Vorinostat impairs the cancer-driving potential of leukemia-secreted extracellular vesicles

Abstract

Background

Leukemia-secreted extracellular vesicles (EVs) carry biologically active cargo that promotes cancer-supportive mechanisms, including aberrant proliferative signaling, immune escape, and drug resistance. However, how antineoplastic drugs affect EV secretion and cargo sorting remains underexplored.

Methods

Leukemia-secreted extracellular vesicles (EVs) were isolated by Differential UltraCentrifugation, and their miRNome and proteomic profiling cargo were analyzed following treatment with SAHA (Vorinostat) in Acute Myeloid Leukemia (AML) and Chronic Myeloid Leukemia (CML). The epigenetic modulation of leukemia-secreted EVs content on interesting key target molecules was validated, and their differential functional impact on cellular viability, cell cycle progression, apoptosis, and tumorigenicity was assessed.

Results

SAHA significantly alters the cargo of Leukemia-derived EVs, including miR-194-5p and its target BCLAF1 (mRNA and protein), key regulators of Leukemia cell survival and differentiation. SAHA upregulates miR-194-5p expression while selective loading BCLAF1 into EVs, reducing the miRNA levels in the same compartment. Additionally, SAHA alters miRNA profile and proteomic composition associated with leukemic EVs, altering their tumor-supportive potential, with differential effects observed between AML and CML. Furthermore, in silico predictions suggest that these modified EVs may influence cell sensitivity to antineoplastic agents, suggesting a dual role for SAHA in impairing oncogenic signaling while enhancing therapeutic responsiveness.

Conclusions

In conclusion, the capacity of SAHA to modulate secretion and molecular composition of Leukemia-secreted EVs, alongside its direct cytotoxic effects, underscores its potential in combination therapies aimed to overcoming refractory phenotype by targeting EV-mediated communication.

Introduction

Leukemia is characterized by extensive biological heterogeneity that often leads to relapse/refractory disease, making treatment extremely challenging [1]. In addition, the communication between cancer cells and surrounding stromal components plays a key role in different stages of the disease, further complicating the scenario [2]. Part of these interactions are mediated by extracellular vesicles (EVs), a heterogeneous population of vesicles secreted by all nucleated cells in the extracellular environment, previously regarded as a means to discard waste material [3, 4]. EV heterogeneity is the result of the different mechanisms and cellular compartments involved in vesicle biogenesis. According to current literature, EVs comprise: (i) exosomes (30–200 nm in diameter) originating in the endosomal compartment, where various mechanisms and pathways are involved, thus contributing to vesicle diversity [3, 4]; (ii) microvesicles (100–1000 nm in diameter) generated via direct outward budding of the plasma membrane [4]; (iii) apoptotic bodies (1000–5000 nm in diameter) deriving from apoptotic stimuli, which play a crucial role in maintaining tissue homeostasis and immune response yet their involvement in intercellular communication is generally considered less pronounced compared to exosome and microvesicles [3]. Therefore, we excluded such EV subtype from our study. Despite these classifications, the overlapping size range and the lack of reliable methods to distinguish between EV subtypes represent significant challenges [5, 6]. To address this, we adopted a size-based distinction between small vesicles (SVs) and large vesicles (LVs), acknowledging that vesicles of similar size (e.g. 150 nm) may include both exosomes and microvesicles. EVs contain various cargo molecules in both their lumen and membrane, including proteins, RNAs, miRNA, DNA, lipids, and metabolites, all of which can influence recipient cell phenotypes [7, 8]. As a result, accumulating evidence have highlighted the role of cancer-derived EVs in promoting tumor-supportive functions, including proliferation [9], angiogenesis [10], immune escape [11], and virtually every aspect of cancer biology [12]. Moreover, these vesicles are implicated in the development of refractory phenotypes in recipient cancer cells [13,14,15,16,17], further complicating treatment outcomes. Beyond investigating the mechanisms underlying this type of intercellular communication, growing efforts are focused on harnessing EVs for therapeutic applications. However, comparatively less attention has been paid to how therapeutics alter EV secretion and molecular composition, changes that can variably influence the phenotype and functionality of recipient cells.

In previous studies, we demonstrated that the imbalance between miR-194-5p and its target Bcl-2-associated transcription factor 1 (BCLAF1) drives the differentiation block and apoptosis resistance observed in Acute Myeloid Leukemia (AML) cell lines—hallmarks of cancer progression [18]. Treatment with SAHA (Vorinostat)—an FDA-approved HDAC inhibitor(NDC code: 0006–0568-40, CAS number: 149647-78-9, ATC code: L01XH01)—restored this imbalance, leading cancer cells toward apoptosis. Notably, SAHA not only downregulated BCLAF1 expression, at mRNA and protein level, but also affected its intracellular localization and function.

Here, we investigated whether SAHA modulates the cargo sorting of leukemia-derived EVs and examined the biological impact of these altered vesicles on surrounding tumor cells. We assessed SAHA-mediated effects in both acute myeloid leukemia (AML) and chronic myeloid leukemia (CML) [22]. Our study focused specifically on myeloid leukemias, given their poorer prognosis, lower survival rates, high relapse rates, and resistance to treatment compared to lymphoid leukemias [19, 20].

