Microparticle-Mediated Transfer of the Viral Receptors CAR and CD46, and the CFTR Channel in a CHO Cell Model Confers New Functions to Target Cells

Cell microparticles (MPs) released in the extracellular milieu can embark plasma membrane and intracellular components which are specific of their cellular origin, and transfer them to target cells. The MP-mediated, cell-to-cell transfer of three human membrane glycoproteins of different degrees of complexity was investigated in the present study, using a CHO cell model system. We first tested the delivery of CAR and CD46, two monospanins which act as adenovirus receptors, to target CHO cells. CHO cells lack CAR and CD46, high affinity receptors for human adenovirus serotype 5 (HAdV5), and serotype 35 (HAdV35), respectively. We found that MPs derived from CHO cells (MP-donor cells) constitutively expressing CAR (MP-CAR) or CD46 (MP-CD46) were able to transfer CAR and CD46 to target CHO cells, and conferred selective permissiveness to HAdV5 and HAdV35. In addition, target CHO cells incubated with MP-CD46 acquired the CD46-associated function in complement regulation. We also explored the MP-mediated delivery of a dodecaspanin membrane glycoprotein, the CFTR to target CHO cells. CFTR functions as a chloride channel in human cells and is implicated in the genetic disease cystic fibrosis. Target CHO cells incubated with MPs produced by CHO cells constitutively expressing GFP-tagged CFTR (MP-GFP-CFTR) were found to gain a new cellular function, the chloride channel activity associated to CFTR. Time-course analysis of the appearance of GFP-CFTR in target cells suggested that MPs could achieve the delivery of CFTR to target cells via two mechanisms: the transfer of mature, membrane-inserted CFTR glycoprotein, and the transfer of CFTR-encoding mRNA. These results confirmed that cell-derived MPs represent a new class of promising therapeutic vehicles for the delivery of bioactive macromolecules, proteins or mRNAs, the latter exerting the desired therapeutic effect in target cells via de novo synthesis of their encoded proteins.

In general, MPs carry with them membrane and cytosolic components specific of their cellular origin [26], including proteins and nucleic acids, such as mRNAs and microRNAs [7,11,[27][28][29][30], and are capable of transferring their cargo to recipient cells [16,29,[31][32][33][34]. During their extracellular release, MPs can also embark components which are foreign to the cells, such as nucleic acids, proteins or glycoproteins expressed transiently or constitutively by a plasmid or viral vector. The latter scenario is reminiscent of the process of virus or virus-like particles (VLPs) pseudotyping by foreign glycoproteins [35][36][37][38][39][40]. MPs are not only considered as circulating biomarkers for the molecular profiling of certain cancers [41], but their therapeutic potential as conveyors of bioactive factors, proteins, RNAs including miRNAs, is being evaluated for personalized medicine and for the treatment of a number of diseases and cellular dysfunctions [5,7,9,11,27,30,42].

Isolation and Recovery of MPs from CAR-and CD46expressing CHO Cells
The CHO-CAR and CHO-CD46 cells were derived from the parental CHO-K1 cell line to constitutively express and display CAR [44] and CD46 [61][62][63] at their surface, respectively. The recovery of MPs from stressed cells is usually higher than from nonstressed cells. However, since MPs issued from stressed cells could transfer stress signals and induce the apoptosis of recipient cells, our starting material was the culture medium of nonstressed CHO-CAR and CHO-CD46 cells. Extracellular MPs were recovered by a two-step ultracentrifugation procedure which separated MPs according to their size, consisting of a first sedimentation at 30,0006g, followed by a second at 100,0006g. Two populations were thus obtained, abbreviated MP 30 and MP 100 , respectively (Fig. S1). The MP 30 fraction contained large MPs characterized by their heterogeneity in shape and size, ranging from 50-500 nm in diameter, consistent with plasma membrane-shedded MPs (Fig. 1A). On the other hand, the MP 100 population consisted of homogenous and regular particles, rather spherical and relatively small in size (50-100 nm; Fig. 1B), reminiscent of exosomes or exosome-like particles [2,28]. Antibodies directed against human and mouse TSG101 and CD63, two surface markers of exosomes, were used in attempts to further characterize this population. However, no clear reaction was obtained with these antibodies on CHO cell lysates or MP pellets by Western blot or flow cytometry, respectively. Extracellular MPs released by CHO-CAR and CHO-CD46 cells were therefore differentiated only by their sedimentation properties, and referred to as MP 30 and MP 100 in the present study.
The yields of MPs spontaneously recovered from 10 7 cells (CHO-CAR or CHO-CD46) ranged from 1610 7 to 3610 7 for MP 30 after 72 h culture, and a similar recovery was obtained for MP 100 . After resuspension of the MP pellets in PBS, the total MP concentration (titer in physical MPs) of our working stocks usually ranged between 4610 7 to 7610 7 MPs/ml for MP 30 or MP 100 , corresponding to total protein concentrations in the range of 700-850 ng/ml. Considering their size and composition heterogeneity, an average of 100 ng protein corresponded to a total number of 5610 6 to 7610 6 MP 30 or MP 100 . Efficiency of Incorporation of CAR and CD46 by MPs MPs isolated from CHO-CAR and CHO-CD46 cells by our two-step ultracentrifugation procedure were referred to as MP 30 CAR and MP 100 CAR, and MP 30 CD46 and MP 100 CD46, respectively. The proportion of CAR-positive and CD46-positive MPs was determined by flow cytometry using specific antibodies, as this technique detected the CAR and CD46 molecules exposed at the MP surface, and potentially active as adenoviral receptors. The mean value of the percentage of MP 30 CAR to total MPs was 5.861.4% (mean 6 SEM, n = 3), corresponding to a mean bioactive MP titer of 2610 5 30 and MP 100 was confirmed by immunoelectron microscopy (immuno-EM), using anti-CAR or anti-CD46 mouse monoclonal antibody followed by a secondary 10 nm-gold-labeled anti-mouse IgG. The proportion of MP 30 or MP 100 associated with anti-CD46-bound or anti-CAR-bound colloidal gold grains was found to range between 2 to 3% (Fig. 1A, B). This value was consistent with the flow cytometry data, considering that MP 30 and MP 100 adsorbed on a solid support did not offer a full access to antibodies, compared to MPs in suspension.

