Introduction

Acute promyelocytic leukemia (APL), a subtype of acute myelogenous leukemia (AML), is characterized by a balanced reciprocal translocation between chromosomes 17 and 15 [t(15;17)(q22;q21)]. This genetic event results in oncogenic fusion between promyelocytic leukemia (PML) and retinoic acid receptor alpha (RARa) genes [1], thereby altering their physiological roles. RARa acts as the nuclear receptor for the retinoic acid (RA), a function that triggers transcriptional activation of genes involved in granulocytic differentiation. While PML–RARa and other less frequent RARa fusion products obliterate myeloid differentiation, the exposure of leukemic cells to all-trans retinoic acid (ATRA) can restore RARa-dependent transcription, also leading to degradation of the PML–RARa oncoprotein by proteasome and autophagy pathways [2]. In combination with arsenic trioxide (ATO), which targets PML–RARa in leukemic stem cells, and chemotherapy, ATRA imparts complete and lasting remission in the majority of APL cases [3].

APL can be viewed as a paradigm of not only “oncogene addiction” and the efficacy of oncogene-directed (differentiation) therapy with ATRA, but also a multifaceted nature of vascular involvement in the leukemic process. Thus, coagulopathy is a common and early event in APL patients, as manifested by severe bleeding tendency with a considerable risk of mortality [4, 5]. This diathesis results from disseminated intravascular coagulation (DIC) largely attributed to procoagulant characteristics of leukemic cells themselves, including their Annexin 2 (A2) and tissue factor (TF) expression, but may also involve emission of TF exposing microparticles/extracellular vesicles (EV) into the circulation [6–8]. Moreover, bone marrow of patients with APL exhibits increased features of angiogenic activation, such as elevated microvascular density, while leukemic cells are known to produce proangiogenic and proinflammatory mediators [9]. This phenotype includes PML–RARa-dependent up-regulation of vascular endothelial growth factor (VEGF) [10], and the concomitant activation of hypoxia response pathways [11]. These characteristics are consistent with the proposed role of the angiogenic circuitry in the progression of hematopoietic malignancies [12]. The related mechanisms include bone marrow perfusion, trophic and endothelial niche effects, which support the expansion of leukemic stem cells [11, 13, 14]. These effects may also entail autocrine stimulation of malignant cells by angiogenic growth factors [15], immunomodulation [16] and possibly other processes [17].

ATRA profoundly alters the cellular phenotype, gene expression [18–20] and interactions between APL cells and the vascular system. While the nature of these effects remains controversial [21], it is well documented that the exposure to ATRA triggers a reversal of TF-related procoagulant activity of APL cells and suppresses coagulopathy in vivo [22, 23]. These responses also include down-regulation of VEGF and of other APL-associated angiogenic activities [10]. However, ATRA also possesses potentially proangiogenic activities, such as direct stimulation of endothelial cell responses [21], and production of proangiogenic mediators by APL cells. The latter includes ATRA-dependent deregulation of cytokines and chemokines, such as interleukin 8 (IL-8/CXCL8), IL-6 and IL-1β, some of which may directly interact with endothelial cells (IL-8), while others promote inflammatory responses, potentially with vascular consequences. These paradoxical effects are exemplified by the drug-related differentiation syndrome (DS) associated with respiratory distress, pulmonary infiltration and other perturbations affecting lung microcirculation in ATRA-treated patients [20, 24].

It is now recognized that in addition to cytokines and soluble mediators the secretome of cancer and leukemic cells also contains other more complex bioactive constituents, including heterogeneous membrane structures known as extracellular vesicles (EVs) [25]. EVs are generated through diverse biogenetic mechanisms [25], some of which involve membrane blebbing leading to abscission and shedding of larger vesicles (100–1000 nm) known as ectosomes, microvesicles or microparticles (MPs). In contrast, smaller vesicles (30–150 nm in diameter) may emerge within endosomal multivesicular bodies (MVB) and undergo a regulated release into the extracellular space as exosomes, rich in tertraspanins, integrins, membrane receptors and nucleic acids. Several other vesiculation pathways may also exist and result in formation of multiple distinct EV subsets [25–27].

EVs are believed to mediate molecular transfer and intercellular communication between hematopoietic, stromal, vascular and leukemic cells. This includes the maintenance of cellular equilibrium within the stem cells hierarchy [28], reprogramming cellular differentiation states [29], coagulation system activation, communication with stromal and endothelial cells, and transfer of signaling receptors, nucleic acids and other cargo in various settings of chronic and acute leukemia [29–33]. While such EV-mediated processes have been described in the case of APL, their nature remains relatively unexplored [7].

