Fibroblast growth factor 2 (FGF-2) or basic fibroblast growth factor (bFGF) is a member of the FGF family which controls several physiological and pathological processes; among its multiple functions, FGF-2 is active in embryogenesis and morphogenesis, but it also plays a key role as an angiogenic factor involved in tumor progression and invasion.1, 2 FGF-2 is overexpressed in a wide range of human tumors, including cancer of the bladder,3 esophagus,4 pancreas,5 as well as in non-small-cell lung cancer6 and glioblastoma.7 At least 5 isoforms of different molecular weights arise through alternative translation of a single mRNA transcript: translation of the isoform of lowest molecular weight (18-kDa) is initiated at a canonical AUG codon, whereas translation of the 4 higher molecular weight FGF-2 isoforms (22, 22.5, 24, 34-kDa) initiates at 4 upstream CUG codons.8, 9, 10 The 18-kDa form lacks a nuclear localization signal and is primarily cytosolic, whereas the high molecular weight isoforms are predominantly located in the nucleus.11, 12, 13 Even though all FGF-2 isoforms lack a hydrophobic signal peptide sequence for translocation in the endoplasmic reticulum,8, 13 the 18-kDa isoform can be detected extracellularly and is efficiently secreted by FGF-2 producing cells. Since this release is unaffected by treatment with Brefeldin A, which blocks ER-to-Golgi trafficking and disrupts the Golgi apparatus,14 it appears to occur by an endoplasmic reticulum/Golgi-independent mechanism15, 16 and through a direct translocation from the cytosol into the extracellular space.17
It has been proposed that the mechanism of FGF-2 nonclassical secretion involves an ATP-dependent pathway and requires the activity of the membrane transporter Na+/K+-ATPase.18, 19 Ouabain, an inhibitor of Na+/K+-ATPase activity, partially inhibits FGF-2 release,18, 20 which can be restored by the expression of an ouabain-resistant Na+/K+-ATPase α-subunit mutant.19 Moreover, direct or indirect interaction between FGF-2 and the α-subunit of Na+/K+-ATPase has been detected in co-immunoprecipitation experiments.18
The Epstein-Barr Virus (EBV) is a human tumor virus that causes or is closely associated with several malignancies, including nasopharyngeal carcinoma and EBV lymphoproliferative syndromes, which are characterized by their invasive potential.21 A series of reports published since 1998 have revealed that EBV, and specifically its principal oncoprotein, the latent membrane protein 1 (LMP1), can induce a spectrum of invasiveness and metastasis factors including several that contribute to angiogenesis: MMP-9, COX-2, VEGF and HIF1α22, 23, 24 as well as FGF-2.25 In the case of FGF-2, LMP1 both induces expression of the factor and its release into the extracellular compartment. MUC-126 and other EBV genes also participate in tumor progression (for review see Ref.27). These are the first such findings reported for tumor viruses, and correlate at least in part with observations made in EBV-infected human tumor tissues.28, 29, 30, 31 Subsequently, related findings have been reported for KSHV.32
Exosomes are membrane vesicles of 50–110 nm wide, released by different cell types such as cytotoxic T-lymphocytes, EBV-transformed B-cells, dendritic cells, reticulocytes and numerous tumor cells (for recent reviews, see Refs.33–35). Tumor-derived exosomes contain tumor antigens and may be able to either suppress the immune system or stimulate tumor rejection by T-cell cross-priming.36, 37 Among the different functions thought to be exerted by them, tumor-derived exosomes are supposed to favor tumor growth and progression not only by inducing immune escape, but also by promoting angiogenesis.38 Exosomes correspond to the internal vesicles of multivesicular bodies (MVBs), a compartment of the late endocytic pathway originating from inward budding of the endosomal membrane; they are released outside the cell after the fusion of MVBs with the plasma membrane. The invagination process of the limiting membrane of the endosome seems to require specific sorting, which leads to the unique composition of exosomes. As a result of this process, the exosomal content in membrane proteins and lipids or in cytosolic components is specifically selected and enriched, and the same orientation as on the cell plasma membrane and cell interior is maintained.
It is well known that the endocytic pathway might be involved in an unconventional process of secretion. For example, the cytosolic leaderless protein interleukin-1 beta (IL-1β) is first translocated in specialized vesicles belonging to the endolysosomal compartment, and then released in the extracellular fluid upon fusion of these vesicles with the plasma membrane.39 Exosomes may represent an unconventional secretion pathway for some cytosolic proteins such as galectin-340 and annexin II,41 which do not bear a signal sequence and which have been characterized in extracellular environments. Importantly, LMP1 has been detected both in MVBs and in exosomes released by lymphoblastoid cell lines (LCLs) in conjunction with major histocompatibility complexes (MHCs) and these findings have been related to possible immunosuppressive effects of LMP1 on activated T-lymphocytes42, 43, 44; the mechanism whereby LMP1 and MHC complexes are released into extracellular fluids is unknown.