CD46 and CAR Molecules Displayed on MP 30 were Functional as Adenoviral Receptors
To assess the functionality of CAR and CD46 molecules on MP 30 as adenoviral receptors, samples of MP 30 CAR and MP 30 CD46 were incubated with HAdV5-GFP or HAdV5F35-GFP vectors for 2 h at 37uC, using equal numbers of MPs and vector particles. The samples were then layered over a 20%sucrose cushion, and centrifuged at 30,0006g for 2 h. The material which pelleted through the cushion was fixed and embedded, and ultrathin sections were processed for observation under the electron microscope (EM). Numerous complexes of MP 30 -viral vector particles were observed in the 30,0006gpelletable fraction, as exemplified with MP 30 CD46 and HAdV5F35-GFP (Fig. 1C). These results confirmed that CD46 and CAR molecules displayed on MP 30 were functional as cellular receptors for their specific adenoviruses. MP 30 CHO from unmodified, parental CHO cells were used as negative control. MP 30 CHO and MP 30 CD46 were incubated with aliquots of HAdV5F35-GFP vector, as above, and the mixtures added to monolayers of target CHO cells. Samples were harvested after 2 h at 37uC, and processed for EM. Particles of adenoviral vector were observed in complex with MP 30 CD46 at the surface of the target cells ( Fig. 2A), whereas no vector particle in association with control MP 30 CHO was observed (Fig. 2B).  30 /cell), 2 to 10% cells were found to become GFP-positive (Fig. 3), which corresponded to the background of infection of CHO cells by HAdV5 and HAdV5F35 via other cell surface molecules, such as heparan sulfate glucosaminoglycans, which act as alternative receptors for HAdV5 and HAdV5F35 [43,64]. However, CHO cells pre-incubated with MP 30 CAR or MP 30 CD46, and infected with HAdV5-GFP and HAdV5F35-GFP, respectively, showed a 2-to 3-fold increase in GFP-positive cells. This significant increase in permissiveness to the adenovirus vectors was MP dose-dependent (Fig. 3). These results indicated that MP 30 CAR and MP 30 CD46 were capable of transfering the CAR and CD46 molecules to CHO target cells.

MP-mediated
(iii) Receptor specificity in MP-recipient cells. As additional controls, we assessed the specificity of vector-receptor recognition, by infecting MP 30 CD46-transduced cells with HAdV5-GFP, and vice versa, MP 30 CAR-transduced cells with HAdV5F35-GFP. Only a background GFP signal was detected with these two heterotypic pairs, demonstrating the specificity of the newly acquired CAR and CD46 receptor molecules towards their respective adenoviral vectors (Fig. 3).
(iv) Functionality of CD46 as complement regulatory factor in MP-recipient cells. The antiapoptotic activity of CD46 transferred to CHO cells was analyzed in CHO cells interacted with MP 30 CHO or with MP 30 CD46, then incubated with increased doses of complement fraction C3. A significant effect of protection against complement lysis was observed in MP 30 CD46-transduced CHO cells, by comparison to control cells (Fig. 4A).

Influence of VSV-G Incorporation on MP 30 -mediated Transfer of CAR or CD46
The following experiments were designed to determine whether the transfer of CAR or/and CD46 to target CHO cells would be improved by the incorporation of VSV-G, the envelope glycoprotein G of vesicular stomatitis virus, into the MP envelope. We found that baculovector-mediated expression of VSV-G in CHO-CAR and CHO-CD46 cells did not enhance the production of CAR-or CD46-pseudotyped MPs (data not shown). Likewise, coincorporation of VSV-G with CAR or CD46 into MP 30 did not significantly increase the degree of permissiveness of CHO cells to HAdV5-GFP or HAdV5F35-GFP after transduction by MP 30 CAR-VSV-G or MP 30 CD46-VSV-G (Fig. S2). This implied  30 -cell interaction, cells were infected with HAdV5-GFP or HAdV5F35-GFP vector, at the same MOI (500 vp/cell). The degree of CHO permissiveness to HAdV5-GFP or HAdV5F35-GFP vector was evaluated by flow cytometry analysis of the intracellular GFP signal. In (a), HAdV5F35-GFP, which does not recognize CAR as cellular receptor, was used as the negative control. In (b), HAdV5-GFP, which does not recognize CD46 as cellular receptor, was used as the negative control. MP 30 from nontransduced CHO cells (Control MP) served as the negative controls in both panels. doi:10.1371/journal.pone.0052326.g003 that VSV-G was not a facilitator of MP 30 uptake by the target CHO cells, and VSV-G was not included in the following experiments.

