The Maturation of Murine Dendritic Cells Induced by Human Adenovirus Is Mediated by the Fiber Knob Domain*

We investigated the mechanism of adenovirus serotype 5 (Ad5)-mediated maturation of bone marrow-derived murine dendritic cells (DC) using (i) Ad5 vectors with wild-type capsid (AdE1°, AdGFP); (ii) Ad5 vector mutant deleted of the fiber C-terminal knob domain (AdGFPΔknob); and (iii) capsid components isolated from Ad5-infected cells or expressed as recombinant proteins, hexon, penton, penton base, full-length fiber, fiber knob, and fiber mutants. We found that penton capsomer (penton base linked to its fiber projection), full-length fiber protein, and its isolated knob domain were all capable of inducing DC maturation, whereas no significant DC maturation was observed for hexon or penton base alone. This capacity was severely reduced for AdGFPΔknob and for fiber protein deletion mutants lacking the β-stranded region F of the knob (residues Leu-485–Thr-486). The DC maturation effect was fully retained in a recombinant fiber protein deleted of the HI loop (FiΔHI), a fiber (Fi) deletion mutant that failed to trimerize, suggesting that the fiber knob-mediated DC activation did not depend on the integrity of the HI loop and on the trimeric status of the fiber. Interestingly, peptide-pulsed DC that had been stimulated with Ad5 knob protein induced a potent CD8+ T cell response in vivo.

Adenovirus (Ad) 1 vectors induce a vigorous stimulation of the innate immune system very rapidly after their administra-tion followed by the induction of the cognate immune response mediated by T and B lymphocytes. The early inflammatory response is independent of the transgene expression (1), but the antigen-specific response can be directed to both viral and transgenic proteins (2,3). As a consequence, transgene expression is only transient, and the occurrence of anti-Ad neutralizing antibodies is in part responsible for the low efficiency of gene transfer using Ad vectors in preimmunized subjects (4,5).
Several studies have attempted to elucidate the mechanism of Ad-triggered activation and the antigenic targets of the immune response against Ad vectors. Increasing evidence has emerged which shows that capsid components of the vector play a major role in this process. The Ad virion is composed of capsomers of two different types. Hexon (Hx) capsomers form the facets and edges of the icosahedral capsid, whereas penton (Pn) capsomers, located on the 5-fold axis of the virus, comprise a protruding fiber (Fi) protein (a homotrimer) linked via noncovalent bonds to the capsid-anchored penton base (Pb, a homopentamer). The fiber protein consists of three structural domains, the N-terminal tail, which binds to the penton base, the central shaft, and the C-terminal spherical domain, called the knob. The capsid proteins of the input virus are not only targets for the humoral and T helper responses (5-7), but they are also recognized by cytotoxic CD8 ϩ T lymphocytes (8,9). The capsid components may also directly interact with cells of the immune system, and our previous data indicated that hexon stimulates helper T cell responses (10).
Ad also activates several kinase pathways (including p38/ MAPK, NF-B, and Raf1) during the minutes following vector administration, leading to the synthesis of chemokines and inflammatory cytokines (11)(12)(13)(14). The stimulation of intracellular signaling pathways seems to be required for the early phases of the viral cycle since activated cAMP-dependent protein kinase and p38/MAPK boost microtubule-mediated nuclear addressing of the virus (13). Ad has been found to modulate dendritic cell (DC) maturation by increasing the expression of major histocompatibility complex and costimulatory molecules and, at least for murine DC, by inducing the secretion of IL12 (15)(16)(17)(18). We and others have shown that such activated DC are fully competent for initiating protective immune responses (19 -21). Interestingly, the Ad-mediated activating effect on DC is apparently independent of viral gene expression (15)(16)(17) but correlates with NF-B activation (16).
The purpose of the present study was to identify the viral component(s) of the Ad5 capsid and structural domains thereof, which were responsible for the activation of bone marrowderived murine DC. By using isolated Ad5 capsid proteins and virions deleted of the fiber terminal domain (knobless Ad5), we provide evidence that the fiber knob was the major viral factor of DC maturation. Experimental data obtained with deletion mutants of recombinant fiber protein mapped this function to region F, one of the 10 ␤-stranded regions defined by crystallographic analysis of the knob.