Additionally, previous studies have highlighted the significant role of soluble factors and EV secretion in the malignant progression of myeloid leukemia, particularly in the reprogramming of the immunosuppressive microenvironment [21]. In this context, EVs may represent a promising therapeutic target to disrupt tumor-supportive communication between leukemia cells and their microenvironment [19, 21].

Our aim was to elucidate the modulatory effects of SAHA on EV secretion and cargo composition in both AML and CML, overcoming the distinct genetic, epigenetic, and molecular backgrounds of these two myeloid leukemia models. Specifically, we sought to identify key leukemic processes and targets influenced by SAHA-induced EVs, proposing them as potential therapeutic candidates.

Our findings reveal that SAHA significantly alters EV secretion and content, thereby skewing their tumor-supportive potential in both AML and CML. Moreover, based on molecular observations and in silico predictions, such changes may enhance the sensitivity of residual Leukemia cells to specific therapeutic agents.

Materials and methods

Cell line culture and chemical compounds

Human cell lines used were U937 (AML; RRID: CVCL_0007)) and K562 (CML; RRID: CVCL_0004). Leukemia cell lines were grown in RPMI 1640 medium (EuroClone, Milan, Italy) supplemented with 10% heat-inactivated FBS (Sigma-Aldrich, St Louis, Missouri, USA) and 2% penicillin/streptomycin (EuroClone). EV-depleted FBS was used to culture cells for EV collection. Briefly, heat-inactivated FBS was ultracentrifuged at 100,000 × g for 12 h at 4 °C, without brake. The supernatant was added to 2% penicillin/streptomycin RPMI at a final concentration of 5%. SAHA (Vorinostat; Merck, Kenilworth, NJ, USA) was dissolved in DMSO (Sigma-Aldrich) and used at a final concentration of 5 μM. All human cell lines have been authenticated using STR profiling within the last 3 years. U937 and K562 cell lines were obtained from ATCC, Manassas, VA, USA. All experiments were performed with mycoplasma-free cells.

Small and large vesicle isolation

Differential ultracentrifugation was performed using an Optima XPN-90 ultracentrifuge (Beckman Coulter, Brea, CA, USA) to isolate small and large vesicle subpopulations (SV and LV). A total of 4 × 105 leukemic cells/ml were plated in 30 ml RPMI containing EV-depleted FBS and treated with SAHA for 24 h before EV isolation. DMSO was added to the control (Ctr) groups. SVs were isolated as described previously [18]. The SV pellet was then carefully resuspended in 200 μl PBS for functional assays or 150 μl PBS for molecular analysis, and stored at − 80 °C. The same protocol was used to isolate EVs for vesicle characterization and miRNA profiling.

LV isolation was performed using the same steps as for SV isolation, but ending the protocol earlier, as LVs were pelleted at 10,000 × g. After centrifugation at 10,000 × g, the pellet was washed with PBS, collected, and referred to as LVs. LV preparations were resuspended in 200 μl PBS for functional assays or 150 μl PBS for molecular analysis, and stored at − 80 °C.

EV sorting, staining, and acquisition

EVs were sorted and acquired as previously described [19]. Protocol details and variations are reported below. Reagent mix was prepared by adding 0.5 μl of fluorescein isothiocyanate (1)-conjugated phalloidin and lipophilic cationic dye (both from BD Biosciences, San Jose, CA, USA; #626267 custom) to 195 μl of PBS 1X. Rosetta Calibration (Exometry, Amsterdam, The Netherlands) was used according to the manufacturer’s instructions, to calibrate side scatter, relate side scatter in arbitrary units to standardized units of nm, as well as to the diameter and refractive index of particle. Samples were acquired by flow cytometry (FACSVerse, BD Biosciences). EVs were identified as LCD positive/phalloidin negative events. EV preparations of LCD positive/phalloidin negative events displayed distributions in the range of ~ 100–300 nm in diameter. Instrument performances were monitored by the Cytometer Setup and Tracking Module (BD Biosciences) and further validated by the acquisition of Rainbow Beads (BD Biosciences). The same settings were used for all other measurements. To stain EVs, 300 ml of CytoFix/CytoPerm 1X (BD Biosciences) was added to 200 ml of supernatants. After 10 min, 2 ml of anti-BCLAF1 was added. After 30 min of staining at room temperature (RT) in the dark, 200 μl of PBS 1X was added to each tube. Next, 1 ml of an anti-rabbit FITC-conjugated (Jackson ImmunoResearch, West Grove, PA, USA) was added to each tube and incubated for 30 min (RT, in the dark). Samples were acquired on a FACSVerse flow cytometer (BD Biosciences), as previously described [20, 21].