Here we show that secretion of the 18-kDa isoform of FGF-2 promoted by EBV LMP1 occurs through a hitherto undescribed pathway, namely, via exosomes. Utilization of this secretory pathway by FGF-2 is dependent on Na+/K+-ATPase activity.
Material and methods
PcDNA3-based LMP1 has been previously described.22 In brief, the LMP1 ORF was cut from pSV2 gptMTLM and subcloned downstream of the cytomegalovirus immediate-early promoter in pcDNA3 (Invitrogen, Carlsbad, CA). The FLAG-tagged 18-kDa FGF-2 expression plasmid, in which FLAG epitope tags are inserted at the COOH terminus of FGF-2 cDNA, was a generous gift of Dr. Robert Levenson (Penn State College of Medicine, Hershey, PA).19
Cell lines and transfections
The EBV-negative human nasopharyngeal cell line Ad-AH was kindly provided by Dr. Erik K. Flemington (Tulane University, New Orleans, LA). The Chinese hamster ovary CHO cell line was kindly provided by Dr. Armando Bartolazzi (Azienda Ospedaliera Sant'Andrea, Rome, Italy) and were maintained in Ham's F12 medium (Euroclone Life Sciences Division, Pero, Italy), supplemented with 7.5% fetal bovine serum (FBS) and antibiotics. The LMP1-positive LCL, LCL-AKE, kindly provided by Dr. Pankaj Trivedi (University of Roma “La Sapienza,” Rome, Italy), was maintained in RPMI 1640 (Euroclone) plus 10% FBS and antibiotics. The human breast cancer MDA-MB-231 cells were maintained in RPMI 1640 medium with 10% FBS, 4 μM L-glutamine and antibiotics. The EBV-infected MDA-MB-231 clones C1D12 and C3B424 were maintained in the same medium but with G418 at 700 μg/ml. The human umbilical vein endothelial cells (HUVECs) were grown in EBM-2 medium supplemented with EGM-2 Single Quots (Cambrex Bio Science Walkersville, Walkersville, MD). For chlorate-treatment, cells were grown for 48 hr in presence or absence of 30 mM sodium chlorate (Sigma, Saint Louis, MO) before fixation for immunofluorescence.
To establish either the vector-transfected Ad-AH cells or the Ad-AH cells stably expressing LMP1, 0.5 μg of pcDNA3 (Invitrogen) or LMP1 expression plasmid, respectively, were transfected into Ad-AH cells with Lipofectamine reagent (Life Technologies, Grand Island, NY) following the manufacturer's instructions. Stable cell lines were obtained by cultivating transfected Ad-AH cells in the presence of geneticin (800 μg/ml; Life Technologies).
For transient transfections, Ad-AH and CHO cells, grown in 100-mm dishes, were transfected with 1–2 μg of DNA using Effectene reagent (Qiagen, Valencia, CA), according to the manufacturer's protocol. Forty-eight hours after transfection, cells were held for 24 hr in DMEM without FBS and with or without ouabain 10, 30 or 50 nM (Calbiochem, EMD Biosciences, San Diego, CA). Cell viability was assessed by the trypan blue exclusion test. The difference in ratio of nonviable cells did not exceed >3% in trypan blue exclusion tests.
For RNA interference and TSG101 silencing, the EBV-infected MDA-MB-231 cells, clone C3B4, were transfected with 0.6–1 μg of small interfering RNA (siRNA) for TSG101 (Santa Cruz Biotechnology, Santa Cruz, CA) using siRNA transfection reagent, according to the manufacturer's protocol. Forty-eight hours after transfection, si-RNA transfected and untransfected cells were processed for western blot analysis.
Mouse anti-LMP1 monoclonal antibody (CS1–4) was purchased from DakoCytomation (Glostrup, Denmark); rabbit anti-FGF-2 polyclonal antibodies, mouse anti-HSC70 monoclonal antibody (B-6) and mouse anti-TSG101 monoclonal antibody were from Santa Cruz Biotechnology; mouse anti-CD63 monoclonal antibody was from BD Bioscience PharMingen (San Jose, CA); rabbit anti-Cathepsin D polyclonal antibodies (which recognize the active form of human cathepsin D) and mouse anti-Na+/K+-ATPase a-1 monoclonal antibody were purchased from Upstate Biotechnology (Lake Placid, NY); rabbit anti-FLAG polyclonal antibodies, mouse anti-FLAG M2 monoclonal antibody, mouse anti-γ-tubulin antibody and rabbit anti-α-actin were from Sigma.
Microinjection was performed using an Eppendorf microinjector (Eppendorf, Hamburg, Germany) and an inverted microscope (Zeiss, Oberkochen, Germany). Injection pressure was set at 30–80 hPa and the injection time at 0.3–0.5 sec. siRNA for TSG101 (Santa Cruz Biotechnology) (100 nM) in a mixture with FITC-conjugated secondary antibodies (goat anti-mouse IgG or goat anti-rabbit IgG, 1 mg/ml) or secondary antibodies alone were microinjected in the cytoplasm of the EBV-infected MDA-MB-231 clone C3B4 to induce RNA interference and consequent TSG101 silencing. Cells were left for 24 hr at 37°C, then fixed and processed for immunofluorescence.