EXPERIMENTAL PROCEDURES
Ad Virions and Capsid Proteins-AdE1°(kindly provided by Transgene SA, Strasbourg, France) was a replication-defective serotype 5 adenovirus (Ad5) deleted of both early regions E1 and E3 and carrying wild-type (WT) fibers (10). AdGFP (AdenoCMV5-GFP, obtained from Quantum Biotechnologies, Inc., Montreal, Canada) was an Ad5 vector carrying WT fibers and encoding the reporter gene for the enhanced green fluorescent protein (GFP). Fiber deletion mutant AdGFP⌬knob was derived from the AdGFP vector by insertion of an opale-stop codon in the fiber gene, downstream of an extrinsic trimerization signal introduced into the fiber shaft. This trimerization signal consisted of the neck region peptide from the human lung surfactant protein D, and the fiber construction has been described in detail elsewhere (22,23). The resulting fiber lacked its entire knob domain ("knobless") and contained only the tail domain and the N-terminal seven repeats of the shaft, i.e. residues 1-157. The schematic structure of the fibers of AdE1°, AdGFP, and AdGFP⌬knob is depicted in Fig. 1, a and b. AdE1°, AdGFP, and AdGFP⌬knob virions were isolated by isopycnic ultracentrifugation in continuous CsCl gradient (10). Capsid proteins hexon (abbreviated Hx) and penton capsomers (abbreviated Pn, for penton base ϩ fiber) were isolated from WT Ad5-infected HeLa cells. Recombinant WT Ad5 proteins penton base (Pb), fiber (Fi), and fiber knob domain were isolated from baculovirus-infected Sf9 cells. Four Ad5 fiber mutants carrying deletions of various lengths in the knob domain (24), Fi⌬402-481, Fi⌬HI, Fi⌬EF, and Fi⌬485-486 (Fig. 1c), were produced as recombinant proteins in Sf9 cells. Ad proteins were purified according to a protocol described in earlier studies (10,25,26). Briefly, we used a three-step procedure consisting of (i) ammonium sulfate precipitation, (ii) anion exchange high performance liquid chromatography, and (iii) concentration-ultrafiltration through concentrator membranes with a 100-kDa cut-off. Protein samples were analyzed by conventional sodium dodecyl sulfate 12%-polyacrylamide gel electrophoresis (10) and immunoblotting (25).
The possible occurrence of endotoxins in Ad virion and protein preparations was assayed using amoebocyte lysate from Limulus polyphemus (E-TOXATE™, Sigma), with Escherichia coli serotype 055:B5 lipopolysaccharide (LPS) as standard. Endotoxin concentration in the different batches of proteins never exceeded 30 pg/ml at the doses used for DC stimulation. At this concentration, Hx protein samples never induced DC maturation above the background of non-stimulated (NS) DC (see Fig. 2). CsCl-purified Ad virion preparations were totally free of detectable endotoxin contamination.
Isolation of Dendritic Cells-This method was adapted from Ref. 27. Lymphocyte-depleted bone marrows from C57BL/6 (H2 b ) mice (Harlan, Gannat, France) were cultured overnight in complete medium (RPMI 1640 with 10% fetal calf serum, 2 mM L-glutamine, 50 M ␤-mercaptoethanol, 100 units/ml penicillin, 100 g/ml streptomycin). Non-adherent cells were harvested and resuspended in complete medium with 1000 units/ml recombinant granulocyte-macrophage colony-stimulating factor (R&D Systems, Minneapolis, MN) and 100 units/ml rIL4 (R&D Systems). The medium was replaced at day 4. At day 6, aliquots of non-adherent cells were resuspended at 3 ϫ 10 6 /ml in phosphate-buffered saline containing 1% fetal calf serum. Cells were then incubated for 1 h at 37°C (i) without (non-stimulated DC (NS-DC)), (ii) with Ad5 capsid components, or (iii) with corresponding chromatographic fractions from extracts of Sf9 cells infected with baculoviral empty vector (mock-stimulated DC (MS-DC)), as indicated in the figure legends (expressed as doses/10 6 cells). After 1-h incubation, the cell concentration was adjusted to 3 ϫ 10 5 /ml with complete medium supplemented with granulocyte-macrophage colony-stimulating factor. Control cells received 1 g/ml LPS from E. coli (Sigma). At day 8, cells were harvested and analyzed in flow cytometry, mixed lymphocytic reaction, and immunization. Day 8 supernatants were collected. For the experiments described in Fig. 4a, DC were purified by using magnetic beads conjugated to anti-mouse CD11c mAb (Miltenyi Biotec).