Immunofluorescent staining and acquisition

Cells were plated at a confluence of 200,000 cell/ml in 2.5 ml medium using a 6-well plate. After 24 h incubation with SAHA (or DMSO as Ctr), 30,000 live cells were added to a 12-well plate containing round cover glasses (Thermo Fisher Scientific, Waltham, MA, USA; #10006111) pretreated with poly-L-lysine (0.1 mg/ml; Merck, #P4707) and centrifuged at 350 × g for 5 min. Cells were fixed with PBS 4% formaldehyde (Thermo Fisher Scientific, #PI28906) and incubated for 20 min at RT. Cell permeabilization was performed with PBS 0.1% Triton X-100 (Sigma-Aldrich, #9036-19-5) and incubated for 10 min at RT. Cells were blocked with PBS 10% serum (FBS) for 30 min at RT. Cells were incubated with the following primary antibodies diluted in PBS 1% BSA/PBS: 1:100 CD63 (BD Biosciences, #556019), 1:100 BCLAF1 (Thermo Fisher Scientific; #PA5-55686) and incubated for 1 h at RT. Cells were then incubated for 30 min with the following secondary antibodies: Goat anti-Mouse IgG conjugated with AlexaFluor 488 (Thermo Fisher Scientific; #A11029) 1:1000 and Goat anti-Rabbit IgG conjugated with AlexaFluor 594 (Thermo Fisher Scientific; #A11012) 1:1000, and incubated for 30 min. Lastly,round cover glasses were mounted on a microscope slide using one drop of mounting medium containing DAPI (Thermo Fisher Scientific, #P36935) for nuclei staining. Images were acquired using an SP8 LIGHNTING confocal microscope (Leica Microsystems, Wetzlar, Germany) using an S Plan Fluor ELWD 40X/0.60 objective (Nikon, Tokyo, Japan). Images were acquired with NIS-Elements software (Nikon).

miRNA profiling

cDNA synthesis and real-time qPCR were performed using a miRCURY LNA Universal RT microRNA PCR system (Qiagen, Germany) according to the manufacturer’s instructions, as previously described [22]. Real-time PCRs were run on a 7900HT thermocycler (Applied Biosystems, Thermo Fisher Scientific) using thermal-cycling parameters recommended by Qiagen. Briefly, we adhered to the manufacturer’s protocol, where Raw Ct values were calculated using the RQ Manager software v.1.2.1 (Applied Biosystems), with manual adjustments for the threshold and baseline settings. Data were analyzed by applying a ΔRn threshold of 60 and subtracting the baseline between cycles 1–10. miRNA profiles were then determined using the delta-delta Ct method and EV-miRNA data were normalized on RNU6 levels.

Target prediction analysis and GO enrichment

The prediction of target genes was performed using the miRSystem database as reported in [22], accomplished here by a value HIT ≥ 5, included validated genes, and observed/expected (O/E) ratio ≥ 2. For Gene Ontology (GO) enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was used the Database for Annotation, Visualization and Integrated Discovery (DAVID) (http://david.abcc.ncifcrf.gov/home.jsp). The Bonferroni Correction and the Benjamini–Hochberg Procedure were used in order to calculate the False Discovery Rate (FDR, p < 0.05) and the statistical significance at a p-value of < 0.05 was set.

RNA isolation

Cells were collected by centrifugation at 350 × g for 5 min and lysed in 1 ml TRIzol (Invitrogen, Waltham, MS, USA). Similarly, 50 μl (out of 150 μl) of vesicle preparations were lysed in 1 ml TRIzol. RNA was extracted as previously described 9, with the addition of 3 μl glycogen (Roche) to the aqueous phase from vesicle preparations (but not to the cell lysate) before adding 500 μl of cold isopropylic alcohol, followed by vigorous shaking. RNA was resuspended in 10 μl DEPC-treated H₂O (for vesicles) and 20 μl DEPC-treated H₂O (for cells).

Reverse transcription and real-time qPCR

mRNA was reverse transcribed using the SuperScript VILOTM cDNA Synthesis Kit (Invitrogen), according to the manufacturer’s instructions. A total of 1 μg RNA was reverse transcribed using a Peltier Thermal Cycler (MJ Research, Deltona, FL, USA). Gene expression was evaluated by real-time qPCR. A total of 50 ng of cDNA was amplified using the Power SYBR Green PCR Master Mix (Applied Biosystems, Waltham, MA, USA), according to the manufacturer’s instructions. Gene expression was normalized to GAPDH levels. Real-time qPCR was performed using a CFX96 detection system (Bio-Rad). The analysis was conducted using the ΔΔCt method. Primers used: BCLAF1 forward TCCGATCCATCTTTGACCACA, reverse BCLAF1 TGATACGAAGTGAACCGCTCG; GAPDH forward GGAGTCAACGGATTTGGTCGT, GAPDH reverse GCTTCCCGTTCTCAGCCTTGA.

miRNA real-time PCR

After RNA extraction, the miRNA fraction was converted into cDNA and subsequentially miRNA Real-Time PCR was performed as described in [23].

