Particulate mediators of the bystander effect linked to suicide and interferon-β transgene expression in melanoma cells

Lucrecia Agnetti1 ● Chiara Fondello1 ● María Florencia Arbe1 ● Gerardo C. Glikin1 ● Liliana M. E. Finocchiaro1


In the context of comparative oncology, melanoma cells derived from companion animal tumors are good models for optimizing and predicting their in vivo response to therapeutic strategies. Here, we report that human, canine, and feline melanoma cells driven to death by bleomycin, interferon-β gene, or herpes simplex virus thymidine kinase/ganciclovir suicide gene (SG) treatment significantly increased their internal granularity. This fact correlated with the release of a heterogeneous collection of nano- and micro-sized granules as revealed by transmission electron microscopy. While killing lipofected cells, the expressed transgenes and their derived products were incorporated into these granules that were isolated by differential centrifugation. These particulate factors (PFs) were able to transfer, in a dose- and time-dependent manner, appreciable levels of therapeutic genes, related proteins, and drugs. Thus, when recipient cells of SG-carrying PFs were exposed to ganciclovir, this prodrug was efficiently activated, eliminating them. These PFs kept the functionality of their cargo, even after being subjected to adverse conditions, such as the presence of DNase, freezing, or heating. Since our in vitro system did not include any of the immune mechanisms that could provide additional antitumor activity, the chemo- gene treatments amplified by these delivery bags of therapeutic agents offer a great clinical potential.


Progress in comparative oncology promises advances in clinical cancer treatments for both humans and companion animals [1–3]. Canine and feline malignant melanomas behave clinically similar to human melanoma. These dis- eases share similar metastatic phenotypes and site selec- tivity [1–3]. Unless diagnosed at an early stage and surgically resected, human, feline, and canine melanoma evolve rapidly as a metastatic disease, are highly resistant to therapy, and associated with poor prognosis [1–3]. Melanoma has emerged as a paradigm of a particularly aggressive and plastic cancer, capable of co-opting the tumor stroma to adapt to the hostile microenvironment, by suppressing immunosurveillance, and widely disseminating [2, 4]. Many of these mechanisms could be mediated by extracellular vesicles (EVs) [4–7]. EVs, in higher abundance in the blood of cancer patients, spread out oncogenic infor- mation that promotes malignant growth and progression, and the establishment of (pre)metastatic niches [4–7]. EVs also confer drug resistance and neutralize target antibody-based drugs [8–12]. Thus, melanoma control is frequently short lived even when some drugs are proven to be effective [1–12]. In our veterinary clinical trials that combined local and systemic gene therapy, intratumor treatment with herpes simplex virus thymidine kinase/ganciclovir suicide gene (SG) and canine interferon-β gene (cIFNβ) reduced spontaneous melanoma tumors in vivo [13–15]. This response was reflected by their derived cell lines [16, 17]. In addition, bleomycin (BLM), that enhanced the cyto- toxic effects of both SG and IFNβ genes in human, canine, and feline melanoma cell lines [17, 18], also improved the ability of both genes to control local tumor and increased the survival of canine melanoma patients [19]. The great bystander effect induced by these therapeutic genes would be among the main causes of the successful outcomes [18, 20]. Part of this effect would be mediated by EVs [21– 25]. During in vivo and in vitro chemo-gene treatments, all the components secreted by treated cells to the extracellular space (including EVs and apoptotic vesicles) could spread pro- or antitumor information to the surrounding cells [4–12, 21–25]. In this context, we evaluated the contribution to the bystander effect of all particulate factors (PFs) released by BLM, IFNβ gene, and SG-treated cells to the extracellular space.

Materials and methods

Cell cultures

This research work followed the tenets of the Declaration of Helsinki, and all samples were obtained after informed consent from the patients. The clinical sampling was approved by the institutional review board of the Instituto de Oncología “Ángel H. Roffo”, Universidad de Buenos Aires, Argentina. Primary cell lines derived from surgically excised lymph nodes (hM1) and spleen metastasis (hM4) of human melanomas, oral (Dc) and abdominal (Rn) feline melanoma, and oral (Bsk, Br, Ol, and Rk), ocular (Ak), and liver metastasis (Lo) of canine melanomas were obtained by the mechanical disruption of tumor fragments in culture medium [16–18]. These cell lines, periodically tested for mycoplasma absence, were cultured as described [16–18].


Plasmids carrying the Escherichia coli β-galactosidase gene (psCMVβgal) [16], herpes simplex virus thymidine kinase gene (psCMVtk) [16], human IFNβ gene (psCMVhIFNβ) [18], feline IFNβ gene (psCMVfIFNβ) [17, 24], or canine IFNβ gene (psCMVcIFNβ) [17, 18, 20] in the polylinker site of psCMV (3.3 kb), downstream of the CMV promoter and upstream of poly A sequences, were amplified, purified, and resuspended to a final concentration of 2.0 mg/ml in sterile PBS as described [16–18].

Liposome preparation and in vitro lipofection

DC-Chol (3β[N-(N′,N′-dimethylaminoethane)]-carbamoyl cholesterol) and DMRIE (1,2-dimyristyl oxypropyl-3- dimethyl-hydroxyethylammonium bromide) were synthe- sized by the Instituto National de Tecnología Industrial (INTI, Buenos Aires, Argentina). DOPE (1,2-dioleoyl-sn- glycero-3-phosphatidyl ethanolamine) was purchased from Sigma. Liposomes were prepared at lipid/co-lipid molar ratios of 3:2 (DC-Chol:DOPE) or 1:1 (DMRIE:DOPE) by sonication as described [16–18]. Before lipofection, lipo- somes and plasmid DNA (1:2, v–v) were mixed and allowed to complex at room temperature for 10 min. Optimal lipid mixtures were determined for every cell line [16–18]. In most experiments, cells were seeded into six- well plates at a density of 3–5× 104 cells/cm2 and allowed to adhere overnight. Cells were exposed to lipoplexes (0.5μg plasmid DNA/cm2 and 1.0 μl liposome/cm2) during 4 h in a serum-free medium. Since IFNβ is species specific, human, feline, and canine cells were lipofected with their respective plamids: psCMVhIFNβ, psCMVfIFNβ, and psCMVcIFNβ.

β-galactosidase staining assay

To ensure that they were comparable in different experi- ments, lipofection efficiencies were checked 24 h after lipofection by β-galactosidase staining with 5-bromo-4- chloro-3-indolyl β-D-galactopyranoside (X-GAL, Sigma) and further counting with an inverted phase-contrast microscope [16–18].

SYBR® Green staining of plasmid DNA

Plasmid DNA was incubated for 5 min with a 9× concentra- tion of SYBR® Green I Nucleic Acid Gel Stain (Invitrogen) before the DNA/cationic lipid complex formation, and then lipofection was carried out as described above.

