A fiber-modified adenoviral vector interacts with immunoevasion molecules of the B7 family at the surface of murine leukemia cells derived from dormant tumors

Tumor cells can escape the immune system by overexpressing molecules of the B7 family, e.g. B7-H1 (PD-L1 or CD86), which suppresses the anti-tumor T-cell responses through binding to the PD-1 receptor, and similarly for B7.1 (CD80), through binding to CTLA-4. Moreover, direct interactions between B7-H1 and B7.1 molecules are also likely to participate in the immunoevasion mechanism. In this study, we used a mouse model of tumor dormancy, DA1-3b leukemia cells. We previously showed that a minor population of DA1-3b cells persists in equilibrium with the immune system for long periods of time, and that the levels of surface expression of B7-H1 and B7.1 molecules correlates with the dormancy time. We found that leukemia cells DA1-3b/d365 cells, which derived from long-term dormant tumors and overexpressed B7-H1 and B7.1 molecules, were highly permissive to Ad5FB4, a human adenovirus serotype 5 (Ad5) vector pseudotyped with chimeric human-bovine fibers. Both B7-H1 and B7.1 were required for Ad5FB4-cell binding and entry, since (i) siRNA silencing of one or the other B7 gene transcript resulted in a net decrease in the cell binding and Ad5FB4-mediated transduction of DA1-3b/d365; and (ii) plasmid-directed expression of B7.1 and B7-H1 proteins conferred to Ad5FB4-refractory human cells a full permissiveness to this vector. Binding data and flow cytometry analysis suggested that B7.1 and B7-H1 molecules played different roles in Ad5FB4-mediated transduction of DA1-3b/d365, with B7.1 involved in cell attachment of Ad5FB4, and B7-H1 in Ad5FB4 internalization. BRET analysis showed that B7.1 and B7-H1 formed heterodimeric complexes at the cell surface, and that Ad5FB4 penton, the viral capsomere carrying the fiber projection, could negatively interfere with the formation of B7.1/B7-H1 heterodimers, or modify their conformation. As interactors of B7-H1/B7.1 molecules, Ad5FB4 particles and/or their penton capsomeres represent potential therapeutic agents targeting cancer cells that had developed immunoevasion mechanisms.


Background
Tumor cells express numerous molecules at their surface that may influence their recognition by the immune system. Among them, proteins of the B7 family play important roles in the immunoevasion of tumor cells and can suppress T-cell-mediated immunity by binding to the inhibitory receptor CTLA-4, e.g. B7.1 (or CD80) and B7.2 (or CD86). Tumor cells that express B7.1 may be shielded from direct cytotoxic T-cell (CTL)-mediated killing [1][2][3]. Other members of the B7 family include B7-H1 (PD-L1 or CD274), B7-DC (PD-L2), ICOS-L, B7-H3 and B7-H4, but only B7-H1 and B7-H4 have been indisputably shown to play a role in the immunoevasion of cancer cells [4]. B7-H1 binds to its receptor PD-1, and this binding mediates immunosuppression [5]. B7-H1 also binds to B7.1 [6], but the function of this interaction remains unclear. B7-H1 suppresses the CTLmediated killing of tumor cells, induces T-cell anergy and likely participates in T-cell exhaustion in cancer, as PD-1 is abundantly expressed on T-cells that infiltrate the tumor microenvironment. B7-H1 is constitutively expressed by several human tumors, and is induced when cancer cells are stimulated with interferon-γIFN-γ) and ligands of Toll-like receptors (TLR) [7][8][9]. Using a DA1-3b mouse model of tumor dormancy, we previously demonstrated that a minor population of dormant leukemia cells persists in equilibrium with the immune system for long periods of time. Dormant leukemia cells suppressed CTL-mediated killing by overexpressing B7-H1 and B7.1 [10][11][12]. All these observations suggested that the B7-H1 and B7.1 molecules of the B7family could represent potential targets for new antitumor strategies (reviewed in [13]).