Time-course Analysis of the Acquisition of New Functionality by MP 30 -recipient Cells
The next issue was to determine at which time after MP 30 CD46-cell interaction the CHO target cells acquired the adenovirus receptor function carried by CD46. CHO cells were incubated with a constant dose of MP 30 CD46 (20 MP 30 /cell), and at different times after MP 30 -cell interaction, the cells were incubated with HAdV5F35-GFP vector at constant vector MOI (500 vp/cell) for 2 h at 37uC. At 48 h post-infection (pi), the permissiveness of the cells to HAdV5F35-GFP was determined by flow cytometry analysis. A progressive increase in adenoviral permissiveness was observed, with 10% GFP-positive cells at 48 h pi, reaching a plateau of 20-25% GFP-positive cells at 72-144 h (Fig. 4B). Similar results were obtained with MP 30 CAR (not shown). The kinetic data implied that a larger proportion of cells became permissive to adenovirus with time, and suggested the contribution of additional CD46 molecules. This was likely due to newly synthesized CD46 from CD46-encoding mRNA transferred by MP 30 CD46 to the target cells, as demonstrated below for MPmediated CFTR transfer.

Generation of CHO Cells Stably Expressing GFP-CFTR
A CHO cell line stably expressing GFP-fused human CFTR was generated (CHO-GFP-CFTR) using the pCEP 4 -GFP-CFTR episomal plasmid. The pCEP 4 -GFP-CFTR-directed expression of the GFP-CFTR fusion protein in these cells showed localisation of GFP signal in the cytoplasm, perinuclear and submembrane regions ( Fig. 5A i, ii; and Fig. S3A). A similar fluorescence pattern was observed in CHO-CD46 cells transduced by the HAdV5F35-GFP-CFTR vector which were used as positive control (Fig. 5A iii, iv). In CHO-GFP-CFTR cells, the CFTR molecules were correctly oriented in the plasma membrane, as shown by the labeling of nonpermeabilized CHO-GFP-CFTR cells with an anti-CFTR monoclonal antibody directed against the first extracellular loop of the CFTR ectodomain (Fig. 5B).
The chloride channel function in CHO-GFP-CFTR cells was investigated using the voltage-sensitive probe DiSBAC 2 (3). The fluorescent signal of DiSBAC 2 (3) has been shown to vary with changes of the membrane potential induced by the cAMP/lowchloride-mediated activation of the CFTR [65]. Parental CHO cells were used as negative controls, and CHO-CD46 cells transduced by the HAdV5F35-GFP-CFTR vector were used as positive controls. A positive fluorescent signal was observed in CHO-GFP-CFTR cells in response to the addition of the cAMPbased cocktail of CFTR activators, demonstrating the functionality of the exogenous GFP-CFTR molecules as chloride channels (Fig. 5C). HAdV5F35-GFP-CFTR-transduced cells used as positive control showed a strong enhancement of the fluorescent signal in the presence of CFTR activators (Fig. 5C), consistent with the high transduction efficiency of CHO-CD46 by HAdV5F35 [55,66]. CHO-GFP-CFTR cells were then tested for the production of MPs carrying GFP-CFTR (MP-GFP-CFTR) and the MP-mediated delivery of CFTR to target cells.

Isolation and Recovery of MP-GFP-CFTR
MP 30 GFP-CFTR and MP 100 GFP-CFTR were recovered from the CHO-GFP-CFTR cell culture medium by the ultracentrifugation procedure mentioned above and depicted in Fig. S1. The presence of the GFP-tag allowed the direct tracking of GFP-CFTR molecules in the different cell compartments, such as the vesicular compartment and plasma membrane [55,66]. It also allowed the detection of GFP-CFTR incorporated in MPs derived from GFP-CFTR-expressing cells. The recovery of MP 30 GFP-CFTR and MP 100 GFP-CFTR ranged from 1610 7 to 5610 7 per 10 7 CHO-GFP-CFTR cells. Flow cytometry analysis indicated that the GFP-CFTR signal was detected in both MP 30 and MP 100 populations, although in significantly higher amounts in MP 30   CFTR molecules between intracellular membranal compartments and the cell surface.

Fractionation of MP 30 GFP-CFTR and MP 100 GFP-CFTR Populations and Distribution of GFP-CFTR between MP Subclasses
To obtain the maximum efficiency of GFP-CFTR transfer to target cells, it was important to determine whether a particular MP fraction could be enriched in GFP-CFTR. To this aim, the MP 30 and MP 100 populations from CHO-GFP-CFTR cells were further fractionated into subpopulations differing in density, using ultracentrifugation in isopycnic gradients [67]. Three subpopulations, corresponding to three density classes, were obtained for each MP population (  30 and MP 100 populations was then assayed for the presence of GFP-CFTR by flow cytometry analysis. The GFP-CFTR protein was detected in all MP subclasses, with the highest proportion of GFP-CFTR-positive MPs found in the high-density compared to low-density subclasses: 14.5% for MP 30 HD versus 6.2% for MP 30 LD, and 8.3% for MP 100 HD versus 1.4% for MP 100 LD ( Table 1). This data suggested that CFTR glycoprotein was associated with the higher density MPs.