Isolation of pellets from IFNβ and HSVtk gene- lipofected cells

Human, canine, and feline melanoma cells were transiently lipofected with psCMVβgal, psCMVtk, and the species cor- responding to psCMVIFNβ plasmids in the presence or absence of 3 μg/ml bleomycin (BLM, Gador, Buenos Aires, Argentina). Four hours post lipofection, media with the remaining BLM and lipoplexes were removed. After three washes of the lipofected cells with complete medium, wells were replenished with fresh media without or with 5 μg/ml ganciclovir (GCV, Richet, Buenos Aires, Argentina). These media were removed 0, 24, 48, or 72 h later. Then, all these donor cells conditioned media (DCCM) as well as lipofection and wash media were transferred into new wells and allowed to adhere overnight to prevent the presence of the remaining viable cells in these media. PFs were isolated from all these media using a four-step differential centrifugation process at 4 °C. First, they were centrifuged in a microcentrifuge (Thermo Scientific Sorvall Legend Micro 17 R) at 500 × g (0.5 K) for 20 min to isolate the large PFs. Then, supernatants were centrifuged at 2000 × g (2 K) for 30 min and at 12,000 × g (12 K) for 60 min to pellet medium PFs. Small PFs were sedimented by a subsequent ultracentrifugation at 100,000 × g (100 K) for 70 min using Beckman Quick seal tubes and a 90Ti rotor (Beckman Coulter, Fullerton, CA). The resultant pellets were resuspended in PBS or complete medium for further assays. For temperature stability studies, all fractions were placed in a freezer (−20 °C) for 1 or more days, or in a water bath at 53 °C for 30 min or at 96 °C for 5 min.

Sensitivity to bleomycin, suicide gene, and IFNβ gene assays

Cells were seeded onto 96-well plates at 1–5× 104 cells/ well 24 h after lipofection or 24 h before exposure to PFs from DCCM. Unless otherwise indicated, all PF fractions were used at a ratio of ten donor cells per one recipient cell (10×). Four days after gene lipofection or PF addition in the absence or presence of BLM and/or GCV, cell survival was quantified using the acid phosphatase (APH) assay [17, 18]. Data were normalized as a percentage of the value of the corresponding untreated cells.

Flow cytometry cell cycle analysis of internal granularity and size of granules released by melanoma cells

Granularity and size of cells and PFs were analyzed by physical parameters on a flow cytometer (PASIII, Partec GmbH, Münster, Germany) with analysis of data performed using the WinList 3D 9.0.1 (Verity Software House). For- ward scatter intensity mainly correlates with cell area or size, and side scatter is a measure of the cell refractive index that depends on the cell internal granularity.

Double staining with acridine orange/ethidium bromide (AO/EB)

After 48 h of chemo-gene treatments, nonfixed cells grown over glass coverslips were incubated with AO (10 µg/ml, green) and EB (10 µg/ml, red) for 1 min and evaluated by fluorescent microscopy Nikon Eclipse TE 2000-s.

Transmission electron microscopy

PFs were isolated from the conditioned media by control and HSVtk ±GCV lipofected Bsk cells, using a four-step (0.5 K/2 K/12 K/100 K) differential centrifugation process at 4 °C. Then, 10 µl of each PF fraction in PBS were loaded onto copper grids, fixed for 5 min with freshly prepared 1% glutaraldehyde, washed twice with bidistilled water, and contrasted with 1% uranyl acetate. PFs were examined with a Zeiss EM109T transmission electron microscopy (TEM) at 70 kV (Carl Zeiss Microscopy GmbH, Jena, Germany).

Preparation of Hirt supernatants from particulate factors, transfected, and control cells

Cells were lipofected with pCMVtk and cultured for 48 h. Before harvest, cells were rinsed three times with phosphate-buffered saline to remove excess of ghosts. PFs were isolated from the DCCM using a four-step (0.5 K/2 K/12 K/100 K) differential centrifugation process at 4 °C. After washing with PBS, cells and all PF fractions were lysed by 20-min incubation in lysis buffer (0.6% SDS, 10 mM EDTA, pH 7.5) and overnight incubated with 5 M NaCl at 4 °C. After 12,000 × g centrifugation, DNA was extracted from the supernatants as described by Hirt [26]. After protein extraction with phenol–chloroform–isoamyl alcohol mixture, DNA was overnight precipitated with absolute ethanol at −20 °C. Then, DNA pellets were washed with 70% ethanol, dried, resuspended in TE buffer (10 mM Tris–1 mM EDTA, pH 8), electro- phoresed and undigested on a 0.8% agarose gel, and stained with EB. Transfection grade pCMVtk was used as standard.

DNase pretreatment

PFs were resuspended in 50 µl of 10 mM Tris-HCl/ 2.5 mM MgCl2/0.5 mM CaCl2 reaction buffer, containing 3 µl of DNase I (2 mg/ml). This DNase I dose was able to degrade 4.5 µg of free plasmid, the amount of plasmid that was used for PF donor cell lipofection. After 30 min of incubation at 37 °C, EDTA was added to a final concentration of 10 mM, and the samples were recen- trifuged at 12 K × g, and PFs resuspended in complete media were added to recipient cells with or without 5 µg/ml GCV.

Western blot analysis

Cells and PFs were lysed [in 50 mM Tris-HCl, pH 8, 100 mM NaCl, 1% Triton X-100, 10 mM EDTA, and 0.1% sodium azide with a mixture of protease inhibitors at a concentration of 4 × 106 cells in 100 μl of buffer] for 20 min on ice, centrifuged at 18,500 × g for 10 min, and the supernatant kept for further analysis. Then 20 μg of both cell lysates and PFs (measured using Bradford reagent) were loaded for each sample in 10% SDS-polyacrylamide gels. Separation was performed under nonreducing con- ditions at 100 V. Proteins were transferred to a PVDF membrane (pre-activated with methanol) for 2 h at con- stant current of 0.3 A. After blocking for 1 h at room temperature in TBS-Tween 0.1 + 5% milk, membranes were incubated with a rabbit specific anti-β-gal antibody kindly provided by Dr Rodolfo G. Goya (Universidad de La Plata, Buenos Aires, Argentina) and then with HRP anti-rabbit secondary antibody (1:10,000, Dako, P0448). After 5 min of incubation with ELC reagent, development was performed using ImageQuant LAS 500 chemiluminescence CCD camera (HE Healthcare Life Sciences).