Of particular relevance in this context is the link between EVs and oncogenic pathways [34, 35]. Thus, cellular vesiculation is often increased or otherwise deregulated in cells expressing mutant oncogenes, while the respective oncogenic proteins, transcripts, micro-RNA and DNA are found in the cargo of EVs produced by cancer cells [34, 36–38]. Moreover, the emission of such EVs containing cancer-specific cargo may serve as a mechanism of intercellular horizontal transfer of biological activity leading to responses that may resemble cellular transformation, including increase in survival and angiogenic potential in the recipient cells [34, 36, 38–41]. For example, EV-mediated transfer of oncogenic epidermal growth factor receptor (EGFR) to normal endothelial cells may result in the onset of VEGF production and autocrine reprogramming of these cells [42]. While some of these effects have also been investigated in models of leukemia [32, 43], the links between PML–RARa-mediated transformation, ATRA therapy and vesiculation pathways in APL cells remain relatively unstudied.

Here we report that PML–RARa and ATRA treatment alter the emission profile and cargo of EVs produced by APL cells. APL-derived NB4 cells produce EVs with endothelial stimulating activity. However, unlike in other tumor contexts, we observe no incorporation of the bioactive PML–RARa oncoprotein itself into the EV cargo, or its direct transfer to target endothelial cells. Instead, EVs released by APL (NB4) cells contain several PML–RARa (ATRA)-regulated vascular effector proteins and transcripts (TF, VEGF, IL-8). Thus, we conclude that oncogenic effects of PML–RARa modulate EV productions and angiogenic cargo, but do not lead to production of PML–RARa-containing EVs.

Materials and methods

Cells and ATRA treatment

Human APL cell line NB4 was generously provided by Dr. Minden of the University of Toronto and cultured in RPMI-1640 medium supplemented with 10 % fetal bovine serum (FBS) (Multicell Wisent, Inc., St-Bruno, Quebec, Canada). Human umbilical vein endothelial cells (HUVEC) were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultured in complete endothelial cell growth medium-2 (Lonza, Allendale, NJ). Unless otherwise indicated, NB4 cells were treated with ATRA (Sigma, Oakville, ON, Canada) at the final concentration of 1 umol/L which was established as non-toxic and capable of inducing myeloid differentiation measured by positivity for the CD11b antigen (data not shown).

Preparation of extracellular vesicles

EVs were isolated as described elsewhere [34]. Briefly, the assay medium was prepared using EV-depleted FBS obtained by ultracentrifugation at 150,000×g for 5 h and passaged through 0.2-µm-pore-size filters. NB4 cells were cultured in the assay medium for the lengths of time indicated (usually 2–3 days), and such conditioned medium was collected through centrifugation at 400×g for 10 min. The supernatant was passed through 0.45- or 0.2-µm filters (Thermo Scientific; Mississauga, ON), followed by centrifugation at 100,000×g for 1 h to pellet EVs. The EV pellet was washed extensively in phosphate-buffered saline (PBS) and then centrifuged at 100,000×g for 1 h and analyzed immediately.

Transmission electronic microscope (TEM)

Cells and EVs were fixed 2.5 % glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4). Samples were processed according to standard protocols and imaged on the Tecnai 12 BioTwin 120 kV TEM.

Nanoparticle tracking analysis (NTA)

EV profiling entailed the measurements of size and number of particles shed from cells cultured in EV-depleted medium using the NS500 nanoparticle tracking analysis system (#NS500, Nanosight, Amesbury, UK).

Co-culture assays

NB4 cells and HUVECs were surface labeled with PKH26 (red fluorescent dye) and PKH67 (green fluorescent dye; both from Sigma-Aldrich, Oakville, ON), respectively, as described earlier [44] and according to manufacturer’s instructions. The labeled cells were co-cultured in 4-well chamber slides for 24 h. Suspended NB4 cells were settled on the surface of the slide by gentle cytospin (Thermo Scientific) centrifugation. Upon incubation, the cells were fixed for 10 min in 4 % paraformaldehyde (PF) and imaged using Zeiss LSM 780 laser scanning confocal microscope.

EV transfer assays

NB4 cells were surface labeled with PKH26 (red fluorescent dye; Sigma-Aldrich) [44] and allowed to release fluorescent EVs into the culture media. Labeled EVs were purified as above and incubated for 24 h with HUVECs following the measurement of the fluorescence uptake using FACSCalibur flow cytometer (BD Biosciences, Mississauga, ON). For EV-mediated RNA and protein transfer, recipient cells were lysed and tested for the respective molecular targets. Biological responses to the EV transfer were recorded using MTS assay (Promega, San Luis Obispo, CA) as described earlier [38]. Briefly HUVEC cells were seeded in gelatin pre-coated 96-well plates and then incubated with different concentration of NB4 EVs at 0, 5, 25, 50, 100 and 200 µg/ml, followed by addition of MTS reagents and reading of the metabolic activity at 405 nm. The changes in viability and cell numbers were inferred from the change in MTS signal.