Kinetics of MP-mediated Transfer of GFP-CFTR to Target Cells
The following experiments were designed to determine the capacity of each subpopulation of MPs to transfer GFP-CFTR to target CHO cells. The MP 30 (LD, ID and HD), and MP 100 (LD, ID and HD) subclasses were incubated with CHO cells at a ratio of 5 MPs/cell for 2 h at 37uC. Inocula were removed, and monolayers were postincubated with fresh, prewarmed medium at 37uC for 6 days. At different time-points, cell samples were examined by fluorescence microscopy and flow cytometry for qualitative and quantitative analyses of the GFP signal, respectively. With the MP 30 population, the GFP profiles did not differ significantly for the three subclasses. At 2 h pt, 3.8 to 4.4% CHO cells were GFP-positive, increasing to 5.5-6.6% at 4 h pt, followed by a slight decrease at 6-8 h pt (Fig. 6A). At 8 h pt, the proportion of GFP-positive cells increased progressively for the three MP 30 subclasses, to attain a value of 73-83% GFP-positive cells 5 days pt (Fig. 6A). A similar kinetics of appearance of GFP-positivity was observed with the three MP 100 subclasses: a discrete peak was detected at 4 h pt, followed by a slight decrease at 6-8 h pt, and a progressive increase to 57-76% GFP-positive cells after 5 days (Fig. 6B).
The cell surface expression and proper membrane insertion of GFP-CFTR was investigated by flow cytometry, using the anti-CFTR ectodomain antibody as above. The maximal CFTR display at the cell surface was observed at 2-3 days pt, with 50-60% CFTR-positive cells after MP 30 -mediated transduction (Fig. 7A). Similar levels (60-70% CFTR-positive cells) were The proportion of GFP-positive MPs was determined by flow cytometry, and the GFP-CFTR mRNA content by qRT-PCR analysis. Data shown in the Table are   observed after MP 100 -mediated transduction (Fig. 7B). The presence of CFTR at the cell surface progressively declined after day-4 (Fig. 7A, B).
Interestingly, there was a slight, but significant difference in profile between the GFP signal and the CFTR ectodomain immunoreactivity. The proportion of GFP-positive cells remained as a plateau after 6 days (refer to Fig. 6 A, B) whereas the display of CFTR at the cell surface decreased after day-4 (refer to Fig. 7). Likewise, MP 30 -and MP 100 -recipient cells examined at day-5 pt by confocal microscopy showed that the majority of the GFP-CFTR signal appeared as cytoplasmic dots and speckles corresponding to vesicular compartment(s), and in relatively low proportion at the plasma membrane (Fig. S3B).

Occurrence of GFP-CFTR-encoding mRNA in MPs and MP-recipient Cells
The kinetics of appearance of GFP-positive cells shown in Fig.5 suggested an early (4 h pt) and a late phase (3 days pt) of MPmediated GFP-CFTR transfer. We hypothesized that at early times after interaction of MP 30 and MP 100 with target cells, membrane-inserted GFP-CFTR glycoprotein molecules were transferred to a small number of target cells (less than 10%), following a pathway of MP-cell binding, endocytosis and membrane fusion. At late times, the GFP-CFTR protein detected in the MP-recipient cells was likely due to the neosynthesis of GFP-CFTR from MP-embarked GFP-CFTR-encoding mRNA molecules. To test this hypothesis, total RNA were extracted from the different MP 30 and MP 100 subclasses, and analysed for the presence of GFP-CFTR-encoding mRNA (mRNA gfp-cftr ), using quantitative RT-PCR and specific primers overlapping the junction of the GFP C-terminal and CFTR N-terminal coding sequences [55,66]. All MP subclasses were found to contain comparable amounts of mRNA gfp-cftr (Table 1). However, when the values of the MP content of mRNA gfp-cftr were expressed as the ratio to b-actin mRNA content used as endogenous control, significant differences in composition were observed between the MP subclasses (Fig. 8A, B). Similarly, total RNA were extracted from MP-recipient cells incubated with MP 30 (LD, ID and HD), and MP 100 (LD, ID and HD) at day-5 pt, and qRT-PCR was carried out as above, using bactin mRNA as internal control. As positive controls, CHO-CD46 cells transduced by HAdV5F35-GFP-CFTR were processed for RNA extraction and qRT-PCR assays. We found that the recipient cells incubated with all the different subclasses of of MP 30 and MP 100 contained mRNA gfp-cftr , with a ratio to b-actin content varying from 0.5-to 2-fold (Fig. 8C, D). Interestingly, the variations in the cellular content of mRNA gfp-cftr reflected the respective contents of the different MP subclasses. The observation that almost 80% of the cells expressed GFP-CFTR protein at late times pt suggested that nearly all cells received mRNA gfp-cftr molecules when a bioactive MP-to-cell ratio of 5 MPs/cell was used.
The efficiency of transfer of mRNA gfp-cftr was calculated as follows. The average content of donor cells was found to be 4.4610 8 copies of mRNA gfp-cftr per mg total cellular RNA extracted, with values ranging from 2.2610 8 to 6.6610 8 copies. The average content of MP-recipient cells incubated with MP 100 -GFP-CFTR was 3.5610 7 copies of mRNA gfp-cftr at 48 h after transfer using 5 MPs/cell, with values ranging from 1.7610 7 to 5.3610 7 copies per mg total cellular RNA. We estimated the transfer efficiency as the ratio 3.5610 7 : 4.4610 8 < 8%.
To eliminate the possibility of DNA transfer via MPs, CHO cells were harvested at 48 h after incubation with each of the different subclasses of MP 30 and MP 100 , the DNA extracted and probed for the episomal plasmid pCEP 4 -GFP-CFTR, using PCR amplification of a 196 nt fragment overlapping the GFP-3' and CFTR-5' sequences [66]. No specific DNA fragment was found at this position in all cell samples tested (Fig. S4).