Heparin pretreatment

Recipient cells were pretreated with 10 μg/ml heparin (sodium heparin 5000 UI/ml, Duncan) for 30 min before and during the 4-h incubation with 0.5, 2, and 12 K PFs released by control and HSVtk ±GCV-treated cells. Then, cells were washed with PBS, and wells were replenished with fresh medium without or with 5 µg/ml GCV. Cell viability was quantified 4 days after using the APH assay [17, 18]. Data were normalized as a percentage of the value of the corresponding control cells unexposed to PFs.

Statistical analysis

The results were expressed as mean ± standard error of the mean (s.e.m) (n: number of experiments corresponding to independent assays). Differences between groups were analyzed using unpaired Student’s t test (if two groups), one-way ANOVA followed by Tukey’s test (if more than two groups), or two-way ANOVA followed by Bonferroni test (if two nominal variables). Correlations were deter- mined by Pearson test with GraphPad Prism program (GraphPad Software Inc., USA).

Results and discussion

Bleomycin and gene transfer treatments increased the proportion of cells with high granularity

The increase in intracellular granularity has been described in very varied cellular processes such as apoptosis, necro- sis, senescence, and autophagy, making this phenotypic change a useful marker to identify death or arrest of cell growth [27].
Regardless of their origin, human (hM1 and hM4), feline (Dc and Rn), and canine (Ak, Br, Bsk, Lo, Ol, and Rk) melanoma cell lines significantly raised their internal granularity in response to bleomycin (BLM) and gene transfer treatments as analyzed by side scatter intensity flow cytometry (Fig. 1a, d and b, e). Two days post lipofection of the respective interferon-β genes (IFNβ), there was an increase in the intracellular granularity in two human and one feline melanoma cell lines (Dc). Herpes simplex virus thymidine kinase/ganciclovir (GCV), SG transfer, amplified even more the proportion of human and canine cells with higher granularity. BLM, when combined with gene lipo- fection, raised the internal granularity in all the tested cell lines. Interestingly, the combined BLM + genetic treat- ments led to greater granularity than any of the other treatments. These data were reflected in their respective dot plots of Fig. 1b, e, where a clear displacement of the event cloud (to the right upper quadrant) indicated an increase in cell size and granularity. Supplementary Fig. 1 provides enhanced-form detailed data of four selected cell lines. This granularity was also present in control nontreated or βgal- lipofected cells, albeit at lower intensity (Fig. 1a, b, d, e). Thus, during the process of death driven by BLM and gene transfer treatments, all assayed cell lines, regardless of their source, strongly enhanced their internal granularity as sug- gested by the high correlation between both parameters (Fig. 1c, f).

The increase in intracellular granularity correlated with the number of granules released by chemo- gene-treated melanoma cells

Associated with the increased intracellular granularity, there was a sharp rise in the number of granules released from BLM and gene-treated melanoma cells. The correlation between both processes suggests that this emerging pattern of intracellular granularity was mainly due to the generation of granules released from dying cells (Fig. 2a).
A strengthening of intracellular granularity generated by BLM and gene treatments can be visualized mostly in the perinuclear region and periphery of Bsk cells stained with AO/EB (Fig. 2b). Red and yellow AO/EB-stained granules in human and canine melanoma cells, near the border ready to be released, are clearly visible in the micrographs (Fig. 2c).
TEM analysis revealed a heterogeneous collection of endogenous nano- and micro-sized structures (Fig. 2d). As reported for EVs [23], pellets with components of different sizes ranging from 30 to 2200 nm in diameter were obtained by a four-step differential centrifugation. Regardless of the treatment, all these PFs exhibited a wider size distribution. Larger PFs of 1400–2200 nm precipitated at 0.5 K × g centrifugation. The diameter range of pellets from 2 K × g centrifugation was 500–1000 nm. Those particles isolated at 12 K × g measured about 300–400 nm. Finally, the smaller components obtained by ultracentrifugation at 100 K × g ranged 30–150 nm. PFs had a closed vesicular form of spheroidal morphology. Some of them, displaying a spreading pattern and a less regular structure, seemed to undergo membrane events that altered their appearance (Fig. 2d).

Plasmid DNA was incorporated into particulate factors released by lipofected cells

To test if plasmid DNA was incorporated into the PFs released by lipofected cells, HSVtk-carrying plasmids were stained with SYBR® Green before the DNA/cationic lipid complex formation. Then, canine melanoma Bsk cells were lipofected with these stained plasmids. The same amount of SYBR® Green was added to unlipofected control cells.
After 4 h of incubation, control and the remaining lipofec- tion media were removed. For maximal lipoplex depletion, three consecutive washes of the cells were carried out with complete medium. Then, wells were refilled with fresh medium without or with GCV that was removed 0, 24, 48, or 72 h later.
PFs were isolated from all these cells conditioned media by differential centrifugation. Next, the number and size distribution of PF-carrying SYBR® Green HSVtk gene were analyzed by flow cytometry. SYBR® Green stained material vs forward–light-scat- tered analysis of Fig. 3a revealed a population of small fluorescent events (red colored in the upper-left quadrant of the dot plots) that precipitated in the three centrifuga- tion fractions (0.5, 2, and 12 K). This green fluorescent population was very abundant in the remaining lipoplex fractions and decreased with the washes, suggesting that it was mostly composed of lipoplexes noninternalized by cells (Fig. 3a, b).
Noteworthily, 24 h post lipofection, a novel repertoire of fluorescent events was released by SYBR® Green HSVtk- lipofected Bsk cells. As expected, the proportion of green As shown in Fig. 3d, all fractions from the lipofection remnant, containing the unincorporated HSVtk lipoplexes, produced a significant cytotoxic effect on recipient cells in the presence of GCV. After three successive washes, pellets no longer produced deleterious effects on GCV-treated recipient cells, suggesting that almost all lipoplexes were removed. The same happened with the PFs collected from the DCCM added and removed at time 0, immediately after washes. However, 24 h post lipofection, the cytotoxic effect of the PFs released by HSVtk-lipofected cells reappeared when they were added to recipient cells in the presence of GCV. As time post lipofection elapsed (48 h), the number and cytotoxicity of PF-carrying SYBR® Green HSVtk increased (Fig. 3d), suggesting a continuous and long-term delivery of this transgene and/or its related products to untreated tumor cells. The toxicity of HSVtk-carrying PFs on GCV-treated recipient cells increased with their number and size (Fig. 3b, d). It is worth to note the inverse correlation between the cell survival of the GCV-treated receptor cells and the amount of fluorescent HSVtk-carrying PFs (Fig. 3c).

