All chemicals, unless otherwise stated, were purchased from Merck KGaA.

The human K562 chronic myeloid leukemia cell line was purchased from the German Collection of Microorganisms and Cell Cultures, GmbH and cultured in RPMI medium (Thermo Fisher Scientific, Inc.) supplemented with 10% fetal bovine serum (FBS; Gibco; Thermo Fisher Scientific, Inc.), 100 U/ml penicillin G, 100 µg/ml streptomycin and 7.5% sodium bicarbonate at 37°C in an humidified incubator with 5% CO₂. The K562^IR cells were derived from original imatinib-sensitive K562 cells, that had been cultured in gradually increasing concentrations of imatinib in the culture medium (from 0.1 to 2 µM) for 9 months, as previously described (23). The concentration of imatinib was increased by 0.1 µM at each step and was maintained for 15-30 days, depending on the proportion of surviving cells. The resulting K562^IR cell line was resistant to 2 µM imatinib.

The K562 and K562^IR cell lines were cultured in the presence of different concentrations (0 to 2 µM) of imatinib for 3 days. Imatinib toxicity was determined by measuring cell viability using a Vybrant™ MTT cell proliferation assay kit (Thermo Fisher Scientific, Inc.), following the manufacturer's instructions. Proprietary solvent B containing SDS (from the kit) diluted with 0.01 M HCl was used to solubilize the purple formazan. Absorbance was detected at 570 nm using a microplate reader (Chameleon; Hidex Oy). Data was analyzed using MikroWin 2000 software, v4.0 (Mikrotek Laborsysteme GmbH).

The K562 and K562^IR cell lines were analyzed using a next-generation sequencing (NGS) method, as previously described (24). Briefly, the amplicon library was prepared using a two-step selective amplification of cDNA, including the BCR-ABL1 kinase domain. At the first step, BCR-ABL1 cDNA was amplified, then primers, from the IRON II study (25), were used to prepare four 350 bp amplicons of the kinase domain. Sequencing was subsequently performed on a GS Junior 454 System, and the data was analyzed using the Amplicon Variant Analyzer software (both from Roche Diagnostics). The raw NGS data are freely available via NCBI Sequence Read Archive (https://www.ncbi.nlm.nih.gov/sra) under the project accession number, PRJNA664680, while the imatinib sensitive K562 and the imatinib resistant K562 cells have the accession numbers, SRX9210642 and SRX9210641, respectively.

The number of BCR-ABL1 gene copies was determined using the quantitative droplet digital PCR (ddPCR) method and a K562 specific assay based on the break-point sequence of the BCR-ABL1 gene (26). ddPCR was performed using a QX200 Droplet Digital PCR system and an Auto Droplet Generator (Bio-Rad Laboratories, Inc.) according to the manufacturer's instructions. The albumin gene was used as a control for the DNA reaction load. QuantaSoft™ v1.7.4.0917 software (Bio-Rad Laboratories, Inc.) was used for data analysis, and samples were analyzed in quadruplicate.

Total RNA was isolated from cells using TRIzol^® (Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. cDNA was synthesized using 200 U M-MLV reverse transcriptase (Promega Corporation) and random hexamer primers (Jena Bioscience GmbH) according to the manufacturer's instructions, with incubation at 37°C for 1 h and denaturation at 95°C for 5 min. β-Glucuronidase (GUSB) was used as the control gene (27,28). The primers and probes for BCR-ABL1 and GUSB were designed, and the measurement of the expression levels were performed, according to the Europe Against Cancer protocol (29). The method has been standardized in the project of European Leukemia Net (30). The following primers and probes were used: GUSB forward, 5′-GAA AAT ATG TGG TTG GAG AGC TCA TT-3, reverse 5-CCG AGT GAA GAT CCC CTT TTT A-3 and fluorescein-containing probe, 5′-FAM-CCA GCA CTC TCG TCG GTG ACT GTT CA-BHQ1-3′; BCR-ABL1 forward, 5′-TCC GCT GAC CAT CAA TAA GGA-3′, reverse 5′-CAC TCA GAC CCT GAG GCT CAA-3′ fluorescein-containing probe, 5′-FAM-CCC TTC AGC GGC CAG TAG CAT CTG A-BHQ1-3′ (Integrated DNA Technologies, Inc.). The following thermocycling conditions were used: Initial denaturation at 95°C for 10 min and 45 cycles of 95°C for 15 sec and 60°C for 1 min. For RT-qPCR gene expression analysis, the ERM-AD623 reference material (Join Research Centre, Belgium) (31,32) was used to create the calibration curve for the determination of the number of copies of BCR-ABL1 and GUS gene.