Cell surface molecules in cancer cells have been considered as privileged targets in cancer therapy, but mostly as targets of therapeutic monoclonal antibodies (mAb) [14]. Alternative therapeutic methods include the use of oncolytic viral vectors naturally directed, or genetically retargeted to specific molecules of the cell surface, capable of triggering tumor cell death. Recombinant oncolytic adenoviruses offer several advantages over other oncolytic viral vectors: (i) they have a large cloning capacity, (ii) are relatively easy to produce to high titers, with vector stocks remaining stable over a long period of storage, and (iii) their therapeutic effects do not require the viral DNA insertion into the host genome [15][16][17][18]. However, with the exception of certain members of species B adenoviruses, e.g. HAdV3, which have the natural ability to bind to B7.1 and B7.2 [19] and to efficiently transduce B7.1-and B7.2-expressing malignant glioma cells [20], the usage of adenoviruses in cancer gene therapy is limited, due to the low level (or absence) of expression of high affinity receptor for adenoviruses in cancer cells, or/and their poor accessibility at the cell surface. This is the case for the Coxsackie and Adenovirus human Receptor (hCAR), one of the natural receptors for adenoviral species A, C, D, E and F, which is located in the tight junctions and expressed at low levels in cancer cells [21], and for desmoglein-2, the receptor of HAdV3, also found in tight junctions [22]. Different strategies of adenoviral vectors have been proposed, and the most popular consisted of fiber capsomere modifications, to allow the vectors to attach to newly defined cell targets for efficient virus entry [15,17,18].
A fiber-modified, ß-galactosidase (ß-gal)-expressing adenoviral vector, originally called HAdV5-F2/BAdV4ßgal and abbreviated Ad5FB4 in the present study, was previously constructed and characterized. Ad5FB4 is a human adenovirus serotype 5 which carries chimeric, human/bovine fibers [23][24][25][26][27]. It does not recognize the ubiquitous hCAR, and binds to cells via an attachment receptor different from the heparan sulfate proteoglycans [25]. Transduction of hCAR-negative and hCARpositive cells occurs via a clathrin-independent endocytic pathway involving lipid raft/caveolae [27]. An advantage of the altered tropism of the Ad5FB4 vector is the restriction of its infection repertoire of cells, which therefore limits the vector dissemination [24]. This results in the reduction of adenovirus-associated humoral and innate cytokine immune responses upon intravenous administration of Ad5FB4 vector to mice [26].
In the present study, we tested Ad5FB4 on malignant cells refractory to conventional Ad5-based vectors, and found that the permissiveness of murine leukemia cells to Ad5FB4 correlated with their dormancy time and the expression level of B7.1 and B7-H1 molecules at their surface. Results from in vitro and in vivo experiments suggested that B7.1 and B7-H1 molecules played different roles in Ad5FB4-mediated transduction of murine dormant leukemia cells, with B7.1 involved in cell attachment of Ad5FB4, and B7-H1 in its cellular uptake. Our data also suggested that the interaction between B7.1 and Ad5FB4 was mediated by the penton capsomeres (or penton base-linked fibers) of the vector capsid.
In situ BRET analysis showed that B7.1 interacted with B7-H1 to form heterodimers at the cell surface, and that Ad5FB4 penton capsomeres interfered negatively with the formation of these complexes. Our finding that tumor cell surface molecules of the B7 family implicated in immunoevasion mechanisms were recognised by the adenoviral vector Ad5FB4 offered novel opportunities for cancer therapy, using intrinsically B7-targeted Ad5FB4 vectors for therapeutic gene transfer. Alternatively, Ad5FB4 penton capsomeres, via their negative interference with the B7-H1/B7.1 heterodimer formation, might be used as therapeutic agents to decrease the amounts of these complexes at the tumor cell surface, and hence lower their capacity to resist to antitumor T-cell responses.

Plasmids
Mouse B7-H1 cDNA was kindly provided by M. Azuma [28], and pSelect-B7.1 was purchased from Invivogen (San Diego, CA). For fusion constructs, the stop codons of the murine coding sequences of B7-H1 and B7.1 were removed from the plasmids. PCR products were cloned in phase with either Rluc8 or YPet into the pcDNA3 vector [29].
Cell sorting DA1-3b cells (10 7 ) were suspended in 3 mL of ice-cold PBS containing 1% BSA and 2 mM EDTA, and incubated with monoclonal PE-labeled antibody against B7-H1 or control, irrelevant isotopic antibody, as described above. DA1-3b cells expressing B7-H1 at low and high levels, respectively, were sorted using an Epics Altra Coulter cell sorter.