Functionality of Exogenous CFTR as Chloride Channel in MP-recipient Cells
The next issue to address was whether target CHO cells incubated with MP-GFP-CFTR acquired the CFTR-associated chloride channel activity. CHO cells were taken 3 days after incubation of MP 30 -GFP-CFTR or MP 100 -GFP-CFTR, the time at which a maximum number of cells were observed to be GFPand CFTR-positive (refer to Fig. 6 and 7). The changes of the DiSBAC 2 (3) fluorescent signal in MP-transduced cells in response to the CFTR activator and inhibitor were monitored by quantitative fluorescence microscopy (Fig. 9). As exemplified with MP 100 -GFP-CFTR, the fluorescent signal increased progressively in the presence of CFTR activators, and rapidly decreased to background levels upon the addition of the CFTR inhibitor GlyH-101 (Fig. 9). This effect was reversible, as the fluorescence recovered progressively with the removal of GlyH-101 from the cAMP-containing superfusion medium (Fig. 9). Thus, the target CHO cells showed a gain in biological function, i.e. the chloride channel activity associated with acquired CFTR molecules.

Discussion
The current trends in biotherapy comprise the development of stem cell technology in combination with ex vivo gene therapy, the improvement of synthetic vectors, or the design of immunologically stealthy viral vectors, or safer viral integrative vectors with controlled integration sites in host cells. The present study explored an alternative strategy to transfer therapeutic biomolecules to target cells, using MPs as conveyors. MPs have been reported to successfully mediate the transfer of surface molecules, such as receptors and extracellular matrix proteases, and also intracellular components, including proteins, lipids and RNA molecules of different types [1][2][3]27].
In the present study, the cellular model for MP production and uptake was the Chinese Hamster Ovarian (CHO) cell line, used as MP-producer and MP-recipient cells. This choice was based on the following considerations. (i) CHO cells did not express orthologs of the human CAR, CD46 and CFTR glycoproteins, which made possible the assessment of the capability of MPs to transfer the CAR-, CD46-and CFTR-specified function(s) to these target cells. (ii) Previous experiments of expression of human CFTR, CAR and CD46 in CHO cells have shown the functionality of these proteins and their localisation at the plasma membrane [44,48,63,68]. (iii) A homologous cell system, with the maximal degree of compatibility between MP-donor and MPrecipient cells in terms of membrane composition, would optimize the cell-to-cell transfer of biological material.
Three types of human cell surface molecules with different degrees of complexity were used as prototype bioactive glycoproteins in our study: (i) CAR and (ii) CD46, which are both monospanins and act as high affinity receptors for human adenovirus serotypes 5 and 35, respectively; and (iii) CFTR, a complex transmembrane glycoprotein which functions as a chloride channel. The MP-donor cells were CHO cell lines which constitutively expressed CAR, CD46, or CFTR. MP-target cells were naive CHO cells, which lacked CAR and CD46 and did not express the equivalent of the human CFTR glycoprotein. In all the three cases, the MP-donor cells were grown in the absence of chemical or biological stress, to avoid the occurrence of apoptotic factors within MPs and the possible transfer of stress signals to recipient cells. The populations of MPs issued from CAR-, CD46or CFTR-expressing CHO cells were separated according to their size by velocity ultracentrifugation, yielding MP 30 and MP 100 fractions.
We found that it was possible to transfer CAR and CD46 to target CHO cells using MPs as vehicles, resulting in the acquisition of new biological functions by the target cells. CHO cells incubated with MP 30 CAR and MP 30 CD46 became permissive to HAdV5 and AdV5F35 infection, respectively. In addition, cells incubated with MP 30 CD46 acquired resistance to complement C3-induced apoptosis. The MP-mediated transfer of CD46 might have potential therapeutic applications in the control of allograft rejection, as well as in gene therapy and/or vaccinology using HAdV35-based vectors. The MP-mediated transfer of CAR might be used to confer viral permissiveness to adenovirus-refractory cells, such as tumor cells towards HAdV5-based oncolytic vectors. The viral glycoprotein VSV-G, which has a high fusiogenic activity, has been advantageously used to augment the efficacy of membrane fusion between virus (or VLP) and target cells [35,[37][38][39][69][70][71]. In the present study however, coincorporation of VSV-G at the surface of MP 30 CAR or MP 30 CD46 did not improve the transfer efficiency of CAR or CD46 molecules to target cells.
In the case of CFTR, MPs were isolated from the culture medium of CHO cells which were engineered to constitutively express GFP-CFTR. MP 30 GFP-CFTR and MP 100 GFP-CFTR obtained by velocity ultracentrifugation were further fractionated into three density subclasses: MPs of low, intermediate and high density, respectively. All MP subclasses incorporated significant levels of GFP-CFTR protein as well as mRNA gfp-cftr , with no significant variations between the different subclasses. Likewise, all MP subclasses were competent in transferring GFP-CFTR to target CHO cells. The kinetics of appearance of the GFP signal in these cells showed two separate peaks, which suggested an early (4-6 h) and late phase (3 days) of protein transfer after MP-cell interaction. The early GFP signal detected in target cells likely resulted from the transfer of GFP-CFTR protein via different mechanisms described in the literature [1][2][3][4], such as the direct fusion between MP membrane and target cell plasma membrane, endocytosis of the MPs and recycling of GFP-CFTR to the cell surface, or the coexistence of both mechanisms. Interestingly, the incorrect insertion and orientation of a protein in the plasma membrane of recipient cells have been reported [72]. In the present study, the GFP-CFTR was inserted in the plasma membrane of target cells in the correct in-out orientation, as validated by an antibody recognising an extracellular loop of the CFTR.
At the late phase after MP transfer, almost 80% of the target cells became positive for GFP-CFTR, due to GFP-CFTR molecules issued from de novo protein synthesis directed by exogenous mRNA gfp-cftr (free or/and polysomal) transferred over by MP 30 or MP 100 . Attempts to block the protein synthesis machinery in MP-recipient CHO cells using cycloheximide (CH) were performed to further explore this mechanism. However, the results were inconclusive, due to the unexpected high cytotoxicity of CH towards GFP-CFTR-expressing CHO cells, when treated with the protein synthesis inhibitory doses usually applied to human cells (10 to 20 mg/mL; [73]). Of note, the half-life of the CFTR gene products in human cells has been determined to be ,48 h for the CFTR glycoprotein [74,75], and $20 h for the mRNA cftr [76].
MPs have been shown to carry and transfer mRNAs and microRNAs to target cells, as a general mechanism of intercellular communication [3,5,7,11,27,29]. MP-delivered RNA has been termed shuttle RNA ( [11]), and has been found to be functional in the new cellular context [7,29,[77][78][79][80]. Consistent with the results of these previous studies, our experimental data showed that our shuttling mRNA gfp-cftr was functional, and that neosynthesized GFP-CFTR glycoproteins were addressed to the cell surface of target cells, correctly inserted into their plasma membrane, and metabolically active in anion transport. Using a fluorescent voltage sensitive DISBAC 3 (2) probe assay, the CFTR-associated chloride channel activity was detected in MP-recipient cells at the late phase (day-3), when the translational machinery from the exogenous mRNA gfp-cftr became fully operational. The chloride channel function was not detectable in MP-recipient cells at early times, possibly due to a threshold in the detection of this activity using the fluorescent probe.
Our observation that MP 30 and MP 100 produced by CHO-GFP-CFTR donor cells were equally capable of transferring mRNA gfp-cftr to target cells implied that both populations were competent for CFTR delivery. Fractionation of MP 30 or MP 100 by isopycnic ultracentrifugation did not result in any significant enrichment of a particular subclass in mRNA gfp-cftr molecules. However, in future studies using MPs issued from donor cells of human origin, fractionation of MPs specifically carrying mRNA cftr using MP sorting based on surface markers available in human cells, would be advantageous for a higher efficiency of CFTR delivery. Cell-derived MP 30 or MP 100 appeared therefore as promising biological vehicles of therapeutic mRNAs exerting their therapeutic function(s) indirectly, via their encoded proteins. In particular, in the case of cystic fibrosis, MPs have a potential application in nongenic, cell-to-cell autologous transfer of human mRNA cftr and the restoration of the normal chloride channel function in CFTR-deficient cells.