RNA analysis

EVs were harvested from conditioned medium of NB4 cells at 72 h and surface pre-treated with 100 µg/ml RNase (Sigma) for 15 min at 37 °C to remove non-luminal RNA contaminants. EV-associated RNA was extracted using RNeasy® mini kit (Qiagen) according to manufacturer’s instructions, following conversion into cDNA using QuantiTect® Reverse Transcription kit (Qiagen, Mississauga, ON). The following primers spanning the fusion site were obtained from Integrated DNA Technologies (Coralville, IA): PML–RARa (forward 5′-TCTTCCTGCCCAACAGCAA-3′ and reverse 5′-GCTTGTAGATGCGGGGTAGAG-3′); GAPDH (forward 5′-GAGTCAACGGATTTGGTCGT-3′ and reverse 5′-TTGATTTTGGAGGGATCTCG-3′). For PCR array analysis, reverse transcription was conducted using the RT2 First Strand kit (Qiagen). PCR array (SABiosciences/Qiagen, Mississauga, ON) analyses were carried out according to manufacturer’s instructions, the result of which was illustrated in MultiExperiment Viewer (MEV) and further analyzed by Vennplex [45]. Fold change of ≥2.0 or ≤2.0 was chosen as cutoff values to define significantly increased and decreased gene expression, respectively.

Western blotting

Protein lysates were obtained in RIPA buffer containing protease inhibitors (Roche, Mississauga, ON), or nuclear protein extraction buffer for EVs and cells. Protein concentrations were qualified by BCA Protein Assay (Thermo Scientific) resolved by SDS-PAGE and transferred to PVDF membranes. The membranes were incubated with antibodies against either rabbit anti-human RARa (Sigma), mouse anti-human RARa (R&D Systems/Cedarlene, Burlington, ON), mouse anti-human b-actin (Sigma), mouse anti-human flotillin-1, mouse anti-human TSG101 (both from BD Biosciences, Mississauga, ON), mouse anti-human Rab27a (Abnova, Corp. Walnut CA), rabbit anti-human CREB (Santa Cruz Biotech, Santa Cruz, CA) followed by corresponding horseradish peroxidase-coupled secondary antibodies (Cell Signaling Technology Inc., Danvers, MA and Bio-Rad, Mississauga, ON). The signal was visualized using enhanced chemiluminescence (GE Healthcare, Toronto, ON), imaged and quantified with ChemiDoc software (Bio-Rad).

Elisa

Vascular regulators were quantitatively detected using ELISA assays as described earlier [34]. IMUBIND TF ELISA kit (Sekisui Diagnostics, Stamford, CT), human VEGF ELISA kit (R&D Systems, Minneapolis, MN) and human IL-8 ELISA kit (R&D Systems, Minneapolis, MN) were used to quantify the levels of TF, VEGF and IL-8, respectively. TF procoagulant activity was measured as described earlier [46] by Actichrome® TF kit (Sekisui) according to manufacturer’s instructions.

Statistical analysis

All experiments were reproduced at least twice with similar results and presented as mean value of replicates ± SD. Statistical significance was evaluated using SPSS software, and differences were considered significant for p < 0.05.

Results

Impact of ATRA on vesiculation of cultured APL cells

While APL cells are known to release bioactive EVs [7], the link between this process and malignant transformation remains presently unknown. Given the impact of oncogenic pathways on vesiculation in other cancer contexts [27], we surmised that PML–RARa driver mutation may influence EV emission by APL cells, both quantitatively and qualitatively. Therefore, we chose NB4 cells, a well-characterized model of human APL, to examine the profile of EV emission in the presence or absence of ATRA-mediated inhibition of the PML–RARa transforming activity. We first performed cell viability (MTS) and myeloid differentiation assays (FACS), using different dosing and timing protocols. As expected, following 72 h of exposure to 1 μmol/L of the drug, ATRA induced a measurable reversal of PML–RARa-dependent transformation of NB4 cells, as indicated by the increase in CD11b positivity and in the absence of any apparent cytotoxic effects (Fig. S1).

Interestingly, in the presence of ATRA the number and size of EVs produced by NB4 cells also underwent a remarkable change. Thus, as measured by NTA (Nanosight) assays and revealed by electron microscopy (TEM), untreated NB4 cells constitutively release readily detectable EVs, heterogeneous in size and shape but largely below 300 nm in diameter (Fig. 1a). While there is a preponderance of larger vesicles (100–300 nm) in this material, smaller, exosome-sized EVs can also be visualized, either in conditioned media (inset) or in association with cellular membranes (Fig. 1b). This profile underwent a profound change following the exposure to ATRA, which resulted in a sharp and selective increase in emission of small exosomal-sized EVs (below 100 nm in diameter; p = 0.02, Fig. 1a). The shape of NB4-derived EVs has also changed as a result of this treatment. In this regard, TEM images reveal a greater number of smaller and more vacuolar EVs in the conditioned medium of treated cells (Fig. 1b–e), along with larger EVs containing intraluminal vesicular structures. The surface of ATRA-treated NB4 cells contained fewer protrusions and blebs relative to control cells, and a greater number of pits reminiscent of exocytosis. More apparent and less electron dense vesicular structures were also present within the cytoplasm of these cells along with features of cellular differentiation. Notably nuclei of ATRA-treated cells assumed a more granulocytic polylobular appearance with reduced nuclear-to-cytoplasm ratio (Fig. 1c). These experiments suggest that PML–RARa affects the pattern of cellular vesiculation in APL cells in a manner that relates to their differentiation status and can be reversed in the presence of ATRA.

