TurboID‐Mediated Profiling of Glioblastoma‐Derived Extracellular Vesicle Cargo Proteins

ABSTRACT

Glioblastoma (GBM), the most aggressive primary brain tumour in adults, presents significant challenges due to its universal recurrence and limited survival rates. A key driver of GBM progression is the subpopulation of brain tumour‐initiating cells (BTICs), which contribute to therapy resistance and interact with the tumour microenvironment, particularly the cellular components in the subventricular zone (SVZ). Extracellular vesicles (EVs) are critical mediators of intercellular communication, carrying bioactive molecules, such as proteins and RNAs, that can modulate the behaviour of recipient cells. This study investigates the role of EVs in GBM's communication with non‐cancer cells. We utilised the proximity‐labelling system TurboID to achieve global and unbiased labelling of proteins within GBM‐derived EVs. By inducing TurboID expression in primary‐cultured human BTICs from GBM patients, we achieved efficient biotinylation of EV proteins without compromising vesicle integrity, and performed proteomic analysis of biotinylated proteins, which revealed a diverse cargo within BTIC‐EVs. Our method marks the first implementation of TurboID for unbiased global labelling of EV protein cargo in primary cells. This approach facilitates the investigation of EV‐mediated communication and potential therapeutic targets, contributing to the understanding of GBM's complex interactions with the brain's microenvironment and identification of biomarkers for improved diagnosis and treatment response.

1. Introduction

Glioblastoma (GBM) is the most malignant form of glioma and is the most common and lethal primary brain tumour in adults (Ostrom et al. 2022; Louis et al. 2021). The current standard of care includes surgical resection, radiation therapy, and administration of the oral chemotherapy agent temozolomide (Schaff and Mellinghoff 2023). Despite current treatment, the median survival expectancy of GBM patients remains at 14–16 months after diagnosis (Ostrom et al. 2022). Devastatingly, GBM significantly progresses within 1 year of diagnosis for ∼70% of patients, and less than 5% of patients survive more than 5 years (Ostrom et al. 2022; Davis 2016). A significant challenge in treating GBM is its inevitable recurrence and the presence of a heterogeneous population of cells within the tumour mass (Xie et al. 2024).

Within the bulk of GBM tumours lies a subtype of cells referred to as brain tumour‐initiating cells (BTICs), a type of cancer stem cell, that play a pivotal role in tumour initiation, progression, resistance to therapy, and ultimately, recurrence (Gimple et al. 2019; Zarco et al. 2019; Piccirillo et al. 2006). BTICs interact with the surrounding tumour microenvironment and possess the ability to self‐renew and differentiate, contributing to intratumoural heterogeneity and the dynamic adaptation of the tumour microenvironment. They actively engage in bidirectional communication with surrounding cells, remodelling the microenvironment to favour tumour growth and survival (Lathia et al. 2015; Norton et al. 2024).

Intercellular communication within the GBM tumour microenvironment is facilitated by various mechanisms, including direct cell‐to‐cell contact via gap junctions, ion channels, or tunnelling nanotubes, as well as soluble factors, such as paracrine signals, autocrine signals, or neurotransmitters (Azorín and Winkler 2021; Yang et al. 2021; Crivii et al. 2022; Alberghina et al. 2024; Yoo et al. 2022). For example, we have shown that there is bidirectional communication between GBM tumour cells and resident cells of the subventricular zone, through direct cell contact and soluble factors such as cathepsin B¹⁰ (Ripari et al. 2021; Norton et al. 2022). Among these mechanisms, extracellular vesicles (EVs) have emerged as critical mediators of cell‐to‐cell communication in GBM biology (Bebelman et al. 2018; Mathieu et al. 2019; Krapež et al. 2022; Yekula et al. 2019). EVs are membrane‐bound particles released by virtually all cell types, ranging from 30 to 1000 nm in size, and are classified mainly into small EVs (30–100 nm) and large EVs (also known as ectosomes or oncosomes, 50–1000 nm) (van Niel et al. 2018). They carry a complex cargo of proteins, lipids, RNAs, and other genetic material that reflects the physiological state of their cell of origin (van Niel et al. 2018). Notably, cancer cells tend to release a higher amount of EVs, consequently increasing communication with nearby non‐cancer cells, generating a tumour‐promoting niche (Bebelman et al. 2018). In the context of GBM, BTIC‐derived EVs have been implicated in promoting tumour resistance, metastasis, angiogenesis, maintenance of the stemness phenotype, and tumour immunosuppression (Su et al. 2021).

Studying the specific protein cargo of GBM‐derived EVs is crucial for elucidating the mechanisms underlying tumour progression and microenvironment adaptation. However, characterising the proteins packaged into EVs and determining which are functionally transferred to recipient cells presents significant challenges. Traditional proteomic analyses provide a broad overview but lack the specificity to identify proteins actively involved in intercellular communication. There have been many efforts to detail the cargo of GBM‐EVs at the RNA and protein level (Song et al. 2024; Castellani et al. 2024; Akers et al. 2013; de Vrij et al. 2015; Ma et al. 2022; Pan et al. 2022), but what cargo is being passed to recipient cells is still understudied. To investigate this deeper, there have been many insightful studies that use fluorescent tagging of EVs (Bonsergent et al. 2021; Joshi et al. 2020; Gurrieri et al. 2024; Seo et al. 2024), which have provided some understanding of EV transfer. Although the EV field has advanced tremendously on this front, there has been little to no progress on using a method to unbiasedly and globally label EV proteins.

To address these challenges, we employed TurboID, a novel proximity‐dependent biotinylation enzyme that enables rapid and unbiased labelling of proteins in living cells (Cho et al. 2020; Branon et al. 2018). TurboID uses ATP and biotin to covalently tag proximal proteins in minutes, allowing for efficient capture and identification via mass spectrometry (Cho et al. 2020; Branon et al. 2018). This approach overcomes limitations of previous methods that required long labelling times or could potentially disrupt cell viability (Roux et al. 2012; Choi‐Rhee et al. 2004).

In this study, we successfully expressed TurboID in primary‐cultured human BTICs isolated from GBM patients. We demonstrate that TurboID effectively biotinylates proteins within BTICs and their released EVs without altering EV size or concentration. Through proteomic analysis of the biotinylated proteins isolated from EVs, we identified a diverse array of potential mediators involved in tumour‐microenvironment interactions. Our work represents the first application of TurboID for global, unbiased labelling of EV protein cargo in primary cell‐derived EVs. By dissecting the complex protein landscape of GBM‐derived EVs, our findings provide critical insights into the mechanisms of EV‐mediated communication in GBM. This knowledge advances our understanding of GBM pathophysiology and opens new avenues for the development of targeted therapies and biomarkers to improve patient outcomes.