ELISPOT-IFN␥-Nitrocellulose microplates (Millipore) were coated with rat anti-mouse IFN␥ antibody (R4-6A2, Pharmingen) and then washed and saturated with complete medium. Aliquots (in triplicate) of freshly isolated spleen cells (10 6 cells/well) or peripheral blood lymphocytes (2 ϫ 10 5 cells/well) were added to the wells in complete medium containing 30 units/ml human rIL2 (Roche Applied Science) and 10 Ϫ6 M GP33 or NP366. After 20 h at 37°C, the spot-forming cells were revealed and counted (10). The values obtained with NP366 were subtracted from the mean values of triplicate tests obtained with GP33.
Cytotoxicity Assay-Splenocytes from DC-immunized mice (effector cells) were cultured for 4 days with 10 Ϫ6 M GP33. Effector cells were then assayed on 51 Cr-labeled EL4 target cells (10 Ϫ4 Ci/cell) pulsed with GP33 or NP366 peptide, in duplicate and at different effector-totarget ratios. After 5 h at 37°C, supernatants were collected, and radioactivity was measured (Top Count, Packard Instruments). Target cells were incubated with medium alone to determine the level of spontaneous release of 51 Cr and with 2% alkyl-trimethyl-ammonium bromide (Sigma) to determine the total 51 Cr release. Specific 51 Cr release was calculated using the formula: (experimental Ϫ spontaneous release)/(total release Ϫ spontaneous release) ϫ 100.

Effect of Isolated Ad5
Capsid Components on DC Phenotype-Immature DC were incubated with isolated Pn or Hx capsomers, purified from Ad5-infected HeLa cells. The major subset of Hx-treated DC showed a poorly mature phenotype, not significantly different from that shown by NS-DC, with a low expression of major histocompatibility complex class II, CD40, CD80 (B7.1), and CD86 (B7.2) molecules (Fig. 2a). In contrast, Pn-treated DC massively expressed a mature phenotype, similar to the phenotype of LPS-stimulated DC, with a high level of expression of the different markers (Fig. 2a). Since the Ad Pn capsomer is composed of two structural entities, Pb and Fi, the following experiments were designed to determine which of these two constitutive proteins was responsible for the DC maturation. DC were then treated with recombinant Pb or Fi protein isolated from insect cell extracts. The results indicated that isolated Fi protein reproduced the entire stimulatory effect of Pn, whereas Pb protein alone showed no detectable effect on DC maturation (Fig. 2b).
Role of the Fiber Knob Domain in Fiber-mediated DC Maturation-The C-terminal domain of Ad fiber has been expressed as recombinant knob protein (26,28). DC incubated with the fiber knob protein expressed a mature phenotype similar to that observed in DC stimulated by the full-length Fi (Fig. 3a). Consistent with this, the knob-stimulated DC were able to vigorously induce the proliferation of allogeneic CD4 ϩ T cells in MLR experiments, an effect that occurred in a knob-dependent manner (Fig. 3b). At the higher knob protein dose used (0.5 g), knob-stimulated DC were much more potent stimulators of allogeneic CD4 ϩ T cells than control NS-or Hx-stimulated DC. However, there was an apparent threshold in the knob-medi-  a and b) or recombinant proteins (c). a, WT Ad5 fiber with its three structural domains, the tail, which is non-covalently linked to the penton base, the shaft and the knob, which is responsible for the attachment of Ad5 to the cell receptor CAR. WT Ad5, AdE1°, and AdGFP carried long shafted fibers with 22 repeats of a 15-residue ␤-sheet motif. b, the knobless fiber vector AdGFP⌬knob carried a shorter fiber shaft (7 ␤-repeats), which terminates by the extrinsic trimerization motif neck region peptide from the human lung surfactant protein D, and an opale-stop codon. c, diagrammatic representation of the conformational structure of Ad5 fiber knob (35), showing the different regions in ␤-sheet structure (A to J) and their connecting loops, aligned with a linear representation of the fiber knob deletion mutants Fi⌬402-481, Fi⌬HI, Fi⌬EF, and Fi⌬485-486. The deleted region of the knob sequence is represented by a thin line, and the intact region is represented by a solid box. The numbers indicate the amino acid numbers, the numbering starting from the initiator methionine of full-length Ad5 fiber. The bold letters m and t indicate the monomeric or trimeric status of the recombinant fibers, respectively, and the symbols ϩ, Ϯ, and Ϫ refer to their degree of affinity for CAR (24,34). Knob regions interacting with CAR (30,31) are represented by stippled boxes.