5′-Bromo-deoxyuridine (BrdU) assay was performed on HUVEC cells, grown on coverslips and used in this set of experiments at no more than passage 4. After 24 hr, EBM-2 complete medium was removed and replaced with DMEM nutrient mix F12 1:1 (Euroclone), containing 0.4% FBS. The cells were then treated for 24 hr with 20 ng/ml human recombinant FGF-2 (Sigma), or 20 ng/ml FGF-2 plus 5 μg/ml anti-FGF-2 polyclonal blocking antibodies (Santa Cruz Biotechnology). Alternatively, cells were treated with 10 μg/ml of isolated exosomes or with 10 μg/ml of exosomes plus 5 μg/ml anti-FGF-2 antibodies. During the last 6 hr of the treatments, 100 μM BrdU (Sigma) was added to the medium to allow BrdU incorporation. Cells were then fixed in 4% paraformaldehyde in PBS for 30 min at 25°C, followed by treatment with 0.5% HCl/0.1% Triton X-100 for 1 hr at 37°C to allow permeabilization. After extensive washing in PBS, cells were buffered with 0.1 M Na2B4O7 and incubated with anti-BrdU monoclonal antibody (1:50 in PBS; Sigma) for 1 hr at 25°C, followed by goat anti-mouse IgG-FITC (1:50 in PBS; Cappel Research Products, Durham, NC). The amount of BrdU incorporation in the samples was evaluated as number of positively stained nuclei versus total number of nuclei in randomly taken 50 different areas. Results are expressed as percentage and represent the mean values ± standard deviations from 2 different experiments.
Immunofluorescence and confocal analysis
Cells, grown on coverslips and transfected as earlier, were fixed 48 hr after transfection in 4% paraformaldehyde in PBS for 30 min at 25°C and permeabilized with 0.1% Triton X-100 in PBS for 5 min. LMP1-transfected or LMP1/FLAG-tagged FGF-2 cotransfected Ad-AH cells were incubated with primary antibodies for 1 hr at room temperature, visualized with goat anti-rabbit IgG-TexasRed (1:100 in PBS; Jackson Immuno Research Laboratories, West Grove, PA) or goat anti-mouse IgG-FITC (1:50 in PBS; Cappel Research Products), after appropriate washing with PBS. Fluorescence signals were analyzed either by recording stained images using a cooled CCD color digital camera SPOT-2 (Diagnostic Instruments Incorporated, Sterling Heights, MI) and IAS 2000/H1 software (Delta Sistemi, Roma, Italy) or by confocal vertical (x–z) sections (interval: 0.5 μm) obtained using a Zeiss Confocal Laser Scan Microscope (Zeiss, Oberkochen, Germany). Quantitative analysis of the fluorescence intensity was performed on a minimum of 6 sequential confocal sections per cell by evaluating endosomal/MVB areas, positive for the CD63 marker, versus comparable cytosolic areas: 10 different cotransfected or singly transfected cells, randomly taken from 3 different experiments, were analyzed.
The media from cultured cells were collected, centrifuged at 10,000g for 15 min to remove cells and debris, then placed in Amicon Ultra-15 Centrifugal Filter Units (30 kDa) (Millipore, Billerica, MA) and spinned at 4,000 rpm for 20 min in a swinging bucket rotor. The concentrated media were removed, placed in micro ultracentrifuge tubes (11 mm × 34 mm PC; Beckman Coulter, Fullerton, CA) and spinned at 144,000g for 1 hr using a TLA100 rotor (Beckman Coulter). The supernatant was removed by aspiration.
Exosome-containing pellet obtained from the ultracentrifugation was fixed with 4% paraformaldehyde in PBS for 2 hr at room temperature and then resuspended in 2% paraformaldehyde in PBS. Droplets of this exosomal fraction were directly spotted onto Formvar/carbon-coated copper grids (100 mesh) and incubated with anti-LMP1 monoclonal antibody (diluted 1:10 in PBS-1% BSA). After several washes in PBS-0.1% BSA, the grids were incubated with 18 nm diameter colloidal gold particles (prepared by the citrate method) conjugated with protein A (Pharmacia, Uppsala, Sweden), diluted 1:10 in PBS-1% BSA. After washing in PBS, grids were treated with glutaraldehyde 1% for 5 min, then washed 8 times in distilled water and incubated first with rabbit anti-FLAG polyclonal antibodies (diluted 1:50 in PBS-1% BSA) and then with 10-nm-diameter colloidal gold particles conjugated with protein A (purchased from Dr. George Posthuma, Department of Cell Biology, University Medical Center Utrecht, Utrecht, The Netherlands), diluted 1:10 in PBS-1% BSA. Control experiments were performed by omission of the primary antibodies from the labeling procedure. Finally, grids were stained with a solution of 2% methyl cellulose and 0.4% uranyl acetate before electron microscopy examination.