Adenovirus vector amplification and labeling
The genetic constructions of the E1-deleted adenoviral vectors Ad5 and Ad5FB4 containing the ß-gal reporter gene have been described previously [23,24]. In the chimeric Ad5FB4 fiber, the junction between human serotype 2 fiber (F2) and bovine serotype 4 (BAdV4) fiber was situated in the shaft repeat 7 at the GKL (glycinelysine-leucine) motif, to generate the chimeric fiber F2/ BAdV4 [24], abbreviated FB4 in the present study. The Ad5FB4 and Ad5 vectors were amplified and purified following conventional protocols. Since Ad5FB4 had a lower tropism for epithelial cells, compared to Ad5, it was not possible to compare their infectious titers by conventional plaque assays on HEK-293 cell monolayers. Stocks of purified vectors were titrated by optical measurement of the viral DNA concentration at 260 nm, and the vector titer expressed as vp/mL. Fluorescent labeling of vector particles with carboxyfluorescein succinimidyl ester (FAM; Invitrogen, Cergy-Pontoise, France) was performed as previously described [27].

Adenovirus infection
Ad5FB4 and Ad5 infections were carried out as previously described [27], except for the virus inoculum which was eliminated by low-speed centrifugation of the infected cells (800 × g, 5 min) at 24 h post-infection (pi). Cells were then resuspended in culture medium with 4% FBS and maintained for an extra 48 h for murine cells, or an extra 24 h for human cells. The ß-gal activity was determined using fluorescein-ß-D-galactopyranoside or the colorimetric X-gal staining procedure (Fisher scientific, Belgium), as previously described [24].

Vector-cell binding and internalization
Cell aliquots (1.5 × 10 6 ) were incubated at 4°C for 1 h in suspension with FAM-labeled Ad5 or FAM-labeled Ad5FB4 at 10 11 vp/mL in serum-free medium. Cells were rinsed with PBS containing 1% BSA (PBS-BSA), and cell-bound vector particles were quantitated using flow cytometry (FACS). For internalization assays, cells and vector were incubated at 4°C for 1 h, then transferred to 37°C and further incubated at this temperature for different periods of time, ranging from 5 min to 2 h.
Before FACS analysis, cell samples were incubated for 15 min at 37°C with trypsin at 0.25% in 1 mM EDTA to detach vector particles possibly sequestered at the cell surface [30]. Cells were resuspended in PBS-BSA, and the amounts of internalized vector were quantitated by FACS analysis.

Surface plasmon resonance (SPR)
SPR analyses were carried out using a BIAcore 2000. Recombinant mouse B7-H1, B7.1 and PD-1 were covalently immobilized onto separate flow cells of a CM5 biosensor chip by amine coupling according to the manufacturer's instructions. As a reference, another flow-cell surface was activated and deactivated. Protein samples were diluted in HBS (0.01 M Hepes, pH 7.4; 0.15 M NaCl; 0.005% P20), and the binding analyses were performed at 25°C with HBS as running buffer. A flow rate of 10 μL/min was used to inject B7-H1, B7-H2, B7.1 (R&D Systems Europe, Lille, France) and 20 μL/min to inject viral proteins; all samples were run five times. For vector particles analysis, SPR experiments (BIAcore 3000) were run on a CM4 sensorship at 5 μL/min using HBS-N (GE-Healthcare) supplemented with 2 mM CaCl 2. Immobilisation of both B7.1 and B7-H1 was performed by interaction of these ligands diluted at 1 μg/ mL in 10 mM sodium acetate buffer pH 4.2 on EDC-NHS activated flow-cells for 10 min at room temperature. After ethanolamine deactivation, vector particles were injected (1.10 10 vp in 25 μL of running buffer) and the signal from the ligand flow-cells was automatically subtracted from the background of an ethanolamine deactivated EDC-NHS flow-cell.
