CCGGCGGATCAGTTAAGTGAAGAAACTCGAGTTTCTTCACTTAACTGATCCGTTTTT-3′).

The Qiagen RNeasy kit was used to isolate total RNA from cells at day 15 after lentiviral infection. Single-stranded cDNA was prepared using the high capacity cDNA reverse transcription kit according to the manufacturer's protocol (Applied Biosystems). The reaction was stopped by incubation at 95°C for 5 min. Approximately of 100 ng cDNA was used for each PCR reaction, performed with TaqMan (Applied Biosystems) on an ABI Step One Plus real-time thermal cycler (Applied Biosystems). PCR primers for sortilin, Rab27a and GAPDH were designed and used for the PCR amplification with Taq DNA polymerase (Roche Diagnostics).

For A549–HUVEC exosome transfer, A549 cells grown in 10-cm² Petri dishes were transfected with full-length sortilin for 18 h prior to metabolic labeling; for the pulse-chase time course experiments, cells were grown in six-well dishes. Cells were starved for 20 min in methionine- and cysteine-free DMEM (Invitrogen) with 2 mM glutamine, and then metabolically labeled with the same medium containing 20 µCi/ml of [³⁵S]methionine and [³⁵S]cysteine (1175 Ci/mmol, 11.9 mCi/ml) protein labeling mix for 60 min (time course) or for 300 min (exosome transfer). Cells were rinsed twice with PBS, and fresh DMEM Glutamax medium was added or solubilized in Triton X-100 immunoprecipitation buffer (Wilson et al., 2000; Wilson et al., 2008) containing 1 mM PMSF and 1 mM non-radioactive methionine and cysteine for timecourse experiments at t = 0 h. For pulse-chase time course experiments, cells were incubated for 6 and 24 h, and both cell lysates and medium were harvested. For exosome transfer, cells were then incubated in fresh complete DMEM Glutamax medium for 24 h and metabolically labeled exosomes were harvested from the medium samples as detailed below. Labeled exosomes (25 µg/ml) were added to the medium of cultured endothelial cells HUVECs for a further 24 h before harvesting both the cell lysates and medium. Samples were precleared and immunoprecipitated with specific antisera prior to SDS-PAGE and autoradiography or phosphoimaging.

Cells grown on glass coverslips (14-mm diameter, Menzel-Gläser, VWR, Fontenay-sous-Bois, France) were fixed using methanol or 2% paraformaldehyde for 10 min, washed with PBS solution containing 1% (w/v) BSA and blocked with PBS with BSA solution for 30 min followed by staining with the indicated primary antibody at room temperature for 2 h, and, subsequently, either with anti-rabbit-IgG Alexa-Fluor-594-conjugated or anti-mouse-IgG Alexa-Fluor-488-conjugated antibodies (1∶200; Invitrogen) for 45 min at room temperature. For PKH67 (Sigma) staining, exosomes were incubated for 5 min at room temperature. The staining reaction was stopped after 5 min with exosome-free FCS. Exosomes were then washed in PBS and pelleted by ultracentrifugation (100,000 g, 1 h). For DiI (1,1′-dioctadecyl-3,3,3′3′-tetramethylindocarbocyanine perchlorate) (Molecular Probes, France) staining of A549 cells transfected with CD63–GFP, cells were incubated with 1 µM DiI for 2 h at 37°C. Fluorescence images were obtained using a charge-coupled device camera (Photonic Science) driven by Visiolab 2000 software (Biocom) or by using an epifluorescence microscope (Zeiss Axiovert-200) equipped with a laser-scanning confocal imaging system (Zeiss LSM 510 META).