Cells and Plasmids
HEK-293 cells, obtained from the ATCC (Manassas, VA), were maintained as monolayers in Dulbecco's modified Eagle's medium (DMEM, Gibco-Invitrogen) supplemented with 10% fetal bovine serum (FBS, Gibco-Invitrogen), penicillin (100 U/ml) and streptomycin (100 mg/ml) at 37uC and 5% CO 2 . Chinese hamster ovarian cells (CHO-K1 or simply CHO) was obtained from the ATCC, CAR-expressing CHO cells (CHO-CAR) from Dr. J. Bergelson [44], and CD46-expressing CHO cells (CHO-CD46) from Dr. D. Gerlier [63].They were grown in Iscove's medium supplemented with 10% FBS and 50 mg/ml gentamicin (Invitrogen). To generate the CHO cell line stably expressing GFP-CFTR, the GFP-CFTR gene was excised from the plasmid pGFP-CFTRwt described in a previous study [55], using Nhe I and Sma I digestion, and inserted into pCEP 4 (Invitrogen, Life Technologies). This plasmid vector was designed for high-level, constitutive expression from the CMV promoter, and which contains the EBNA-1 gene for episomal maintenance in human cell lines. Sfi Ilinearized and blunted pCEP4 was further digested with Nhe I. After purification by agarose gel electrophoresis, the Nhe I-cut pCEP4 was ligated with the GFP-CFTR-containing Nhe I-Sma I DNA fragment, generating the pCEP 4 -GFP-CFTR vector. CHO cell monolayers were transfected with pCEP 4 -GFP-CFTR, using Lipofectamine-2000 (Invitrogen), according to the manufacturer's instructions. CHO cells harboring pCEP 4 -GFP-CFTR (referred to as CHO-GFP-CFTR) were selected using culture medium alpha-MEM with ribonucleosides (Invitrogen), supplemented with 5% FBS, penicillin (100 U/ml), streptomycin (100 mg/ml) and hygromycin-B (50 mg/mL). The episomal plasmid pCEP 4 -GFP-CFTR was found to be maintained in transfected cells for at least six months under hygromycin-B selection pressure. . This was carried out by using a previously published centrifugation protocol [81] with the following modifications. Cells were seeded at 70-80% confluence, grown for 48 h to reach confluence (10 7 cells), then maintained for an additional 24 h. Culture medium was clarified from cell debris by centrifugation for 2 min at 13,0006g and 4uC. This clarified supernatant (S0) was the source of MPs. S0 was centrifuged at 30,0006g and 4uC for 2 h. The pellet P1 was kept, and referred to as MP 30 . The supernatant (S1) was centrifuged at 100,0006g and 4uC for 2 h, to obtain pellet P2, referred to as MP 100 . Pellets P1 and P2 were resuspended in 200 ml phosphate buffered saline (PBS) with gentle mixing, for further analysis by flow cytometry, and stored at 4uC before use for MP-mediated protein transfer, or further MP fractionation.