Fig. 1
figure 1

Impact of ATRA on vesiculation of APL cells in culture. a Nanoparticle tracking (NTA/Nanosight) profiling of EV emission by NB4 cells cultured in the presence or absence of ATRA (1 μM) for 72 h. ATRA triggers a change in numbers and size distribution of leukemic EVs; mean particle counts ± SD (n = 3 independent runs); *p < 0.05; **p < 0.01; ***p < 0.001; b–e Transmission electron microscopy of vesiculating NB4 cells and their EVs (insets) in the absence (b, c) and presence (d, e) of ATRA. Features of more differentiated nuclear morphology, vacuolar cytoplasm, changes in cell surface and shape of EV are apparent in ATRA-treated cells (representative images)

Full size image

Responses of endothelial cells to the uptake of NB4-derived EVs

Oncogenes are known as drivers of the angiogenic phenotype in cancer [47], in part though their impact on cellular vesiculation [34]. We asked whether this could be applicable to APL cells. To explore this question, we first examined whether co-culture of NB4 cells with normal endothelial cells (HUVEC) leads to a detectable transfer of membrane material between these two cell populations. Membranes of NB4 cells were labeled with the PKH26, red fluorescent dye, and these cells were seeded in dishes containing HUVEC monolayers pre-labeled with PKH67 (green dye). Following co-culture for 24 h, the cells were visualized using fluorescent confocal microscopy (Fig. 2a–d). Indeed, both cell types could be readily located using the corresponding fluorescent signals (Fig. 2a, c). Interestingly, in the course of co-culture, some HUVECs exhibited a mixture of green (original) and red (NB4-derived) fluorescence on their surfaces (Fig. 2b). In contrast, NB4 cells remained only labeled with the red fluorescence throughout the experiment (Fig. 2d) and clusters of NB4 and HUVEC cells were not readily observed.

Fig. 2
figure 2

Uptake and responses of endothelial cells to leukemic EVs. a–d Uptake of fluorescently labeled cellular membranes from NB4 cells to HUVECs. NB4 cells were surface labeled with PKH26 (red), while adherent HUVECs were labeled with PKH67 (green) and the cells were imaged using confocal microscopy individually (control; a, c) or upon co-culture for 24 h (co-culture; b, d). HUVECs acquired red fluorescence in this co-culture system (yellow; b), suggesting the uptake of EVs. e Treatment of HUVEC cells with fluorescent (PKH26) NB4-derived EVs results in their uptake and cell fluorescence (FACS). f Stimulation of survival (MTS) of starved HUVEC cells by treatment with NB4-derived EVs; mean absorbance ± SD (n = 3 independent runs); *p < 0.05; **p < 0.01, ***p < 0.001

Full size image

We reasoned that this membrane exchange could occur through the uptake of shed EVs and in the absence of direct cell–cell contact. To explore this possibility more directly, fluorescent EVs were purified from cultures of NB4 cells that were pre-labeled with PKH26. HUVEC cells incubated with these EV preparations revealed the acquisition of robust red fluorescence, as documented by FACS analysis indicating the ability of endothelial cells to avidly take up leukemic EVs (Fig. 2e). Since such EVs were previously postulated to possess proangiogenic properties, we used MTS assay to test responses of growth-factor-starved HUVEC to the addition of increasing concentrations of EVs derived from NB4 cell cultures. This treatment increased the viability signal in HUVEC cultures, in a significant and dose-dependent manner with the maximal effect achieved at the EV protein concentration of 100 µg/ml (Fig. 2f). Since these responses were achieved in the absence of exogenous endothelial survival factors, MTS readings likely reflect cell survival in this case. This observation suggests that leukemic cells can interact with endothelium via the exchange of EVs.

EV-mediated emission and intercellular transfer of PML–RARa transcript

It has previously been suggested that horizontal stimulation and transformation may occur by EV-mediated intercellular trafficking of oncogenic mRNA and protein from cancer cells to their untransformed neighbors [36]. A similar process may also be involved in tumor-vascular interactions and angiogenesis [42]. To explore whether such oncogene transfer may occur in the context of APL cells and to explain their aforementioned ability to stimulate endothelial cell survival (Fig. 2f), we first assessed the expression and EV-mediated release of the oncogenic PML–RARa transcript by NB4 leukemic cells (Fig. 3). NB4-derived EVs were externally pre-treated with RNase to remove surface contamination with soluble RNA [48], and the lysates of this material were used to test for the expression of PML–RARa transcript by RT-PCR. Indeed, this analysis revealed the presence of PML–RARa fusion gene product in both NB4 cells and their derived EVs, the latter either tested collectively, or size fractionated by passage through either 0.45-, or 0.22-µm-pore-size filters. This filtration step was intended to enrich the EV preparation for exosome-like small vesicles and also to preclude the contamination with live cancer cells and cellular debris. Furthermore, we found that leukemic EVs are capable of delivering oncogenic mRNA into normal endothelial cells, since HUVECs incubated with NB4-derived EV preparations contained the PML–RARa transcript, which was undetectable in EV-untreated control cultures (Fig. 3b). These experiments document that APL cells release EVs containing oncogenic mRNA, which can be taken up by non-transformed primary endothelial cells.

