ROR1 sustains multivesicular endosomes by interacting with HRS and STAM1


 The receptor tyrosine kinase-like orphan receptor 1 (ROR1) regulates caveolae formation and caveolae-dependent endocytosis by interacting with caveolae components, which in turn sustains pro-survival signaling toward AKT from multiple RTKs, including EGFR, and MET. We report here a novel function of ROR1 as a scaffold for HRS and STAM1, two essential components of ESCRT-0. The present results show that ROR1 facilitates interactions of HRS and STAM1, thereby preventing the lysosomal degradation of HRS. Furthermore, interaction of ROR1 with STAM1 was found to be required to sustain binding of ROR1 to HRS as well as HRS subcellular localization. Additionally, ROR1 localized in both the limiting membrane and intraluminal vesicles (ILVs) of Rab5-induced multivesicular endosomes (MVEs) containing HRS, CD63, and EEA1 was found to regulate the formation of Rab5-induced MVEs by an association with the GTP-bound form of Rab5 in cancer cells. Notably, ROR1 depletion inhibits CD63-positive MVEs formation and reduces exosomes release. Our findings provide the first evidence that the onco-embryonic antigen receptor ROR1 regulates exosome biogenesis via MVE formation in cancer cells.


INTRODUCTION
Endosomes have critical roles in the highly dynamic transport mechanism that operates between the plasma membrane and lysosomes via the biosynthetic secretory pathway 1, 2, 3 .
Early endosomes mature into late endosomes then accumulate intraluminal vesicles (ILVs) in the lumen, and are thereafter referred to as multivesicular endosomes (MVEs) or multivesicular body (MVBs) 4,5,6,7 . ILVs formed by inward budding of early endosomal membrane sequester proteins, lipids, and cytosol are specifically sorted. In most cells, the main fate of MVBs is to fuse with lysosomes for degradation of their contents, while an alternative function is exocytic fusion with the plasma membrane leading to release of ILVs as exosomes into extracellular space 8,9 . Exosomes are generally defined as secreted extracellular vesicles (EVs) and known to play key roles in cell-to-cell communication 10,11 .
However, knowledge regarding the mechanisms that control the alternative fates of ILVs to degradation or secretion is largely unknown. Thus, an important unanswered question is how are MVBs that fuse with lysosomes different from those that fuse with the plasma membrane to release exosomes, which the present study attempted to address.
The endosomal sorting complexes required for transport (ESCRT) components involved in MVB and ILV biogenesis 12,13 . This is an intricate protein machinery composed of four separate protein ESCRTs (-0, -I, -II, -III) that function cooperatively to facilitate MVE formation, vesicle budding, and protein cargo sorting 14 . ESCRT-0, -I, and -II contain ubiquitin-binding subunits that function to sort ubiquitylated membrane proteins to specific domains of endosomes and into endosomal invaginations, while ESCRT-III subsequently drives vesicle scission 15 . ESCRT-0 contains HRS (hepatocyte growth factor-regulated tyrosine kinase substrate, gene symbol HGS), which recognizes monoubiquitinated cargo proteins and is associated with STAM (signal transducing adaptor molecule, another ESCRT-0 component) 12 . Evidence is accumulating indicating that sorting of proteins into ILVs can 4 also occur in a manner independent of ubiquitination. The transferrin receptor (TfR) in reticulocytes generally undergoes exosome secretion, though does not become ubiquitylated 16 . It has also been shown that ALIX is involved in exosome biogenesis and exosomal sorting of syndecans in a manner independent of ubiquitin 17 . However, some studies have found that the ESCRT-0 members HRS and STAM are required for exosome secretion, as demonstrated by decreased exosome secretion following HRS or STAM inhibition in various cell types 18,19,20,21 . Since exosomes correspond to ILVs, the same mechanisms are thought to be involved in their biogenesis. Nevertheless, the precise details regarding MVE/MVB formation for exosomes remain unclear.
We previously reported that the receptor tyrosine kinase (RTK)-like orphan receptor 1 (ROR1) is a transcriptional target of the lineage-survival oncogene NKX2-1/TTF-1 in lung adenocarcinomas 22,23 . In addition to its kinase-dependent role, ROR1 serves as a scaffold protein to facilitate interaction between CAV1 and CAVIN1, and consequently maintains caveolae formation in lung adenocarcinoma cells, which in turn sustains pro-survival signaling toward AKT from multiple RTKs, including EGFR and MET 24 . Therefore, ROR1 is an attractive target for overcoming EGFR-TKI resistance caused by various mechanisms such as EGFR double mutation and bypass signaling from other RTKs. Recently, we also showed that ROR1 has a novel scaffold function that is vital for efficient caveolae-dependent endocytosis 25 . CAVIN3 was demonstrated to bind with ROR1 at a site distinct from site for CAV1 and CAVIN1, a function necessary for proper subcellular localization of CAVIN3 and caveolae-dependent endocytosis. Additionally, evidence demonstrating a mechanistic link of ROR1-CAVIN3 interaction and consequential caveolae trafficking with RTK-mediated prosurvival signaling towards AKT in early endosomes was obtained.
The present findings revealed a novel function of ROR1 as a scaffold for HRS and STAM1. ROR1 was shown to facilitate interactions of HRS and STAM1, thereby preventing 5 lysosomal degradation of HRS. Furthermore, we also found that interaction of ROR1 with STAM1 is required to sustain binding of ROR1 to HRS as well as HRS subcellular localization. Together, these results show that ROR1 regulates formation of Rab5-induced MVEs by an association with the GTP-bound form of Rab5. Additionally, ROR1 depletion inhibits CD63-positive MVEs formation and reduces exosome release in cancer cells.