Isolation of MPs
(ii) Isopycnic ultracentrifugation. MP 30 and MP 100 were further fractionated by ultracentrifugation of flotation in isopycnic gradient, to separate the MP 30 and MP 100 fractions into subclasses differing by their apparent density [67]. Samples of MP 30 or MP 100 in PBS were placed at the bottom of a preformed linear sucrose-D 2 O gradient (10 ml total volume; 0.25 to 2.5 M sucrose). The 2.5 M sucrose solution was made in D 2 O buffered to pH 7.2 with NaOH, and the 0.25 M sucrose solution was made in 10 mM Tris-HCl, pH 7.2, 150 mM NaCl, 5.7 mM Na 2 EDTA. The gradients were centrifuged for 18 h at 100,0006g in a Beckman SW41 rotor. Fractions of 0.4 ml were collected from the top, and density measured by weighing 100 ml-aliquots, using a precision scale. Fractions were pooled according to their apparent density, and three pools were constituted : MPs of low density (MP 30 LD and MP 100 LD, respectively) corresponded to pooled fractions 1-7, with densities (r) ranging from 1.06 to 1.14 (mean r : r m = 1.

MP Titration
(i) Titer in total MPs. Quantification of the total MP content in samples resuspended in PBS was determined by flow cytometry, using a FACSCanto TM II cytometer and the DIVA6 software (Becton Dickinson Biosciences). The CountBright TM Absolute Counting Beads kit (Invitrogen Catalog # C-36950) and the Flow Cytometry Size Calibration kit (Nonfluorescent Microspheres; Invitrogen Catalog # F-13838) were used for the calibration of MP number and size, respectively. A total of 20,000 events was acquired for each sample for the calculation of the titer in total MPs.

MP-mediated Transfer of CAR, CD46 or GFP-CFTR into Target Cells
Samples of CHO cell monolayers (5610 5 cells/well) were incubated with 200 ml-aliquots of MP-CAR, MP-CD46 or MP-GFP-CFTR suspensions in PBS mixed with prewarmed serumfree medium, at a constant transducing dose of 5 MPs per cell. After 2 h MP-cell interaction at 37uC, the 200 ml-mix was removed and replaced by 500 ml of prewarmed complete medium. The cells were further incubated for 48 h at 37uC, and assayed for newly acquired functions, i.e. permissiveness to adenovirus, resistance to complement induced apoptosis, GFP signal, or chloride channel activity. MP-transduced cells were taken at 72 h after MP interaction, and incubated with HAdV5-GFP or HAdV5F35-GFP at 500 vp/cell. GFP-expression was monitored in live cells using a Zeiss Axiovert-135 inverted microscope (magnification: 620) equipped with an AxioCam digital camera. Cells were then harvested at 48 h pi, and GFP expression quantitated using flow cytometry.

CD46 Anti-complement Activity
MPs isolated from the culture medium of CHO cells (negative control MP-CHO) or CHO-CD46 (MP-CD46) were incubated with aliquots of recipient CHO cells for 2 h at 37uC and 5% CO 2 , as above described. The culture medium was then removed and replaced by fresh medium containing complement fraction C3 (Sigma-Aldrich) at the concentrations of 1, 10, and 100 mg/mL, and cells further incubated for 48 h. The degree of cell apoptosis was measured by flow cytometry, using the Annexin V-FITC apoptosis detection kit (Sigma-Aldrich) according to the manufacturer's instructions.

PCR Analysis
(i) Real-time RT-PCR quantification of GFP-CFTRencoding mRNA. Total RNA was extracted from MPs released from GFP-CFTR-expressing cells, or from MP-GFP-CFTRtransduced cells, using the Nucleospin RNA II kit (Macherey Nagel). Aliquots of 1 mg RNA was reverse transcribed using the SuperScript TM III First Strand Synthesis SuperMix kit (Invitrogen) and real-time PCR was performed using the LightCycler and a LightCycler DNA Master SYBR Green I kit (Roche). Quantitative real-time PCR was performed using an antisense primer designed from the 59 end of the CFTR gene (nucleotide position 78 in the CFTR gene; 59-GCGCTGTCTGTATCCTTTCCTCAA) and a forward primer designed from the 39 end of the GFP gene (nucleotide position 621; 59-AACGAGAAGCGCGATCA-CATG). The PCR-amplified fragment was 196 nucleotides in length and overlapped the GFP and CFTR junction sequence [66].
(ii) GFP-CFTR DNA. The GFP-CFTR gene carried by the pCEP 4 -GFP-CFTR plasmid vector was detected using the same primers as above, on the host cell genomic DNA substrate obtained using the DNeasy Blood and Tissue kit (Qiagen).

CFTR Channel Function Assayed by Membrane Potentialsensitive Oxonol Probe and Cell Imaging
The fluorescent voltage sensitive probe bis(1,3-dialkylthiobarbituric acid)oligomethine oxonol (DiSBAC 2 (3)) was used as previously described [65]. In brief, cells were loaded for 30 min with 100 nM DiSBAC2(3) in a normal chloride solution containing 136 mM NaCl, 4 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 2.5 mM glucose and 10 mM Hepes (pH 7.4). Cells were then superfused with DiSBAC 2 (3) in a low chloride solution, whereby NaCl was replaced by sodium gluconate, and supplemented with a cAMP cocktail consisted of 200 mM dibutyryl-adenosine 39:59-cyclic monophosphate, 200 mM of 4-chlorophenylthio)-cAMP, 20 mM Forskolin and 50 mM 3-isobutyl-1-methylxanthine. Fluorescent cells were viewed on an inverted TMD300 microscope (Nikon AG, Kürsnacht, Switzerland) equipped with a high-sensitivity black and white CoolSNAP HQ2 CCD camera (Visitron Systems GmbH, Puchheim, Germany). DiSBAC2(3) was excited at 546 nm with a 100-W xenon lamp and the emitted fluorescence was collected through a 580 nm barrier filter. Images were captured every 10 sec, stored and processed using Metafluor version 8.01 software (Universal Imaging Corp., Downington, PA). Regions of interest were delineated for up to 30 cells and changes in the fluorescent signal measured in each region were expressed as the F t /F 0 ratio, in which F t and F 0 were the fluorescence values at the time t and at the time when the cAMP cocktail was added, respectively. The cell-permeable glycinyl hydrazone compound (GlyH-101; Calbiochem) was used as a selective inhibitor of CFTR at 20 mM.