LMP1/FLAG-tagged FGF-2 cotransfected Ad-AH cells were fixed with a mixture of 2% paraformaldehyde and 0.2% glutaraldehyde in 0.1 M phosphate buffer for 2 hr at room temperature, washed and embedded in 12% gelatin (Sigma) in 0.1 M phosphate buffer that was solidified on ice. Gelatin blocks were infused with 2.3 M sucrose overnight at 4°C, frozen in liquid nitrogen and then cryosectioned. Ultrathin cryosections were collected with sucrose and methyl cellulose and the double immunostaining was performed as earlier, using rabbit anti-FLAG polyclonal antibodies conjugated with protein A–10-nm-diameter colloidal gold particles and anti-LMP1 monoclonal antibody, anti-CD63 monoclonal antibody or anti-Na+/K+-ATPase a-1 monoclonal antibody (diluted 1:10 in PBS-1% BSA) conjugated with protein A–18-nm-diameter colloidal gold particles. Control experiments were performed by omission of the primary antibodies from the labeling procedure. Finally, ultrathin cryosections were stained with a solution of 2% methyl cellulose and 0.4% uranyl acetate before electron microscopy examination.
Western blot analysis
Cell lysates were extracted in 200 μl of lysis buffer [50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% NP40, 2 mM EDTA, 0.50% sodium deoxycholate, 0.10% SDS, 0.2 mM sodium orthovanadate, 0.1 M NaF, 5 μg/ml leupeptin, 10 μg/ml aprotinin and 1 mM phenylmethylsulfonyl fluoride], and protein concentration was determined by Bradford protein assay (Bio-Rad Laboratories, Hercules, CA). Protein (100 μg) was boiled in SDS sample buffer for 5 min, electrophoresed on 15% SDS polyacrylamide gels, and transferred onto a nitrocellulose membrane (GE Osmonics Labstore, Minnetonka, MN). Nonspecific reactivity was blocked by incubation for 30 min in Tris-buffered saline solution containing 0.1% Tween 20 and 5% nonfat dried milk. The membrane was incubated overnight at 4°C with the primary antibody, then washed with Tris-buffered saline solution containing 0.1% Tween 20 (TBST) and incubated with horseradish peroxidase-conjugated anti-mouse (for primary monoclonal) or anti-rabbit (for primary polyclonal) secondary antibody, at room temperature for 1 hr. After washing the membrane 3 times for 15 min with TBST, the peroxidase activity was detected by enhanced chemiluminescence (Pierce Biotechnology, Rockford, IL).
For western blot of exosomal fraction, the exosome pellet obtained from ultracentrifugation was resuspended in 30 μl of urea-containing lysis buffer (5% SDS, 9 M urea, 10 mM EDTA, 2.5% β-mercaptoethanol, 120 mM Tris-HCl pH 6.8) and incubated at 65°C for 15 min. The amount of exosomal protein recovered was measured by RC DC protein assay (Bio-Rad). Five microgram of total exosomal protein was electrophoresed as described earlier.
Densitometric analysis was performed using Quantity One program (Bio-Rad). Briefly, the signal intensity for each band was calculated and the background subtracted from experimental values. The resulting values were then expressed as arbitrary densitometric units.
LMP1 induces expression and relocalization of FGF-2 in epithelial cells
LMP1 is known to localize in part at the cell plasma membrane, but mostly in intracellular compartments, where it associates with lipid rafts whence it signals.45 First localization of endogenous LMP1 was analyzed by immunofluorescence staining in latently EBV-infected LCLs, or in EBV-infected MDA-MB-231 breast cancer cells (clone C3B4) as well as following transfection in EBV-negative nasopharyngeal epithelial cells (Ad-AH). Immunostaining of LMP1 either in LMP1-transfected cells or in the LCLs and C3B4 clone appears in intracellular spots mostly distributed beneath the cell-plasma membrane (Fig. 1a, arrows), but also perinuclearly (Fig. 1a, arrowheads). The vector-transfected cells and the nontransfected among the transfected cells are unstained, which demonstrates the specificity of the immunolabeling (Fig. 1a).
