Plasmids and cell lines

Trastuzumab light chains 1 and 2 were obtained by Tebubio Srl and cloned in the CD81-GFP vector (OriGene, 7268 bp), obtaining the antiHER2 construct (CD81-antiHER2-GFP, 7901 bp) still fused to the turbo GFP reporter. Human embryonic kidney HEK293T (ATCC, CRL-3216), human breast cancer MDA-MB-231 (ATCC, HTB-26) and SK-BR-3 (AMSBIO, Abingdon, UK) cell lines were cultured under standard conditions in DMEM supplemented with 10% Fetal Bovine Serum, 2 mM L-Glutamine, and 100 U/ml penicillin–streptomycin (all Gibco). SK-BR-3 HER2-knockout (SK-BR-3 KO) were obtained using pSpCas9 BB-2A-Puro (PX459) V2.0 (Addgene, 9200 bp) containing a sgRNA sequence (5ʹ-TCATCGCTCACAACCAAGTG-3ʹ) targeting exon 7 of ERBB2 (cloned by Twin Helix) and selected with puromycin (Sigma-Aldrich) for 4 days (1 μg/ml puromycin for 72 h, followed by 24 h at 2 μg/ml). MDA-MB-231 cells expressing HER2 (MDA-MB-231 HER2 OE) were obtained by transient transfection of the pCMV3-SP-N-HA vector (SinoBiological, 6086 bp) for 24 h. Cells were transfected using either Lipofectamine 3000 (Invitrogen) or polyethylenimine (PEI, Sigma Aldrich) according to manufacturer’s protocols.

Cell fractionation and immunoblotting

Cell fractionation experiments were performed as already described [40] using lysis buffer (50 mM HEPES pH 8, 10 mM NaCl, 10 mM MgCl2, 1 mM DTT, 10% glycerol, 1X protease inhibitor cocktail) supplemented by 25 µg/ml Digitonin (buffer A), 1% Igepal (buffer B) or 1% Triton X-100 and 1% Sodium deoxycholate (buffer C) for the sequential incubation and centrifugation protocol. Input samples corresponded to 2% of whole cell lysate from all the conditions analyzed. The first supernatant (cytosolic fraction) was collected after incubation of cells with buffer A on a rotary shaker for 10 min at 4 °C, then centrifuged at 2,000 rcf for 10 min at 4 °C. The obtained pellet was resuspended in ice-cold buffer B and vortexed before incubation on ice for 30 min and centrifuged at 7000 rcf for 10 min at 4 °C. The resulting supernatant corresponded to the organelle-enriched fraction, while the pellet was resuspended in ice-cold buffer C with the addition of benzonase (Novagen) and incubated on a rotary shaker for 30 min at 4 °C. Next, samples were sonicated at 4 °C for 45 s at 35 Amplitude (three cycles of 10 s on and 5 s off) within a ultrasonic bath sonicator (Q700, QSonica) and centrifuged at 7,800 rcf for 10 min at 4 °C to collect the nuclear fraction. Each fraction was loaded on 13% polyacrylamide gel for SDS-PAGE.

Protein concentration was measured in triplicate with the bicinchoninic acid (BCA) protein assay kit (Thermo Fisher Scientific). Immunoblotting experiments were performed as already described [41] using the following antibodies: tGFP (TA150041, OriGene), SYNTENIN (ab133267, Abcam), CD9 (ab236630, Abcam), CALNEXIN (ab22595, Abcam), HER2 (ab237715, Abcam), RAB5 (C8B1, 3547, Cell Signaling Technology), GAPDH (GTX627408, GeneTex), H3 (GTX122148, GeneTex), SERCA2 (ab2861, Abcam), TSG101 (GTX70255, GeneTex), secondary antibodies Peroxidase AffiniPure Goat Anti-Mouse and Anti-Rabbit IgG (H + L) (Jackson ImmunoResearch).