Western blotting

Cells and vesicles were lysed with RIPA buffer containing protease inhibitor cocktail (Roche, Switzerland), and the protein concentration was determined by BCA assay (Pierce, Etten-Leur, The Netherlands). Cell and vesicle lysates were diluted in different types of loading buffer according to the protein of interest. Blotting for vesicle-associated markers (CD63, CD81) required loading buffer without reducing agents, while blotting for other proteins was performed using loading buffer containing reducing buffer (beta-mercaptoethanol, Sigma- Aldrich, St. Louis, MO, USA), in both cases boiled at 95 °C for 5 min. Cell lysates and EV preparations in sample buffer were run on a 10% SDS gel and blotted on a nitrocellulose membrane (GE Healthcare, Eindhoven, The Netherlands). Membranes were incubated with antibodies against CD63 (BD Biosciences, San Jose, CA, USA; #556019) 1:300, BCLAF1 (Thermo Fisher Scientific, Waltham, MA, USA; #PA5-55686) 1:1000, and horseradish peroxidase-conjugated goat anti-mouse (Bio-Rad, Hercules, CA, USA; #1706516) 1:3 000 and goat anti-rabbit (Bio-Rad, #1706515) 1:3 000 antibodies.

Transmission electron microscopy

EV pellets were fixed with 2.5% glutaraldehyde and all analysis were performed as described in [24].

EV immunogold-labeling

Isolated EVs were loaded onto carbon-coated grids, fixed in 2% paraformaldehyde, washed, and then immune-labeled with anti-CD63 antibody (Abcam, United Kingdom; #ab59479), anti-CD9 antibody (Abcam, #ab92726), and anti-Tsg-101 (Abcam, #ab125011), followed by a 10 nm gold-labeled secondary antibody (Sigma-Aldrich). EVs were post-fixed in 2.5% glutaraldehyde, washed three times, contrasted with 2% uranyl acetate, and then examined with a JEOL 100CX transmission electron microscope (JEOL, Peabody, MA, USA).

Cell viability assay

Cell viability was evaluated using a CCK8 kit (Sigma-Aldrich, #96992) following the manufacturer’s instructions. A total of 100 ul containing 10 000 cells/well were plated in a 96-well plate followed by the addition of 10 ul/well of WST-8 solution (1:10) and incubated overnight. The absorbance was then measured at 460 nm using an Infinite M200 plate reader (Tecan, Switzerland). Data were analyzed and presented as a percentage of live cells normalized to the relative control.

Propidium iodide staining and cell cycle assessment

A total of 400,000/ml cells were transferred to polypropylene tubes, centrifuged at 350 × g, and the pellet was resuspended in 500 μl PBS. Propidium iodide (Sigma-Aldrich) was then added at a final concentration of 1 μg/ml, and cells were immediately analyzed by FACS Celesta Flow Cytometer (BD, Franklin Lakes, NJ, USA). For cell cycle assessment, cell cycle buffer was prepared by dissolving sodium citrate (10%; Sigma-Aldrich), NP-40 (10%; Sigma-Aldrich), and propidium iodide (2 mg/ml, Thermo Fisher Scientific) in PBS. Cells were resuspended in 500 μl cell cycle buffer, incubated for 15 min at room temperature, and then analyzed by FACS Celesta Flow Cytometer.

Tumorigenicity assay

The tumorigenic capability of leukemic cells upon EVs w/wo SAHA treatment was evaluated by colony-forming unit assay using MethoCult (Stem cells, #SF H4236). A total of 500,000 leukemic cells/well were plated in a 6-well plate in 2.5 ml medium, to which 50 μl EV preparation were added, and incubated for 24 h. Subsequently, SAHA was added to the appropriate wells and incubated for 24 h. Following completion of the treatments, 10,000 cells were resuspended in 100 μl medium and added to 900 μl MethoCult medium. The mixture was then carefully pipetted to a new 6-well plate and incubated for five days. Images were acquired using a Cytation Cell Imaging Reader (Biotek, Winooski, VT, USA) at 4× magnification.

Label-free proteomics analysis

To evaluate the protein cargo of sorted EVs from U937 medium, 700,000 purified EVs were used for proteomics analysis by comparing EVs of the two cell lines treated with SAHA to related controls. Samples were prepared for the Filter Aided Sample Preparation(FASP) protocol. EVs were lysed by sonication on ice (U200S sonicator control, IKA, Staufen, Germany) at 70% amplitude in a lysis buffer (urea 6 M in 100 mM Tris/HCl, pH = 7.5) for overnight tryptic digestion at 37 °C. As previously reported, the number of separated EVs was used as a normalization parameter for protein label-free identification and quantification [25]. Tryptic peptides were analyzed in triplicate by LC–MS/MS using the UltiMate3000 UPLC chromatographic system (Thermo Fisher Scientific, Milan, Italy) coupled to the Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific). Details of LC–MS/MS parameters are reported in our previous works [25, 26]. Briefly, the flow rate was set at 300 nL/min, with a total run time of 65 min using a chromatographic gradient from 2 to 90% of acetonitrile/water. Peptides were acquired in positive-ion polarity with data-dependent acquisition mode and MS2 sequence, using N2 as collision gas for HCD fragmentation.

Proteomics LC–MS/MS raw data were then processed using the free computational platforms MaxQuant version 1.6.10.50 and Perseus version 1.6.10.50 (Max-Planck Institute for Biochemistry, Martinsried, Germany).

and the UniProt database (released 2020_06, taxonomy Homo Sapiens, 20 588 entries), as previously described in full [27]. Proteins with a − 0.05 < p-value > 0.05 and FC > 1 were considered differentially loaded into EVs. Lastly, protein ratios (EV SAHA/EV Ctr) were used for Gene Ontology and functional enrichment analysis using the Ingenuity Pathway Analysis (IPA) tool (Qiagen, Hilden, Germany). Briefly, IPA can predict the activation (z-scores ≥ 2.0) or inhibition (z-scores ≤ − 2.0) of transcriptional regulators or downstream for the loaded dataset based on prior knowledge of expected effects from published literature citations stored in the IPA system. The proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [28] partner repository with the dataset identifier PXD042168.