Since LMP1 has recently been reported to induce FGF-2 expression and secretion,25 we analyzed the localization of endogenous FGF-2 in cells positive for LMP1 by double-immunofluorescence labeling for LMP1 and FGF-2 in LMP1-transfected Ad-AH cells, followed by confocal microscopic analysis. Concordant with the biochemical data,25 in the LMP1-positive cells the FGF-2 signal is increased and distributed not only in the nucleus and in the central cytoplasm, but also at the cell periphery and along cellular projections, colocalizing as dots with LMP1 in the periphery (Fig. 1b, arrows). This intracellular localization of FGF-2 differs from that observed in LMP1-negative cells in the same microscopic field; in fact, the FGF-2 signal is not only quantitatively lower in intensity compared to that in LMP1-positive cells, but appears mostly nuclear and centrally located in the cytoplasmic portion of the cells (Fig. 1b, arrowheads). Colocalization of LMP1 and FGF-2 in peripheral dots was observed also in the C1D12 cells that express endogenous LMP1 (Fig. 1c), demonstrating that the redistribution of the FGF-2 signal from diffuse cytosolic to punctate is not a consequence of LMP1 overexpression. These observations suggest that the peripherally located FGF-2 visible in LMP1-expressing cells is the growth factor in the process of being released. Treatment of the vector-transfected cells with PMA induces an increase in FGF-2 signal in all cells, but this increased expression does not result in a peripheral distribution of FGF-2, which instead displays the usual nuclear and central cytoplasmic distribution (Fig. 1d), consistent with the lack of secretion of the growth factor when analyzed biochemically.25
Secretory 18-kDa FGF-2 colocalizes with LMP1 in multivesicular bodies/late endosomes
To follow selectively the secretory 18-kDa form of FGF-2 by confocal microscopy, we transfected Ad-AH cells with FLAG-tagged FGF-2: the FGF-2 signal appears intense and diffuse throughout the entire cell cytosol (Fig. 2a). To search for possible colocalization of FGF-2 and LMP1, we carried out cotransfections with pcLMP1 and FLAG-tagged FGF-2. Interestingly, the 18-kDa form of FGF-2 appears to colocalize with LMP1 visualized in both peripheral and perinuclear intracellular dots (Fig. 2a, arrows). To identify these intracellular spots that stained for LMP1, we doubly labeled the transfected cells with antibodies directed against markers of endosomal compartments and found that the dots are also positive for the CD63, a specific marker of late endosomes/MVBs, and for the endolysosomal hydrolase cathepsin D (CD), used as a marker for late endosomes and lysosomes, which suggests that they are late endosomal structures (Fig. 2b, arrows). Such structures may fuse with the plasma membrane and participate in a nonconventional mode of secretion of proteins.
To exclude the possibility that FGF-2 localization in endocytic structures could be ascribed to its secretion at the plasma membrane, followed by binding to low- and high-affinity receptors and internalization, we performed experiments in which the low-affinity binding of FGF-2 to the membrane heparan sulphate proteoglycans (HSPGs) was inhibited by treatment of Ad-AH cells with sodium chlorate which removes endogenous sulfate groups and inhibits HSPG biosynthesis46; in addition, parallel experiments were performed on CHO cells which do not express the specific high-affinity receptor FGFR1.47 Immunostaining with anti-FLAG antibody was performed before permeabilization to detect only FGF-2 released and bound to surface receptors: in chlorate-treated cells, the extracellular FGF-2 signal appeared drastically reduced in Ad-AH cotransfected cells (Fig. 3a) compared to the untreated ones, and absent in CHO cotransfected cells (Fig. 3b), indicating that HSPGs are mainly responsible for the binding of extracellular FGF-2 to the cell surface. When performed after permeabilization, immunostaining of FGF-2 appeared still in spots also in chlorate-treated cells (Figs. 3a and 3b), ruling out the possibility that the endosomal signal may correspond to growth factor secreted and then bound to surface receptors and internalized. Quantitative analysis of the fluorescence intensity was also performed to evaluate the possible concentration of FGF-2 in endosomal structures respect to its cytosolic distribution: to this aim, endosomal/MVB areas, positive for the CD63 marker, and comparable extensions of cytosolic areas were analyzed in 6–8 sequential confocal sections of 10 different cotransfected or 10 singly transfected cells, randomly taken from 3 different experiments. The results revealed that in Ad-AH cells transfected with 18-kDa FGF-2 alone the FGF-2 signal in MVBs is similar to that detected in the cytosol (MVB intensity/cytosolic intensity ± SE: 1.09 ± 0.07), whereas in doubly transfected cells the FGF-2 signal in MVBs is increased (MVB intensity/cytosolic intensity ± SE: 2.47 ± 0.13), revealing concentration of the growth factor in endosomes when LMP1 is expressed.
To demonstrate that FGF-2 concentration in late endosomal structures occurs as a consequence of the inward budding responsible for the formation of the internal vesicles of MVBs and, therefore, of the released exosomes, we have inhibited the process of MVB biogenesis by RNA interference to knockdown the expression of the TSG101 protein, which is known to be involved in the invagination of the endosomal membrane.48 The C3B4 clone of EBV-infected MDA-MB-231 cells, which express endogenous LMP1 and FGF-2 as shown earlier, were transfected with siRNA for TSG101 and the expression of the TSG101 protein was then evaluated by western blot analysis: as shown in Figure 4a, the TSG101 protein expression appeared drastically down-regulated in siRNA-transfected cells. To directly evaluate the effect of the inhibition of TSG101 expression and MVB biogenesis on FGF2 intracellular localization, C3B4 cells were microinjected with a mixture of siRNA for TSG101 and IgG-FITC to identify the injected cells or with the IgG-FITC alone as control. After microinjection, cells were left for 24 hr before fixation. In cells microinjected with siRNA for TSG101, the CD63-positive MVBs appeared enlarged and altered in their distribution (Fig. 4b, arrows) as a consequence of TSG101-depletion and the FGF-2 signal was diffuse and did not localize in dots (Fig. 4b, asterisk). In contrast, in uninjected cells among the injected ones or in cells injected with the tracer only, numerous dots positive for FGF2 were visible (Fig. 4b, arrowheads).