The K562 and K562^IR cell lines were cultured for 5 days in RPMI medium with 10% exosome-depleted FBS. Exosomes were isolated from 200 ml cell culture media, as previously described (33). Briefly, conditioned medium was obtained as the supernatant from the centrifugation of live cells (300 × g for 10 min at 4°C). Subsequently, dead cells were removed using a further round of centrifugation (2,000 × g for 10 min at 4°C). The supernatant was filtered [using a Filtropur S syringe filter (0.22-µm membrane); Sarstedt AG & Co. KG] to remove larger vesicles and cell debris. The resultant suspension was then ultracentrifuged (100,000 × g for 70 min at 4°C) using a Sorvall™ WX+ ultracentrifuge and a T647.5 fixed angle rotor (Thermo Fischer Scientific, Inc.). Exosome pellets were resuspended in PBS, and sedimented again (100,000 × g for 70 min at 4°C).

Isolated exosomes were resuspended in Trump's 4F:1G fixative, comprising 86 ml distilled water, 10 ml 40% formaldehyde (Merck KGaA), 4 ml 25% glutaraldehyde (Polysciences, Inc.), 1.16 g monosodium phosphate and 0.27 g sodium hydroxide (34), and adsorbed on Formvar/carbon coated grids conditioned with 1% Alcian blue in 1% acetic acid. The adsorbed particles were embedded in a layer of 2% phosphotungstic acid. Grids were viewed at 100 kV using a JEM 2000 CX microscope (JEOL, Ltd.) equipped with an Olympus MegaviewTM II digital camera (Olympus Corporation).

An aliquot of exosomes resuspended in PBS was placed in the nanopore NP150 (qNano; Izon Science). All samples were measured at defined membrane stretch and with the same applied voltage at two different pressure levels (5 and 11 mbar). Calibration particles were measured directly after the sample measurement, and under identical conditions.

A total of 5 independent isolations of exosomes, from both the K562 and K562^IR cell lines, were subjected to MS/MS analysis. The filter-aided sample preparation method was used, with some modifications (35). Exosomes in PBS were resuspended in 100 mM ammonium bicarbonate, transferred to spin columns (Amicon Ultra 0.5 ml 10 kDa MWCO centrifugal filters; Merck KGaA) and centrifuged at 26,000 × g for 20 min at 4°C. The samples were then washed twice with 400 µl 100 mM ammonium bicarbonate and centrifuged again (26,000 × g for 20 min at 4°C). RapiGest™ (0.1%; Waters Corporation) dissolved in 100 µl 50 mM Tris/HCl, (pH 7.5) was subsequently added to the samples in the spin columns, then the samples were incubated at 95°C 10 min. After allowing the samples to cool down, 200 µl 0.1% RapiGest™ in 50 mM Tris/HCl, (pH 7.5) with 8 M guanidinium chloride was added, and the samples were subsequently incubated for 20 min at room temperature. The samples were then centrifuged at 18,000 × g for 25 min at 25°C. Aliquots of the samples (10 µl) were taken, and the protein concentration was quantified using a QuantiPro™ BCA Assay kit (Sigma-Aldrich; Merck KGaA). Subsequently, the samples were reduced with 100 µl 100 mM (Tris)2-carboxyethyl phosphine hydrochloride for 30 min at 55°C in a thermoshaker (Biosan, Ltd.) set at 600 rpm, alkyl-ated with 100 µl 300 mM iodoacetamide at 37°C for 30 min in the dark, then centrifuged at 12,000 × g for 35 min at 25°C. Next, the samples were digested overnight at 37°C using 2 µg sequencing-grade trypsin (Promega Corporation). The digested samples were then transferred into a new microtube for subsequent centrifugation (12,000 × g for 35 min at room temperature). Empore™ Solid Phase Extraction cartridges (C18; standard density, bed I.D., 4 mm) (3M Company) were used to desalt the peptide mixtures. Peptides were eluted in 60% acetonitrile (ACN)/0.1% trifluoroacetic acid (TFA), then dried in a SpeedVac. Prior to MS analysis, the samples were resuspended in 30 µl 2% ACN/0.1% TFA.