2. Materials and Methods

2.1. Cell Cultures

All cell culture was performed in sterile conditions using aseptic technique. Cells were monitored for mycoplasma contamination once per month and were discarded if positive. With the support of the Mayo Clinic Department of Neurosurgery, our group has established a brain tumour biobank containing over 100 patient‐derived tissue samples. From these, we have established, characterised, and maintained primary GBM BTIC lines as previously described (Garcia et al. 2021) (Quiñones‐Hinojosa et al. 2024). For each cell line used in this manuscript, the BTICs have been analysed to evidence their self‐renewal, pluripotency, and tumour initiation when cells are implanted in vivo. Self‐renewal was measured using an extreme limiting dilution assay, pluripotency was evaluated by differentiation assay, followed by immunocytochemistry for markers such as glial fibrillary acidic protein (GFAP) and beta‐tubulin III (TUBB3) or Tuj1, and tumour initiation in vivo was tested by orthotopic implantation of GFP labelled tumour cells in immunosuppressed mice (Garcia et al. 2021). For the purposes of this paper, we utilised cell lines: GBM612, GBM120, GBM1a, and GBM965. BTICs were cultured in DMEM/F12 supplemented with 1x NeuroPlex (Gem21 NeuroPlex Serum‐Free Supplement) (GeminiBio 400‐161) and 1X Antibiotic‐Antimycotic (Invitrogen 15240062) (deemed base media). Base media was filtered and sterilised using a 0.22 µm filter. To make BTIC complete media, we added human epidermal growth factor (hEGF) (PeproTech) and human fibroblast growth factor (hFGF) (PeproTech) for a final concentration of 20 ng/ml for each. Complete media was kept at 4°C for up to a week. For multiple experiments, we utilised human embryonic kidney (HEK) 293T cells (ATCC #CRL‐3216). These cells are cultured using DMEM +GlutaMAX (Thermofisher 10569010), 10% Fetal Bovine Serum (FBS) (GeminiBio 900‐208‐500), and 1x Antibiotic‐Antimycotic. Additionally, we utilised a commercially available adipose‐derived human mesenchymal stem cell line (ATCC PCS‐500‐011) to compare with our GBM cell line EV extractions. These cells were cultured using low glucose DMEM, + GlutaMAX (1 g/L) + sodium pyruvate (110 mg/L) (Gibco 10‐567‐014) with 10% FBS and 1x Antibiotic‐Antimycotic. All media was filtered and sterilised using a 0.22 µm filter.

2.2. EV Enrichment Protocol

We seeded our BTICs on pre‐laminated 10 cm dishes (Thermo Scientific 150350) at a density of 2 × 10⁶ cells per dish. The cells were plated in complete BTIC media. After waiting 16–24 h for the cells to properly attach, cells were rinsed with PBS and complete media was replaced with BTIC base media, depleted of growth factors, to avoid confounding effects. Culture media was conditioned for 24 h, collected and placed on ice for 5 min. To secure the cell lysate, we washed the cells gently with sterile 1x PBS twice to get rid of any remaining cell culture media. Then, we added 200 uL of lysis buffer (1x RIPA buffer with 1% protease and phosphatase inhibitors) to each dish, scraped the cell lysate, and transferred it to a 1.5 mL Eppendorf tube. We incubated the tubes on ice for 5 min and then stored them at −80°C until used for further analysis. After a 5 min incubation on ice, we took the conditioned media and centrifuged twice at 1000 × g for 5 min at 4°C, to remove any remaining cell debris or larger apoptotic bodies. In between spins, we transferred the supernatant to a new tube without disturbing the pellet between spins. After the second spin, we transferred the supernatant into a 50 mL conical tube and attach a 0.22 µm Steriflip‐GV poly (vinylidene fluoride) (PVDF) radio‐sterilised filter (EMD Millipore) to a vacuum source, turned it on slowly, and allowed for one drop to pass at a time. This step is important to retain and remove the larger sized vesicles, as they do not pass through the pores of the filter. We transferred the flow through media into an Amicon ultra‐15 centrifugal filter unit with a 10 or 100 kD cut off (EMD Millipore). We spun the tubes at 4000 × g for 15 min at 22°C in a swinging bucket rotor. We collected the flowthrough, deemed vesicle free media (VFM), and stored it at −80°C until needed for further downstream analysis. Next, we took the concentrated media caught in the filter tube and transferred it to a 1.5 mL Eppendorf tube. We added a 1:2 ratio, half volume of Total Exosome Isolation reagent (TEI) (Thermo Fisher Scientific), to the concentrate and pipetted up and down to mix the viscous TEI into the concentrate. The TEI reagent forces less soluble components, such as EVs of out solution to allow for their collection by centrifugation. The protocol established for the TEI reagent does not include filtration or concentration steps before mixing the reagent with EVs overnight to begin the precipitation process. Our protocol uses a 0.22 µm filter to exclude larger EVs, in addition to a 100 kD centrifugal filter unit to concentrate the EVs before the overnight precipitation step. The tubes are then incubated at 4°C overnight. The next day, we spun the tubes at 10,000 × g for 1 h at 4°C, and we carefully removed the supernatant without disturbing the EV pellet. Finally, we resuspended the pellet in 200µL of filtered 1X PBS and stored the EV suspension at −80°C until needed for further downstream analysis (Figure 1A).

FIGURE 1.

Successful extracellular vesicle isolation and characterisation using nanosight tracking analysis (NTA), immunogold electron microscopy, and western blot. (A) Schematic of extracellular vesicle (EV) isolation protocol from cell culture supernatant. (B) NTA histograms of vesicles plotting vesicles size to the concentration of vesicles per milliliter of sample. Size mode is indicated at the top of the peak. E8 stands for a scale of 1 × 10⁸, E6 is 10⁶, and E7 is 10⁷. The red shading on each plot represents the standard deviation. All EV samples display one single peak whereas vesicle free media (VFM) samples all display multiple peaks of various sizes. (C) Summary plot of EVs size modes in nanometers between the three BTIC lines used in the study. All size modes fall into the category of smaller EVs and are <140 nm. n = 9–18. (D and E) Summary plot of particles/mL and particles/frame gathered from NTA for EV and VFM samples across the three BTIC lines. In all conditions, EV samples have significantly more particles/mL and particles/frame compared to VFM. n = 6–18. (F) Immunogold electron microscopy against CD63 shows gold particles labelling the surface of EVs, and no labelling or presence of EVs in the VFM fraction. Scale bars= 50 nm. (G) Western blot for brain tumour‐initiating cell (BTIC) lines (612, 120, 1a) WCL, EV, and VFM probing for positive (CD81, CD63, CD9), negative (calnexin) EV markers. We observe positive markers in EV lanes, negative markers in WCL lanes only, no EV protein contamination in VFM. (H‐ I) Silver stain analysis for BTIC cell lines WCL and VFM comparing the VFM proteins when using 10 kD filter units and 100 kD filter units, respectively. Unpaired t‐test, **p < 0.001, ***p < 0.0008, ****p < 0.0001.

2.3. Ultracentrifugation EV Enrichment Protocol

To compare our EV enrichment protocol to a commonly used method, we performed EV isolation using ultracentrifugation (UC). We plated the BTIC cells at the same density and conditioned the cells similarly to our protocol described in Section 2.2, followed by similar low‐speed spins to get rid of debris. After the 0.22 µm filtering step, we transferred the EV solution to UC tubes and spun for 16 h at 100,000 × g as suggested in the described protocol by Jeppesen et al. 2014, suspending the pellet in PBS after.

2.4. Nanosight Tracking Analysis (NTA)

BTIC‐EVs and VFM, and 612 wt and 612 turbo EVs were diluted at 1:20 in filtered 1x PBS and analysed for total particle count and size on the Nanosight NS300 (Malvern Panalytical). For each sample, we used a manual injection system set at 25°C and collected three separate 60 s videos to count three independent replicates. Videos were analysed using the Nanosight software v3.4 and a camera detection threshold of five. For all samples described in this paper, the above parameters remained consistent.