18-kDa FGF-2 and LMP1 are both present in exosomes released by the cells
Exosomes are membrane vesicles released by cells and may be involved in a nonconventional secretory process. To analyze if exosomes could be the vehicle for secretion of FGF-2 induced by LMP1, we purified the exosomal fraction from the supernatant fluids of Ad-AH cells cotransfected with 18-kDa FGF-2 and LMP1. The purity of the fraction was assessed by the presence of the HSC70 protein, which is highly enriched in exosomes.33 Biochemical analysis with LMP1 and FLAG antibodies in western blots of total cell lysate and purified exosomal fraction disclosed 18-kDa FGF-2 in the fractionated lysate (Fig. 5a). Loading of increasing concentrations of the exosomal fraction corresponded to a parallel increase in the content of both HSC70, as expected, and 18-kDa FGF-2 (Fig. 5b), confirming the presence of the secreted form of FGF-2 in exosomes.
To demonstrate the possible concentration of FGF-2 in exosomes induced by LMP1, we compared the content of LMP1 and FGF-2 in cell lysates and in exosomal fractions, loading equal amount of them, followed by densitometric analysis: from cotransfected cells, a 7.5-fold increase of FGF-2 was detected in exosomes respect to the corresponding lysate, whereas no increase was observed in cells transfected with 18-kDa FGF-2 only (Fig. 6a). LMP1 protein content appeared highly enriched (more than 35-fold increase) in exosomes compared to cell lysates (Fig. 6a).
Loading of the entire protein content isolated from an equal amount of culture supernatant fluid of untransfected, vector-transfected and cotransfected cells suggested an increased release of exosomes containing HSC70 and TSG101 from cotransfected cells, possibly induced by LMP1 expression and FGF-2 secretion (Fig. 6b). Parallel control experiments of exosomal isolation from cells transfected with FLAG-tagged 18-kDa FGF-2 and the pcDNA3 vector revealed that the single overexpression of 18-kDa FGF-2 by itself did not lead to an increase in exosome formation (Fig. 6b). Moreover, western blot of total cell lysates from singly and doubly transfected cells showed that HSC70 and TSG101 expression are not modified by either LMP1 or FGF-2 (Fig. 6b).
To verify morphologically the composition of the purified fraction, we used negative-staining electron microscopy. Ultrastructural examination of purified exosomes isolated from Ad-AH cells revealed a homogeneous pellet composed of membrane vesicles 60–100 nm in diameter (Fig. 7a). To analyze if these membrane vesicles contained both LMP1 and secreted FGF-2, we performed immunoelectron microscopy of purified exosomes isolated from Ad-AH cells cotransfected with LMP1 and FLAG-tagged 18-kDa FGF-2. The pellet was first labeled with LMP1 monoclonal antibody followed by incubation with 18-nm-diameter colloidal gold particles conjugated with protein A and then doubly labeled with rabbit anti-FLAG polyclonal antibodies followed by incubation with 10-nm-diameter protein A–colloidal gold particles. In both singly and doubly immunolabeled exosomes, the gold granules corresponding to LMP1 appeared associated with the exosomal membrane, consistent with the expected membrane localization of the protein (Figs. 7b, 7d and 7e). In contrast, most of the gold particles corresponding to 18-kDa FGF-2 were localized inside the exosomes, consistent with a cytosolic distribution of the protein (Figs. 7c–7e). In Figures 7d and 7e, the double immunogold labeling showed that both LMP1 (large golds, arrows) and 18-kDa FGF-2 (small golds, arrowheads) were present in the exosomal vesicles. Double gold immunolabeling was also performed on ultrathin cryosections of the Ad-AH cotransfected cells, since this method provides unequivocal detection of LMP1 and FGF-2 inside membrane vesicles, and the electron microscopic examination revealed the presence of 18-kDa FGF-2 (Figs. 7f–7h: small golds, arrows) in MVBs highly positive for LMP1 (Fig. 7g, large golds) and identified by their positivity for the CD63 marker (Fig. 7h: large golds).
To evaluate if the FGF-2-containing exosomal fractions may exert a proliferative activity comparable to that induced by the growth factor, we performed a BrdU incorporation assay on endothelial HUVEC cells and observed the labeled nuclei by immunofluorescence. Cells were serum starved and incubated with the isolated exosomes (10 μg/ml) or with FGF-2 (20 ng/ml) for 24 hr in the presence or absence of anti-FGF-2 blocking antibodies. Cells were then incubated with BrdU for the last 6 hr and fixed; incorporated BrdU was then visualized with anti-BrdU antibody, followed by fluorescein-conjugated secondary antibodies. In cell cultures incubated with both isolated exosomes and FGF-2, the number of positive nuclei was increased compared to the untreated cells (Fig. 8), demonstrating that both treatments were able to elicit a similar stimulation of DNA synthesis and cell proliferation. The reduction of the number of stained nuclei observed in the presence of anti-FGF-2 blocking antibody (Fig. 8) suggests that the activity of the isolated exosomes could be ascribed to their FGF-2 content.