Particulate factors were able to carry plasmid DNAs and their translated proteins

Figure 4a confirmed that, like HSVtk-lipofected cells, every centrifugation fraction (0.5, 2, 12, and 100 K) of PFs carried detectable amounts of episomal HSVtk DNA normally absent in mammalian cells. Conversely, the DNA extracted from control cells and their PF derivatives did not contain the bands corresponding to the psCMVtk plasmid (Fig. 4a). With regard to their cytotoxicity (Fig. 3d), the amount of HSVtk of PFs diminished with their size (Fig. 4a).
As shown in Fig. 4a, right panel, DNase I was able to completely degrade all the free HSVtk plasmids and the corresponding lipoplexes used for lipofection of PF donor cells. However, DNase I digestion of PFs released by HSVtk-lipofected cells did not modify their cytotoxic effect on GCV-treated recipient cells (compare + and – DNase- treated PFs of Fig. 4a, b). Therefore, these PFs were able to transfer their genetic cargo, in a structure protecting from DNAse. Interestingly, all the assayed fractions of PFs (0.5, 2, and 12 K) kept the functionality of their genetic and proteinic cargo, even after being subjected to adverse con- ditions, such as freezing (some days at –20 °C) or heating (30 min at 56 °C or 5 min at 96 °C) (Fig. 4c).
As expected, the protein translated from plasmid DNA was also incorporated into PFs (Fig. 4d–f). Western blot analysis of β-gal-lipofected cells and their respective PF lysates proved the presence of β-gal protein in all the fractions of PFs (Fig. 4d). The deep-blue β-gal-stained PFs of a wide range of sizes strongly support these data (Fig. 4e). It is noteworthy that the bluish color seems to be spreading from stained PFs to cells contacting them, sug- gesting PF transfer of this protein (Fig. 4e). The β-gal rise in recipient cells as increasing their contact time with the β-gal-carrying PFs from 4 to 48 h reinforced this hypothesis (Fig. 4f). Probably, there was also an increasing production of the reporter protein over time. It is also possible that PFs also carried mRNA derived from plasmid DNA. This hypothesis was not evaluated in this work because of the difficulties for completely eliminating the high background generated by the remaining plasmid DNA after DNase I digestion. For the sake of simplicity, hereafter genes and related proteins will be generically called as the transgenes: cIFNβ or HSVtk.

Particulate factors contributed to the bystander effect induced by interferon-β and suicide genes

In the case of local cancer gene therapies, it is important that the therapeutic genes kill both lipofected and neighboring or distant nontreated cells. Previous reports demonstrated the release of various cytotoxic factors as mediators of the bystander effect produced by both genes [20]. Here we found that most of the PFs could transfer appreciable levels of therapeutic genes and their related proteins to the surrounding cells (Figs. 3 and 4). Therefore, we evaluated the contribution of all the PFs secreted by treated cells into the extracellular space, in the dissemination of the cytotoxicity of our chemo- gene treatments from affected to neighboring cells.
The SG system was very useful to test the PF ability to transfer the genetic treatment to the surrounding or distant cells. This system requires the presence of both the HSVtk gene-encoded enzyme and the prodrug (GCV) to generate its cytotoxic effect. This allowed combining the absence/ presence of each of them in the PF-releasing cells to elu- cidate if one or both components were transferred by PFs to recipient cells.
As shown in Fig. 5a, Bsk canine cells were very sensitive to both cIFNβ and HSVtk/GCV gene transfer. The Figure 5b shows the cytotoxic effect of the PFs collected from DCCM 2 days after BLM-gene treatments. All the PF fractions (0.5, 2, 12, and 100 K) were able to transfer these transgenes to untreated tumor cells, achieving significant expression of functional protein in these recipient cells. Thus, when recipient cells of HSVtk-carrying PFs were exposed to GCV, the prodrug could be efficiently phos- phorylated to its active form, eliminating these cells. This capacity decreased with the size of the PFs. Therefore, smaller HSVtk-carrying PFs isolated by ultracentrifugation (100 K) could not generate a significant cytotoxic effect on GCV-treated recipient cells (except those from GCV + BLM-treated donor cells in GCV-treated recipient cells).
The cytotoxicity induced in recipient cells by PFs iso- lated from HSVtk-lipofected donor cells in the presence of GCV was lower than that induced by HSVtk plus GCV direct treatment of donor cells, but significantly higher than that of HSVtk without GCV and control cells. These results suggest that while killing donor cells, appreciable HSVtk and GCV were incorporated into the PFs to efficiently eliminate the recipient cells. Interestingly, the addition of GCV to the recipient cells of PF-carrying HSVtk + GCV, enhanced even more their cytotoxicity. Undoubtedly, enough levels of functional HSVtk were transferred by PFs to recipient cells to convert the added prodrug in the active drug that further enhanced the PF cytotoxicity on recipient cells.
The cIFNβ gene-induced cytotoxicity on donor cells (about 50%) was reflected in their respective PF recipient cells. However, the PFs released by cIFNβ or cIFNβ + BLM-treated donor cells produced equal recipient cell response.
On the other hand, all the PF fractions released during the BLM treatment (4 h) of unlipofected (Fig. 5c), or HSVtk-lipofected (Fig. 5d) donor cells, produced deleter- ious effects on recipient cells. These data suggest that BLM was incorporated into the PFs released by donor cells or co- precipitated with the serum PFs or unincorporated lipoplexes [28]. BLM cytotoxicity disappeared during the three successive washes, and did not reappear with the PFs released by these cells 48 h post treatment (Fig. 5b–d). BLM, as other chemotherapeutic agents [9, 10], stimulated donor cells to produce and release PFs (Figs. 1 and 2). However, they only contributed to an early destruction of treated and surrounding melanoma cells, especially when combined with lipoplexes (Fig. 5a, c, d). The PFs released 2 days after the treatment of donor cells with BLM, in the absence or presence of lipoplexes, no longer produced harmful effects on recipient cells (Fig. 5b–d). Possibly this is one of the main causes of contradictory bystander effects observed post treatment with BLM [29, 30]. Conversely, the PFs released 48 h post IFNβ- and HSVtk lipofection, produced significant cytotoxic effects on their respective untreated and GCV-treated recipient cells (Fig. 5b, d). These data suggest that prolonged transgene expression, during the process of lipofected cell death, allowed a long- term PF-mediated delivery of these transgenes and their related products to untreated tumor cells.