EV isolation and characterization

For EV collection, HEK293T were plated in 100 mm TC-treated Culture Dish (Corning) and transfected with 3 µg plasmid/dish when at 75% of confluence. Forty-eight hours later, cells were washed with PBS and incubated for additional 24 h with serum-free medium. Media were centrifuged at 2,800 rcf for 15 min to eliminate cell debris and bigger particles. The supernatant was transferred to ultracentrifuge tube (38.5 ml-Beckman Coulter Ultra-Clear™) for ultracentrifugation at 100,000 rcf for 70 min at 4 °C using a Optima XE-90 ultracentrifuge with a SW 32 Ti rotor (Beckman Coulter). The EV pellet was resuspended in 0.22 μm-filtered sterile PBS and freshly-used or stored at -80 °C until further use. Nanoparticle tracking analysis (NTA) was performed using a NanoSight NS300 instrument (Malvern Panalytical, Malvern, UK) with a 532 nm laser. Each sample was subjected to 3–5 consecutive 60 s videos recorded at camera level 15. Particle concentration and size distribution were determined using NanoSight NS300 software NTA 3.4 Build 3.4.003 (Malvern Panalytical), setting 4 as a detection threshold.

Imaging flow cytometry was performed using an ImageStreamx (ISX) MKII instrument (Luminex Corporation) at 60 × magnification, high gain mode, and low flow rate. EV samples in PBS were labeled with 1 μg/ml (final concentration) of Cell Mask Deep Red (CMDR, Invitrogen) in ratio 1:1 (v/v) and incubated at RT for 20 min. Then, samples were diluted in PBS to obtain a final concentration lower than 10^10 objects/ml before acquisition. All samples were analyzed using INSPIRE® software (Luminex Corporation, Seattle, WA, USA), with a minimum of 3,000 events collected. Data analyses were performed using ISx Data Exploration and Analysis Software (IDEAS®, Luminex Corporation, Seattle, WA, USA).

Cryogenic electron microscopy (Cryo-EM) acquisitions were performed at Fisher Scientific (Eindhoven, The Netherlands). QuantiFoil 1.2/1.3 cu 200 grids, pretreated with 30 s glow discharging, were used. Vitrobot parameters were set as follows: 3.0 μl of each sample, temperature 4 °C, humidity 95%, blot time 7 s, blot force 0. Fifteen cryo-electron images were collected per each sample using a transmission electron microscope Glacios (Thermo Fisher Scientific), equipped with a Falcon 4i Selectris camera, at a nominal magnification of 49,000x. Imaging parameters were set as follows: Pixel size (Å) 2.4, Dose rate (e/pix/sec) 12.4, Total dose (e/Å^2) 20, Exposure time (sec) 9.3, Energy filter (eV) 10, defocus -1.9 μm. For image analysis, vesicular structures were manually selected as regions of interest (ROIs) using FIJI software and the size measured in pixels. From the ROI's pixel size, the actual area and diameter of EVs were calculated.

AlphaLISA interaction assay

The Amplified Luminescent Proximity Homogeneous Assay (ALPHA Assay) was performed in white 384-well Optiplates (PerkinElmer) in a final volume of 20 μl. The antiHER2 protein was produced from CD81-antiHER2-tGFP vector (600 ng) using TNT® Quick Coupled Transcription/Translation Systems (Promega), following manufacturer’s instructions. The resulting product was pre-incubated with 30 nM HER2-DDK (TP322909, OriGene) for 30 min before addition to the other components: 10 nM tGFP Ab (TA150041, OriGene), anti-FLAG donor and protein G coated acceptor beads (10 ng/μl final concentration; AS103D and AL102C, PerkinElmer). For the competitive assay with CD81-GFP and antiHER2 EVs, 40 nM HER2-DDK and 9 nM anti-HER2 Ab (ab237715, Abcam) were used. All the components were diluted in the Alpha Screen Control Buffer (Perkin Elmer). The signal was detected using an EnSight® multimode plate reader (PerkinElmer) after 1 h of plate incubation at RT in the dark under rotation at 70 rpm.