Fig. 3
figure 3

EV-mediated emission and intercellular transfer of PML–RARa transcript. a Detection of PML–RARa mRNA in lysates of NB4 cells and their EVs (RT-PCR). EVs were surface pre-treated with RNAse to remove external contamination and used to extract RNA, either directly or upon passage thorough either 0.45- or 0.22-µm-pore-size filters, to remove debris and cells, and to restrict the sizes of EVs to near exosomal range. In all cases PML–RARa transcript was readily detected. b HUVEC cells were used as recipients of NB4-derived EVs, resulting in the transfer of PML–RARa mRNA signal (RT-PCR; as in “Methods” section)

Full size image

The absence of functional PML–RARa protein in leukemic EVs or endothelial cells following vesicle transfer

Stimulation of endothelial cells with leukemic EVs (Fig. 1), and their content of PML–RARa transcripts are reminiscent of earlier findings suggesting the horizontal transfer and expression of active oncogenes [42]. For PML–RARa oncogene to exert such an effect would necessitate the expression of the bioactive oncoprotein in recipient endothelial cells. This could occur either through translation of the EV-transferred mRNA in recipient cells, or through the intercellular transfer of preformed oncoprotein (regardless of mRNA), as demonstrated for oncogenic membrane receptors [34, 39]. In either case, the recipient cells would be expected to become responsive to PML–RARa inhibition by ATRA. To test these possibilities, we first determined that PML–RARa protein is readily detected in NB4 cells and its expression could be inhibited somewhat by the exposure to ATRA for extended periods of time (72 h; Fig. 4a), while the cells underwent myeloid differentiation (data not shown).

Fig. 4
figure 4

Absence of PML–RARa protein expression in leukemic EVs and in endothelial cells following the vesicle transfer. a PML–RARa (top band) and RARa proteins (bottom band) are readily detected in nuclear lysates of NB4 cells (Western blotting). Addition of ATRA for different lengths of time results in a slight diminution of the PML–RARa signal after 72 h of treatment. b The absence of PML–RARa and RARa proteins in lysates of NB4-derived EVs. The respective protein loading is documented by signal for Creb (cells) and Flotillin-1 (EVs). c NB4 cells and their derived EVs express vesicular, raft and exosomal proteins (Flotillin-1, TSG101, RAB27a), but EVs are negative for PML–RARa; beta (b)-actin was used as loading control. d Transfer of NB4-derived EVs to HUVECs does not result in expression of PML–RARa by recipient endothelial cells, suggesting that neither functional protein, mRNA nor DNA is expressed in the latter cellular background (Western blotting; as in “Methods” section)

Full size image

We next examined the NB4 EVs in a similar manner. However, no PML–RARa protein could be detected by Western blotting in this material with or without pre-treatment with ATRA (Fig. 4b). This suggests that this particular oncoprotein is not released from APL cells as cargo of EVs. Of note is the fact that our EV preparations contained abundant protein cargo and were positive for several markers associated with exosomes (TSG-101, RAB27a) and other vesicles (Flotillin-1), suggesting that PML–RARa protein is selectively excluded from the vesiculation process (Fig. 4c). To explore the possibility that PML–RARa transcript may, nonetheless, trigger the expression of the bioactive oncoprotein in EV recipient endothelial cells, we cultured HUVECs in the presence or absence of NB4-derived EVs for 24, 48 or 72 h (Fig. 4d). Once again, no PML–RARa protein was found in HUVEC lysates treated in this manner. Collectively, these results suggest that, unlike in the case of other oncogenes [42], PML–RARa is not incorporated into the EV cargo [34], or directly transferred to recipient endothelial cells.

Indirect effects of PML–RARa on vascular mediators present in the leukemic secretome

Since stimulation of HUVECs with leukemic vesicles occurs without a direct intercellular transfer of the bioactive PML–RARa, we reasoned that this oncogene may impact the biological activity of EVs more indirectly, namely through reprogramming the repertoire of vascular mediators incorporated into these vesicles prior to their release. To this end, we explored the known PML–RARa targets with vascular activity, namely VEGF, IL-8 and TF. Their levels in the cargo of NB4 EVs and cellular supernatants were analyzed using ELISA and activity assays.