Statistical analysis

All experiments were repeated three times (biological replicates). Statistical analysis was performed using GraphPad Prism 7 software; data were expressed as means of values and error bars represent standard deviation of the means. A two-tailed unpaired t-test was applied to assess the effects of independent variables on quantitative results. Significance was defined as a p-value < 0.05, indicated by asterisks in the figures (ns = not significant, * = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001).

Results

SAHA modulates secretion and miRNA content in leukemia-derived EVs

We previously conducted a comprehensive analysis of SAHA’s effect on cell viability and apoptosis across various leukemia cell lines [11]. Briefly, we demonstrated that SAHA increased the proportion of cells in the pre-G1 phase, while enhancing Caspase-8 and −9 activity after 24 h of treatment, resulting in approximately 50% cell death in U937 cells and 40% in K562 cells (Fig. S1A). Building on these observations, we then shifted our focus to investigating the impact of SAHA on the EV secretion and composition. We isolated and characterized EVs using electron microscopy (Fig. 1A) and Western blotting (Fig. 1B). Flow cytometry analysis, optimized for stained EVs within the 100–300 nm size range, revealed that SAHA treatment increase4d EV secretion from U937 (Fig. 1C). Although SAHA exerted a cytotoxic effect on Leukemia cells [18], the enhanced EV secretion could not be attributed to an increase in apoptotic bodies, given the applied sorting threshold. Conversely, SAHA reduced EV secretion in K562 cells (Fig. 1C). Though changes in EV yield may significantly influence the tumor microenvironment, SAHA-induced intracellular molecular rearrangements [18] may be reflected in the extracellular compartment further contributing to phenotypic alterations in recipient cell, highlighting the need for molecular evaluations to investigate the malignant potential of Leukemia-derived EVs following treatment.

Fig. 1

Characterization of leukemia-derived EVs and SAHA affects leukemia EV-associated miRNA profile. A Top, transmission electron microscopy micrographs of EVs isolated from U937 cells. Scale bar, 100 nm; Bottom, CD63, CD9, and Tsg101 analysis by immunoelectron microscopy in EVs isolated from U937 cells; B Western blot for CD63 and cytochrome C in U937 and K562 cells and corresponding EVs; C Left, SAHA-induced effect on EV secretion in U937 cells assessed by FACS; Right, SAHA induced effect on EV secretion in K562 cells assessed by FACS. Data were normalized to viable cell count and conditioned medium volume and expressed as fold change relative to the experimental control. Error bars indicate standard deviation; D Heat map of differentially expressed miRNAs in EVs derived from untreated vs SAHA-treated cells (Supplementary Table 1 shows the miRNA list); E Venn diagram of miRNAs carried by EVs derived from untreated vs SAHA-treated cells; F miR-194-5p sorting in U937 EVs modulated by SAHA expressed as SAHA mean. Error bars indicate standard deviation

Accordingly, we assessed the miRNA profile of AML cells and their corresponding EVs to determine how SAHA affects miRNA loading. Treatment-induced changes in miRNA composition were observed both intracellularly and within EVs (Fig. S1 B-C). Notably, SAHA significantly altered the miRNA profile of U937 EVs (Fig. 1D, E, Table S1). Among 257 miRNAs shared between cells and vesicles, several tumor suppressors—such as miR424-5p, miR455-3p, miR-16-5p, miR-19a-3p, miR-19b-1-5p, miR-15a-5p [29,30,31,32,33,34,35,36]—were lowered in EVs following treatment (Table S1). In contrast, SAHA enhanced the loading of miRNAs associated with drug sensitivity, including miR424-5p, miR-451a, miR-153-3p, miR-17-5p, miR-21-5p, miR-27a-3p [30, 37,38,39]. Intriguingly, these miRNAs exhibit diverse effects on drug responsiveness, suggesting that SAHA-altered EVs may support pro-tumorigenic functions while potentially enhancing the efficacy of specific therapies. Gene Ontology (GO) analysis for the predictive targets of miRNAs differentially expressed in SAHA-modulated EV in AML, distinguishing between up- and down-regulated, reveals some peculiar biological processes and key pathways such as miRNA involvement in cancer, regulation of the G1/S transition in the mitotic cell cycle, the positive induction of apoptosis and differentiation, and the regulation of transcription by RNA pol II. These findings underscore the critical role of miRNAs in U937-derived EVs and their modulation by SAHA, prompting us to speculate on the potential impact of these changes on recipient cells. Furthermore, miR-194-5p was upregulated intracellularly upon SAHA treatment, yet its EV-associated levels were reduced (Fig. 1F; Table S1).