Anti-human-CD63 antibody was coupled to 4.5-µm aldehyde-sulfate beads (Invitrogen) by incubating 35 µg of antibody with 1×10⁸ beads followed by blocking of remaining activated groups with PBS containing 4% BSA. Cell culture supernatants were cleared by centrifugation at 18,000 g (for 20 min). Cleared supernatants (100 µl) were incubated with 20,000 anti-CD63-coupled beads overnight at room temperature with 200 rpm horizontal shaking. Beads were washed twice in PBS containing 2% BSA and were incubated with FITC-conjugated anti-CD81 (Biolegend) or anti-EGFR antibody (Serotec) for 1 h at room temperature. Beads were washed twice in PBS containing 2% BSA. Beads were acquired on a FACSCalibur (BD) and data were analyzed with CellQuest software (BD). The threshold of negative staining was obtained with beads incubated with unconditioned medium, and for each culture condition, the amount of exosomes was calculated in arbitrary units (AU) as a percentage of positive beads.

Exosomes were isolated from the cell culture medium by differential centrifugation. Briefly, cells were cultured for the time periods as indicated in figure legends in exosome-free medium and centrifuged at 300 g for 5 min at 4°C to remove cell debris. Supernatants were further centrifuged at 16,500 g for 30 min at 4°C. Next, exosomes were pelleted by ultracentrifugation at 100,000 g for 2 h at 4°C, washed in PBS and re-centrifuged at 100,000 g for 2 h at 4°C. For each exosome preparation, the concentration of total proteins was quantified with the Bio-Rad protein assay kit. Exosomes were then resuspended in an appropriate volume of PBS for further analysis or used for protein isolation.

Exosome particles harvested from culture medium were quantified using either the CD63 or CD81 ELISA kit (System Biosciences, Ozyme, France). Exosome pellets were resuspended in 100 µl Exosome Binding buffer and analyzed according to the manufacturer's instructions.

To investigate phosphorylated kinases or the angiogenesis pathway we used Proteome Profiler™ arrays (R&D Systems, Wiesbaden, Germany). In brief, cells grown in 10-cm² Petri dishes were either transfected or not with full-length sortilin for 18 h and the medium was refreshed with fresh complete DMEM Glutamax medium for a further 24 h. Exosomes were purified by differential ultracentrifugation and were added (25 µg/ml) or not to the medium of cultured endothelial cells HUVECs for a further 24 h before harvesting the cell lysates. Samples were lysed in RIPA buffer supplemented with protease inhibitor cocktail. Total protein from HUVEC lysates (300 µg) was analyzed using the human phospho-MAPK or the human angiogenesis antibody array (R&D Systems). The arrays were scanned and quantified with ImageJ software (NIH, Bethesda, MD). All of the arrays were performed according to the manufacturer's recommendations.

Invasion assays were performed in a BD Biocoat Matrigel invasion chamber with an 8-µm diameter pore size membrane and a thin layer of Matrigel, in a 24-well plate. Inserts were rehydrated for 2 h in EGM-2 medium. After detachment of confluent cells with trypsin and EDTA, HUVECs (5×10⁴) were seeded in the upper surface of the transwell plates with A549, and sortilin- or Rab27a-knockdown A549 cells were added to the lower chamber and cultured for 24 h at 37°C under 5% CO₂. After incubation, the lower chamber was treated with 50 µM calcein AM fluorescent dye for 30 min at 37°C and under 5% CO₂. Non-invading cells in the upper part of the insert were carefully removed. Three independent assays were performed, and cells were seeded in triplicate for each cell line. Invading cells were measured using a Twinkle LB 970 Microplate Fluorometer (Berthold Technologies) and images were obtained using a fluorescence microscope M2FLIII (Leica). Results are presented as mean±s.d. for each sample, and invasion levels obtained were compared with the pKLO control cell line.

Treatments, relative fluorescence intensities, antibody arrays and western blotting experiments were compared with control using Statview software (Statview v.5.0). Data shown are representative of at least three independent experiments. Error bars represent by s.e.m. Results were analyzed for statistical significance by one-way ANOVA or Student's t-test. P<0.05 was considered as significant, with actual values represented by asterisks (*P<0.05; **P<0.01; ***P<0.001).

We are grateful to all our colleagues who have contributed their time and materials towards this paper. We especially thank the technical support of the electron microscopy service of CHU Poitiers and Imaging Cytometry Platform of the University of Limoges.