The cytotoxicity of HSVtk-carrying particulate factors was dose-dependent

As a preliminary characterization of the PFs, we determined the range of HSVtk-carrying PFs needed to attain significant cell death when treating naive recipient cells. In Fig. 6a, Bsk recipient cells were treated with different amounts of PFs released by untreated or HSVtk-lipofected Bsk cells in the absence or presence of GCV. The PF concentration range extended from 50× (PFs from 50 donor cells over 1 receptor cell), up to 0.1× (PFs from one donor cell over ten recipient cells). Regardless of their number or size, all PFs from HSVtk-lipofected cells had no effect on recipient cell survival in the absence of GCV, but produced significant cell death after prodrug administration.
The upper panel of Fig. 6a shows the cytotoxic effect of the most powerful PFs, those obtained by 0.5-K cen- trifugation. These PFs released by HSVtk-lipofected cells in the presence or absence of GCV, generated a strong cyto- toxic effect on GCV-treated recipient cells. All the assayed PF concentrations showed cytotoxicities superior or near to 50%. Interestingly, there were no significant differences in the effect produced by the different ratios of the wide range studied. These buffering effects suggest a saturation of the uptake or later processes. On the other hand, except the two largest dilutions (1× and 0.1×), these powerful 0.5-K PFs released by HSVtk plus GCV-treated donor cells, exhibited significant cytotoxicity on receptor cells also in the absence of additional GCV added to the recipient cells.
In the middle panel of Fig. 6a, the buffer effect of the 2-K PFs was lower than that of 0.5-K ones, and the consequences of their dilution were more evident. The cytotoxicity produced by the PF-HSVtk ±GCV did not show differences between the 50×, 25×, and 10× ratios when the prodrug was present in recipient cells. The 1× PFs generated a lower but significant cytotoxic effect. Only the highest dilution (0.1×) lost the ability to generate harmful effects on the recipient cells. On the other hand, when the prodrug was absent in the receptor plate, none of the PF concentrations released by HSVtk ±GCV donor cells was cytotoxic.
In the lower panel of Fig. 6a, only the 50×, 25×, and 10× concentrations of 12-K PFs released by HSVtk-lipofected cells in the presence or absence of GCV, were enough to kill the recipient cells in the presence of the prodrug.

The internalization of particulate factors occurred speedily

Next, we carried out an analysis of the kinetics of PF effects on recipient cells. For this purpose, untreated Bsk cells were incubated with the PFs released by HSVtk-lipofected donor cells, in the absence or presence of GCV. At indicated time points, the PFs containing media were removed from reci- pient cells. After three washes, fresh media without or with GCV were added to recipient cells whose survival was quantified 4 days later.
Time-course experiments (Fig. 6b) indicated a clear impact of size and contact time on the PF-HSVtk cyto- toxicity on GCV-treated recipient cells. The number of new vesicles entering the cell would increase over time. GCV- treated recipient cell survival decreased linearly as the contact time increased, suggesting a continuous PF uptake and expression of HSVtk by recipient cells over at least 4 h. The internalization and release of the PF content into the recipient cell cytoplasm would occur rapidly. Only 15-min contact of bigger (0.5 K) PF-carrying HSVtk was enough to kill significant Bsk recipient cells after treatment with GCV. Double contact time (30 min) was necessary for the HSVtk- carrying 2 K PFs to produce significant effects on recipient GCV-treated cells (Fig. 6b). Despite this increase in the minimum incubation time, both fractions (0.5 and 2 K) were extremely efficient, triggering the process that eventually led recipient cells to death. In the case of 12 K-isolated PFs, it was necessary to increase the concentration to 20× (instead of 10×) to visualize this effect. Only after 2 h of 12 K PF-HSVtk contact with the cells, there was a sig- nificant cytotoxic effect.
When cells were treated with the HSVtk-loaded PFs during 4 days, the results highlighted the common observed satura- tion response for bystander effects, suggesting that a steady state between uptake and turnover had been reached (Fig. 6b). In all the cases, cell death of GCV-treated recipient cells decreased with smaller PFs. At tested concentrations, ultracentrifugation-isolated HSVtk-PFs could not generate a significant cytotoxic effect on GCV-exposed recipient cells (data not shown). Regardless of their size, Bsk cells exposed to PFs from HSVtk plus GCV-treated donor cells did not reduce the recipient cell viability, but resulted in significant cell death after additional prodrug administration (Fig. 6b).

Heparin reduced the cytotoxic effect of particulate factors

The interaction of EVs with their target cells is not only mediated by membrane–membrane contact, but often results in EV uptake with subsequent transfer of EV cargo. The recognition between EVs and the target cell has been reported to involve proteins present at the cell surface of both EVs and target cells [31, 32]. Heparan sulfate pro- teoglycans (HSPG) are cell surface receptors that are structurally related to heparin, and are important in a variety of biological processes. Since heparin has been shown to inhibit the uptake of EVs [31], here we analyzed if the addition of 10 μg/ml heparin 30 min before and during 4 h of incubation with the PFs, could inhibit the transfer of their cargo to recipient cells.
As shown in Fig. 6c, heparin generated a strong “pro- tective” effect on PF-HSVtk recipient cells, whose survival was significantly higher in the presence (dotted pattern) than in the absence (without pattern) of heparin. In contrast, this molecule did not modify the viability of control cells that did not receive toxic PFs. These results suggest that heparin inhibited PF-HSVtk uptake, possibly by blocking HSPG receptors in the recipient cells [31, 32].
At the used concentration, heparin partially blocked the 0.5- and 2-K PF cytotoxicity and totally that of the 12-K PFs, suggesting that other internalization mechanisms could be involved.

Suicide gene-loaded particulate factors from diverse melanoma cell lines displayed a remarkable cytotoxic effect on canine melanoma cells