Immunoprecipitation and immunofluorescence

Serum-free media from transfected HEK293T was concentrated to about 2 ml with Amicon Ultra-15 Centrifugal Filter 10 kDa MWCO (Merck Millipore) at 3,000 rcf for 15–20 min, and diluted according to the relative concentration of GFP-positive particles. Input samples were prepared from 15 μl of undiluted conditioned medium. Anti-FLAG M2 magnetic beads (Sigma-Aldrich) or protein G dynabeads (Invitrogen) were washed three times in PBS before pre-incubation, at 4 °C for 1.5 h under rotation, with HER2-DDK or tGFP antibody, respectively. Bead-antibody mixtures were added to the media for 20–30 min at 4 °C under gentle rotation. For competitive HER2 binding, 1 μg/sample of Trastuzumab (anti-HER2-Tra-hIgG1, InvivoGen) was added directly to the media before incubation with beads. After the final incubation, beads were washed with PBS and resuspended with 1X Laemmli sample buffer for subsequent denaturation at 98 °C for 5 min and SDS-PAGE.

Immunofluorescence for HER2 detection was performed in breast cancer cell lines fixed with 4% paraformaldehyde for 15 min at RT and washed three times with cold PBS. Blocking (10% FBS, 0.05% Triton X-100 in PBS) was performed for 1 h at RT followed by primary Ab (ab237715, Abcam, diluted 1:1000, 0.1% FBS in PBS) incubation for 1 h at RT. After three washes with PBS, cells were incubated for 1 h at RT with Goat anti-Rabbit Alexa Fluor™ 633 (Invitrogen, diluted 1:1000, 0.1% FBS in PBS). Three washes with cold PBS 5 min each were performed before the addition of Hoechst for 15 min. One final wash with cold PBS was performed before acquisition. For CD81 and RAB5 IF, HEK293T were seeded on optical coverslips and IF was similarly performed, with the addition of a 5 min permeabilization step (0.1% Triton X-100 in PBS) before primary Ab incubation overnight at 4 °C (0.1% FBS and 0.05% Triton X-100 in PBS). CD81 Ab (MA5-13,548, Invitrogen) or RAB5 Ab (C8B1, 3547, Cell Signaling Technology). Goat anti-Mouse or anti-Rabbit IgG (H + L) Cross-Adsorbed Secondary Antibody Alexa Fluor™ 568 (Invitrogen) were used for the assays. Coverslips were mounted using ProLong Diamond Antifade Mountant (Invitrogen).

Confocal microscopy and image analysis

Time-lapses and images of EV uptake by recipient breast cancer cells were acquired at the Optical Imaging Centre (OIC) at Erasmus MC (Rotterdam, The Netherlands) with a LEICA TCS SP8 AOBS confocal microscope, with Galvo Z stage and Adaptive Focus Control, using a HC Plan Apo CS2 40x/1.3 oil immersion objective. Live cell confocal imaging was performed under humidified conditions with 5% CO2 at 37 °C. MDA-MB-231 and SK-BR-3 cells were seeded in glass-bottom dishes (CELLview™ Culture dish, 35 mm, four chambers) and, before acquisition, 340 μl DMEM already containing 1.25 nM LysoTracker™ Red DND-99 (L7528, Invitrogen) and 1 µg/ml Hoechst (62,249, Thermo Scientific) were added to each chamber for 15 min. Before starting acquisition CD81-GFP or antiHER2 EVs were added in a ratio of 20,000–50,000 per seeded cell, in 350 μl as final volume. For fixed cell acquisitions, MDA-MB-231 (WT and HER2 OE) and SK-BR-3 (WT and KO) were seeded in the same dishes and incubated with EVs for 4 h considering the relative abundance of GFP-positive EVs, then washed with PBS before fixation and immunofluorescence (IF) for HER2 receptor. Fourteen Z-stacks were acquired within around 11 μm of total Z size, with voxel size 0.1623 × 0.1623x0.7991 μm3. Images were processed and analyzed using FIJI/ImageJ software.

To quantify EV-cell interactions, an automated pipeline was applied using CellProfilerTM version 4.0.7 on the Maximum Intensity Projection of 14 z-stacks for each acquired channel. Cell nuclei were identified as primary objects with a threshold in the Hoechst blue channel, the cytoplasms were defined as secondary objects using a low threshold in the HER2 red channel that allowed the segmentation of a cytoplasmic region also in the HER2 negative cells. EVs were defined as green spots in a range of diameters from 0.8 to 5 µm and the interaction within the cells has been established overlapping the EVs object mask to the cytoplasmic region using the object processing function “RelateObject”. Objects number, AreaShape and Red Intensity features were calculated for the identified Nuclei, Cytoplasms and EVs. The final Spreadsheets have been combined and the dataset has been analyzed using KNIME Analytics Platform v 4.7.5.