2.5. Determination of Protein Concentration

Cell lysates collected during EV isolations and stored at −80°C were vortexed, centrifuged at 12,000 × g for 2 min, and the protein supernatant was collected in a new tube. For each VFM and EV sample, 50 µL of RIPA buffer was added (with protease and phosphatase inhibitors) to each tube, vortexed for 30 s, and placed on ice for 20 min while vortexing every 5 min. This was followed by sonication (15 s on, 10 s off, 60% amplitude, repeated 3 times for a total of 45 s), centrifuged at 12,000 × g for 2 min, and the supernatant was transferred to a new tube. The concentration of all protein samples was quantified using the BCA assay (Pierce 23225) and read on a plate reader at 562 nm and plotted using a linear regression curve fit model against an albumin standard curve. To calculate the ratio of particles/mL to total protein, we divided each EV and VFM concentration by the total protein amount and displayed the data using a bar graph.

2.6. Western Blot

For BTIC cell lines GBM612, GBM120 and GBM1a, 10 µg of proteins obtained from cell lysate, VFM, and EV were added to 4x Laemmli non‐reducing buffer (Invitrogen NP0007) and incubated for 5 min at 96°C. For HEK293T cell lysates and EVs, 5 µg of protein was obtained, and cell lysate proteins were added to 6x reducing Laemmli buffer (Thermo Scientific J61337.AD), while EV proteins were added to non‐reducing buffer, and incubated for 5 min at 96°C. Then, samples were electrophoresed on a 4%–12% gradient Bis‐Tris 1.5m gel (NuPAGE, Thermofisher) for 1.5 h at 150 V. The gel was transferred to a 0.45 µm PVDF membrane using the Bio‐Rad Trans‐Blot turbo transfer system for 12 min at 2.5A, 25 V. Membranes were blocked using 5% BSA in 0.1% Tween in TBS (TBST) for 1 h at room temperature. Membranes were then incubated with the following primary antibodies overnight at 4°C: CD63 (1:1000, EMD Millipore #CBL553), CD81 (1:500, Santa Cruz #SC‐166029), CD9 (1:500, Cell Signaling #D801A), Calnexin (1:1000, Cell signaling #2433S), Annexin 1 (1:500, Santa Cruz #SC‐12740), HSP90B1 (1:1000, Cell Signaling #7411), and Streptavidin HRP (1:10,000, Cell Signaling #3999) all diluted in 5% BSA. V5 (1:1000, Thermo Scientific #R960‐25) and anti‐albumin (1:1000, Fortis life sciences #A90‐134A) and CD44 (1:1000, abcepta #AM1901b) were diluted in 5% milk. After primary antibody incubation, membranes were washed with TBST for 5, 10, and then 15 min. Incubation with secondary antibodies was done for 1 h at room temperature with goat anti‐mouse HRP or goat anti‐rabbit HRP (1:5000, Invitrogen). Membranes were again washed as described above with TBST. Next, they were exposed to Enhanced chemiluminescence (ECL) (Cytiva Amersham ECL Prime Western Blotting Detection Reagent) for a 3‐min incubation and imaged on the Chemidoc MP Imaging System (Bio‐Rad).

2.7. Silver Stain

10µg of protein was run in an electrophoresis 4%–12% Bis‐Tris gel as described above. After electrophoresis, the gel was stained with silver stain using the SilverQuest Silver Stain Kit (Invitrogen), per manufacturer's instructions.

2.8. Immunogold Electron Microscopy (EM)

Immunogold electron microscopy was utilised to characterise the presence of EV surface markers. EVs were fixed in paraformaldehyde (PFA) at a final concentration of 2% in dPBS (Corning 21‐031‐CV) to a total of 8 µL, for 25 min at room temperature. An 8 µL drop of fixed EVs or VFM was then placed on a 200‐mesh formvar‐coated copper grid (Electron Microscopy Sciences) for 20 min. Grids were washed three times for 3 min each in filtered dPBS and then four times for 3 min each with 50 mM glycine. Next, we used a blocking solution containing acetylated BSA (BSAc) at 0.3% in dPBS for 10 min. Samples were then incubated in 25 µL of primary antibody, mouse anti‐CD63 at 1:50 in 0.3% BSAc in dPBS for 30 min at room temperature. Following primary antibody incubation, six washes for 2 min each were performed using 0.1% BSAc. Next, samples were incubated with 0.3% BSAc for 10 min, followed by incubation in 6 nm Colloidal Gold‐conjugated Goat Anti‐Mouse IgG (Jackson ImmunoResearch 115‐195‐146) 1:20 in 0.1% BSAc for 20 min. Grids were then rinsed with 1x filtered dPBS 8 times, followed by fixation with 1% glutaraldehyde (Electron Microscopy Sciences 16210) for 5 min and washing with ultrapure water. Finally, negative staining with 2% uranyl acetate was performed for 5 min, followed by a final wash in 0.4% methylcellulose for 2 min. All solutions were filtered and sterilised using a 0.22 µm filter syringe, except for the uranyl acetate and the glutaraldehyde. All solutions were dissolved in dPBS, except for the methylcellulose and the uranyl acetate, which were dissolved in MilliQ H₂O. The grids were visualised in a JEM‐1400Flash electron microscope (JEOL).

2.9. TurboID Establishment In Vitro

We used HEK293T cells and primary BTIC cell lines GBM612, GBM120, GBM1a, and GBM965. HEK293T cells were transfected with the TurboID plasmid (V5‐TurboID‐NES‐pCDNA3 (Addgene plasmid #107169) and expression was validated using immunofluorescence as described below. For lentiviral production, cloning was performed to incorporate the TurboID plasmid into a lentiviral backbone (pLenti‐puro (Addgene plasmid #39481). Lentiviral particles were produced in HEK293T cells using a standard second‐generation lentiviral production protocol. All BTIC primary cell lines were transduced with TurboID‐puro lentivirus and selected for stable expression with 0.5 µg/mL puromycin in complete BTIC media for over one week. Expression and function of the lentiviral construct were validated using immunofluorescence and western blotting.

2.10. Immunocytochemistry (ICC)

Immunocytochemistry was performed to validate the transduction of TurboID and the incorporation of biotinylated proteins. 30,000 cells per condition were seeded in each well of an 8‐well chamber (ibidi 80826). The conditions for the experiments were as follows: Wildtype ‐biotin, Wildtype +biotin, Turbo −biotin, and Turbo +biotin. After cells attached overnight, the media was aspirated from all dishes, and 50 µM of biotin was added to the media depleted of growth factors of the 2 conditions that were to receive biotin (Wildtype +biotin and Turbo +biotin). The other two conditions did not receive biotin, but did receive media depleted of growth factors. The 8‐well chamber was placed back in the incubator for 1 h. After this time point, all media were aspirated from the wells, and each well was washed twice with cold sterile 1x PBS (Corning 21‐040‐CV), then fixed with cold sterile 4% paraformaldehyde (Santa Cruz Biotechnology sc‐281692) for 20 min at room temperature. Next, cells were washed 3 times for 5 min each with PBS‐0.1% Triton (PBS‐t) and blocked with 10% goat serum (Sigma‐Aldrich G9023) in PBS‐t for 1 h at room temperature. Cells were then incubated with the primary antibody V5 in 2% NGS in PBS‐t overnight at 4°C. The next day, cells were washed 3 times for 5 min each with PBS‐t and then incubated with the following secondary antibodies diluted in PBS for 1 h at room temperature: goat‐anti‐mouse Alexa Fluor 488 (1:500, Invitrogen A11008) and streptavidin HRP Alexa Fluor 647 (1:500, Thermo Scientific S21374). Cells were then washed 3 times for 5 min each with PBS and then incubated with DAPI (1:1000, Life Technologies62248) diluted in PBS for 5 min. All washes and incubations were done with very gentle shaking. All wells were imaged using a confocal microscope (Zeiss LSM800). Laser power and detector voltage were conserved across each condition to ensure that each well was imaged using the same parameters.