Thus, as expected NB4 cells produced appreciable amounts of VEGF protein, which was mostly associated with the soluble fraction of the conditioned medium and only in minor amounts retained in cell lysates or EVs (Fig. 5a). Treatment of NB4 cells with ATRA led to a significant reduction in the residual EV content of this angiogenic growth factor (Fig. 5b). Similarly, NB4 cells also produced considerable amounts of cell-associated and soluble IL-8. While some of the secreted IL-8 immunoreactivity could be removed from the conditioned medium by ultracentrifugation, suggesting the association with the fraction of EVs, the recovery of IL-8 from purified EVs was minimal (as in the case of VEGF), possibly through losses during filtration and preparation steps or by association with cell debris or larger vesicles. Nonetheless, unlike in the case of VEGF, treatment of donor NB4 cells with ATRA resulted in a robust, near ten-fold up-regulation of the IL-8 content within leukemic EVs (Fig. 5d). Thus, angiogenic factors produced by NB4 cells are mostly contained in the soluble fraction and the immediate biological effects related to their content in EVs are likely negligible, especially in the case of VEGF. However, the EV content of angiogenic factors reflects ATRA-induced reprogramming of the vascular phenotype of leukemic cells, from proangiogenic (VEGF) to angiogenic/inflammatory (IL-8) state, a notion with conceivable implications for the use of EV as biomarkers of the vascular phenotype.

Fig. 5
figure 5

Secretion of angiogenic growth factors and EVs from cultured leukemic cells. a VEGF release by cultured NB4 leukemic cells (ELISA). Leukemic cells produce appreciable amounts of mostly soluble VEGF, of which only a small proportion was depleted by ultracentrifugation of conditioned media (w/o EVs). b The residual expression of EV-associated VEGF is reduced by NB4 cell pre-treatment with ATRA (1 μmol/L). c IL-8 release by cultured NB4 cells (ELISA). While some IL-8 appears to be cell-associated, a large proportion is secreted into the culture media. A minor fraction of IL-8 can be depleted from conditioned medium by ultracentrifugation (w/o EVs) and is poorly recovered from purified EVs, possibly due to association with cell debris and larger vesicles. d ATRA induces a dramatic increase in the IL-8 content of NB4-derived EVs; mean absorbance ± SD (n = 3 independent runs); *p < 0.05; **p < 0.01, ***p < 0.001

Full size image

Impact of ATRA on procoagulant properties of leukemic EVs

Unlike in the case of angiogenic growth factors, TF, the potent procoagulant and signaling transmembrane receptor with multiple vascular activities [49] is mostly cell-associated in the case of NB4 leukemic cells (Fig. 6a). However, these cells also release measurable amounts of TF into the culture media, of which the vast majority can be recovered with the EV fraction. We observed no obvious release of soluble or alternatively spliced TF (asTF) in this setting [46, 50]. Treatment with ATRA markedly reduced the EV-associated TF pool, in keeping with earlier reports linking the expression of this receptor with the oncogenic activity of PML–RARa [22] (Fig. 6b). While the presence of TF in the cargo of tumor-related EVs is often thought to contribute to cancer coagulopathy, this area is mired in controversy over the extent of the actual procoagulant activity of these circulating particles in various disease settings [51]. One way to reconcile these observations is to consider the role of TF taken up (as EV cargo) by normally non-coagulant endothelial cells [52]. Indeed, we observed that cultured HUVECs readily incorporate NB4 EVs (Fig. 2) leading to the increased exposure of the TF antigen on their surfaces and the elevation of the related procoagulant activity (TF-PCA). Notably, this uptake was markedly diminished by pre-treatment of NB4 cells with ATRA, suggesting that this procoagulant conversion of the recipient endothelial cells is ultimately influenced by the oncogenic PML–RARa activity within tumor cells (NB4) acting as EV and TF donors (Fig. 6).

Fig. 6
figure 6

ATRA-modulated expression of tissue factor in NB4 cells and their EVs. a Tissue factor (TF) is released into NB4 cell-conditioned medium in relatively small quantities, mostly within the EV fraction (ELISA). b The EV-associated TF content is markedly reduced (threefold) by NB4 cell pre-treatment with ATRA (1 μmol/L). c HUVECs treated with NB4-derived EVs acquire the expression of TF antigen (ELISA), which is reduced if donor cells are pre-treated with ATRA. d HUVECs become highly procoagulant upon the uptake of NB4-derived EVs, and ATRA pre-treatment of donor cells dramatically suppresses this conversion; mean absorbance ± SD (n = 3 independent runs); *p < 0.05; **p < 0.01, ***p < 0.001

Full size image

Impact of PML–RARa on angiogenesis-related gene expression profile in NB4 cells and EVs

To obtain more comprehensive insights as to vascular mediators affected by the PML–RARa-induced transformation of APL cells, we generated profiles of angiogenesis-related transcripts expressed in NB4 cells, or in their EVs, and in the presence or absence of ATRA. Using targeted RT-PCR array platform, we evaluated the levels of 84 key genes implicated in angiogenesis and vascular regulation including growth factors and their receptors, adhesion molecules, cytokines, chemokines, proteases, enzyme inhibitors and other regulatory molecules (Fig. 7). Several notable differences in distribution of these transcripts were observed between NB4 cells and their EVs, and between the cells and EVs following the treatment with ATRA. Thus, in comparison with their parental NB4 cells, 17 angiogenic transcripts were significantly up-regulated and 13 down-regulated in NB4 EVs. After ATRA treatment, 42 angiogenic transcripts were up-regulated while 18 transcripts were down-regulated in NB4 EVs compared to the corresponding cells (Fig. S2–S4).