Na+/K+-ATPase activity is required for FGF-2 secretion through exosomes
Since it has been proposed that the sodium potassium ATPase (Na+/K+-ATPase) activity is responsible for FGF-2 translocation,18 and that ouabain, a specific inhibitor of the α1-subunit of this enzyme, is able to block, although only partially, FGF-2 release,18, 19, 25 we analyzed if Na+/K+-ATPase could also be involved in the exosome-mediated mechanism of FGF-2 export. First we analyzed by confocal immunofluorescence microscopy the localization of Na+/K+-ATPase in Ad-AH cells transiently cotransfected with LMP1 and FLAG-tagged 18-kDa FGF-2. Cells were permeabilized and incubated with antibodies to Na+/K+-ATPase α1 subunit and FLAG. Whereas in untransfected cells the Na+/K+-ATPase staining was mostly associated with the plasma membrane, partial relocalization of the Na+/K+-ATPase signal from the plasma membrane to intracellular dots visible as dots positive for 18-kDa FGF-2 was clearly evident in cells cotransfected with LMP1 and FGF-2 (Fig. 9a, arrows). Double immunogold electron microscopy of ultrathin cryosections of cotransfected cells, performed with anti-Na+/K+-ATPase antibody and large protein A–gold particles (18 nm) followed by anti-FLAG antibodies and small protein A–gold particles (10 nm), showed the presence of both Na+/K+-ATPase and FGF-2 in endosomal/MVB structures (Fig. 9b), which appear virtually unlabeled in control experiments performed omitting the primary antibodies from the labeling procedure (Fig. 9b).
To demonstrate the involvement of Na+/K+-ATPase in exosomal FGF-2 release, we treated the cells with ouabain (50 nM) for 48 hr and performed the biochemical analysis described earlier. Western blot analysis of purified exosomes from Ad-AH cells transfected with control vector (pcDNA3) or cotransfected with LMP1 and FLAG-tagged 18-kDa FGF-2 and treated with ouabain showed that FGF-2 secretion by exosomes is dramatically reduced by ouabain treatment. The presence of an equal amount of HSC70 in the fractions from untreated or ouabain-treated cells demonstrated that the treatment did not influence exosomal release (Fig. 10a). Ouabain does not change the amount of exosomes, but influences the amount of FGF-2 protein recovered inside the vesicles. In addition, treatment of cotransfected cells with increasing concentrations of ouabain (10 and 30 nM) demonstrated that the effect of ouabain on FGF-2 secretion though exosomes is dose-dependent; in fact, the higher doses of the drug appeared to induce a clear reduction in the amount of FGF-2 protein released in exosomes (Fig. 10b).
In this article, we demonstrate that FGF-2 is packaged in exosomes for secretion and that the EBV oncoprotein LMP1 promotes this secretory process selectively. FGF-2 is one of the known secretory proteins lacking an N-terminal leader signal peptide for ER translocation. The unconventional mechanisms for secretion of such “leaderless” proteins are only partially understood, but seem to require the involvement of different molecular machinery and cellular pathways; moreover, it is likely that the same leaderless protein can use different nonclassical mechanisms.49, 50, 51, 52 For some of the proteins exported by unconventional secretion, such as IL-1β, it has been proposed that vesicular structures of the endocytic compartment represent the intracellular sites for translocation and may be used for secretion; in fact, cytosolic IL-1β is first translocated in such structures, which correspond to late endosomes and lysosomes, and then released extracellularly upon fusion of these vesicles with the plasma membrane.39 Other proteins, such as galectins, are believed to exit the cells by different nonclassical pathways, according to their structure or to the cell type. Galectin 1 is generally thought to be directly secreted through the plasma membrane,17 whereas galectin 3 export seems to involve blebbing of the plasma membrane.53 For the latter protein it is also hypothesized that there is an exosome-mediated release.40 In contrast, the leaderless proteins FGF-1 and FGF-2 appear to translocate and be released into the extracellular space directly at the cell plasma membrane.17, 54, 55
Na+/K+-ATPase has been proposed to play a role in FGF-2 translocation18 because ouabain, a specific inhibitor of Na+/K+-ATPase, appears to inhibit, although only partially, FGF-2 export.18, 19, 25 The exact role of this transporter is unknown, but there are various hypothesis: it may be a component of a translocation apparatus involved in FGF-2 export; it may regulate this export through the maintenance of electrochemical gradients; or it may function in recruiting FGF-2 to the cytoplasmic side of the plasma membrane. Co-immunoprecipitation experiments also provided evidence of a functional interaction between FGF-2 and the α subunit of the enzyme.18 Recently, however, it has been demonstrated through an in vitro secretion assay that FGF-2 is able to translocate directly at the plasma membrane, but in a Na+/K+-ATPase-independent manner.17 Moreover, a second recent report suggested that FGF-2 release may occur by shedding of membrane vesicles that do not correspond to exosomes because of their larger dimensions and the absence of heat-shock proteins for which exosomal vesicles are usually enriched.56 Therefore, these conflicting results suggest that FGF-2 may use different mechanisms for secretion, which may depend on cell type and conditions, and that its release can be variously regulated. Our current hypothesis, based on the results presented here, is that FGF-2 may use a third nonconventional pathway for secretion mediated by the release of exosomes.