ROR1 knockdown reduced HRS and STAM1 proteins, and altered late endosome localization.
Based on our previous results showing that ROR1 sustains caveolae formation and caveolaedependent endocytosis 24,25 , we speculated that ROR1 may be involved in cell membrane organization and dynamics in cancer cells. Initially, we analyzed the effects of siROR1 treatment on membrane dynamics and trafficking using electron microscopy, which revealed a large number of abnormal vesicles in the cytoplasm of NCI-H1975 cells (Fig. 1a). Thus, we examined whether ROR1 was required for Golgi or ESCRT protein. Western blot (WB) analysis revealed significantly decreased expressions of HRS and STAM1 proteins, but not of TGN38 or GM130, in siROR1-treated NCI-H1975 cells (Fig. 1b), after which siROR1induced reduction of HRS and STAM1 was confirmed in three lung cancer cell lines, NCI-H1975, PC-9, and SK-LC-5 (Fig. 1c). Following siROR1 transfection, HRS expression was decreased, while that of STAM1 gradually decreased though remained readily detectable for up to 24 hours (Fig. 1d). These findings led us to speculate that abnormal vesicles in cytoplasm induced by siROR1 may be caused by impairment of the ESCRT-0 machinery. Consistent with previous investigations 4,26,27 , HRS-and STAM1-depleted cells showed typically large and ring-shaped vacuoles with one or two intralumenal vesicles, similar to that seen in NCI-H1975 cells following ROR1 knockdown ( Supplementary Fig. 1). We also found that the subcellular localization of Rab7 (late endosome marker) and CD63 (exosome marker) was markedly altered in siROR1-treated cells ( Fig. 1e and 1f). The present findings clearly indicate that ROR1 is required to sustain HRS and STAM1 expression, as well as resultant proper endosome formation in cancer cells.

7
To gain more in-depth insight regarding the involvement of ROR1 in this process, we next investigated whether ROR1 interacts with HRS and/or STAM1. Immunoprecipitation-western blotting (IP-WB) analysis using octylglucoside 28 as a detergent revealed an interaction of endogenous ROR1 with endogenous HRS and STAM1 in PC-9 and NCI-H1975 lung cancer cells (Fig. 2a). IP-WB analysis using A549 cells transfected with an ROR1-GFP expression vector verified interactions of ROR1 with endogenous HRS and STAM1 (Fig. 2b). Pull-down assay findings obtained using purified glutathione S-transferase (GST)-tagged ROR1 protein as well as lysates of NCI-H1975 cells also demonstrated an interaction of ROR1 with HRS or STAM1 (Fig. 2c). Additionally, immunofluorescence analysis revealed partial colocalization of punctate signals of ROR1-GFP with those of HRS and STAM1 (Fig. 2d).