To explore SAHA’s impact on BCLAF1 (mRNA and protein) sorting into EVs, we analyzed specific EV subpopulations, by specifically analyzing different vesicle subpopulations (small vesicles, SVs, and large vesicles, LVs). EVs from untreated cells were designed as EV Ctr (LV Ctr and SV Ctr), while those from SAHA-treated leukemic cells were referred to as EV SAHA (LV SAHA and SV SAHA). Such deeper investigation highlighted that the reduced miR-194-5p loading was observed in both SV and LV (Supplementary Fig. 1 C-D). These findings highlight that EV content modifications induced by treatment may not directly mirror intracellular expression changes, warranting further investigation into other EV-associated molecules.

SAHA enhances BCLAF1 secretion via EVs in leukemia cells

In U937, SAHA downregulated BCLAF1mRNA in cells (Fig. 2A, left) yet it increased its loading into SVs. However, BCLAF1 protein levels in EVs were not significantly altered, aligning with trends observed for EV-associated markers (CD63 for small vesicles—being it a recognized endosomal marker—and α-tubulin, indicative of larger ones)(Fig. 2A, right). Particularly, LV SAHA showed increased α-tubulin levels, suggesting enhanced LV secretion (Fig. 1C, left) and resulting in greater BCLAF1 circulation.

Fig. 2

BCLAF1 quantification in leukemic cells and related EVs. A BCLAF1 and EV marker expression in U937 cells and cognate vesicles after 24 h SAHA treatment. Top left, BCLAF1mRNA quantification in U937 cells normalized to GAPDH expression; Bottom left, BCLAF1 protein quantification normalized to tubulin expression and CD63 expression in U937 cells with or without SAHA; Top right, BCLAF1mRNA quantification in U937-derived EVs normalized to Y-RNA; Bottom right, BCLAF1 and EV-associated marker quantification in U937-derived EVs; B BCLAF1 expression in K562 cells after 24 h SAHA treatment. Top left, mRNA quantification in K562 cells normalized to GAPDH mRNA expression; Bottom left, protein quantification normalized to β-actin and CD63 expression in K562 cells treated with or without SAHA; Top right, bclaf1mRNA quantification in K562-derived EVs normalized to Y-RNA; Bottom right, BCLAF1 and EV-associated marker quantification in K562-derived EVs. Error bars indicate standard deviation; C Immunofluorescence-based co-staining of BCLAF1 and CD63 (with DAPI for nucleus staining) in U937 cells untreated or treated with SAHA; D BCLAF1 and CD63 co-staining in K562 cells untreated or treated with SAHA; 40X magnification; E FACS-based quantification of BCLAF1-positive vesicles in EVs derived from SAHA treated vs untreated U937 cells. Data were normalized to viable cells and conditioned medium volume. Results are expressed as percentage of BCLAF1 + EVs to the total number of events. Error bars indicate standard deviation

On the other hand, in K562, SAHA induced BCLAF1 loading (both mRNA and protein) in all EV subpopulations, despite reduced EV marker detection (Fig. 2B, right), thus further proving that the treatment differently affected intra- and extra-cellular compartment content.

Furthermore, imaging approach revealed increased BCLAF1/CD63 overlap in both U937 (Fig. 2C) and K562 cells (Fig. 2D), therefore confirming the enhanced association of BCLAF1 with the extracellular compartment following SAHA treatment (observation confirmed in U937 also at cytometry level in Fig. 2E). Based on these results, we speculated that BCLAF1 might be either regarded as waste for disposal or exert functional activity within the tumor microenvironment, potentially contributing to treatment resistance.

SAHA induces prominent changes in leukemia EV proteomics

Besides BCLAF1, we also explored the entire proteomic landscape. Proteomics analysis of U937-derived EVs detected 140 proteins in EV Ctr and 150 in EV SAHA, with 130 proteins (81.3%) shared between the two groups (Table 1). EV SAHA exhibited lower levels of actin (ACTG1) and heat shock protein 90 (HSP90) (Fig. 3), both implicated in drug resistance [40,41,42]. Conversely, proteins such as Cullin-3 (Cul3), associated with drug sensitivity [43], and Interleukin Enhancer Binding Factor 2 (ILF2), linked to resistance to genotoxic drugs [44], were uniquely identified in EV SAHA. Moreover, Ingenuity Pathway Analysis highlighted differential biological functions induced by the EV groups, predicting that EV SAHA could inhibit tumor invasion and viability (Table 2, Fig. 3).

Table 1 Proteomics in U937 EVs w/wo SAHA

Fig. 3

SAHA-induced changes in AML EV content and resulting biological functions predicted to be inhibited in recipient cells. A Volcano plot showing differentially loaded proteins in AML EVs upon SAHA treatment. Proteins most detected in EV SAHA compared to EV Ctr are in red; proteins less detected in the SAHA group compared to related Ctr are in blue. Proteins involved in drug resistance are circled (ACTG1, fold change = − 0.76; HSP90AA1, fold change = − 1.13); B Ingenuity Pathway Analysis showing functions predicted to be modulated in recipient cells by U937 EVs. Bottom, pie chart of all functions predicted to be modulated by the vesicle groups. Functions not significantly modulated in the comparison between EV SAHA vs EV Ctr are in gray. Functions predicted to be significantly inhibited in EV SAHA are in orange (“Cell viability of tumor cell lines”, 4.3%) and blue (“Invasion of tumor”, 4.3%). Top, proteins involved in the reported functions. The function “Invasion of tumor” showed a p-value of 6.5 × 10^–6 and z-score of − 2.155. The function “Cell viability of tumor cell lines” showed a p-value of 1.10 × 10^–9 and z-score of − 2.391. Activation z-score £ − 2 means the function is inhibited; z-score ≥ + 2 means the function is induced