2.11. Biotinylation of EV Proteins

For each EV isolation, 4–6 million cells (2 million per dish) per condition were seeded following the same plating protocol as above. After the cells attached overnight, the media was aspirated from all dishes, and 50 µM of biotin (lyophilised powder 100 mM stock, Sigma‐Aldrich B4501) was added to the media depleted of growth factors of the 2 conditions that were to receive biotin (Wildtype +biotin and Turbo +biotin). The other two conditions did not receive biotin, but did receive media depleted of growth factors. Cells were left in the incubator for 48 h before beginning the EV isolation protocol as described above. Cell lysate and EV proteins from this incubation time were saved for further validation through western blot, following the same methodology described below.

2.12. Streptavidin‐conjugated Bead Pulldown

The streptavidin bead pulldown method was adapted from Cho et al. For each condition, 25 µL of streptavidin Dynabeads (Invitrogen 11205D) was added to 1 mL of RIPA buffer, placed on an end‐over‐end rotator for 2 min, then on a magnetic rack, and supernatant was removed. This washing process was repeated once more for each condition. In a separate tube, 100 µL of protein diluted up to 200 µL in PBS was prepared per condition. An additional 500 µL of RIPA buffer was added to the 200 µL protein solution, and this new mixture was added directly onto the streptavidin magnetic beads. Each tube was incubated for 5 h with 800 rpm shaking at 8°C on a ThermoMixer (Eppendorf). Tubes were removed from the ThermoMixer and placed on the magnetic rack to collect the flow‐through fraction. Beads were then washed twice with 1x RIPA buffer for 2 min. Next, the beads were washed twice with 1 M potassium chloride (SIGMA P‐3911‐500 g) for 2 min. This is followed by one wash with 50 mM sodium carbonate (Thermo Scientific 011552.18) and one wash with 2 M urea (SIGMA U5378‐500G) dissolved in Tris HCl (SIGMA T3253‐100G) adjusted to pH8, for less than 10 s (to ensure not to denature the beads) each to remove any unspecific binding to the beads. The final two washes are done with 1x RIPA buffer. Lastly, the biotinylated proteins are eluted from the beads with 30 µL of 3x elution buffer (6x Laemmli buffer, 20 mM DTT (Novex B0009), and 2 mM biotin) by boiling for 10 min at 95°C. Tubes are added back to the magnetic rack, and protein elute is collected and stored at −80°C until ready for downstream analysis.

2.13. LFQ Mass Spectrometry Sample Preparation

The streptavidin bead pulldown elute samples are electrophoresed for 8 min on a 4%–12% gradient Bis‐Tris 1.5m gel, and each lane was excised for processing. Excised gel bands in 1.5 Eppendorf tubes are washed with 400 µL MeOH/H₂O/Acetic acid (45%/45%/10%) for 15 min, then with 1 mL H₂O three times for 5 min each time. Gel bands are then stored at −20°C until further prep. Next, gel bands are washed with 1 mL H₂O for 10 min, then washed with 1 mL 100 mM ammonium bicarbonate (ABC) in 50% acetonitrile for 20 min. Proteins in the gel are then reduced with 150 µL of a 4.5 mM dithiothreitol (DTT) in a 25 mM ABC buffer for 20 min at 37°C and allowed to cool to room temperature. Proteins are then alkylated in the dark with 150 µL of a 10 mM iodoacetamide (IAN) in a 25 mM ABC buffer for 20 min at room temperature. Gel bands are then washed with 1 mL of a 50% CH3CN/100 mM ABC solution for 20 min, followed by a 20‐min wash with 1 mL of a 50% CH₃CN/10 mM ABC solution. Gel pieces are then centrifuged in a SpeedVac to dryness. Ingel tryptic digestions are performed overnight with 150 µL of a 1:200 dilution of a 0.5 µg/µL stock trypsin (Promega Trypsin Gold MS grade) solution in 25 mM ABC. Digests are transferred to new tubes, and peptides in gels are extracted with 450 µL 80% CH3CN/0.1% trifluoroacetic acid (TCA) for 15 min. The tubes are then centrifuged, and the supernatant is removed and combined with the transferred digests. The combined peptide mixtures are dried in SpeedVac and stored at −20°C until LC MS/MS analyses.

2.14. LFQ Data Collection

Label‐free quantitation (LFQ) was performed on a Thermo Scientific Q‐Exactive HFX, mass spectrometers connected to a Waters nanoACQUITY UPLC system equipped with a Waters Symmetry C18 180 µm × 20 mm trap column and a 1.7‐µm, 75 µm × 250 mm nanoACQUITY UPLC column (35°C). 5 µL of each digest at 0.05 µg/µL concentration was injected in block randomised order. To ensure a high level of identification and quantitation integrity, a resolution of 120,000 and 30,000 were utilised for MS and MS/MS detection, respectively. TopN of 20 MS/MS spectra were acquired per MS scan using HCD with a normalised collisional energy of 30, an isolation window of 1.4 m/z, scan range of 200–2000 m/z, charge state exclusion for unassigned, 1+, and >8+, minimum AGC target of 1.0e4, and a dynamic exclusion setting of 15 s. All MS (Profile) and MS/MS (centroid) peaks were detected in the Orbitrap. Trapping was carried out for 3 min at 5 µL/min in 99.5% Buffer A (0.1% FA in water) and 0.5% Buffer B (0.075% FA in acetonitrile (ACN) prior to eluting with linear gradients that start at with a 300 µL/min flowrate at 3%B and reach 6% B at 2 min, 25% B at 175 min, 40% B at 195 min, and 90% B at 200 min; then maintain at 90% B for 10 min before dropping down to 3% at 212 min. Column is then re‐equilibrated for 13 min at 3% B, until the trap and column are washed with four blanks following each sample injection to ensure against sample carry over.

2.15. Proteomic Data Analysis

LC MS/MS collected data are analysed using Proteome Discoverer software v2.5 (Thermo Scientific) for label‐free quantification (LFQ) analysis. Briefly, there are two distinct processes: processing workflow (used for protein identification) and consensus workflow (used to carry out LFQ quantification). In protein identification component, data searching is performed using the SEQUEST HT (Thermo Scientific) against the SwissProt Homo sapiens protein database (20,385 sequences). The search parameters included tryptic peptides with up to 2 missed cleavages, 10 ppm precursor mass tolerance, 0.02 Da fragment mass tolerance, and variable (dynamic) modifications of oxidation on methionine and carbamidomethylated on cysteine. An additional decoy database was searched to gauge the false discovery rate (FDR) for the identification, and the proteins used in LFQ analyses were filtered based on a confidence level of 95% (e.g., FDR of 5%; p < 0.05) and at least two unique peptides. Protein abundances were normalised with total peptide amount prior to being used to calculate for fold change. A t‐test was performed to obtain fold change and p value between 612 turbo +B EVs and 612 turbo −B EVs. Proteins that had a fold change ANOVA FDR < 0.05 were considered highly biotinylated and were analysed using Ingenuity Pathway Analysis (IPA) (version 24.0.2, Qiagen) to interpret the molecular networks, biological processes, and pathways associated with the highly biotinylated proteins. Pathway enrichment scores and z‐scores were calculated to determine the activation or inhibition status of each pathway. Data was filtered to include proteins with a minimum fold change of 1.5 and a p value of 0.05. To understand the role of the EV proteins, the FunRich application was used to compare our highly biotinylated proteins to the Vesiclepedia database (Keerthikumar et al. 2016; Chitti et al. 2024). STRING analysis was also performed using version 12.0 (Szklarczyk et al. 2019; Szklarczyk et al. 2023). Clusters were created using MCL clustering with a minimum confidence interaction score of 0.7.