The PF internalization was a saturable, time-dependent process, partially prevented by heparin (Fig. 6). However, it is still controversial whether vesicular uptake is cell type or species specific. Thus, we studied the cytotoxic effects of the PFs released by HSVtk-lipofected melanoma cells of other species on canine recipient cells. Unlike the IFNβ gene, the SG system is not species specific.
In the three canine cell lines, direct HSVtk lipofection in the presence of GCV was more cytotoxic than HSVtk-PF effects on GCV-treated recipient cells. On the other hand, the PFs released by both, Dc feline and hM1 human HSVtk- lipofected cells, when added to both canine recipient cells, were more (Bsk) or as active (Rk) as direct HSVtk lipo- fection of GCV-treated autologous cells (Fig. 7).
In a previous work [18], we demonstrated that, regardless of BRAF oncogene V600 mutational status (V600E in hM1 and hM2) or wild type (hM4), these human melanoma cell lines responded to both genes (IFNβ and HSVtk/GCV) and even more to their combinations with BLM. Here we found that, independently of their BRAF mutational condition, HSVtk/GCV direct cytotoxicity on hM1 and hM4 donor cells was reflected by their respective PF recipient cells in the presence of GCV (Supplementary Fig. 2).
The treatment of both canine cancer cell lines (Bsk and Rk), with HSVtk (±GCV)-loaded PFs, also resulted in sig- nificant cell death after prodrug administration. While PFs from HSVtk (±GCV)-lipofected Rk melanoma cells were highly cytotoxic for Bsk untreated cells, it was slightly but significantly cytotoxic for their own untreated recipient cells (Fig. 7).
Regardless of their source, these PFs, as cell-derived “lipoplexes,” were able to enter into melanoma cells to effi- ciently deliver genetic tools for therapeutic intervention. The extension of cytotoxic effects produced on treatment-naive recipient cells depended on PF size. Bigger (0.5 and 2 K) HSVtk (±GCV)-loaded PFs resulted in significant recipient canine cell death after prodrug administration (Fig. 7).
These data suggest that direct or PF-mediated cytotoxi- city of HSVtk gene lipofection in the presence of GCV was a dynamic process inherent of each individual tumor inde- pendent of the donor cell sensitivity to the transfected gene. Prolonged HSVtk expression, during the process of death of transfected canine, feline, and human melanoma cells, allowed the incorporation of enough levels of the transgene and derived products to change the PF information. Thus, whatever the species from which they come, PFs released by HSVtk-lipofected melanoma cells, were cytotoxic for the two GCV-treated recipient canine cells.
It would be interesting to know if lipofected normal melanocytes and melanocyte stem cells (MCSCs: potential ancestors of the melanoma tumor-initiating cells) [18, 33] produce the same kind of PFs than melanoma cells. In human skin, melanocytes are intricately regulated by kera- tinocytes and the surrounding stroma [34]. However, taking into account the small amount of these cells with respect to other cell types in the tumor stroma or surrounding tissues, their contribution to the bystander effect would not be crucial. In the human epidermal basal layer, the melanocyte- to-keratinocyte ratio is about 1:10 [35], and it is even lower in other tissues like mucosa [3].
Melanocytes are highly differentiated cells that comprise a very stable population whose proliferation is extremely low under normal circumstances [35]. The HSVtk/GCV SG system relies on cell proliferation at two levels: (i) lipo- fection is much more efficient in proliferating cells, and (ii) it needs DNA replication for GCV triphosphate incorpora- tion [36]. Thus, its effects are naturally targeted to highly proliferating cells such as tumor cells.
If this work is considered in a translational frame, when treating canine melanoma in a veterinary setting, we were using (i) HSVtk/GCV [14], (ii) both genes (HSVtk/GCV and IFNβ) [15], and (iii) both genes (HSVtk/GCV and IFNβ) plus bleomycin [19]. These treatments were used locally (peri- and/or intra-tumorally) in areas where mela- noma cells could remain after surgery. Although our plas- mids contain the therapeutic genes under a nontissue- specific promoter (i.e., CMV) that could be efficiently expressed in different kinds of cells, we did not detect any negative effects on the surrounding tissues or on the post- surgical healing ability.
There are in vitro data about the growth-inhibitory effects of IFNβ protein in human melanocytes [37], the cytotoxic effects of HSVtk/GCV in human lens epithelial cells [38], and the genotoxic effects of bleomycin in somatic cells of Drosophila melanogaster [39]. Nevertheless, in our veter- inary clinical trials, these effects were not noticed, while the antitumor effects clearly suppressed or delayed local relapse (in the case of complete surgery), or allowed disease control (in the case of partial surgery) [14, 15, 19]. On the other hand, engineered nontumor cells were proposed as a source of EVs carrying mRNA and/or proteins of SGs for cancer treatment, such as human mesenchymal stem cells [40] and embryonal kidney cells [41]. These findings suggest that during local SG therapy treatments, the EVs produced by the surrounding nontumor cells could provide additional antitumor activity.


Successful treatment against complex cancers depends on our full understanding of the intricate interactions between different components within tumors [4–6]. The large blood and lymphatic irrigation of the membranes, promoting early spread and metastasis of mucosal melanoma, poses unre- solved challenges and the urgent need for effective, safe, and accessible therapies [3]. It has been shown that tumor and stromal cells may talk to each other via extracellular vesicles (EVs) to establish a favorable tumor niche, and to promote tumor growth, invasiveness, and progression [4–12]. However, such EV dependency also reveals an intrinsic vulnerability of melanoma cells, which can be exploited therapeutically. The same particulate factors (PFs) that spread malignancy through the lympho-vascular supply of the mucous membranes [4–12] could be transformed into cell-derived vectors to efficiently transfer genetic tools for therapeutic intervention [21–25, 42–44].
During the process of death driven by bleomycin (BLM), suicide (SG), and interferon-β (IFNβ) genes, melanoma cells enhanced their internal granularity possibly due to the generation and release of a wide variety of granules, as suggested by the high correlation between both parameters (Figs. 1, 2). This heterogeneous collection of nano- and micro-sized structures, provided by the own treated cells, carrying these chemo-genetic agents, could become an interesting therapeutic strategy.
Specific protein markers of EVs in general (Hsp70 in all centrifugation pellets), microvesicles (α-actinin and CD9 in 2- and 12-K pellets), and exosomes (CD81 in 12- and 120-K pellets) (data not shown) strongly suggest that EVs were part of these granules. During in vivo and in vitro chemo-genetic treatments, all the components (including EV subpopulations and apoptotic vesicles) secreted by treated donor cells to the extracellular space, could transfer appreciable levels of ther- apeutic genes and their translated proteins to the surrounding cells (Figs. 3 and 4). Therefore, here we evaluated the con- tribution of all the PFs released by cells driven to death by BLM, IFNβ gene, and SG treatments, in the transfer of the cytotoxicity of these treatments from affected to surrounding cells. Thereby, the bystander effect could be transformed from being an adaptive element of neighboring cells to the tumor, into a sensitizing factor to our chemo-gene therapy.
An optimal treatment would reduce native melanoma- derived PFs while increasing the genetically modified ones. Consistent with this, melanoma cells, lipofected with canine IFNβ (cIFNβ) and HSVtk genes, expressed high levels of these therapeutic genes and derived proteins that not only killed these cells but were also incorporated into PFs (Figs. 3–5). These PFs (supplied by the same lipofected cells) delivered in a dose- and time-dependent manner appreciable levels of these functional genes, related proteins, and drugs (BLM and GCV) to efficiently eliminate untreated recipient cells (Fig. 5 and 6). Such capacity increased with the size of these pellets. The PFs-mediated delivery of the HSVtk gene and protein led to the elimination of recipient cells upon treatment with the GCV prodrug, which is converted within tumor cells into an anticancer agent (Fig. 5b).
The successful clinical outcome of our veterinary clinical trials in canine spontaneous melanoma patients [13–15] suggests that such therapeutic PFs release by SG and cIFNβ gene-treated tumor cells, would also occur in vivo. In line with this assumption, a significant in vivo growth inhibition was reported of both schwannoma and glioblastoma mice tumor models when EVs carrying cytosine deaminase fused to uracil phosphoribosyl transferase (CD::UPRT) were injected into tumors in combination with systemic delivery of the prodrug, 5-fluorocytosine [42, 43].
Here, the in vitro lipofection of our chemo-gene ther- apeutic agents applied only once produced a significant cytotoxicity. Therefore, repeated injections of HSVtk plus GCV and/or cIFNβ gene in the tumor bed could provide a continuous and sustained release of these PFs that carry cytotoxic activity to surrounding and distal cells. The pow- erful local antitumor response of our canine spontaneous melanoma patients is consistent with this premise [13, 14]. In addition, BLM, that enhances the cytotoxic effects of both SG and cIFNβ genes in human, canine, and feline melanoma cell lines [17, 18], also improved the ability of both genes to control local tumor increasing disease-free and overall sur- vival of canine melanoma patients [19].
Our in vivo and in vitro findings also suggest that the same prodrug-activating scheme, that empowers in vivo tumors to become their own worst enemies, also provides an immune system adjuvant strategy for preventing or treating metastatic lesions through the injection of acces- sible primary tumors. The high percentage of metastasis- free patients and the complete remission of a pulmonary metastasis in a grade IV patient (bearing metastasis when entering the study) by repeated intratumor injections of the SG system plus cytokine-secreting xenogenic cells, support this hypothesis [13]. Possibly, once released from chemo- gene-treated tumors, these “Trojan horses” endowed with target-homing specificity and the ability to surmount in vivo biological barriers could deliver suicide and cIFNβ genes and derived products to bystander and metastatic cells [44]. The great advantage of the PFs with respect to secreted signaling molecules is that they send signals over short and long distances without dilution or degradation of the bio- molecules they carry, since they are protected within the vesicular structure. As shown in Fig. 4b, c, the PFs were able to keep the functionality of their genetic and proteinic cargo, even after being subjected to adverse conditions, such as the presence of DNase, freezing, or heating. This could explain (i) the high proportion of tumor responses, (ii) the suppression or delay in metastasis appearance, (iii) the complete remission of noninjected distant metastasis, and (iv) the significant increase in patients’ metastasis-free and overall survival [13–15, 19].
Here, our in vitro system did not include any of the immune mechanisms that could provide additional antitumor activity, possibly partially mediated by PFs [45, 46]. There- fore, these chemo-gene treatments amplified by these delivery bags of therapeutic agents have a great clinical potential.