Fig. 7
figure 7

Impact of ATRA on the expression of angiogenesis-related transcripts in NB4 cells and their derived EVs. Angiogenesis-targeted PCR-based gene expression array (SAB) analysis was conducted on lysates of intact or ATRA-treated NB4 cells and their corresponding EVs, and the fold change was computed using untreated NB4 cells as baseline. Fold change of ≥2.0 or ≤2.0 was chosen as the basis for cutoff values (n = 2 independent runs, see text for details)

Full size image

To assess the influence of ATRA on the expression of angiogenic transcripts in NB4 EVs specifically, we compared the profiles of EVs collected from untreated or ATRA-treated NB4 cells. In this case, 37 genes were significantly up-regulated and 18 genes down-regulated in EVs as a result of ATRA treatment. In this regard, the transcript for chemokine (C–C motif) ligand 2 (CCL2) was found to be the most remarkably up-regulated, and EPH receptor B4 (EPHB4) was the most strongly ATRA down-regulated gene under these experimental conditions (Fig. S2-4). We also observed that IL-8 mRNA was up-regulated in ATRA-treated EVs with the fold change of 72.15, ranking as fourth among the most up-regulated genes. VEGF and TF were found to be down-regulated by ATRA at the mRNA level (in EVs), with the fold change of 1.83 and 4.32, respectively. These results are consistent with our analysis at the protein level and, in a wider sense, demonstrate that EVs are a rich source of information as to the reprogramming of the angiogenic phenotype of APL cells following the inhibition of the PML–RARa oncogene by ATRA.

Discussion

Our present study brings forward several novel observations. First, unlike in the case of oncogenic receptor tyrosine kinases (EGFR, HER2, MET) or GTP-ases (RAS) [27, 34, 38], the transforming PML–RARa oncoprotein is not included in the vesicular cargo of APL cells, neither intact nor upon treatment with ATRA. Moreover, PML–RARa transcript present in the cargo of leukemic EVs does not induce any detectable expression of the PML–RARa oncoprotein upon the EV uptake by recipient endothelial cells. Thus, from the functional standpoint, NB4 cells do not produce oncogenic EVs (oncosomes) capable of horizontal transfer of the APL driver gene, at least to endothelial cell recipients. Second, PML–RARa transformation does influence cellular vesiculation profile, by inducing a wider size range and lower emission rates of leukemic EVs. This can be inferred from the opposite trend in the NTA profile in the presence of ATRA. Whether this is due to the effect of the PML–RARa inhibition on effectors of EV biogenesis in NB4 cells, or represents a secondary reflection of their myeloid differentiation remains presently unclear. It is also plausible that the change in EV profile results from some off-target effects of ATRA. This is unlikely given well-documented specificity of this agent, but all possibilities remain to be carefully investigated. Third, ATRA (PML–RARa) profoundly changes the molecular cargo of leukemic EVs, especially as it relates to multiple vascular effectors. These influences could reveal novel PML–RARa-dependent interactions between APL cells and the vascular system and the related consequences of the ATRA therapy [3, 10, 11, 17, 53].

These results may be informative as one element in considering the merits of EV-based biomarker development to monitor vascular effects in leukemia and other malignancies [27]. For example, even though VEGF is released from NB4 cells predominantly as a soluble protein, its content in defined EV subsets may reflect the change in production of this factor by a specific cell type (APL cells in this case) and thereby could help distinguish this from other sources of the angiogenic activity. Similar inferences can be made in the case of TF or the corresponding (vascular) EV-associated mRNA, as indicated by our results involving ATRA-induced shift in the profile of transcripts contained in NB4-derived EVs.

The oncogene and therapy-related changes in the cargo of leukemic EVs are also of interest in view of their possible biological activity. In this regard, interactions of APL cells with the vascular system are complex and unlikely reducible to angiogenesis stimulation or inhibition, or solely reflected by endothelial growth responses in vitro. For example, while we observed stimulation of HUVEC survival by NB4-derived EVs, this activity was not significantly affected by ATRA (Fig S5). This is not surprising, as decline in VEGF content following ATRA exposure was paralleled by the up-regulation of IL-8, another endothelial and inflammatory agonist, along with multiple other changes. While the exceedingly low content of VEGF and IL-8 associated with EVs may argue against the involvement of these factors in biological effects of EVs, they illustrate the underlying trends. We posit that at least some of these effects could be relevant to the ATRA-induced inflammatory side effects, such as differentiation syndrome (DS), some of which may be linked to the biological activity of APL-related EVs. It also cannot be presently excluded that EVs may possess a unique ability to transport certain cytokines, such as IL-8 and others (even in low amounts) to remote organs and cellular populations. In this manner EVs could, perhaps, still contribute to the recruitment of inflammatory cells or exert other effects qualitatively different from those attributable to extracellular diffusion gradients of the respective soluble mediators.