2.16. Transfer of Biotinylated EV Proteins in Recipient Cells

To prepare TurboID EVs, GBM612 cells were cultured as described above for extraction of TurboID EVs. HEK293T cells were plated in 10 cm dishes at 80% confluency and left overnight to attach. The next day, EVs were added at a concentration of 6x10⁸ to recipient HEK293T cells with a concentration of 50 µM biotin in serum‐free media for 1 h at 37°C. Biotin was supplemented in the media when EVs were added to the HEK cells, similar to the experimental design in Li et al. 2023. Downstream protein extraction from cell lysates and validation of successful transfer of biotinylated EV proteins was assessed using western blot as previously described.

2.17. Statistical Analysis

Statistical analysis was performed using GraphPad Prism 10.4.0. For all EV characterisation of concentration and particles per frame comparing two conditions, unpaired t‐test using Mann–Whitney was performed. For all other analyses on multiple groups, non‐parametric Kruskal–Wallis ANOVA was performed.

3. Results

3.1. Optimised EV Isolation Yields Intact and Pure Vesicles From BTICs

We first established an effective method for isolating EVs from BTICs derived from GBM patients. By combining differential centrifugation and overnight precipitation using a Total Exosome Isolation (TEI) reagent, we developed an optimised protocol designed to gently enrich EVs while preserving their integrity and minimizing contamination from non‐vesicular components (Figure 1A). We isolated EV populations from three of our established BTIC cell lines (Garcia et al. 2021) (GBM612, GBM120 and GBM1a), then performed nanoparticle tracking analysis (NTA) to assess vesicle count and size mode after each EVs isolation and compared them to the vesicle‐free media (VFM) fractions from each cell line (Figure 1B). The NTA results showed that the isolated EVs ranged in size from 86–136 nm (n = 6–18) (Figure 1C). Representative NTA graphs for each BTIC line demonstrated that EV samples displayed single, uniform peaks at a 1 × 10⁸ per mL order of magnitude, indicating a homogeneous population of vesicles (Figure 1B). In contrast, the VFM fractions exhibited multiple peaks of varying sizes at a 1 × 10⁶‐Zarco et al. 2019) particles per mL order of magnitude, reflecting the presence of non‐vesicular particles. Quantitative analysis confirmed that the concentration of particles (particles/mL) as well as particles per frame for EVs to be 1–2 orders of magnitude higher than in VFM fractions (n = 5–18) (Figure 1D,E). The VFM fractions did not show presence of EV markers, confirming the successful separation of EVs from soluble proteins. Additionally, we observed no albumin, which could originate from serum contamination in any EV fraction (Figure S1A). Additionally, the protein concentrations (mg/mL) and ratio of particles/mL to total protein in EVs was significantly greater than their VFM counterparts (Figure S1B,C). To confirm the presence and integrity of EVs, we conducted immunogold EM using an antibody against the EV surface marker CD63. The EM showed intact vesicle membranes with gold particles labeling their surfaces in the EV fractions isolated from all three primary BTIC lines (Figure 1F). No vesicle‐like structures or gold labelling was detected in any of the VFM fractions, confirming the absence of EVs in these samples. Western blot analysis revealed EV markers (CD81, CD63 and CD9) in EV fractions and the absence of the endoplasmic reticulum marker calnexin, suggesting minimal contamination from cellular organelles (Figure 1G). We also analyzed the protein content of the VFM fractions to demonstrate that they contained free soluble proteins. When we used a 10 kDa filter to concentrate the EVs, we found soluble proteins that were <10 kDa by silver stain (Figure 1H). When we used a 100 kDa filter, we observed higher molecular weight proteins (Figure 1I). These results suggest that our method is not only efficient at enriching EVs, but also separating free soluble proteins from those contained in EVs.

3.2. Superior EV Integrity Compared to Ultracentrifugation Methods

To compare the effectiveness of our EV enrichment protocol with the conventional ultracentrifugation (UC) method (Jeppesen et al. 2014), we isolated EVs from GBM612 cells using both methods. The EVs isolated by UC method had a particle concentration of 7.43 × 10⁹ particles/mL and a mode size of 136 nm, similar to the EVs isolated using our method (Supplemental Figure 1D,E). However, when we examined the UC‐isolated EVs by immunogold EM, we observed disrupted vesicle membranes and irregular dark particles without gold labeling (Figure S1F). This suggests that UC may damage EV membranes and compromise EV integrity. Additionally, we calculated the ratio of particle concentration to total protein concentration for the UC‐isolated EVs and found it to be lower than that obtained using our method (Figure S1G). Overall, our optimised protocol not only preserved EV integrity, as evidenced by intact membranes and proper marker labelling across cancer and non‐cancer cell lines, but also provided a unique VFM fraction for further analysis of soluble factors.

3.3. Applicability of the EV Isolation Protocol to Non‐Cancer Cell Types

To assess the versatility of our EV isolation and TurboID labelling methods, we applied them to non‐cancerous cell lines, including HEK293T cells and human mesenchymal stem cells (MSCs). We isolated EVs from these cells using our optimised protocol and performed NTA to characterise them. The EVs isolated from HEK293T cells and MSCs had size distributions and concentrations comparable to those from BTICs, with significantly higher particle concentrations in the EV fractions compared to the VFM fractions (Figure S1H). This indicates that our EV isolation method is effective across different cell types. We also transduced HEK293T cells with the TurboID lentivirus and confirmed TurboID expression and protein biotinylation by immunofluorescence and Western blot (Figure S2A,B). Biotinylated proteins were successfully enriched from both cell lysates and EVs of TurboID‐expressing HEK293T cells, demonstrating that our approach is applicable to various cell types and can be used to study EV‐mediated communication in different biological contexts.

3.4. Establishment of TurboID Expression and Biotinylation in BTICs

To investigate the protein cargo of GBM‐derived EVs, we employed the TurboID proximity protein labelling system due to its ability to rapidly label proximal proteins. TurboID uses ATP and biotin to covalently label proximal proteins in as little as 10 min in cell culture (Cho et al. 2020; Branon et al. 2018) (Figure 2A). We cloned the TurboID sequence, tagged with V5 epitope and a nuclear export signal (NES) for cytoplasmic localisation, into a lentiviral vector (Figure 2B). By targeting TurboID to the cytoplasm rather than to an EV tetraspanin surface marker, we aimed to biotinylate proteins that are packaged into EVs, capturing both internal and surface EV proteins (Figure 2C). We validated the system first in HEK293T cells (Figure S2A,B), then transduced four BTIC lines (GBM612, GBM120, GBM1a and GBM965) with the TurboID lentivirus and assessed TurboID expression and protein biotinylation. Immunofluorescence microscopy revealed strong expression of TurboID (detected by V5 expression) and extensive biotinylation (detected by streptavidin staining) in all TurboID‐transduced BTIC lines upon biotin supplementation (Figure 2D and Figure S2 C–E). Cells were incubated with 50 µM biotin for 1 h before fixation. The co‐localisation of V5 and streptavidin signals confirmed that TurboID was active and effectively biotinylating proximal proteins in the cytoplasm. Western blot analysis of whole cell lysates (WCLs) further confirmed TurboID expression and biotinylation activity. We observed strong expression of the V5‐tagged TurboID and robust biotinylation of proteins in the TurboID‐transduced BTIC lines, with GBM612 cells showing the highest levels of biotinylated proteins (Figure 2E and Figure S2F). Based on these results, we selected the GBM612 cell line for subsequent experiments.