The raw files were further analyzed using MaxQuant software, v1.5.3.30 (36) [with Andromeda as the search engine (37)] against the Homo sapiens subset of the SwissProt database (downloaded on 4th July 2019; 26,468 sequences). Only tryptic peptides, that were at least 7 amino acids in length, with up to two missed cleavages were considered. Mass tolerance was set to 4.5 ppm at the MS level, and 0.5 Da at the MS/MS level. The oxidation of methionine was set as a variable modification, and the carbamidomethylation of cysteine was set as a fixed modification. A false discovery rate (FDR) of 1% was used for peptide spectrum matches and protein identification using a target decoy approach. Relative quantification was performed using the default parameters of the MaxLFQ algorithm (38), with the minimum ratio count set to 2.

The 'proteinGroups.txt' MaxQuant output file was uploaded into Perseus (39) v1.5.2.6. Decoy hits, proteins only identified by site, and potential contaminants were removed. Protein groups quantified in at least four replicates out of five were considered for further log₂ transformation of the LFQ intensities. Missing values were imputed from a normal distribution [Gaussian distribution width, 0.3 standard deviation (SD) and downshift 1.8 SD of the original data]. Data was normalized using the open source tool, Normalyzer (http://quantitativeproteomics.org/normalyzerde) and the variance stabilization normalization method (40). A Student's t-test (permutation based FDR 0.05, S0=0.1) was used for statistical analysis. Finally, proteins from this group with a fold change at least 1.5 were considered as being significantly different (P<0.05). Pearson's correlation test was performed to evaluate the inter-run reproducibility of individual LC-MS analyses. Proteins with known or expected cell-surface localization were selected using GenieScore, an algorithm for the prediction of surface localization (41). The Exocarta database (www.exocarta.org) was used to compare the proteins identified with those already found in exosomes.

The exosome pellets were lysed in 150 µl lysis buffer containing 140 mM sodium chloride, 10 mM HEPES, 0.15% Triton X100 and a protease inhibitor cocktail (1 tablet/10 ml; Roche Diagnostics), and subsequently incubated on ice for 20 min. The exosome samples were pooled and concentrated in an Amicon Ultra 0.5 ml 3 kDa MWCO centrifugal filter (Amicon Ultra; Merck KGaA) from 3 or 4 individual isolations. The protein concentration was determined using a Micro BCA™ protein assay kit (Pierce; Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions, and protein samples were immediately frozen and stored at −80°C.

The cell pellets were lysed in CelLytic™ M lysis buffer containing protease inhibitor cocktail (1 tablet/10 ml; Roche Diagnostics) on ice for 20 min. The cleared cell lysates were collected by centrifugation at 15,000 × g for 20 min at 4°C, and the protein concentration was then determined using the Bradford method (Bio-Rad Laboratories, Inc.).