It is also presently unclear whether angiogenic transcripts found in APL-derived EVs merely reflect the state of parental cells and responses to ATRA or are capable of generating functional proteins in recipient cells. In some cases, these changes, indeed, parallel the levels of the corresponding bioactive mediators. One of the most striking functional effects of this kind was the ability of ATRA to diminish the ability of leukemic EVs to trigger procoagulant conversion of endothelial cells through the transfer of TF.

Notably, a somewhat different pattern of vesiculation and intercellular TF trafficking was observed in the case of the ATO treatment. In this case, greater cytotoxicity of this compound at higher doses resulted in undiminished coagulant phenotype of NB4 cells and of endothelial recipients receiving NB4-derived EVs (Fig. S6). This observation exemplifies the diversity and drug specificity of EV responses to PML–RARa inhibitors, a hitherto unknown detail which could be diagnostically and biologically informative. It is possible that the uptake and not the mere presence of TF-containing EVs in the circulation is among mechanisms driving coagulopathy in APL and other malignancies, as we suggested earlier [40, 44, 52]. Some of these effects may also depend on EV-unrelated mechanisms of membrane exchange between cells (e.g., trogocytosis) [54], processes that still remain to be investigated in the context of vascular regulation.

We reasoned that the absence of PML–RARa oncoprotein in EVs either may represent a common feature associated with transformation mediated by fusion gene products with nuclear location, or be unique to this particular gene and/or to APL cells. Indeed, recent studies revealed the presence of DNA, RNA and protein corresponding to other fusion oncogenes in the cargo of EVs. For example, the breakpoint cluster region–Abelson leukemia gene human homolog 1 (BCR–ABL1) was found in EVs isolated from K562 cells derived from blast crisis of human chronic myelogenous leukemia CML [32, 55]. BCR–ABL1-positive EVs were implicated in the induction of genetic instability and leukemia-like malignant growth in mice exposed to this material [55]. This finding is reminiscent of the phenomenon known as donor cell leukemia, a rare hematopoietic malignancy occurring in recipients of therapeutic bone marrow transplants [56]. These observations may suggest that, unlike in our study, the BCR–ABL1 oncogene may possess a direct horizontal transforming ability due to its EV-mediated intercellular transfer.

While ATRA has revolutionized the treatment of APL, the effects of monotherapy are not curative. Interestingly, the increased efficacies of agents capable of inducing long-term APL remission, such as ATO, can be recapitulated in the NB4 model in which ATO (but not ATRA) markedly decreases cell survival. This effect is paralleled by a distinctive pattern of cellular vesiculation, which is dominated by the upsurge in release of larger particles (Fig. S6). It is of interest to determine the biological basis of this differential EV release profile and whether measurements of the EV output and cargo could be informative as to responses or resistance of APL cells and their stem cell subsets to different targeted therapeutics [3].

We suggest that EV-mediated cell–cell communication may be a part of a larger network. EVs are critical in the intercellular transfer of MET receptor to myeloid cells during dissemination of melanoma [39]. AXL receptor is known to traffic as cargo of EVs in chronic B cell leukemia (CLL) and influence bone marrow stroma [30]. Exosomal trafficking of WNT regulates stem cell equilibrium in the context of diffuse large B-cell lymphoma (DLBCL) [28]. In a similar manner, perturbations in the EV content in the context of APL could impact various tissue responses and side effects of APL-directed agents, such as ATRA, ATO or chemotherapy. Whether such effects contribute to angiogenesis, inflammation or DS syndrome associated with ATRA therapy is presently unclear. DS mediators were postulated to be involved in chemotactic transmigration of ATRA-treated APL cells toward alveolar epithelial cells [57], and it is possible that the EV uptake by epithelial, inflammatory or stromal cells may play an important role in these events.

In conclusion, our study documents a novel link between oncogenic PML–RARa and vesiculation of APL cells. These observations are not without limitations and await extension through the use of multiple APL models and patient material. Future studies will explore whether EVs contribute to the intercellular regulatory network that likely exists between PML–RARa-transformed leukemic cells and the various facets of the vascular system, including endothelial and inflammatory cells, stem cell niches, as well as components of the coagulation system. The effects of ATRA, ATO and other therapeutic agents on these interactions require further study, while EVs may present themselves as a conceivable source of predictive or pharmacodynamic biomarkers in this context [31].