FIGURE 2.

Validation of TurboID system in glioblastoma primary cell lysates and extracellular vesicles. (A) Adaptation of reaction for TurboID to covalently label proximal proteins. (B) TurboID plasmid used for lentiviral transduction. (C) Depiction of an extracellular vesicle with TurboID and biotinylated protein cargo. Created with BioRender.com. (D) Immunofluorescence validation of TurboID system in GBM612 cells after the addition of 50 µM biotin in the media for 1 h. V5 was used to see expression of TurboID in the cells and streptavidin AF647 to see biotinylation. Scale bars = 50 µm. (E) Western blot validation of the TurboID system in cell lysates of GBM612 cells after the addition of 50 µM biotin in the media for 48 h. V5 was used to probe for the TurboID tag, Streptavidin HRP for biotinylated proteins, and B‐actin for loading control. (F) Summary plot of EVs size modes in nanometers of the GBM612 wt and TurboID expressing cells. All size modes fall into the category of smaller EVs and are <150 nm, and there is no significant difference in size across the conditions. n = 4–6. (G) Summary plot of particles/mL of the GBM612 wt and TurboID expressing cells, showing no significant difference in concentration of EVs across the conditions. n = 4–6. (H) Western blot validation of the TurboID system in EVs of GBM612 cells after the addition of 50 µm biotin in the media for 48 h. V5 was used to probe for the TurboID tag, Streptavidin HRP for biotinylated proteins, and CD63 for EV loading control.

3.5. TurboID Expression Does Not Affect EV Production or Characteristics

To determine whether TurboID expression and protein biotinylation affected EV production or characteristics, we isolated EVs from GBM612 cells under three conditions: wild‐type cells with biotin supplementation (612 wt + biotin), TurboID‐expressing cells without biotin (612 turbo−biotin), and TurboID‐expressing cells with biotin (612 turbo + biotin). NTA showed that EVs from all conditions showed peaks that were clean and uniform (Figure S3A), with modes within the small EV range (<150 nm), and there were no significant differences in EV size across the conditions (Figure 2F). The EV concentrations (particles/mL) were consistent across conditions, indicating that neither TurboID expression nor biotin supplementation affected EV production (Figure 2G). Western blot analysis of EVs confirmed the presence of the V5‐tagged TurboID and biotinylated proteins only in the EVs from TurboID‐expressing cells exposed to biotin (Figure 2H). We probed for the EV tetraspanin marker CD63 to ensure equal loading control and verify the EV identity of the samples (Figure 2H). The presence of TurboID and biotinylated proteins in EVs indicates that proteins biotinylated in the cytoplasm were successfully packaged into EVs.

3.6. Biotinylated Proteins Are Efficiently Enriched From Cell Lysates and EVs

To isolate and analyse the biotinylated proteins, we performed streptavidin‐conjugated magnetic bead pull‐downs from cell and EV lysates. We adapted a previously published protocol (Cho et al. 2020) for our primary BTIC lines. After pull‐down, we separated the enriched proteins by SDS‐PAGE and visualised them using silver staining. In cell lysates, we observed strong enrichment of biotinylated proteins in the sample from TurboID‐expressing cells treated with biotin (612 turbo + biotin), as evidenced by multiple protein bands on the silver stain gel (Figure S3B). Control samples (612 wt +biotin, 612 turbo −biotin) showed greatly reduced protein enrichment, confirming the specificity of the TurboID‐mediated biotinylation and streptavidin pull‐down. Similarly, in EV samplesby silver stain and western blot, we detected strong enrichment of biotinylated EV proteins only in the EVs from TurboID‐expressing cells treated with biotin (612 turbo + biotin) (Figure S3C,D). Overall, these results demonstrate that our approach effectively enriches biotinylated proteins from both cell lysates and EVs from patient‐derived primary cell line, enabling downstream proteomic analyses to identify and characterise the EV protein cargo.

3.7. Proteomic Analysis Identifies Key Proteins in GBM‐Derived EVs

We conducted mass spectrometry‐based proteomic analysis on the enriched biotinylated proteins from TurboID‐expressing WCLs and EVs to identify the specific proteins present in GBM‐derived EVs. The analysis revealed 1135 total proteins, 544 significant proteins (FDR < 0.05), in the EVs from the 612 turbo + biotin condition. When comparing the significant EV proteins to the Vesiclepedia database (Keerthikumar et al. 2016; Chitti et al. 2024) filtering for human glioblastoma cell line proteins only, we saw there were 455 overlapping proteins, and 89 unique proteins to our 612 turbo + biotin group (Figure 3A). Ingenuity Pathway Analysis (IPA) of the 1135 proteins 612 turbo + biotin proteins revealed a comprehensive profile of turbo + B EV cargo proteins (compared to turbo—B EV cargo proteins) related to cancer, neurological disease, tumour morphology, among other diseases (Figure 3B). Further analysis using Vesiclepedia revealed the EV cargo cellular components most related to cytoplasm and exosomes, aligning with the targeting of the TurboID system to the cytoplasm and the enrichment of intact EVs (Figure 3C). Molecular function showed the highest percentage of genes related to transporter activity cytoskeletal protein binding, and ubiquitin‐specific protease activity (Figure 3D). We then performed STRING analysis on the top 100 proteins in the turbo +B EVs and examined reactome pathways, which showed terms such as axon guidance, and regulation of expression of SLITs and ROBOs (Figure 4A). STRING analysis also revealed 22 clusters, the largest cluster containing 21 proteins involved in eukaryotic translation elongation and cytoplasmic translation. Other major clusters were involved in SLIT‐ROBO signaling, chaperone complex activity, and the signalosome (Figure 5A). Canonical pathway, signaling by ROBO receptors, was the most upregulated pathway and contained the largest number of proteins (Figure 5A). Other canonical pathways such as mitotic and cell signaling pathways and eukaryotic translation initiation were also among the most upregulated pathways; all which are associated with uncontrolled cancer cell growth. Two of the most downregulated signaling pathways were semaphorin neuronal repulsion and RHO GDP dissociation inhibitors, both which are involved in cancer progression by mediating differentiation and migration. Additionally, IPA analysis revealed EV cargo proteins involved in cancer cell proliferation and invasion (CCT8, ANXA1, HSP90AB1), epithelial‐mesenchymal transition (CD44, DPYSL3), and multiple growth factors that can trigger cancer signaling pathways (MAPK/ERK, EIF2, PI3K) (Figure 5B). We validated some of the top biotinylated EV proteins such as annexin 1 (ANXA1), heat shock protein 90 beta (HSP90AB1), and CD44 (Figure S3E). To ensure that the turboID method did not affect native protein enrichment, GBM 612 EV proteins were analyzed using proteomics. This revealed 288 proteins captured (Figure S3F). When comparing this number of proteins to the number of significantly enriched proteins from the GBM 612 turbo + B EVs, there were 163 overlapping. Lastly, we did pathway analysis for cellular component for the 163 overlapping proteins and found 80.6% of these proteins associated with exosomes (Figure S3G). Overall, this suggests that our TurboID‐based approach effectively captures a broad spectrum of EV cargo relevant to tumour microenvironment interactions. A detailed list of all the identified proteins and their functional annotations is provided in Table S1 and their abundances are provided in Table S2.

FIGURE 3.