Table 2 Downstream analysis by IPA tool—U937 EVs

Regarding K562, EV SAHA contained 234 proteins compared to 124 in EV Ctr, with 100 proteins (38.8%) shared (Table 3). Bioinformatic predictions suggested that EV SAHA might enhance pro-tumoral functions (Table 4). However, proteins like 60S ribosomal protein L11 (RPL11), which sensitize cells to therapeutic agents [36], were more abundant in EV SAHA.

Table 3 Proteomics in K562 EVs w/wo SAHA

Table 4 Downstream analysis by IPA tool—K562 EVs

These observations suggest the potential therapeutic value of SAHA in particular when combined with complementary therapies.

SAHA-inducedEVs affect tumorigenicity but not resistance to SAHA in leukemic cells

Following molecular assessments and in silico predictions, we interrogated the functional impact of EV SAHA on recipient leukemic cells. Specifically, we evaluated whether EV SAHA influences tumorigenicity and modulates resistance to SAHA treatment.

In U937 cells, neither EV Ctr nor EV SAHA significantly affected cell viability or cell cycle (Fig. S3A-B).In a tumorigenicity assay, however, AML-derived EV SAHA elicited pro-tumoral effects comparable to those observed with EV Ctr (Fig. 4A, B). Notably, EV treatments increased BCLAF1mRNA levels in recipient cells, gain that was abrogated by SAHA treatment (Fig. 4C Left). Expectedly, the EV intervention did not enhance the miRNA levels in recipient AML cells; instead, SAHA induced miR-194-5p overexpression regardless of EV exposure (Fig. 4C Right). Furthermore, neither EV Ctr nor EV SAHA conferred resistance to the treatment in recipient cells (Fig. 4D, E).

Fig. 4

Tumorigenic capability, drug resistance and molecular delivery induced by EV SAHA in U937 and K562. A Images acquired with Cytation of U937 colony formation upon treatment with PBS, LV Ctr, LV SAHA, SV Ctr, and SV SAHA; B Left, number of newly formed colonies (per cm²); Right, area of the colonies (indicated as pixel density). Error bars indicate standard deviation; C Left, BCLAF1mRNA yield in U937 cells exposed to EVs from different sources and subsequently untreated or treated with SAHA (5 µM at 24 h), normalized to GAPDH expression; Right, miR-194-5p yield in U937 cells exposed to EVs from different sources and subsequently untreated or treated with SAHA (5 µM at 24 h), normalized to RNU6 expression; D Survival and tumorigenic capability of U937 cells pretreated with EVs from different sources followed by SAHA treatment (5 µM at 24 h). Images acquired with Agilent BioTek Cytation 5; E Propidium Iodide-based evaluation of cell survival following SAHA treatment induced by EVs from different sources. Error bars indicate standard deviation; F Images acquired with Cytation of K562 colony formation upon treatment with PBS, LV Ctr, LV SAHA, SV Ctr, and SV SAHA; G Left, number of newly formed colonies (per cm²); Right, area of the colonies (indicated as pixel density). Error bars indicate standard deviation; H BCLAF1mRNA yield in K562 cells exposed to EVs from different sources and subsequently untreated or treated with SAHA (5 µM at 24 h), normalized to GAPDH expression; I miR-194-5p yield in K562 cellsexposed to EVs from different sources and subsequently untreated or treated with SAHA (5 µM at 24 h), normalized to RNU6 expression

Comparable evaluations were performed in K562 cells (Fig. 4). While EVs did not affect viability or cell cycle (Fig. S3C-D), EV SAHA (both SV and LV) produced fewer new tumor colonies relative to EV Ctr (Fig. 4F, G), indicating an anti-neoplastic effect. Consistent with the AML findings, EV treatments delivered BCLAF1 mRNA to recipient CML cells, with only LV Ctr showing a restrained BCLAF1accumulation (Fig. 4H). Regarding the miRNA expression, no significant differences were observed between cells treated with EV SAHA and those receiving EV Ctr, although LVs appeared to deliver greater amounts of miR-194-5p than SVs. Interestingly, EV exposure prior SAHA treatment resulted in an overall enhancement of miR-194-5p expression compared to SAHA treatment alone (Fig. 4I).

In summary, our findings suggest that EV SAHA does not exacerbate tumor progression in AML, whereas it may inhibit malignant development in CML. These results underscore the context-dependent therapeutic potential of SAHA in leukemia.

Discussion

EV-mediated communication between cancer cells and their microenvironment is gaining increasing recognition as a pivotal process in tumor progression, as it significantly contributes to numerous cancer-related processes, such as tumorigenesis, tumor growth, angiogenesis, immune escape, metastasis, and drug resistance [45, 46]. However, albeit their relevance, the impact of anti-cancer treatments on vesicle-mediated communication is often underestimated and therefore under explored.