ROR1 localized in both the limiting membrane and ILVs of Rab5-induced MVEs
To determine whether ROR1 is present in MVEs, we used the constitutively active mutant Rab5 Q79L , which is known to induce formation of enlarged early endosomes by promotion of their homotypic fusion 29 . As expected, co-expression of mCherry-Rab5 Q79L with ROR1-GFP resulted in the presence of both proteins in enlarged early endosomes. Furthermore, ROR1-GFP was found colocalized with HRS, CD63, and EEA1 ( Fig. 3a and Supplementary Fig. 2).
In addition, partial colocalization of ROR1-GFP and mCherry-Rab5 Q79L was verified at a much higher resolution in super-resolution structured illumination microscopy (SIM) 30 findings (Fig. 3b), while line intensity profiles of individual endosomes clearly showed ROR1 localization in the lumen and membrane of Rab5-induced endosomes (Fig. 3c). Electron microscopy was also employed and confirmed those immunofluorescence data, with the presence of ROR1-GFP showing an association with both the limiting membrane and ILVs of Rab5-induced MVEs in immunoelectron microscopy analysis (Fig. 3d). An ascorbate peroxidase (APEX2) system was then employed to better visualize the presence of ROR1 in 8   MVEs, as APEX2 has improved enzyme activity for biotinylating proteins within 20 nm and   can increase electron microscopy contrast by catalyzing 3,3′-diaminobenzidine (DAB) polymerization 31,32,33 . Biotinylated proteins were predominantly detected in ROR1-APEX2expressed HeLa cells ( Supplementary Fig. 3a), with those results confirmed by WB analysis with a streptavidin-horseradish peroxided (HRP) process ( Supplementary Fig. 3b). ROR1-APEX2 was clearly localized in both the limiting membrane and ILVs of endogenous MVEs, findings consistent with our immunofluorescence results (Fig. 3e).

ROR1 required for prevention of HRS being routed to lysosomes
HRS and STAM become stabilized by forming a complex via their coiled-coil regions and HRS-STAM interaction is required to prevent degradation of each protein 34,35 . Interaction of HRS with STAM leads to ESCRT-0, which has a crucial role for initiation of the ESCRT pathway 36 . Therefore, we performed ROR1 knockdown and examined its effect on the interaction between HRS and STAM1 by taking advantage of the delayed reduction of HRS and STAM1 that occurs after siROR1 transfection. A significantly decreased association of HRS with STAM1 was observed in PC-9 and NCI-H1975 cells at 24 hours after siROR1 treatment (Fig. 4a). In addition, subcellular localization of HRS was markedly altered at that time point, resulting in significant colocalization with the endosome/lysosome marker LAMP1 (Fig. 4b). These findings indicated that ROR1 has a novel function as an indispensable scaffold protein of HRS and STAM1, thus preventing lysosomal HRS degradation and sustaining HRS expression.

ROR1-STAM1 interaction required for HRS-STAM1 binding and HRS subcellular localization
An interesting finding was inhibition of the interaction of ROR1 with HRS in NCI-H1975 9 cells caused by STAM1 knockdown at 24 hours after siSTAM1 transfection under conditions that did not have effects on HRS expression (Fig. 5a). Consistent with that result, two-color immunofluorescence analysis revealed significant loss of ROR1-GFP and HRS colocalization ( Fig. 5b). Next, we used various ROR1 deletion mutants to examine the STAM1 binding regions. IP-WB analysis revealed a requirement of the C-terminal proline-rich domain of ROR1 for its binding to STAM1 (Fig. 5c). We also investigated whether the STAM1-binding region of ROR1 has effects on formation of an HRS and STAM1 complex, and/or HRS subcellular localization. NCI-H1975 cells stably reconstituted with either siRNA-resistant ROR1-WTm or ROR1-Pm were subjected to treatment with siROR1, then IP-WB analysis was performed. Different than ROR1-WTm, STAM1-binding deficient ROR1-Pm did not sustain the interaction of HRS with STAM1 (Fig. 5d). On the other hand, similar to treatment with siROR1, HRS subcellular localization was altered in ROR1-Pm-expressing cells, resulting in colocalization of HRS with LAMP-1 (Fig. 5e). These results indicate a requirement of the STAM1-binding region of ROR1 to sustain the interaction of ROR1 with HRS, as well as stabilize the HRS protein for prevention of being routed to lysosomes.