The lysate samples (30-60 µg) were mixed with Laemmli sample buffer (Bio-Rad Laboratories, Inc.) containing 2-mercaptoethanol and separated on 4-15% or 7.5% (in the case of Bcr-Abl separation) precast gels (Mini PROTEAN^® TGX™; Bio-Rad Laboratories, Inc.). The separated proteins were transferred onto polyvinylidene fluoride membranes using the iBlot system, according to manufacturer's instructions (Thermo Fisher Scientific, Inc.). The membranes were blocked for 1 h in SuperBlock™ blocking buffer (Thermo Fisher Scientific, Inc.) and incubated overnight at 4°C with primary antibodies diluted to 1:1,000 in PBST (PBS, 0.1% Tween-20). The following primary antibodies were used: Anti-c-Abl rabbit polyclonal antibody (cat. no. 2862S; Cell Signaling Technology, Inc.), anti-Bcr-Abl mouse monoclonal antibody (7C6) (cat. no. ab187831; Abcam), anti-GAPDH rabbit monoclonal antibody (cat. no. SAB5600208; Merck KGaA), anti-IFITM3 (cat. no. 59212; Cell Signaling Technology, Inc.), anti-CD146 mouse monoclonal antibody (cat. no. 563619; BD Biosciences), anti-CD36 rabbit monoclonal antibody (cat no. 14347S; Cell Signaling Technology, Inc.) and EXOAB antibody kit 1 (Systems Biosciences, LLC) containing rabbit polyclonal antibodies against CD63, CD81, anti-CD9 and HSP70. After extensive washing in PBST, the membranes were incubated with secondary horseradish peroxidase-conjugated anti-rabbit antibody (cat. no. 7074P2) or anti-mouse antibody (cat. no. 7076P2) (both at 1:20,000 and from Cell Signaling Technology, Inc.) for 90 min at room temperature. Protein bands were detected with an enhanced chemiluminescence detection reagent (Cytiva) using a G:BOX imager (Syngene Europe), and quantified using ImageJ software, v1.8.0 (National Institutes of Health).

Fresh exosomes were washed and resuspended in PBS at room temperature. A 10 mM stock solution of carboxyfluorescein succinimidyl ester (CFSE) (Invitrogen; Thermo Fisher Scientific, Inc.) was diluted to a final concentration of 20 µM and added to the exosomes. The suspension was subsequently mixed and incubated for 25-30 min at room temperature in the dark. The labeling process was stopped by adding 4 ml of cold complete media, containing 10% FBS on ice for 5 min. CFSE-labeled exosomes were diluted in 60 ml PBS, collected by ultracentrifugation (100,000 × g for 70 min at 4°C) and resuspended in 1.5 ml cell culture media with K562 cells (500,000 cells/ml). CFSE-positive cells were observed under a FluoView FV1000 confocal laser scanning microscope (Olympus Corporation) using an UPlanSAPO 60× NA1.35 oil immersion objective (magnification, ×60). A 488 nm laser was used for CFSE excitation, and fluorescence emission was detected with a high sensitivity GaAsP detector at 500-600 nm. Fluorescent images were processed using the FluoView software|(FV10 ASW v3.1; Olympus Corporation).

The K562 cells were co-cultured with either K562^IR-derived or K562-derived exosomes for 4 h, and then treated with 2 µM imatinib for 48 h. Cell viability was measured using a Vybrant^® Cell Proliferation Assay kit (Thermo Fisher Scientific, Inc.); proprietary solvent B containing SDS was mixed with 0.01 M HCl and used to solubilize the purple formazan. The absorbance was measured at 570 nm using a microplate reader (Chameleon; Hidex Oy).

The K562 and K562^IR cells (1.5×10⁶ cells) were washed in washing buffer (PBS supplemented with 0.1% BSA and 2 mM EDTA), centrifuged at 300 × g for 10 min at room temperature, then resuspended in washing buffer. Aliquots (50 µl) of the cell suspension (100,000 cells/tube) were transferred to FACS tubes and 1 µl anti-IFITM3 AF405 (cat. no. ITA8095; 1:50; G-Biosciences; Geno Technology, Inc.), 1 µl anti-CD146 (cat. no. 563619; 1:50; BD Biosciences) and 2 µl anti-Hu CD36 FITC (cat. no. 1F-451-T100; 1:25; EXBIO Praha, a.s.) were added. The samples were incubated in the dark for 30 min at room temperature, then washed again with 1 ml of washing buffer, prior to centrifugation (300 × g for 5 min at 25°C). Washing buffer (250 µl) was added to the cell pellet, and the samples were analyzed using flow cytometry, in triplicate, using BD FACSCanto™ II Cell Analyzer (BD Biosciences). The data was analyzed using BD FACSDiva software (v6.1.3; BD Biosciences). MFI was determined for the whole sample, and the fraction of positively stained cells (P2) was determined as the percentage of the parent population.