This study provides valuable insights into the role of SAHA in modulating EV secretion and composition, while also offering interesting speculations on the potential function of SAHA-modified EVs in Leukemia. By altering both the quantity and molecular content of EVs, SAHA affects intracellular dynamics and extracellular communication, presenting new avenues for therapeutic exploration.

Our previous efforts revealed that SAHA reverses the miR-194-5p/BCLAF1 imbalance responsible for the differentiation arrest and apoptosis resistance characteristic of Leukemia cells. In this study, we uncover that SAHA, while upregulating miR-194-5p expression at the expense of BCLAF1 intracellularly, it selectively fosters BCLAF1 (at both mRNA and protein level) loading into EVs and reduces the miRNA levels in the same compartment. This highlights a previously unrecognized role for SAHA in regulating selective cargo sorting into vesicles, potentially reshaping the tumor microenvironment. Regardless of whether this is traceable to waste disposal dynamics or communication means, EV-associated content may influence recipient cells and considering the anti-apoptotic role of BCLAF1, we speculated whether EV-BCLAF1 endows cancer cells with a phenotype refractory to SAHA. Noteworthy, the overall increase in extracellular BCLAF1 did not confer recipient cancer cells with resistance to SAHA treatment. To explain this result, we initially hypothesized a reduction in EV uptake yet the BCLAF1mRNA levels are increased in AML cells treated with EV SAHA, suggesting that the vesicles are internalized by recipient cells and avoid lysosome degradation. Interestingly, treatment with SAHA impaired even BCLAF1gained via EV communication. This effect may result from SAHA’s dual role, whereby it enhances miR-194-5p expression while simultaneously limiting its incorporation into EVs, leading to elevated intracellular miRNA levels which could further restrict the accumulation of BCLAF1.

In addition to our molecules of interest, SAHA significantly modulated the overall miRNA profile and proteomic landscape of Leukemia-derived EVs. Such altered vesicular composition was predicted—and confirmed for some biological functions—not to exacerbate the aggressive phenotype of recipient cancer cells, as demonstrated by our functional analysis. Moreover, SAHA modulated the loading of many factors linked to drug resistance and sensitivity, underscoring a dual role for SAHA who has the potential to limit tumor progression while enhancing therapeutic responsiveness, thereby suggesting new options for potential combinatorial therapies.

Trials exploring the use of SAHA in combination with other drugs are ongoing (NCT00392353, NCT01522976) or concluded (NCT00948064, NCT00875745, NCT00479232, NCT00275080, NCT01550224, NCT01534260), and, despite showing promising results in some early-phase studies, the overall success of these trials varies. Our findings suggest the possibility of testing SAHA in combination with drugs for which trials have not been designed yet. Moreover, previous trials investigated SAHA in combination with other therapeutics, while our results suggest adopting a sequential approach, wherein SAHA is used prior to other drugs. In this regard, pre-treatment with SAHA might directly kill malignant cells meanwhile increasing the efficacy of cisplatin and paclitaxel, due to impaired delivery of actin and HSP90 as well as increased transportation of miR-424-5p via EV SAHA. Similarly, these vesicles have the potential to increase the sensitivity to doxorubicin, because of the unique presence of Cul3 in EV SAHA. Moreover, the ability of SAHA to enhance sensitivity to other treatments could allow for dose reductions, potentially minimizing toxicity-related side effects—one of the leading causes of treatment-associated mortality [47]. Additionally, SAHA pre-treatment may influence the efficacy of cell-based and antibody-based therapies, for instance by upregulating surface antigen expression or attenuating immunosuppressive signaling, thereby increasing tumor susceptibility to both the host immune system and CAR-T cells [48, 49].

These findings highlight SAHA’s potential to modulate the extracellular compartment as part of a broader anti-cancer strategy thereby supporting its potential clinical relevance as a component of combination therapies aimed at overcoming resistance in Leukemia.

Importantly, these predictions were only observed in AML cells, as EV SAHA from CML cells were predicted to enhance some tumor-supportive function. SAHA-induced EVs demonstrated a context-dependent impact also at functional level. In CML, EV SAHA treatment hindered the process of colony formation, while no differences between EV Ctr and EV SAHA were observed in AML cells. These different outcomes may be attributed to the diverse differentiation potential, genetic, epigenetic and molecular background, as well as cell lineage and function of acute versus chronic leukemia cells [50]. Moreover, SAHA also exerted differential effects on EV secretion in AML and CML cells, likely reflecting cell-specific differences invesicle biogenesis pathways or sensitivity to SAHA-induced stress, therefore emphasizing the importance of tailoring therapeutic approaches to specific Leukemia subtype.

In conclusion, SAHA’s effects on leukemic EV secretion and molecular composition, along with its direct cytotoxicity toward tumor cells, support its application as a therapeutic agent, especially in combination regimens customized to specific leukemic subtypes. Our findings provide a speculative yet promising foundation for future research into novel therapeutic strategies, encouraging further investigation into the mechanisms underlying these effects and their potential clinical implications.