Rab5 identified as ROR1-HRS and ROR1-STAM1 binding protein
To further evaluate the interaction of ROR1 with HRS/STAM1 and its function, a split-APEX2 system 37 was used by taking advantage of ROR1 binding with HRS or STAM1. ROR1-AP, HRS-EX and STAM-EX fragments are inactive on their own, though are reconstituted to provide peroxidase activity when driven together by a molecular interaction ( Fig. 6a and Supplementary Fig. 4a). The split-APEX2 system enables biotin labelling on proteins that are located within 20 nm from active reconstituted APEX2. Biotinylated proteins are then affinity-enriched with streptavidin beads, with the eluted products analyzed by mass spectrometry (LC-MS/MS). Fluorescence analysis results showed that V5-tagged ROR1-AP and HA-tagged HRS-EX or STAM1-EX complexes formed puncta, and were tightly colocalized with biotinylated proteins (Fig. 6b and Supplementary Fig. 4b). Using WB analysis with streptavidin-HRP, H2O2-and APEX2-dependent biotinylation of protein complexes surrounding ROR1-HRS or ROR1-STAM1 interactions were confirmed ( Fig. 6c and Supplementary Fig. 4c). Biotinylated proteins are enriched by streptavidin selection and can be analyzed with LC-MS/MS. Among the proteins which were detected only in ROR1-AP+HRS-EX or ROR1-AP+STAM1-EX, RAB5A and RAB5B were identified as ROR1/HRS and ROR1/STAM1 binding proteins, respectively (Fig. 6d, Supplementary Fig.   4d, and Supplementary Table 1). Evidence obtained in previous studies strongly suggests that HRS is recruited to endosomes 12,36 . In the present study, a proximity ligation assay (PLA) was performed to detect protein-protein interactions in close proximity (<40 nm) 38,39 , which revealed a significant foci of signals arising from ROR1-GFP and endogenous HRS in the proximity of the membrane of mCherry-Rab5 Q79L -expressed endosomes (Fig. 6e). These results indicate recruitment of HRS to Rab5-positive early endosomes by ROR1.

ROR1 required for formation of Rab5-induced MVEs
Rab5 Q79L causes formation of enlarged endosomes (MVEs) that have a large number of ILVs, while those containing both early and late endocytic markers are also frequently observed 29,40 . To investigate the role of ROR1 in MVEs in greater detail, the effects of siROR1 treatment on Rab5-induced MVE formation were subjected to analysis. Interestingly, Rab5 Q79L -induced MVE enlargement was reduced by ROR1 knockdown (Fig. 7a and 7b). Furthermore, in siROR1-treated HeLa cells, small sized-Rab5 Q79L MVEs were surrounded by lysosomes (Fig.   7c). The Rab5 Q79L mutant has been shown to stimulate fusion between endosomes and is known to more active because of its low GTPase activity, causing a higher proportion to be in a GTP-bound state, which leads to increased fusion of early endosomes and results in oversized MVEs 29,40 . We found that the amount of GTP-bound Rab5 was significantly decreased in ROR1-depleted cells (Fig. 7d). Also, IP-WB analysis revealed an interaction of ROR1 with the Rab5 Q79L form, but not the Rab5 WT form (Fig. 7e). These results indicate that ROR1 interacts with and stabilizes GTP-bound Rab5, which then sustains Rab5-induced MVE formation.

ROR1 depletion inhibits CD63-positive MVEs formation and reduces exosome release
Findings obtained in the present experiments led us to speculate that the HRS/STAM1/Rab5 interacting function of ROR1 might also be involved in exosome biogenesis via formation of MVEs in cancer cells, thus we investigated MVE formation. Using immunoelectron microscopy, localization of CD63, an exosome marker, was examined in cells knocked down for ROR1. In contrast to CD63 labelling observed in typical MVE structures in siControltreated cells, few CD63 labels were found in the vestige of MVE in siROR1-treated cells (Fig.   8a), suggesting ROR1 to be essential for formation of CD63-positive MVEs in cancer cells. In addition, ROR1 was found in isolated exosomes in amounts proportional to the amount of ROR1 expression in cells from which the exosomes were derived (Fig. 8b). We also examined exosomes from lung cancer cells using nanoparticle tracking analysis (NTA).
Consistent with previously reported findings 18,19,20,21 , exosome secretion was significantly decreased in HRS-as well as STAM1-depleted NCI-H1975 lung cancer cells. However, it was interesting to note that while inhibition of ROR1 did not have an effect on exosome size, siROR1-treated NCI-H1975 lung cancer cells also showed reduced exosome secretion ( Fig.   8c and Supplementary Fig. 5). Together, the present findings clearly show that ROR1 facilitates the interaction of HRS with STAM1 in Rab5-induced MVE membranes of cancer cells, which results in sustained MVE formation and exosome secretion via its scaffold function for HRS and STAM1 (Fig. 8d).