BRET At 24 h before transfection, cells (2 × 10 5 -aliquots) were plated in 6-well plates and transfected with increasing amounts of B7.1-YPet-, B7-H1-YPet-or IR-YPet-expressing plasmids (10 to 500 ng/well), and constant amounts (10 ng-aliquots) of plasmid expressing B7-H1-Rluc8 or B7.1-Rluc8 fusion protein. 48 h later, cells were collected and washed twice with PBS, and aliquots were placed in 384-well plates. Coelenterazine H substrate was added at a final molarity of 5 μM, and BRET was measured immediately. To analyze the effect of the penton, cells were incubated for 5 min without or with penton protein solution at 0.33, 0.66 or 132 ng/μL, followed by Coelenterazine H addition and BRET measurement. BRET was monitored using a lumino/fluorometer (Mithras; Berthold Technologies, France), allowing for the sequential integration of luminescence with two filter settings (Rluc filter, 485 ± 10 nm; YFP filter, 530 ± 12.5 nm). The emission signal values obtained at 530 nm were divided by the emission signal values obtained at 485 nm. The BRET ratio was calculated by dividing the emission signal value obtained with coexpressed donor and acceptor by that obtained with the donor protein expressed alone. Data from at least three independent experiments were averaged, and results expressed as milliBRET (mBRET), corresponding to the BRET ratio multiplied by 1,000. Donor saturation curves were determined as previously described [35,36].

Statistical analyses
Data were presented as the mean of triplicate experiments (m ± SEM), and were representative of the results obtained from three independent experiments that produced similar results. Statistical analyses were performed using the Mann-Whitney test.

Cell and vector nomenclature
Murine dormant leukemia cell lines have been previously established from long-term persistent tumor cells isolated from mice in a state of tumor dormancy [12,13]. In brief, C3H/HeJ mice have been immunized with irradiated, interleukin-12-treated or CD154-transduced DA1-3b cells, challenged with parental DA1-3b cells, and randomly sacrificed during a one-year follow-up period at day 35, 90 and 365, respectively. Dormant leukemia cells were collected from the spleen at these different times, and the amount of BCR/ABL mRNA quantitatively assayed using real-time PCR, leading to the following cell lines: DA1-3b/d35, DA1-3b/d90 and DA1-3b/d365 [12]. The ß-gal-expressing, fiber-modified adenoviral vector HAdV5-F2/BAdV4-ß-gal comprises of the human Ad5 DNA backbone containing a chimeric fiber gene. The vector particle consisted of serotype 5 capsid carrying chimeric fibers, each formed by the human serotype 2 fiber tail fused to the shaft and knob domains of bovine serotype 4 (BAdV4) [23][24][25][26][27]. For reasons of simplification, this human/bovine chimeric fiber vector was referred to as Ad5FB4 in the present study, and the control, ß-gal-expressing Ad5 vector with homotypic serotype 5 capsid proteins was abbreviated Ad5.

Permissiveness of dormant leukemia cells to the chimeric fiber vector Ad5FB4
The three cell lines DA1-3b/d35, DA1-3b/d90 and DA1-3b/d365 were incubated with equal physical particle inputs of the chimeric vector Ad5FB4 or control vector Ad5. The efficiency of cell transduction was assayed by the percentage of ß-gal-positive cells and the level of ßgal expression, determined by the mean fluorescence intensity (MFI) using a fluorescent ß-gal substrate. We found that Ad5FB4 transduced murine dormant leukemia cells with a higher efficiency, compared to Ad5 vector ( Figure 1A). Interestingly, the percentage of ß-galpositive cells was not not significantly different for the various DA1-3b cell lines derived from in vivo passages. Rather, the level of Ad5FB4-mediated transgene expression correlated with increased dormancy: the lowest ßgal activity was observed in DA1-3b/d35, the highest in DA1-3b/d365 cells, with an intermediate value in DA1-3b/d90 cells ( Figure 1B). Of note, Ad5 transduction, as determined by the percentage of transduced cells, also correlated with the length of in vivo passage ( Figure  1A). Interestingly, the transgene expression progressively increased with the period of time after transduction, as exemplified with Ad5FB4-transduced DA1-3b/d365 cells, which showed a 30-fold enhancement of ß-gal activity between 72 and 120 h posttransduction ( Figure  1C). These results suggested that the DA1-3b/d365 cells provided a more favorable environment for the expression of the transgene transduced by Ad5FB4 or Ad5. However, the possibility of an increased expression of mouse CAR (mCAR) in the different cell lines, and/or modifications of intracellular factors, e.g. transcription factors, was envisaged. All cell lines were found to express mCAR at their surface at various levels, as assayed by flow cytometry (Figure 2A, B). The pattern of mean fluorescence intensity in the different lines roughly paralleled that of the percentage of mCAR-positive cells, and both data clearly showed no direct correlation between mCAR levels, dormancy time and increased Ad5 (or Ad5FB4) transduction.