The data are expressed as the mean ± SD, from at least three replicates. Statistical analysis was performed using GraphPad Prism v8.0 software (GraphPad Software, Inc.). Relative resistance of K562 and K562^IR cells to imatinib was evaluated using an unpaired Student's t-test. For the investigation of cell survival following exosome exposure, one-way ANOVA with Tukey's post hoc test was used to determine the statistical significance. P<0.05 was considered to indicate a statistically significant difference.

Imatinib-resistant K562^IR cells were derived from the originally imatinib-sensitive K562 cells [half-maximal inhibitory concentration (IC₅₀), 0.25-0.35 µM]. The K562^IR cells proliferated in imatinib concentrations exceeding 2 µM (Fig. 1).

A mutation in the kinase domain of BCR-ABL1 is the most common mechanism by which imatinib (and other TKIs) resistance develops in patients with CML (7). To investigate the mutational status of the BCR-ABL1 kinase domain in the K562^IR cells, NGS sequencing was performed; however, no mutations in the kinase domain, with variant allele frequency >1% were found. In patients with TKI resistance but without a kinase domain mutation, an amplification of the BCR-ABL1 gene is typically found (7). Therefore, ddPCR was used to identify the number of BCR-ABL1 gene copies, and it was revealed that the K562^IR cells harbored a 2-fold higher number of BCR-ABL1 gene copies compared with that in imatinib sensitive K562 cells (data not shown).

As the BCR-ABL1 gene amplification may result in overexpression of the BCR-ABL1 gene and increase the levels of the Bcr-Abl protein, the mRNA and protein expression levels were compared between the K562 and K562^IR cell lines. The results revealed increased expression levels of BCR-ABL1 mRNA (Fig. 2A) and of the Bcr-Abl protein (Fig. 2B).

The exosomes were isolated from the K562 and K562^IR cell culture supernatants using ultracentrifugation. The exosome purity was verified using transmission electron microscopy, revealing round or cup-shaped vesicles, ranging between 50-150 nm in diameter (Fig. 3). The qNano/TRPS analysis of the exosomes confirmed comparable size distributions of the vesicles, with the highest peak occurring at 100-110 nm. The number of exosomes isolated from the cell media of the K562^IR cells (1.67×10¹¹ particles/ml) was ~2 times higher compared with that in the K562 cells (8.14×10¹⁰ particles/ml) (Fig. S1).

To further confirm that the pelleted material represented exosomes, the presence of 'exosomal markers' i.e. proteins commonly found in exosomes (CD63, CD9, CD81, HSP70 and GAPDH) (42), was confirmed using western blot analysis (Fig. 4). As the CML landmark fusion protein, Bcr-Abl was overexpressed in the K562^IR cells, it was also possible to show the presence of the Bcr-Abl protein in the corresponding exosomes.

K562-derived exosomes have been previously shown to be taken up by various cell types, such as bone marrow stromal cells, macrophages, endothelial cells and leukemic cells (16,17,19-21,43). Notably, exosomes derived from imatinib-resistant CML cells have been recently shown to fuse with, and confer drug resistance traits to, imatinib-sensitive CML cells (22). In the present study, the fusion of K562^IR-derived exosomes (labelled with fluorescent CFSE) with K562 cells, and the fusion of K562-derived exosomes with K562 cells, was confirmed following a 4-h incubation (Fig. 5). CFSE-positive K562 cells were detected using confocal microscopy, as soon as 1 h after the addition of the labelled exosomes, with the maximum uptake occurring after 4 h, as determined in a pilot time-course experiment (data not shown).