Viral Vectors
(i) Adenoviral vectors. HAdV5-GFP and HAdV5F35-GFP vectors have been described in previous studies [36,55,64]. The capsid of HAdV5F35-GFP consisted of hexon and penton base capsomers of HAdV5, and of chimeric fibers made up of the shaft and knob domains of serotype 35 fiber (F35) fused to HAdV5 fiber tail. Chimeric vector HAdV5F35-GFP-CFTR has been previously described [66]. HAdV5F35-GFP-CFTR encoded the wild-type (wt) allele of the CFTR gene fused to the 39 end of the GFP gene. Vector stocks were produced and titrated on HEK-293 cell monolayers [82].
(ii) Baculoviral vector AcMNPV-VSV-G. The coding sequence for the glycoprotein G of vesicular stomatitis virus (VSV-G) was inserted into the genome of Autographa californica Multi-Capsid NucleoPolyhedrosis Virus (AcMNPV) in the polyhedrosis gene locus, under the control of the CMV immediate-early promoter. The infectious titer was determined by the plaque assay method in Sf9 cells, and expressed as plaque-forming units per mL (pfu/mL). The titer in physical virus particles of AcMPV-VSV-G vector was determined by adsorbance measurement at 260 nm (A 260 ) on 1-mL samples of SDS-denatured virions (0.1% SDS for 1 min at 56uC) in 1-cm pathlength cuvette, using the following formula : A 260 of 1.0 = 0.3610 12 vp/mL, considering the length of 134 kbp for the viral genomic DNA. Infectious titers of stocks of AcMPV-VSV-G concentrated by ultracentrifugation were usually 5610 9 to 1610 10 pfu/mL, and the corresponding physical particle titers ranged between 1610 12 and 5610 12 vp/mL [36].

Electron Microscopy
(i) Negative staining of MPs. Samples were applied to carbon-coated grid and negatively stained with 1% uranyl acetate, pH 7.5. They were examined under a Jeol JEM-1400 electron microscope (EM), equiped with an ORIUS TM digital camera (Gatan France, 78113-Grandchamp).
(ii) Immunogold staining of MPs. Samples of MP suspension (10 ml) were deposited on top of carbon-coated grids. 30 sec later, the excess of liquid was removed by blotting with filter paper. 10 ml of a 50-fold diluted solution of primary antibody (anti-CAR or anti-CD46) was placed on the grid and incubated for 2 min at room temperature. The antibody solution was then removed by filter paper adsorption, and replaced by 10 ml Tris-buffered saline (TBS). After three steps of rinsing with TBS, grids were postincubated with 10-nm colloidal gold-conjugated goat anti-mouse IgG antibody (British BioCell International, Cardiff, UK; diluted to 1:50 in TBS) for 2 min at room temperature. The secondary antibody solution was then removed by filter paper adsorption, and replaced by 10 ml of stain (2% uranyl acetate, pH 7.4). After a further 1 min, the grid was dried on filter paper, and examined under the EM as above.
(iii) EM analysis of MP-adenovirus complexes and cell sections. Samples of MPs incubated with adenoviral vectors, with or without postincubation with target CHO cells, were fixed with 2% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4), pelleted, and post-fixed with osmium tetroxide (1% in 0.1 M cacodylate buffer, pH 7.4). The specimens were dehydrated and embedded in Epon resin and sectioned. Sections were stained with 7% uranyl acetate in methanol, post-stained with 2.6% alkaline lead citrate in H2O, and examined under the EM as above.

Statistics
Results were expressed as mean 6 SEM of n observations. Sets of data were compared with an analysis of variance (ANOVA) or a Student's t test. Differences were considered statistically significant when P,0.05. Symbols used in figures were (*) for P,0.05, (**) for P,0.01, (***) for P,0.001, and ns for no significant difference, respectively. All statistical tests were performed using GraphPad Prism version 4.0 for Windows (Graphpad Software). Figure S1 Schematic procedure for MP isolation and fractionation.

Supporting Information
(TIFF) Figure S2 Influence of VSV-G-pseudotyping on the efficiency on the efficiency of MP-mediated transfer of CAR and CD46. (a), The effect of the fusiogenic VSV-G glycoprotein on the efficiency of CAR transfer was evaluated by the degree of CHO permissiveness to the HAdV5-GFP vector. Aliquots of CHO cells were incubated with MP 30 CAR at different MP doses per cell, as indicated in the x-axis. At 72 h after MP 30 CAR interaction, cells were infected with HAdV5-GFP vector. Adenoviral vector HAdV5F35-GFP, which does not recognize CAR as cellular receptor, was used as the negative control. (b), The effect of VSV-G on the efficiency of CD46 transfer was evaluated by the degree of CHO permissiveness to the HAdV5F35-GFP vector. Aliquots of CHO cells were incubated with MP 30 CD46 at different MP doses per cell, as indicated in the x-axis. At 72 h after MP 30 CD46 interaction, cells were infected with HAdV5F35-GFP vector. Adenoviral vector HAdV5-GFP, which does not recognize CD46 as cellular receptor, was used as the negative control. In (a) and (b), MP 30 lacking CAR or CD46 and carrying VSV-G alone were used as negative controls for CAR and CD46 receptor activity, respectively.