Images in Fig. 1A, B were acquired with a Nikon AX laser scanning inverted confocal microscope, using a Plan Apo H 60x/1.4 oil immersion objective. Images in Fig. 4A and Fig. S3B were acquired with a Nikon Ti2 inverted microscope equipped with a Crest X-light V2 Spinning Disc system and an Andor iXon Ultra 888 EMCCD camera, using a Plan Apo 20x/0.75 objective. Images in Fig. S1B were acquired with a Nikon Ti2 inverted microscope equipped with a Crest X-light V2 Spinning Disc system and an Andor Zyla 4.2 PLUS sCMOS camera, using a Plan Apo 20x/0.75 objective. For all the acquisitions, settings were kept constant within the same experiment and linear adjustments for brightness and contrast were equally applied to the reported images.

Fig. 1

CD81 fusion proteins are expressed in HEK293T upon transient transfection and co-sediment with organelle-enriched sub-cellular fractions. A, B GFP detection and immunofluorescence staining of endogenous CD81 and RAB5 proteins in transfected HEK293T cells. Cell were subjected to confocal microscopy after 48 h of transfection with CD81-GFP and antiHER2 plasmids. Recombinant proteins are visualized in green (GFP), endogenous CD81 or RAB5 in magenta (Alexa Fluor 568), and cell nuclei in cyan (Hoechst). Scale bar is 20 μm in A and 10 μm in B. C Immunoblotting of sub-cellular fractions obtained through a sequential lysis buffer-centrifugation protocol. Separation of subcellular fractions was confirmed by the enrichment of corresponding protein markers: Cytosol (GAPDH), nuclei (histone H3), and organelles (SERCA2 for endoplasmic reticulum, RAB5 for early-endosomes). GFP-positive chimeric proteins were detected at the expected molecular weight (45 for CD81-GFP and 75 kDa for antiHER2). The histogram reports the densitometric quantification normalized over CD81-GFP condition, with mean and SD of two independent experiments. Significance is *P < 0.05

RNA isolation and digital droplet PCR

Total EV-RNA was extracted with TRIzol reagent (Ambion, Life Technologies) and chloroform precipitation, followed by single-cell RNA purification (Norgen kit) including on-column DNAse I (Qiagen) treatment for 10 min at RT. cDNA synthesis was performed following manufacturer’s instructions (SensiFAST™, Meridian Bioscience™) starting from 15 μl of RNA template. To assess the integrity of the fusion protein-encoding transcripts (CD81-GFP, antiHER2), the corresponding cDNA samples were amplified prior DNA electrophoresis. ddPCR experiments were carried out using EvaGreen, following manufacturer’s instructions (Bio-Rad). EV-derived cDNA samples (5.5 μl each) were mixed with 11 μl of 2X QX200™ ddPCR™ EvaGreen Supermix (Bio-Rad) and 5.5 μl of primers (35 nM each). The following primers were used to specifically amplify the fusion sequences of interest: A) 5’-CTTCAAGGAGGACTGCCAC & 5’-GCTCGCGCTATAAATCAGCAGT; B) 5’-TGACCAAAAGCTTTAACCGTG & 5’-TGGGGTAGGTGCCGAAGT; C) 5’-CTTCAAGGAGGACTGCCAC & and 5’-TGGGGTAGGTGCCGAAGT) and target concentration was determined using QuantaSoft Software™ (Bio-Rad).

RNA extraction from tumor xenografts was performed with TRIzol and RNeasy Mini Kit (Qiagen), following the manufacturer’s instructions. cDNA synthesis was carried out using WarmStart reverse transcriptase (New England Biolabs) from 6 μg of RNA as starting material, including couple B primers (160 nM each). RNAse H from E. coli (Illumina) was added (10U/sample) for 30 min at 37 °C. cDNA purification was performed using the Nucleospin gel and PCR clean-up kit (Macherey–Nagel) following manufacturer’s instructions. ddPCR reactions were performed with cDNA diluted 1:300.