1. Prouteau A, André C. Canine melanomas as models for human melanomas: clinical, histological, and genetic comparison. Genes. 2019;10:501.
2. Chamel G, Abadie J, Albaric O, Labrut S, Ponce F, Ibisch C. Non- ocular melanomas in cats: a retrospective study of 30 cases. J Feline Med Surg. 2017;19:351–7.
3. Lerner BA, Stewart LA, Horowitz DP, Carvajal RD. Mucosal melanoma: new insights and therapeutic options for a unique and aggressive disease. Oncology. 2017;31:e23–32.
4. Nishida-Aoki N, Gujral TS. Emerging approaches to study cell- cell interactions in tumor microenvironment. Oncotarget. 2019;10:785–97.
5. Zhang HG, Grizzle WE. Exosomes: a novel pathway of local and distant intercellular communication that facilitates the growth and metastasis of neoplastic lesions. Am J Pathol. 2014;184:28–41.
6. Peinado H, Alečković M, Lavotshkin S, Matei I, Costa-Silva B, Moreno-Bueno G, et al. Melanoma exosomes educate bone mar- row progenitor cells toward a pro-metastatic phenotype through MET. Nat Med. 2012;18:883–91.
7. Demirsoy S, Martin S, Maes H, Agostinis P. Adapt, recycle, and move on: proteostasis and trafficking mechanisms in melanoma. Front Oncol. 2016;6:240.
8. Lv MM, Zhu XY, Chen WX, Zhong SL, Hu Q, Ma TF, et al. Exosomes mediate drug resistance transfer in MCF-7 breast can- cer cells and a probable mechanism is delivery of P-glycoprotein. Tumour Biol. 2014;35:10773–9.
9. Samuel P, Mulcahy LA, Furlong F, McCarthy HO, Brooks SA, Fabbri M, et al. Cisplatin induces the release of extracellular vesicles from ovarian cancer cells that can induce invasiveness and drug resistance in bystander cells. Philos Trans R Soc Lond B Biol Sci. 2018;373:20170065.
10. Emam SE, Ando H, Abu Lila AS, Kobayashi S, Shimizu T, Okuhira K, et al. Doxorubicin expands in vivo secretion of cir- culating exosome in mice. Biol Pharm Bull. 2018;41:1078–83.
11. Martinez VG, O’Neill S, Salimu J, Breslin S, Clayton A, Crown J. et al. Resistance to HER2-targeted anti-cancer drugs is associated with immune evasion in cancer cells and their derived extra- cellular vesicles. Oncoimmunology.2017;6:e1362530
12. Ciravolo V, Huber V, Ghedini GC, Venturelli E, Bianchi F, Campiglio M, et al. Potential role of HER2-overexpressing exo- somes in countering trastuzumab-based therapy. J Cell Physiol. 2012;227:658–67.
13. Finocchiaro LM, Fiszman GL, Karara AL, Glikin GC. Suicide gene and cytokines combined nonviral gene therapy for spon- taneous canine melanoma. Cancer Gene Ther. 2008;15:165–72.
14. Finocchiaro LM, Glikin GC. Cytokine-enhanced vaccine and suicide gene therapy as surgery adjuvant treatments for sponta- neous canine melanoma: 9 years of follow-up. Cancer Gene Ther. 2012;19:852–61.
15. Finocchiaro LM, Fondello C, Gil-Cardeza ML, Rossi Ú, Villa- verde MS, Riveros MD, et al. Cytokine-enhanced vaccine and interferon-β plus suicide gene therapy as surgery adjuvant treatments for spontaneous canine melanoma. Hum Gene Ther. 2015;26:367–76.
16. Gil-Cardeza ML, Villaverde MS, Fiszman GL, Altamirano NA, Cwirenbaum RA, Glikin GC, et al. Suicide gene therapy on spontaneous canine melanoma: correlations between in vivo tumors and their derived multicell spheroids in vitro. Gene Ther. 2010;17:26–36.
17. Agnetti L, Fondello C, Villaverde MS, Glikin GC, Finocchiaro LME. Therapeutic potential of bleomycin plus suicide or interferon-β gene transfer combination for spontaneous feline and canine melanoma. Oncoscience. 2017;4:199–214.
18. Fondello C, Agnetti L, Villaverde MS, Simian M, Glikin GC, Finocchiaro LME. The combination of bleomycin with suicide or interferon-β gene transfer is able to efficiently eliminate human melanoma tumor initiating cells. Biomed Pharmacother. 2016;83:290–301.
19. Finocchiaro LME, Agnetti L, Fondello C, Glikin GC. Combina- tion of cytokine-enhanced vaccine and chemo-gene therapy as surgery adjuvant treatments for spontaneous canine melanoma. Gene Ther. 2019;26:418–31.
20. Rossi Ú, Gil-Cardeza ML, Villaverde MS, Finocchiaro LM, Gli- kin GC. Interferon-β gene transfer induces a strong cytotoxic bystander effect on melanoma cells. Biomed Pharmacother. 2015;72:44–51.
21. Wu K, Xing F, Wu SY, Watabe K. Extracellular vesicles as emerging targets in cancer: recent development from bench to bedside. Biochim Biophys Acta Rev Cancer. 2017;1868:538–63.
22. Turturici G, Tinnirello R, Sconzo G, Geraci F. Extracellular membrane vesicles as a mechanism of cell-to-cell communication: advantages and disadvantages. Am J Physiol Cell Physiol. 2014;306:C621–33.
23. Minciacchi VR, Freeman MR, Di Vizio D. Extracellular vesicles in cancer: exosomes, microvesicles and the emerging role of large oncosomes. Semin Cell Dev Biol. 2015;40:41–51.