DISCUSSION
The present study revealed an unanticipated function of ROR1, an onco-embryonic antigen receptor present in cancer cells. ROR1 was shown to have a role as a scaffold protein for HRS and STAM1, and facilitate their association in the membranes of Rab5-induced early endosomes. That interaction of ROR1 with STAM1 was found to maintain HRS expression by preventing lysosome-dependent degradation. Also, ROR1 regulates Rab5 as well as subsequent MVE formation, which in turn is involved in exosome biogenesis in cancer cells. In the present study, ROR1 was shown required to sustain MVE formation via the interaction of HRS and STAM1 in cancer cells. It is possible that ROR1 is involved in formation of MVEs in various fetal tissues including the lungs, which have abundant ROR1 expression, because ROR1 functions as an onco-embryonic antigen 41,42 . However, we 13 consider it unlikely that ROR1 is invariably required for stabilization of HRS and STAM1 expression, because of its negligible level or absence of expression in normal human adult tissues 41,42 . In this regard, UBPy (USP8) is essential for stability of the HRS-STAM complex in MEF cells 43 , which do not express ROR1. As for the existence of marked tissue-type specificity related to HRS and STAM expression, there may be distinct molecular mechanisms involved in sustainment of MVE formation for some cell states or lineage, which might potentially have effects on the function of MVEs.
In the present investigation, electron microscopy revealed that ROR1 knockdown resulted in a marked number of abnormal vesicles in cytoplasm of lung cancer cells. It is also notable that this specific morphological impairment was repeated by knockdown of HRS or STAM1.
Previous studies showed that S. cerevisiae and D. melanogaster mutants have an impaired Hrs function, as well as a reduced number of ILVs in endosomes and vacuoles 6,44 , and that the same effect can be reproduced in mammalian cells treated with siHRS 26 . Also, siHRS-treated HeLa cells show typically large and ring-shaped vacuoles, and contain few internal vesicles 26 .
Our study obtained similar results with NCI-H1975 lung cancer cells treated with siRNA against ROR1, suggesting that ROR1 might be indirectly responsible for intralumenal vesicle formation by initiating recruitment of HRS and STAM1 to endosome membranes, and also indicate that ROR1 controls not only controls MVE formation but also generation of ILVs in cancer cells. Furthermore, lipid microdomains and their components may be involved in ILV generation, or function together with other proteins with an affinity for lipid raft domains 45 .
Our previous study showed a portion of ROR1 residing in detergent-resistant membranes (DRMs) that contained the caveolae-specific protein CAV1, consequently ROR1 is considered responsible for maintaining formation of caveolae and also caveolae-dependent endocytosis in lung cancer cells 24 . Therefore, we speculate that high amounts of ROR1 in exosomes indicate that recruitment of other membrane proteins from the limited membranes 14 of endosomes into ILVs of MVBs is involved in their early incorporation into ROR1containing DRM domains.
HRS overexpression is known to be associated with malignancy and poor prognosis 5,46 , and previous studies shown an association of several types of human cancer, such as colorectal, gastric, and hepatocellular carcinomas, with a significantly elevated expression of HRS 47,48 , though the underlying related molecular mechanism remains unclear. Additionally, HRS-deficient mice, or mice with double knockout of STAM1 and STAM2 show embryonic mortality 49,50 . HRS and STAM1 are not likely to serve as suitable molecular targets for cancer therapy due to their crucial physiological functions in various normal organs. In this regard, it is important to note that ROR1 is an onco-embryonic antigen showing tumorspecific expression in adults 41,42 . As a result, the scaffold function of ROR1 is an attractive target for dealing with tumor-specific exosome secretion in cancer cells. In the present study, we found ROR1 expression in exosomes from lung cancer cells, which led us to speculate that ROR1-positive exosomes may have crucial roles in tumor progression, metastasis, signaling transduction, and/or tumor-to-stroma communication. These are exciting area for future research. Finally, ROR1 is suggested to play important roles in development of lung cancer as well as other types of human malignancy, including that in the breast 51 , pancreas 52 , stomach 53 , colon 52 , ovary 54 , and skin 41 , and also acute and chronic leukemia 42,55,56,57 .
In summary, ROR1 was found to possess an unanticipated function as a scaffold protein of HRS and STAM1, and thus sustains MVEs formation and exosome secretion from cancer cells. Future investigations of ROR1 functions are needed to elucidate the molecular mechanisms involved in regulation of cancer cell membrane dynamics and trafficking.