Min et al (22) demonstrated that exosomes released from imatinib-resistant K562 cells were able to transfer drug-resistant traits to imatinib-sensitive K562 cells. To confirm this observation, K562^IR- or K562-derived exosomes were isolated and incubated with K562 cells for 4 h, prior to the addition of imatinib (2 µM) for 2 days. As shown in Fig. 6, exosomes derived from the K562^IR cells significantly increased the survival rate of imatinib-sensitive K562 cells in the presence of imatinib, compared with the K562 cells incubated with exosomes derived from K562 cells, and also compared with the K562 cells treated with no exosomes. The control exosomes derived from the K562 cells; however, had no measurable effect on cell viability (Fig. 6).

Therefore, it could be hypothesized that the specific composition of the K562^IR-derived exosomes was responsible for the enhanced survival of the K562 cells. To characterize the unique protein content of the exosomes and to identify their specific (preferably surface) markers, a detailed proteomic analysis was subsequently performed.

Exosomes derived from the K562 and K562^IR cells were subjected to LFQ proteomic analysis. A total of 10 samples of the respective exosomes, obtained from five independent isolations of both the K562^IR and K562 cells, were analyzed using LC-MS/MS (Q-Exactive). With the FDR set to 0.01, between 1,072 and 1,751 exosomal proteins were identified in each sample; in total, 3,218 unique exosomal proteins were identified. Of those, 2,693 proteins were listed in the Exocarta database of exosomal proteomes (www.exocarta.org); therefore, the present study identified 525 novel exosomal proteins.

To determine the quantitative robustness of the label-free analysis, the quantitative similarity of all the LC-MS/MS runs was examined. Correlation analysis revealed good inter-sample reproducibility, with Pearson's correlation coefficients in the range of 0.723-0.971 (Fig. S2). The LFQ analysis provided semi-quantitative data for 1,241 proteins (Table SI).

A total of 35 proteins with significantly different quantities were identified in the K562^IR-derived exosomes compared with the K562-derived exosomes (fold change >1.5) (Fig. 7 and Table I). Of these 35 proteins, 28 were found to be upregulated, while 7 were downregulated, in the K562^IR exosomes. The most upregulated proteins in the K562^IR exosomes included interferon-induced transmembrane protein 3 (IFITM3), desmoglein-2 (DSG2), cell-surface glycoprotein MUC18 (CD146) and platelet glycoprotein 4 (CD36). Among the most downregulated were collagen α-1 (XV) chain, galectin-3-binding protein (LGALS3BP), laminin subunit β-1 (LAMB1), bone marrow stromal antigen 2 (BST2) and epidermal growth factor-like protein 7 (EGFL7).

The most highly upregulated proteins (at least 5-fold) were further investigated with special focus on proteins localized on the cell or exosomal membrane. Among the candidates selected by the LFQ proteomic analysis, three such proteins were identified using GenieScore, an algorithm for the prediction of surface localization (41): IFITM3, CD146 (MUC18) and CD36. The presence of these proteins was confirmed in the exosomes and also in the cells of their origin using western blot analysis with specific antibodies (Fig. 8).

The western blot data confirmed the upregulation of all three proteins in the K562^IR-derived exosomes, as well as in the 'source' K562^IR cells. Therefore, the upregulated surface membrane proteins could potentially serve, not only as exosomal markers, but also as cell-surface markers specific for a resistant population of the K562 cells.

To determine the potential of the identified proteins as cell-surface markers of imatinib resistance, flow cytometric analysis of the K562 and K562^IR cells was performed using antibodies against IFITM3, CD146 and CD36. Live cells were used to determine the cell-surface expression of the putative markers.

Flow cytometry revealed markedly increased surface expression levels of IFITM3 and CD146, and to a lesser extent also of CD36, in K562^IR cells compared with the K562 cells (Fig. 9). CD146 was detected in 96.2% of the K562^IR cells compared with 0.7% of the K562 cells, while IFITM3 was detected in 51.6% of the K562^IR cells compared with 4% of the K562 cells. Anti-CD36 antibody stained 10.5% of the K562^IR compared with 3% of the K562 cells. The analysis demonstrated that CD146 expression, in particular, clearly distinguishes K562 from K562^IR cells, and therefore it may be used as a reliable cell-surface marker of imatinib resistance in these cells. In addition, promising results were also obtained with IFITM3.