Our results show also that secretion of FGF-2 through exosomes is controlled by Na+/K+-ATPase activity and blocked by ouabain. We demonstrated that ouabain treatment does not change the amount of exosomes, but changes the amount of FGF-2 protein in them. These data suggest the possibility that Na+/K+-ATPase activity could be required for recruitment of the cytosolic FGF-2 to the endosomal membrane of the multivesicular bodies; therefore, ouabain could affect the intake of FGF-2 protein while in the process of binding to the endosomal membrane or in the process of cytosolic sequestering during inward budding.
We propose further that LMP1 protein, which induces processes that lead to promotion of angiogenesis (and ultimately tumor progression22, 23, 24, 25, 27), also does so by enhancing FGF-2 secretion through this unusual secretory mechanism. In fact, in our experiments, LMP1 seems to increase the quantity of exosomes that are secreted by the cell. LMP1 signaling in intracellular compartments, possibly of the endocytic pathway,45 may play a role in the sorting and budding of vesicles at the lumina of multivesicular bodies responsible for exosomal formation, which may control the number of exosomes. Consistent with our present results, the presence of LMP1 in exosomes released by EBV-positive cell lines has been previously reported.42, 43, 44 Taken together, our results indicate that FGF-2 release in both lymphoblastoid and epithelial cells may be controlled through 2 distinct pathways: LMP1 can upregulate the number of exosomes secreted, and Na+/K+-ATPase can increase the FGF-2 concentration in the exosomes. These findings may also be relevant to the colocalization of LMP1 and MHC molecules and immune evasion by EBV.
In vitro reconstitution of FGF-2 secretion with the use of “inside-out” vesicles has recently demonstrated that the translocation of FGF-2 across the plasma membrane is unidirectional and requires cytosolic factors.17 Our results with FGF-2 secretion by exosomes are in accord with the postulated mechanism of translocation; however, we propose that in EBV-positive tumor cells such a mechanism of secretion may be optimized by the concentration of FGF-2 in small and enriched fractions of the cell. In fact, due to the exosomal membrane orientation and cytosolic content, exosomes may represent selected cellular portions characterized by high concentration and sequestration of either membrane and cytosolic components, such as HSPs or MHC,34, 57 necessary for specific functions in intercellular communications. Therefore, the mechanism of FGF-2 release from exosomes may require its selective translocation across the exosomal membrane, similar to that proposed to occur at the cell plasma membrane17; alternatively, FGF-2 secretion could be simply a consequence of rupture of the exosomal membrane as suggested for the export of other cytosolic proteins found in exosomes. The exosome-mediated secretion may improve the angiogenic potency of FGF-2 by carrying the factor at a distance from the plasma membrane of the producing cells and preventing autocrine interactions with its own cell-surface HSPGs, thus favoring heterotypic interactions in the neighborhood of the tumor necessary for induction of angiogenesis and vasculogenesis.
The possible function(s) of tumor-derived exosomes is still unknown. It has been recently highlighted that exosomes may be able either to suppress or stimulate the immune system, reflecting the behavior of the releasing cells.36, 37 Our results with LMP1/FGF-2 doubly-positive exosomes suggest that, besides the likely role in escape from antitumor immune responses due to the presence of LMP1 or LMP1 fragments,42 exosomes may act in promoting angiogenesis. It is interesting to note that release of an angiogenic factor, the developmental endothelial locus-1, from mesothelioma cells has been recently reported,38 providing further evidence for a general angiogenic function of tumor-derived exosomes and suggesting their use in developing antiangiogenic factors for therapeutic purposes.
In addition, the high and selected concentration of tumor antigens in exosomes opens the possibility of the use of exosomes for possible vaccination strategies for cancer therapy. In fact, LMP1 is expressed in many EBV-associated malignancies and, depending on the stage of viral infection, is known to either increase or decrease the immune response and to positively modulate the presentation of host antigens as well as of viral antigens.58 Therefore, strategies for active immunization using such exosomes as a target for CTL-based immunotherapy and for antiangiogenic therapy may lead to the development of new multifunctional approaches and tools.
We thank Dr. Leslie Grab (University of North Carolina at Chapel Hill) for kindly providing reagents and protocols for the exosome isolation, Dr. Pankaj Trivedi (University of Roma La Sapienza) for the generous gift of the LCL-AKE cell line and Dr. Laura Leone (University of Roma La Sapienza) for performing the immunogold labeling.