Cell lines
The derivation, characteristics, and culture conditions of the human lung adenocarcinoma cell lines (NCI-H1975, PC-9, SK-LC-5) utilized, and the immortalized human lung epithelial cell line HPL1D were previously reported 22 . HeLa cells were purchased from Japanese Collection of Research Bioresources Cell Bank. All cell lines were authenticated by short tandem repeat (STR) DNA profiling and were free of mycoplasma contamination. and anti-rabbit IgG (#7074) from Cell Signaling Technology.

Primers
The following primers were used to construct various ROR1 mutants by in vitro mutagenesis: siControl (AllStars Negative Control siRNA) was also obtained from QIAGEN.

Constructs
Constructions of pCMVpuro-ROR1 was previously reported 22

RNA interference
Cells were seeded at 5.0x10 4 cells per well in 12-well plates with coverslip for immunofluorescence microscopy; at 1.0x10 5 per well in 6-well plates for WB analysis and electron and immunoelectron microscopy; and at 1x10 6 in 10 cm dishes for IP-WB analysis.
On the next day, the cells were transfected with siRNAs (each at 20 nM) using Lipofectamine RNAiMAX (Invitrogen), according to the manufacturer's instructions. The cells were fixed or harvested 48 hr for immunofluorescence analysis or 72 hr for WB or IP-WB analysis after transfection. In the IP-WB analysis of the interaction between HRS and STAM1 or ROR1 and HRS in siROR1 or siSTAM1-treated cells respectively, the cells were harvested 24 hr after siRNA transfection. In the time course analysis, the cells were harvested or fixed at indicated time points after siROR1 or siControl transfection for WB analysis.

WB and IP-WB analyses
WB and IP-WB analyses were performed according to standard procedures using Immobilon-P filters (Millipore) and an enhanced chemiluminescence system (GE Healthcare). To analyze the physical interactions between ROR1 and HRS or STAM1, as well as those between HRS and STAM1, whole-cell lysates of NSCLC cell lines were solubilized in octylglucoside buffer

Clarification of ROR1-RNAi effects
Stable transfectants expressing siRNA-resistant forms of ROR1-WT or ROR1-P were generated by introducing the respective plasmids (pCMVpuro-ROR1-WTm or pCMVpuro-ROR1-Pm) using FuGENE6 (Promega), followed by puromycin selection (1.0 μg ml -1 ). The resultant stable clones were then seeded at 1x10 5 cells per well in 6-well plates for immunofluorescence analysis and at 1x10 6 cells in 10 cm dishes for IP-WB analysis, introduced with siControl or siROR1 on the next day, and harvested 2 days after siRNA transfection.

Immunofluorescence microscopy
A total of 5.