Correlation between cell surface levels of B7-H1 and B7.1 molecules and cell permissiveness to Ad5FB4 The apparent correlation between the level of Ad5FB4mediated ß-gal expression and the dormancy time, as shown in Figure 1B, suggested that the degree of permissiveness of DA1-3b cells to Ad5FB4 depended on the level of expression of B7-H1 and B7.1 molecules at the cell surface. To test this hypothesis, we analyzed the permissiveness to Ad5FB4 of various mouse and human cell lines differing by their absolute and relative levels of surface-expressed B7-H1 and B7.1 ( Figure 3A). We found that the Ad5FB4-mediated cell transduction required both B7.1 and B7-H1 molecules, but the transduction efficiency seemed to correlate with the B7-H1 levels ( Figure 3B).
Since intrinsic properties of individual cell lines could influence the cell transduction levels independently of The two subclones isolated, referred to as DA1-3bLow and DA1-3bHigh, respectively ( Figure 3A), were assayed for ß-gal activity after Ad5FB4 transduction. The results indicated that the subpopulation of high B7-H1-expressors was transduced with a 7-fold higher efficiency, compared to the subpopulation of low B7-H1-expressors ( Figure 3B). This confirmed that the transduction efficiency by Ad5FB4 correlated with the level of expression of the B7-H1 protein and the ratio of B7-H1 to B7.1. Interestingly, the fact that both the original DA1-3b and the DA1-3bLow subclone were low expressors of B7-H1 and B7.1 but exhibited marked differences in ß-gal expression ( Figure 3B) suggested the existence of a threshold of B7-H1 at which the influence on ß-gal expression became apparent. This threshold was consistent with a cooperative effect between B7-H1 and B7.1 for Ad5FB4 uptake, with B7-H1 being the limiting factor: a certain concentration of B7-H1 molecules would be required to form B7.1-B7-H1 complexes at the cell surface and mediate the cell entry of the vector, as described below.
A similar correlation between permissiveness to Ad5FB4 and B7.1/B7-H1 pattern was observed for human cell lines Raji, Jurkat and A549 ( Figure 3A, B). Unfortunately, no control data could be provided using mouse cell lines expressing only one or the other B7 molecule, since tumor dormancy was consistently associated with the expression of both B7 products at the cell surface. In the case of human cells however, we found no cell line that coexpresses B7-H1 and B7.1. This was the reason why indirect methods, e.g. specific RNA interference and single or double expression of B7 molecules via plasmid transfection, were used (Cf. below).
Respective roles of B7.1 and B7-H1 molecules in the cellular uptake of Ad5FB4 To determine whether B7.1 or/and B7-H1 could be used by Ad5FB4 as receptors for attachment and/or entry into murine cells   Figure 4A). DA1-3b/d365 cells electroporated with siRNA were then transferred to 4°C, and incubated with aliquots of FAM-labeled, fluorescent Ad5FB4 particles for 90 min at 4°C. Cell-bound Ad5FB4 particles were quantitated by flow cytometry analysis of the fluorescent signal. A significant reduction (ca. 50%) in cell binding was observed in B7.1-siRNA-treated cells, compared to control siRNA-treated cells ( Figure 4B). This suggested that B7.1 played the role of attachment receptor for Ad5FB4 at the surface of DA1-3b/d365 cells.
(ii) B7-H1 knockdown and cellular internalization of Ad5FB4 Specific siRNA was also used to decrease the surface expression of B7-H1 ( Figure 5A). After B7-H1 silencing, there was no detectable change in the binding of FAMlabeled Ad5FB4 to DA1-3b/d365 cells at low temperature ( Figure 5B). For cell internalization assays, fluorescent-labeled Ad5FB4 particles were incubated with DA1-3b/d365 cells at 4°C to allow for vector-cell attachment, then samples transferred to 37°C. Vector particles remaining trapped at the cell surface were removed by digestion with trypsin [30], and the amounts of fluorescent-labeled Ad5FB4 particles internalized by DA1-3b/ d365 cells were measured by flow cytometry at different times posttransfer to 37°C (5 to 120 min). The cell transduction efficiency, measured by the ß-gal activity, was also determined at vector doses, ranging from 500 to 10,000 vp/cell. The cellular internalization of Ad5FB4 was significantly reduced in B7-H1-silenced DA1-3b/ d365 cells at all time points, with a maximum 40% inhibition at 60 min ( Figure 5C). Likewise, the transduction efficiency significantly decreased after B7-H1 knockdown at all vector doses used: a maximum 50% inhibition was observed at 1,000 to 2,000 vp/cell, but the inhibitory effect almost plateaued at higher vector doses ( Figure 5D).