Proteomic analysis of TurboID GBM‐derived EV proteins. (A) Venn diagram showing 455 overlapping proteins between the significant EV proteins identified in our GBM612 Turbo+B dataset and GBM cell line–specific proteins from the Vesiclepedia database. (B) Ingenuity Pathway Analysis of upregulated diseases and disorders and their p values in Turbo +B EVs compared to Turbo ‐B EVs. (C) The percentage of proteins associated with cellular components and their p values in Turbo +B EVs compared to Turbo ‐B EVs. (D) The percentage of proteins associated with molecular functions and their p values in Turbo +B EVs compared to Turbo ‐B EVs.

FIGURE 4.

Reactome pathway enrichment for top 50 TurboID GBM‐derived EV proteins. (A) STRING analysis for reactome pathways of top 50 612 turbo + B EV proteins. Proteins were grouped by similarity with a score of 0.8.

FIGURE 5.

STRING and IPA reveal pathways associations for TurboID GBM‐derived EV proteins. (A) STRING analysis on the top 100 612 turbo + B EV proteins. (B) Ingenuity Pathway Analysis of canonical pathways in Turbo +B EVs, shown by the number of proteins in each pathway and their z‐score, compared to Turbo ‐B EVs.

3.8. Transfer of Biotinylated EV Proteins to Other Cells

We then tested the applicability of this TurboID labelling system in EVs and how it can be applied to tracing EV proteins from one cell to another. HEK293T cells were used as recipient cells for 612 wt + biotin EV and 612 turbo + biotin EV treatment (Figure 6A). Successful transfer of biotinylated EV proteins was validated using western blot, as evidenced by detection of V5 and strongest biotinylation in HEK293T cells treated with 612 turbo + biotin EVs (Figure 6B) (Li et al. 2023). These results show successful transfer of biotinylated EV proteins and open an avenue for tracing intravesicular protein cargo from one cell to another.

FIGURE 6.

Successful transfer of biotinylated EV proteins to recipient cells. (A) Experimental workflow for treating HEK293T cells with EVs derived from 612 wt + biotin and 612 turbo + biotin EVs. (B) Western blot showing transfer of V5 and biotinylated proteins to recipient HEK293T cells.

4. Discussion

In this study, we leveraged the TurboID proximity labelling system to achieve global and unbiased biotinylation of proteins within GBM‐derived EVs. Our findings provide insights into the complex protein cargo of EVs released by BTICs and the mechanisms by which GBM cells interact with and manipulate the tumour microenvironment. The application of TurboID in primary human BTICs represents a significant advancement in EV research. Traditional methods for studying cell‐cell interactions often face challenges such as low specificity, potential cell toxicity (Cho et al. 2020; Roux et al. 2012; Gingras et al. 2019; Rhee et al. 2013) and lengthy labelling times (Roux et al. 2012). Using TurboID, we achieved rapid and effective biotinylation of cytoplasmic proteins, capturing a comprehensive snapshot of proteins packaged into EVs. This approach overcomes previous limitations, such as the addition of harmful reagents to cells (Rhee et al. 2013) or a long labelling time (Roux et al. 2012), and allows for the identification of both surface and internal EV proteins, enhancing our understanding of the molecular constituents involved in GBM progression.

There are many studies in the EV field that revolve around understanding how EVs are transferred intercellularly. Most of these papers utilise the fluorescent tagging of EVs or an EV membrane protein (Bonsergent et al. 2021; Joshi et al. 2020; Gurrieri et al. 2024; Seo et al. 2024; Luhtala et al. 2017) to look at just a singular donor and recipient cell interaction. In 2018, Luhtala et al. reported there to be no transfer of a specific protein from donor EVs to recipient cells (Luhtala and Hunter 2018). By using cell‐specific protein labelling, our study effectively labels EV protein cargo, not just membrane proteins, in an unbiased manner. Our method was able to capture both EV tetraspanins, such as CD81, CD63 and CD9, as well as proteins involved in the ESCRT pathway, such as TSG101, and proteins reflective of the cell of origin, such as CD44. Additionally, when performing proteomic and STRING analysis on naïve GBM‐EVs, we found an expected overlap of 80% of proteins associated with exosomes and 81% of the proteins associated with cytoplasm as well.

TurboID is a method that has been exclusively used in mammalian cells (Cho et al. 2020) and only once previously in 2023 to label an EV surface protein (Li et al. 2023). We directed the expression of this biotin ligase, TurboID, to the cytoplasm of the cell so that we can covalently label proteins that get packaged into EVs as the proteins are being moved through the endosomal pathway and shuttled to the surface of the cell to be released into extracellular space (van Niel et al. 2018; Trajkovic et al. 2008). This allows for global and unbiased labelling of EV proteins and opens the door for future research to utilize this method to get a better understanding of the communication of multiple EV proteins to recipient cells and what effects they have. When comparing the enrichment of proteins from GBM‐EVs, there were less proteins in native GBM‐EVs when compared to GBM‐EVs with the turboID tag. This suggests that the addition of biotin to the EV proteins can allow for better enrichment of proteins from EVs, along with a wider range of proteins inside the EV. This can be advantageous, particularly for EV proteins, because it can allow for greater capture of proteins inside EVs that may not be captured normally due to low expression.

In recent years, there has been an emergence of using proximity‐labelling to map cell‐specific protein activity in cells (Cho et al. 2020; Gingras et al. 2019). Proximity labelling systems utilise an enzyme to covalently label proximal molecules in an unbiased manner (Cho et al. 2020). The use of a covalent label, most often biotin, allows for a strong and stable way to perform downstream enrichment using streptavidin‐conjugated beads or anti‐biotin beads, and proteomic profiling of labelled molecules. Engineered peroxidases, such as APEX2, have been emerging as a way to label proteins for fluorescent‐based live‐cell proteomics (Lam et al. 2015). Although APEX2 has been shown to have strong temporal resolution, it requires exogenous H₂O₂ to activate the labelling, which is toxic to cells and organisms (Qu et al. 2025). Biotin ligase‐based proximity labelling techniques, such as BioID (Roux et al. 2012), are advantageous because they do not utilise reagents that may be harmful to cells, but they have been known to need long labelling times, sometimes greater than 18 h (Roux et al. 2012; Choi‐Rhee et al. 2004). To address these challenges, we employed TurboID, a novel proximity‐dependent biotinylation enzyme that enables rapid and unbiased labelling of proteins in living cells. TurboID uses ATP and biotin to covalently tag proximal proteins in minutes, allowing for efficient capture and identification via mass spectrometry (Cho et al. 2020; Branon et al. 2018). Additionally, this is the first study to use TurboID in primary patient‐derived cell lines, whereas reported use of this system in the past has been in HEK293T cells (Cho et al. 2020; Branon et al. 2018; Li et al. 2021), Drosophila melanogaster or Caenorhabditis elegans (Branon et al. 2018), or commercially available cell lines (Yuan et al. 2024; Medica et al. 2023), or murine‐derived cells (Day et al. 2023). Patient‐derived cell lines can be heterogeneous, so testing the TurboID system in multiple cell lines to identify which has high transduction efficiency was necessary.