UHPLC-MS analysis of secreted doxorubicin

To collect doxo-EVs from drug-treated cells, transfected HEK293T were incubated with 10 μM doxorubicin (BD32885, BLD Pharmatech GmbH) for 3 h at 37 °C and 5% CO2. After a PBS wash, EVs were isolated from serum-free DMEM as previously described. Metabolite extraction was carried out by adding 80% cold methanol to the EV stocks. Samples were then vigorously shaken (5 min) and kept at -80 °C overnight. Finally, the samples were vacuum-dried using a SpeedVac concentrator. Dried extracts were equilibrated to RT, resuspended in 30 µl of acetonitrile:water:formic acid (5:95:0.1%, v/v/v), and thoroughly mixed. Seven serial drug dilutions, ranging from 8,000 to 1.95 nM, were used to prepare the standards for the calibration curves. Ten μl of standards and samples were injected onto Ultimate 3000RS (Thermo Scientific) UHPLC system coupled online with an Orbitrap Fusion Tribrid (Thermo Scientific) mass spectrometer. A Hypersil Gold C18 column (Thermofisher, 100 × 2.1 mm, particle size: 1.9 µm) was used for separation. The LC method consisted of a linear gradient from 5 to 100% B (B: acetonitrile 0.1% of formic acid; A: water + 0.1% formic acid) over 15 min, followed by 4 min at 100% at the flow of 0.2 ml/min. The MS spray voltage was set at + 3500 V with the ion transfer tube temperature set at 300 °C (sheath and auxiliary gasses were set at 20 and 5 Arb, respectively). The MS data were acquired in full scan in the Orbitrap at 120.000 FWHM (200 m/z), in the scan range of 100–1000 m/z. Software FreeStyle ver.1.6 (Thermo Scientific) was used to inspect mass spectra. The area under the peak of each precursor ion and the total ion current (TIC) were extracted using Skyline (MacCoss Lab Software) [42]. The software derived precise m/z values as well as isotope distributions for each precursor. Each analyte was also investigated for common adducts, [M + H] + , [M + K] + , [M + NH4] + , [M + Na] + , and the [M − H2O + H] + ions, for each considering the three most abundant isotopes. The total area of precursor ions was calculated by summing all precursor levels. The log2-transformed value of the total area of precursor ions of each standard sample was plotted as function of the concentration to construct the standard curves of each compound.

Cell viability assay

The 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide (MTT, Thermo Scientific) assay was performed on cells seeded and treated in 96 plates for 72 h with the free doxorubicin or doxo-EVs. Doxorubicin concentration in doxo-EVs was estimated from previous LC–MS analysis. After incubation, the medium was removed and MTT solution (0.5 mg/ml in DMEM) incubated for 4 h at 37 °C and 5% CO2, before cell lysis in DMSO. Absorbance (570 nm) was measured at a Varioskan LUX Multimode Microplate Reader (Thermo Fisher Scientific) and cell viability was calculated as % with respect to untreated cells (DMEM only).

In vivo study

The in vivo study was performed in collaboration with Reaction Biology Europe GmbH (Freiburg, Germany). Each experimental groups contained five female athymic nude mice (Crl:NU(NCr)-Foxn1nu). On Day 0, 5.0 × 106 JIMT-1 human breast carcinoma tumor cells in 100 µl PBS were implanted into the left mammary fat pad of each mouse. After animals had been randomized on Day 12, treatments of the test samples were initiated. All treatments were administered at a dose of 0.5 µg/kg and a dosing volume of 5 ml/kg subcutaneously at the tumor implantation site. Animal weights were measured three times, one time on day of randomization (Day 12) and two times after the start of therapy (Days 13 and 15). During the study, the growth of the intramammary implanted JIMT-1 primary tumors was determined twice by caliper measurement on Days 12 and 15. All animals reached the end of the study as scheduled. The primary tumor samples were collected during the final necropsy on Day 15, the study endpoint.

Statistical analysis

Statistical analysis was performed using GraphPad Prism 9, as well as data visualization, applying non-parametric Anova Kruskal–Wallis or Student’s t-test. A minimum of 95% confidence level was considered significant. The significance level was set at *P < 0.05; **P < 0.01; ***: P < 0.001; ****P < 0.0001; ns: not statistically significant. Mean and standard deviation of independent experiments are reported in the graphs and detailed in figure legends. Schematics were created in biorender.com.