Interestingly, when Ad5FB4 particles were preincubated with the B7.1-Fc protein prior to injection onto surface-immobilized B7-H1-Fc, no extra signal over the basic signal of B7.1-B7-H1 interaction was detected on the sensorgrams (data not shown), suggesting that the binding of Ad5FB4 to B7.1 did not increase the affinity of B7.1 to B7-H1. The next experiments were designed to further explore the respective roles of B7.1 and B7-H1 in the cell binding and entry of Ad5FB4 in the context of the DA1-3b/d365 plasma membrane.
Cellular internalization of B7-H1 molecules upon Ad5FB4 uptake DA1-3b/d365 cells in suspension were incubated with Ad5FB4 vector particles for 90 min at 4°C, a temperature which allows cell attachment of the vector but blocks its endocytosis. Cells were then transferred to 37°C to induce the vector internalization, and the status of B7.1 and B7-H1 molecules at the cell surface examined by flow cytometry at 10-15 min after transfer. No modification of the B7.1 signal was detected upon Ad5FB4 endocytosis, even at high vector doses (10,000 vp/cell). By contrast, a discrete but significant decrease was observed in the levels of B7-H1 protein at the cell surface upon Ad5FB4 uptake, and in a vector dose-dependent manner: 15-17% at 5,000 vp/cell, and 22-25% at 10,000 vp/cell (Figure 8). However, the possibility existed that cell-bound vector particles masked the epitope of the B7-H1 molecules recognized by the specific antibody used in flow cytometry, and biased the results. To address this issue, cells were analyzed by flow cytometry immediately after incubation with Ad5FB4 at 4°C. The immunoreactivity of B7 molecules was found to be similar at the surface of control cells without Ad5FB4 and cells incubated with Ad5FB4 at 4°C (Figure 8 compare open bars and grey bars). The data clearly showed that the B7-H1 epitope was still accessible after Ad5FB4-cell binding, and suggested a cointernalization of B7-H1 molecules with Ad5FB4 particles. This supported the above-mentioned hypothesis that B7-H1 would be involved in Ad5FB4 endocytosis. In order to further dissect the mechanism of cellular entry of the Ad5FB4 vector and the contribution of the B7 molecules in this pathway, interactions between B7.1, B7-H1 and the adenoviral penton capsomeres were analyzed in situ within the context of live cell plasma membrane, using a recently developed enhancement of the bioluminescence resonance energy transfer (BRET) technique [29].
(iii) Effect of Ad5FB4 on B7.1-B7-H1 interaction We then applied the BRET analysis to explore the molecular events implicated in, or resulting from, the binding of Ad5FB4 to B7-H1/B7.1-expressing cells. When HeLa cells coexpressing donor B7.1-Rluc8 and acceptor B7-H1-YPet, as in the configuration of Figure 9A, were incubated with increasing doses of Ad5FB4 vector particles, a modest effect in BRET signals was observed at the highest dose of 10,000 vp/cell (not shown). We then used Ad5FB4 penton protein, the components of the adenoviral capsid responsible for cell attachment and endocytosis [17,18]. The rationale for the use of penton capsomeres instead of vector particles in BRET analysis was based on the following arguments: (i) like adenovirus particles, penton capsomeres are capable of cell attachment and entry [43][44][45][46][47]; (ii) in terms of macromolecules, isolated capsomeres have a higher solubility and dispersity compared to virus particles, of which suspensions are prone to aggregate; (iii) there are 12 pentons per adenoviral capsid, which represent ca. 5% of the total protein content of the virion. Considering that 3.4 × 10 12 adenovirus particles correspond to 1 mg protein [48], a solution of penton protein at the concentration of 132 ng/μL (the maximum concentration used in our dose-dependent curves) would correspond to a theoretical number of 9 × 10 9 vp/μL, viz. as many as 45,000 vp/ cell in our standard BRET assays. Such high doses } * * might provoke nondesired cytotoxic effects interfering with the metabolism of the B7 proteins, and justified the use of penton protein as opposed to vector particles. HeLa cells coexpressing donor B7.1-Rluc8 and acceptor B7-H1-YPet were incubated with increasing doses of Ad5FB4 penton protein, with Ad5 penton used as the control. Donor saturation curves showed an Ad5FB4 penton-dependent, dose-response decrease of the BRET signal ( Figure 9C), whereas control Ad5 penton did not induce any change in the BRET signal, even at the maximum concentration of 132 ng/μL ( Figure 9D).