Our optimised EV isolation protocol demonstrated superiority over conventional ultracentrifugation methods. Standard ultracentrifugation methods have evolved to include a size exclusion chromatography step followed by a short ultracentrifugation spin (Cappe et al. 2023). This method yields intact vesicles but requires a large volume of starting media (100 mL), purchasing of the size exclusion columns, and relies on having an ultracentrifuge. Our protocol does not require any expensive materials or large pieces of equipment. Additionally, by combining differential centrifugation with gentle precipitation, we preserved EV integrity and obtained higher yields from smaller volumes of conditioned media. Our protocol begins with 7–14 mL of conditioned media, whereas commonly used tangential flow filtration (TFF) method requires 500 mL of cell culture media to extract EVs (Zeng et al. 2022). Moreover, Tang et al. used ∼100 mL of conditioned media before they tested the efficiency of precipitation, ultracentrifugation, and ExoQuick to extract EVs, and in every case, the particle concentration was lower than the one we obtained using our optimised protocol (Tang et al. 2017). Additionally, TFF requires a lengthy procedure involving cleaning, flushing, and passing the media slowly through the TFF system, whereas our protocol is more time‐efficient and does not involve costly equipment. Ultracentrifugation, either by density gradient or sucrose gradient, can be used to generate an EV pellet (Lobb et al. 2015), and relies on the use of an ultracentrifuge. Our gentle EV isolation method uses differential centrifugation at low speed to ensure the recovery of intact EV membranes, which is crucial for functional studies and for accurately characterising EV content. We also did not observe any contamination with albumin, and NTA revealed uniformly sized EVs when using our method. This method also allowed for the separation of vesicle‐free media (VFM), providing an opportunity to study soluble factors independently of EVs. Although we have shown an optimised method to maintain intact EVs for downstream profiling, it is important to note that there remains a gap in the field for a gold standard protocol that can be applied to bodily fluids, cell culture, and tissue.

Our proteomic analysis of biotinylated EV proteins revealed the presence of key molecules implicated in tumour growth, invasion, migration, and proliferation. GBM tumours are characterised by being highly aggressive, with stem cells that are invasive and resistant to therapy (Xie et al. 2024; Gimple et al. 2019; Bao et al. 2006; Chen et al. 2012). More recently, EVs have been explored as a method of intercellular transferring for these tumourigenic factors (Mathieu et al. 2019; Balakrishnan et al. 2020). In this study, we chose to focus on identifying EV protein cargo that may be involved in these malignant processes, as RNA cargo has been heavily studied (Akers et al. 2013; Xia et al. 2019; Holdhoff et al. 2013), with few publications highlighting proteins and only as possible biomarkers (Greco et al. 2021; Redzic et al. 2014). Although, the technique used here can also help to complement existing studies focused on RNA, to have a deeper understanding of the intravesicular cargo at a cell‐specific level. The identification of molecules such as transcription regulators, kinases, enzymes and cytokines suggest that GBM‐derived EVs play a pivotal role in promoting cell proliferation, cell movement and invasion capabilities, and immune response modulation. These findings align with the concept that EVs act as vehicles for intercellular communication, enabling GBM cells to influence neighboring cells and create a microenvironment conducive to tumour survival and expansion.

The comprehensive profiling of EV protein cargo provides a valuable resource for further understanding biological mechanisms and identifying potential therapeutic targets. Proteins that are selectively enriched in GBM‐derived EVs and are involved in critical cancer‐promoting pathways may serve as candidates for targeted therapy. This system can be applied to co‐culture and in vivo models to trace specific protein origin and communication to nearby cells. Similar techniques have been used in vivo and in co‐culture models in GBM with other cells of the tumour microenvironment (Norton et al. 2024), but there have been no studies to elucidate EV protein cargo tracing. Here, we show the robust applicability of this technique to trace EV protein cargo from GBM‐derived EVs to recipient cells. Additionally, the stability and accessibility of EVs in bodily fluids make them attractive candidates for biomarker development (Greco et al. 2021; Redzic et al. 2014; Li et al. 2017). The detection of specific EV proteins in patient samples could aid in early diagnosis, monitoring of treatment response, and detection of recurrence. Additionally, the successful application of our methods to non‐cancerous cell lines, such as HEK293T cells and MSCs, highlights the versatility of the TurboID system and our EV isolation protocol. This suggests that our approach can be generalised to study EV‐mediated communication in various biological contexts, broadening its impact beyond GBM research.

Future studies are essential to deepen our understanding of the protein cargo within GBM‐derived extracellular vesicles (GBM‐EVs). Functional validation of identified proteins is crucial to elucidate their roles in tumour microenvironment adaptation. Mechanistic investigations are needed to determine how EV proteins influence recipient cells within the tumour microenvironment. While our methods were applied in an in vitro setting, in vivo validation is necessary to confirm their physiological relevance, particularly in non‐cancer cell communication. Additionally, although we demonstrate the transfer of biotinylated EV proteins, comprehensive proteomic analysis of both isolated GBM‐EVs and recipient cells following TurboID EV treatment is required. This approach would enable a more precise identification of EV proteins transferred from cancer‐derived EVs to non‐cancerous recipient cells. Investigating EV protein cargo in patient‐derived biofluids, such as cerebrospinal fluid or blood, could further enhance the translational potential of our findings. Finally, while TurboID enables rapid labelling, its high reactivity and the presence of endogenous biotin necessitate stringent experimental controls to mitigate non‐specific biotinylation. Optimising biotin concentrations and labelling durations can help reduce background signal and improve the specificity of protein identification.

In conclusion, we have successfully utilised the TurboID proximity labelling system to achieve global and unbiased profiling of proteins within GBM‐derived extracellular vesicles from primary human BTICs. Our findings reveal a diverse array of EV proteins involved in critical pathways of tumour progression and microenvironment adaptation. The optimised EV isolation protocol ensures the integrity and purity of vesicles, facilitating accurate characterisation of their cargo. This work represents a significant step forward in understanding the mechanisms of EV‐mediated communication in GBM. By identifying potential therapeutic targets and biomarkers, our study lays the groundwork for future research aimed at improving diagnosis, monitoring, and treatment of GBM. The methodologies developed herein have broad applicability and may be leveraged to investigate EV roles in other cancers and diseases, ultimately contributing to the advancement of personalised medicine and targeted therapies.

Author Contributions

TuKiet Lam T: methodology, funding acquisition, validation, writing–review and editing, formal analysis, project administration. Marissa N. Russo: conceptualization, investigation, writing–original draft, methodology, validation, visualization, writing–review and editing, formal analysis, data curation. Emily S. Norton: conceptualization, investigation, methodology, writing–review and editing. Maria José Ulloa‐navas: methodology, writing–review and editing. Natanael Zarco: conceptualization, methodology, writing–review and editing. Weiwei Wang: methodology, validation, writing–review and editing, formal analysis. Alfredo Quiñones‐hinojosa: writing–review and editing, resources, funding acquisition. Veronique V. Belzil: writing–review and editing, methodology, conceptualization. Hugo Guerrero‐Cázares: conceptualization, investigation, funding acquisition, writing–original draft, visualization, writing–review and editing, formal analysis, project administration, resources, data curation, supervision.

Ethics Statement

Samples were collected after informed consent. All the sample collection procedures were approved by the Mayo Clinic Institutional Review Board.

Conflicts of Interest

The authors declare no competing interests.

Supporting information

Supporting Figure: jev270158‐sup‐0001‐FigureS1.pdf

Supporting Figure: jev270158‐sup‐0002‐FigureS2.pdf

Supporting Figure: jev270158‐sup‐0002‐FigureS2.pdf

Supporting Table: jev270158‐sup‐0004‐TableS1.xlsx

Supporting Table: jev270158‐sup‐0005‐TableS2.xlsx

Russo, M. N. , Norton E. S., Ulloa‐Navas M. J., et al. 2025. “TurboID‐Mediated Profiling of Glioblastoma‐Derived Extracellular Vesicle Cargo Proteins.” Journal of Extracellular Vesicles 14, no. 9: e70158. 10.1002/jev2.70158

Data Availability Statement

Proteomics data will be deposited in the PRoteomics IDEntifications Database (PRIDE) from EMBL‐EBI accession number: PXD063142.