Split-APEX2 Proximity biotin labelling
A total of 1.0x10 6 of HeLa cells were plated in 10 cm dishes 24 hours before DNA transfection.
Cells were co-transfected with the 5 ug of ROR1-AP_pLX304 and 5 ug of HRS-EX_pLX304 or STAM1-EX_pLX304 vectors using FuGENE HD reagent (Promega). 24 hr after transfection, Hemin was added (final concentration 5 uM) to culture medium and incubated for 90 min. After incubation, the cell culture medium was changed to DMEM with 500 uM biotin-phenol and incubated for 30 min at 37°C. Then H2O2 solution was added (final concentration 1 mM) and incubated for 1 min to initiate Proximity biotin labelling. To stop this labelling reaction, the cells were washed three times with a quencher solution (10 mM sodium azide, 10 mM sodium ascorbate, and 5 mM Trolox in PBS). Subsequently, the cells were lysed with RIPA buffer (50 mM Tris, 150 mM NaCl, 0.1% SDS, 0.5% sodium deoxycholate, 1.0% NP40) containing 10 mM sodium azide, 10 mM sodium ascorbate, 5 mM Trolox, and protease inhibitor cocktail (Roche) on ice. The lysates were incubated for 10 min at 4°C and clarified by centrifugation at 13,000 rpm for 10 minutes at 4°C. The lysates were mixed with equilibrated NeutrAvidin coated magnetic beads for overnight at 4°C with gentle rotation. NeutrAvidin beads were then washed twice with RIPA buffer, once with 1M KCl, once with 0.1 M Na2CO3, once with 2M urea in 10 mM Tris-HCl, pH 8.0, and twice with RIPA buffer. Biotinylated proteins were eluted by incubating the beads with 1x SDS Sample Buffer (2% SDS, 10% Glycerol, 0.005%BPB, 62.5 mM Tris) supplemented with 5% 2-mercaptoethanol and 2 mM biotin and heating to 95 °C for 10 min. Biotinylated proteins were separated on a SuperSep Ace 12.5% gel (#199-14971, Wako Pure Chemical, Osaka, Japan) run for about 10 min at 20 mA. The gel was stained with SimplyBlue SafeStain (#LC6060, Thermo Fisher) and subsequently destained with water.

Mass Spectrometry
Biotinylated proteins were separated on a SuperSep Ace 12.5% gel (#199-14971, Wako Pure Chemical, Osaka, Japan) run for about 10 min at 20 mA. The gel was stained with SimplyBlue SafeStain (#LC6060, Thermo Fisher) and subsequently destained with water for in-gel digestion. The gel containing proteins was excised, cut into approximately 1mm sized pieces.
Proteins in the gel pieces were reduced with DTT (#20291, Pierce/Thermo Fisher), alkylated with iodoacetamide (A39271, Pierce/Thermo Fisher), and digested with Trypsin/Lys-C Mix (V5073, Promega) in a buffer containing 40-mM ammonium bicarbonate (018-21742, Wako), pH 8.0, overnight at 37°C. The resultant peptides were analyzed on an Advance UHPLC system (AMR/Michrom Bioscience) coupled to a Q Exactive mass spectrometer (Thermo Fisher) processing the raw mass spectrum using Xcalibur (Thermo Fisher Scientific). The raw LC-MS/MS data was analyzed against the UniProtKB Homo sapiens database using Proteome Discoverer version 1.4 (Thermo Fisher) with the Mascot search engine version 2.5 or 2.6 (Matrix Science). A decoy database comprised of either randomized or reversed sequences in the target database was used for false discovery rate (FDR) estimation, and Percolator algorithm was used to evaluate false positives. Search results were filtered against 1% global FDR for high confidence level. The resulting datasets were further analyzed with Scaffold 4 (Matrix Science/Proteome Software Inc.).

Electron microscopy
The siControl-, siROR1-, siHRS-, siSTAM1-treated NCI-H1975 cells were fixed with 2% paraformaldehyde (PFA) and 2% glutaraldehyde (GA) in 0.1 M phosphate buffer (PB) pH 7.4 at incubation temperature and then they were put into a refrigerator for 30 min in order to lower the temperature at 4˚C. Thereafter, they were fixed with 2 % GA in 0.1 M PB at 4˚C overnight.
After these fixations the samples were washed 3 times with 0.1 M PB for 30 min each, and were postfixed with 2% osmium tetroxide (OSO4) in 0.1 M PB at 4˚C for 1 hour. The samples were dehydrated in graded ethanol solutions (50%, 70%, 90%, and 100%). The schedule was as follows: 50% and 70% for 5 min each at 4˚C, 90% for 5 min at room temperature, and 3 changes of 100% for 5 min each at room temperature. The samples were transferred to a resin (Quetol-812; Nisshin EM Co., Tokyo, Japan), and were polymerized at 60˚C for 48 hours. The polymerized resins were ultra-thin sectioned at 70 nm with a diamond knife using an ultramicrotome (Ultracut UCT; Leica, Vienna, Austria) and the sections were mounted on copper grids. They were stained with 2% uranyl acetate at room temperature for 15 min, and then they were washed with distilled water followed by being secondary-stained with Lead stain solution (Sigma-Aldrich Co., Tokyo, Japan) at room temperature for 3 min. The grids were observed by a transmission electron microscopy (JEM-1400Plus; JEOL Ltd., Tokyo, Japan) at an acceleration voltage of 100kV. Digital images were taken with a CCD camera.
For determining the ROR1-GFP localization in Rab5 Q79L -induced MVEs or CD63-GFP localization in MVEs, the cells on the gold disks were frozen in liquid propane at -175˚C. Once the samples were frozen, they were freeze substituted with 2% tannic acid in ethanol and 2% distilled water at -80˚C for 24 hours. Afterwards they were transferred where they can keep -20˚C for 4 hours, then they were warmed up to 4˚C for 1 hour. The samples were dehydrated, infiltrated, and then transferred to a fresh 100% resin, and were polymerized at 50˚C overnight.
The polymerized resins were ultra-thin sectioned at 90 nm with a diamond knife using an ultramicrotome and the sections were placed on nickel grids. The grids were incubated with the primary antibody (rabbit anti-GFP pAb) in 1% BSA, PBS at 4˚C overnight, then they were washed with 1% BSA, PBS 3 times for 1 min. They were subsequently incubated with the secondary antibody conjugated to 10 nm gold particles (goat anti-rabbit IgG pAb) for 2 hours at room temperature. And after washing with PBS, the grids were placed in 2% glutaraldehyde