The BRET data indicated that less B7.1/B7-H1 heterodimers were present at the cell surface after contact with Ad5FB4 penton, which suggested that the binding of Ad5FB4 penton to B7.1 prevented the formation of the B7.1/B7-H1 heterodimeric complexes, or/and that the complexes dissociated upon Ad5FB4 penton interaction. However, we could not exclude another mechanism, which consisted of a conformational change within the B7.1/B7-H1 complex upon Ad5FB4 penton binding. This structural modification might result in (i) a reorientation of the two partner proteins less favorable to generate a BRET signal, or/and (ii) in an increased distance between the donor and acceptor moieties of the two fusion proteins, without a complete dissociation of the complex [36]. Whatever the molecular mechanism, our BRET analysis demonstrated that a significant modification occurred in the B7.1-B7-H1 interaction or/and in their three-dimensional conformation and respective topology within the heterodimeric complex, upon Ad5FB4 penton interaction with the cell surface.

Discussion
The interaction of adenovirus with host cells leading to a productive infection and to viral progeny represents a complex, multifactorial process which depends on both viral and cellular functions. The cell permissiveness to the virus is the result of a fine balance between intrinsic and extrinsic factors, but the efficacy of the primary event of virus-cell attachment is one major parameter which conditions the subsequent steps, and notably the cell entry of the virus and the transcription of its genome. In the present study, we showed that the efficiency of cell transduction by the chimeric fiber-pseudotyped Ad5FB4 vector depended on the coexpression of two molecules of the B7 family, B7-H1 and B7.1, at the surface of cells otherwise refractory or poorly permissive to Ad5FB4. This was observed in both human cells and murine dormant leukemia cells. Ad5FB4 efficiently bound to and transduced DA1-3b/d365 leukemia cells, a murine cell line derived from long-term dormant leukemia cells which overexpressed B7-H1/B7.1 molecules. Our data suggested that B7.1 was involved in Ad5FB4cell attachment, but both B7-H1 and B7.1 were required for cell entry and efficient Ad5FB4-mediated transduction. BRET experiments demonstrated that B7-H1 and B7.1 formed homodimers and also heterodimers at the cell surface, and that B7-H1/B7.1 heterodimeric complex formation was altered upon cell interaction with Ad5FB4 penton capsomeres, the capsid components which are responsible for the steps of vector-cell attachment and entry. A tentative model for the cell attachment and entry pathway of Ad5FB4 is presented in Figure 10.
Under steady-state conditions, B7.1 is present as a heterogenous population of monomers and noncovalent dimers at the cell surface, and the same pattern has been described for B7-H1 [40,41]. However, interaction between B7-H1 and B7.1 molecules has also been observed [41]. B7.1 has a relatively low affinity for B7-H1 with, an intermediate affinity for CD28 and CTLA-4, and a high affinity for PD-1. Using T-cells deficient for different combinations of PD-1, B7.1, CD28 and CTLA-4, Butte et al. found a bidirectional inhibitory interaction between B7-H1 and B7.1 [6], with some overlapping of the binding domains of B7.1 and PD-1 on the B7-H1 molecule [6,9]. In the present study, we demonstrated and confirmed the occurrence of B7-H1/B7.1 heterodimeric complexes in live cells.