24. Maas SLN, Breakefield XO, Weaver AM. Extracellular vesicles: unique intercellular delivery vehicles . Trends Cell Biol. 2017;27:172–88.
25. Jella KK, Rani S, O’Driscoll L, McClean B, Byrne HJ, Lyng FM. Exosomes are involved in mediating radiation induced bystander signaling in human keratinocyte cells. Radiat Res. 2014; 181:138–45.
26. Hirt B. Selective extraction of polyoma DNA from infected mouse cell cultures. J Mol Biol. 1967;26:365–9.
27. Haynes MK, Strouse JJ, Waller A, Leitao A, Curpan RF, Bologa C, et al. Detection of intracellular granularity induction in prostate cancer cell lines by small molecules using the HyperCyt high- throughput flow cytometry system. J Biomol Screen. 2009;14:596–609.
28. Gil-Cardeza ML, Rossi Ú, Villaverde MS, Glikin GC, Finocchiaro LM. Cationic lipid:DNA complexes allow bleomycin uptake by melanoma cells. Biomed Pharmacother. 2013;67:269–75.
29. Chinnadurai M, Chidambaram S, Ganesan V, Baraneedharan U, Sundaram L, Paul SF, et al. Bleomycin, neocarzinostatin and ionising radiation-induced bystander effects in normal diploid human lung fibroblasts, bone marrow mesenchymal stem cells, lung adenocarcinoma cells and peripheral blood lymphocytes. Int J Radiat Biol. 2011;87:673–82.
30. Savu D, Petcu I, Temelie M, Mustaciosu C, Moisoi N. Com- partmental stress responses correlate with cell survival in bystander effects induced by the DNA damage agent, bleomycin. Mutat Res. 2015;771:13–20.
31. Atai NA, Balaj L, van Veen H, Breakefield XO, Jarzyna PA, Van Noorden CJ, et al. Heparin blocks transfer of extracellular vesicles between donor and recipient cells. J Neurooncol. 2013; 115:343–51.
32. Christianson HC, Svensson KJ, van Kuppevelt TH, Li JP, Belting M. Cancer cell exosomes depend on cell-surface heparan sulfate proteoglycans for their internalization and functional activity. Proc Natl Acad Sci USA. 2013;110:17380–5.
33. Moon H, Donahue LR, Choi E, Scumpia PO, Lowry WE, Grenier JK, et al. Mela-nocyte stem cell activation and translocation initiate cutaneous melanoma in response to UV exposure. Cell Stem Cell. 2017;21:665–78.e6.
34. Santiago-Walker A, Li L, Haass NK, Herlyn M. Melanocytes: from morphology to application. Skin Pharmacol Physiol. 2009;22:114–21.
35. Cichorek M, Wachulska M, Stasiewicz A, Tymińska A. Skin melanocytes: biology and development. Postepy Dermatol Alergol. 2013;30:30–41.
36. Karjoo Z, Chen X, Hatefi A. Progress and problems with the use of suicide genes for targeted cancer therapy. Adv Drug Deliv Rev. 2016;99:113–28.
37. Krasagakis K, Garbe C, Krüger S, Orfanos CE. Effects of inter- ferons on cultured human melanocytes in vitro: interferon-beta but not-alpha or –gamma inhibit proliferation and all interferons sig- nificantly modulate the cell phenotype. J Investig Dermatol. 1991;97:364–72.
38. Yang J, Liu TJ, Lu Y. Effects of bicistronic lentiviral vector- mediated herpes simplex virus thymidine kinase/ganciclovir sys- tem on human lens epithelial cells. Curr Eye Res. 2007;32:33–42.
39. Danesi CC, Dihl RR, Bellagamba BC, de Andrade HH, Cunha KS, Guimarães NN, et al. Genotoxicity testing of combined treatment with cisplatin, bleomycin, and 5-fluorouracil in somatic cells of Drosophila melanogaster. Mutat Res. 2012;747:228–33.
40. Altanerova U, Jakubechova J, Benejova K, Priscakova P, Pesta M, Pitule P, et al. Prodrug suicide gene therapy for cancer targeted intracellular by mesenchymal stem cell exosomes. Int J Cancer. 2019;144:897–908.
41. Erkan EP, Saydam N, Chen CC, Saydam O. Extracellular vesicles as carriers of suicide mRNA and/or protein in cancer therapy. Methods Mol Biol. 2019;1895:87–96.
42. Mizrak A, Bolukbasi MF, Ozdener GB, Brenner GJ, Madlener S, Erkan EP, et al. Genetically engineered microvesicles carrying suicide mRNA/protein inhibit schwannoma tumor growth. Mol Ther. 2013;21:101–8.
43. Erkan EP, Senfter D, Madlener S, Jungwirth G, Ströbel T, Saydam N, et al. Extracellular vesicle-mediated suicide mRNA/protein delivery inhibits glioblastoma tumor growth in vivo. Cancer Gene Ther. 2017;24:38–44.
44. Villagrasa A, Álvarez PJ, Osuna A, Garrido JM, Aránega A, Rodríguez-Serrano F. Exosomes derived from breast cancer cells, small Trojan horses? J Mammary Gland Biol Neoplasia. 2014;19:303–13.
45. Hsu DH, Paz P, Villaflor G, Rivas A, Mehta-Damani A, Angevin E, et al. Exosomes as a tumor vaccine: enhancing potency through direct loading of antigenic peptides. J Immunother. 2003; 26:440–50.
46. Zhang H, Huang B. Tumor cell-derived microparticles: a new form of cancer vaccine. Oncoimmunology. 2015;4:e1017704.