Preparation of recombinant proteins
GST-tagged ROR1 (intracellular domain) was expressed in Sf9 insect cells using a Gateway system (Invitrogen) according to the manufacturer's instructions. Recombinant GST-tagged ROR1-WT protein was purified by glutathione-affinity chromatography. GST was purchased 26 from Abnova (Taipei). Purification was performed according to the manufacturer's instructions, and the purified proteins were stored at -80°C for GST-pull down assay.

GST pull-down assay
The NCI-H1975 cells were solubilized in octylglucoside buffer (60 mM octylglucoside, 150 mM NaCl and 50 mM EDTA). The cell extracts were mixed with glutathione beads coated with recombinant GST-tagged ROR1. After several rounds of washing, the bound proteins were eluted and subjected to SDS-PAGE followed by WB analysis using anti-GST, anti-HRS or anti-STAM1 antibodies.

Rab5 activation assay
For Rab5 GTPase activation assay, the cells were treated with siROR1, and amount of GTPform Rab5 was measured 48 h later with a Rab5 activation assay kit (NewEast Biosciences, Malvern, PA) according to the manufacturer's instructions. Briefly, the cells were washed twice with ice-cold PBS and lysed in an ice-cold 1× assay/lysis buffer for 20 min on ice. The lysates were transferred to appropriately sized tubes and cleared at 12,000 g for 10 min at 4°C. The supernatants were then collected to a microcentrifuge tube, and anti-active Rab5 MAb (1:1,000) and 20 μl of resuspended bead slurry were added. The tubes were incubated at 4°C for 1 h with gentle agitation, followed by aspiration and discarding of the supernatant by centrifugation for 1 min at 5,000 g. We then resuspended the bead pellet in reducing SDS-PAGE sample buffer, boiled it for 5 min, and then processed it for SDS-PAGE and immunoblotting detection.

EVs purification from conditioned media
For conditioned media, the cells seeded in T300 flask. Next day, the cells were washed with phosphate-buffered saline (PBS), and the culture medium was replaced with advanced RPMI 27 1640 medium for NCI-H1975, H23 and PC9 cells. For HPL1D, the EVs producing culture was started with HPL1D complete medium without FBS. After incubation for 48 h, the culture supernatant was collected and centrifuged at 2,000 g for 10 min at 4˚C. To thoroughly remove cellular debris, the supernatant was filtered through a 0.22 mm filter (#SLGVR33RS, Millipore). The filtrated culture supernatant was then used for EV isolation. To prepare EVs, culture supernatant was ultracentrifuged at 110,000 g for 70 min at 4˚C. The pellets were washed with 35 ml of PBS, ultracentrifuged at 110,000 g for 70 min at 4˚C and resuspended in PBS. RNA inference was using reverse transfection protocol and collect culture medium after 48hr.

Particle size and concentration by NTA
Nanoparticle tracking analysis (NTA) was carried out using the Nanosight system on purified EVs diluted 4-fold with PBS for analysis. A 60s video recorded all events for further analysis by NTA software.