Targeting the molecules of the B7 family is potentially a promising strategy for cancer therapy for the following reasons. (i) Dormant tumor cells have developed several in vivo mechanisms to ensure their long-term persistence in their hosts (reviewed in [13]). One of these mechanisms is the overexpression of B7-H1 and B7.1, which shields these cells from CTLs [10][11][12]. This suggested that the equilibrium between minimal residual disease and host immune response could be modified to prevent disease recurrence. (ii) B7-H1 is frequently observed in human cancers and has a prognostic role for renal cell carcinoma [49]. (iii) B7-H1 and B7.1 play a role in immunoevasion through their expression in dendritic cells present in tumor draining lymph nodes. The delivery of various transgenes that may antagonize immunoevasion mechanisms, such as the chemokine CXCL10, which activates NK cells to kill B7-H1-overexpressing dormant leukemia cells [11], could help the host in the clearance of the cancer cells.
In this context, our finding that Ad5FB4 acted as a ligand of B7.1 monomer, and that Ad5FB4 penton negatively interferred with B7.1/B7-H1 heterodimer formation, made Ad5FB4 a unique adenoviral vector for cancer gene therapy, as it targeted cell surface molecules involved in the immunoevasion mechanisms. In the light of our observation in BRET analysis using Ad5FB4 penton protein, one could envisage to antagonize immunoevasion mechanisms by using Ad5FB4 penton as monovalent, single capsomeres, or as multivalent, double chimeric dodecamers (or dodecahedrons) formed by twelve penton base subunits of serotype 3 adenovirus (Ad3Dd) linked to twelve chimeric FB4 fibers  Figure 10 Hypothetical model for the cell entry pathway of Ad5FB4.
Step 4: B7-H1-mediated endocytosis and internalization of Ad5FB4. The data obtained in vitro (SPR) and in vivo (cell binding assays and RNA interference) support the mechanisms proposed for steps 1 and 2; the molecular events depicted in step 3 are based on BRET analysis; the endocytic step 4 is supported by data of flow cytometry.
(Ad3DdFB4). Ad3 dodecahedrons have been used as efficient protein vectors to bind to and enter mammalian cells [43][44][45][46]. It would be expected that Ad3DdFB4 would impair the formation of B7-H1/B7.1 heterodimers in tumor cells, induce their dissociation, or would cointernalize with one or the other molecule. The lower expression of immunoevasion complexes at the cell surface would in turn confer to tumor cells a higher susceptibility to tumor-reactive T-cells. Further characterization of the interaction between chimeric Ad3DdFB4 and B7-H1/B7.1 molecules in vitro and in vivo will be necessary to optimize the conditions for potential applications to cancer therapy.

Conclusions
Tumor cells express specific molecules at their surface which may negatively affect their recognition by the immune system. This is the case for proteins of the B7 family, which play important roles in the immunoevasion of tumor cells. In the present study, we showed that leukemia cells DA1-3b/d365 derived from longterm dormant tumors, which are refractory to conventional adenovirus serotype 5 (Ad5)-based vectors, were permissive to Ad5FB4, an adenoviral vector carrying chimeric fibers. We found that the permissiveness of DA1-3b/d365 cells to Ad5FB4 correlated with the level of expression of B7.1 and B7-H1 molecules at their surface, and that the permissivity to Ad5FB4 could be reverted by RNA silencing of one or the other B7 gene transcript. Results from in vitro and in vivo experiments suggested that B7.1 and B7-H1 molecules played different roles in Ad5FB4-mediated transduction of DA1-3b/d365 cells, with B7.1 involved in Ad5FB4-cell attachment, and B7-H1 in Ad5FB4 internalization. We showed that the interaction between B7.1 and Ad5FB4 was mediated by the penton protein, the capsid component carrying the fiber projection. In situ BRET analysis showed that B7.1 and B7-H1 form heterodimeric complexes at the cell surface, and that Ad5FB4 penton capsomeres interfered negatively with the formation of B7.1/B7-H1 heterodimers. Our observation that the adenoviral vector Ad5FB4 interacted with cell surface molecules of the B7 family known to be implicated in immunoevasion mechanisms offers novel opportunities for cancer therapy using B7-H1/B7.1 heterodimers as cell surface targets, and Ad5FB4 vectors or Ad5FB4 penton capsomeres as therapeutic agents. Two different strategies might be envisaged: (i) naturally B7.1-targeted Ad5FB4 vectors can be designed for transferring therapeutic genes to B7.1/B7-H1-overexpressing cells, (ii) whereas soluble Ad5FB4 penton capsomeres would act via their negative interference with the B7-H1/B7.1 heterodimer formation.