ated DC activation since a detectable effect was only observed at the dose of 0.2 g of knob protein, i.e. 5 ϫ 10 6 molecules of fiber knob per DC. Interestingly, the MLR values for 0.2 g of knob protein were similar to that of 1 g of Fi, a result consistent with the fact that the quantity of knob protein present in the 0.2-g knob sample roughly corresponded to the knob protein content of the 1-g Fi sample (180 amino acid residues for the knob domain versus 581 amino acid residues for the fulllength fiber). The knob protein also induced IL12 and TNF␣ secretion in a dose-dependent manner (Fig. 3c).
We then investigated whether the knob was capable of directly targeting DC and induce their maturation, without the participation of any other intermediate cells. To this aim, CD11c ϩ cells were purified using anti-CD11c magnetic beads and tested in maturation assays. Although NS-DC showed some up-regulation of expression of co-stimulatory molecules, knob-stimulated DC expressed a significantly more mature phenotype than control cells (Fig. 4a). Moreover, recombinant Fi protein was able to bind to immature DC, with 63% fiberpositive cells at day 6 of culture (Fig. 4b). In this assay, cellbound Fi was detected using anti-fiber tail mAb, implying that the epitope in the tail was accessible and that the fiber attachment to the cell surface had occurred via the knob domain. These data strongly suggested that DC were direct targets for the fiber knob domain.
To assess whether the effect of Ad5 virion on DC activation was also directly mediated by the fiber knob domain, comparative experiments were carried out using virions of AdE1°, carrying WT Fi projections, and of AdGFP⌬knob, a knobless Ad5 vector (Fig. 1, a and b). Upon incubation with AdE1°, there was an up-regulation of the expression of maturation markers on DC (Fig. 4c), as expected from previous observations (15,16). With the same doses of AdGFP⌬knob virions, the DC showed a significantly less mature phenotype. These results confirmed that the fiber knob was the common stimulating factor in all the active viral components assessed, Ad5 virion, Pn capsomer, Fi, and knob proteins, and that the knob domain carried most of the determinants responsible for the stimulation of murine DC.
Absence of CAR Expression at the Surface of Murine DC and Low Degree of Permissiveness to Ad5 Infection-The cell entry of Ad5 results from the attachment of the fiber knob to a high affinity primary receptor, identified as the CAR in most Adpermissive cells (24,30,31), followed by the binding of the RGD motifs in Pb to ␣ v ␤ 3 / 5 integrins, which promotes virus endocytosis (32). It has already been shown that human and murine DC do not express CAR (17, 33) but do express ␣ v ␤ 3 / 5 integrins at their surface, and this probably explains their low permissiveness to Ad. We verified by FACS analysis the level of CAR expression at the surface of DC differentiated in our culture conditions. HeLa cells, which are known to be high CAR expressors, were used as positive control, whereas Chinese hamster ovary cells, which do not express CAR, were used as negative control. As expected, no CAR signal above the background level could be detected on immature DC collected before stimulation (Fig. 5a), confirming previous observations. We thus checked whether knob stimulation of DC was associated with effective cell entry of Ad and Ad-mediated  (Fig. 5b), confirming that murine DC were poorly permissive to Ad5.

Mapping of the Fiber Knob Region Involved in DC Maturation-Four Ad5
Fi protein mutants carrying deletions in the knob domain (Fig. 1, c and d) were assayed for DC maturation. and Fi⌬485-486, carried deletions involving the short double ␤-stranded region EF, and the two residues Leu-485 and Thr-486, which form the ␤-sheet F (34, 35), respectively. Both Fi⌬EF and Fi⌬485-486 occur as fiber trimers, whereas ⌬402-481 and ⌬HI are defective in fiber trimerization (24). We compared the DC maturation activity of WT knob and fiber mutant proteins by evaluating their capacity to induce allogeneic CD4 ϩ splenocyte proliferation using MLR assays, as above. Since the knob domain represents approximately onethird of the whole fiber protein bulk, we used three to four times less WT knob protein than fiber mutant proteins to stimulate DC, to normalize the samples to similar knob contents. As shown in Fig. 6, mutant Fi⌬HI retained its full DC maturation activity, implying that the distal HI loop and the trimeric structure of the knob were not required for induction of DC maturation. The three other mutants, Fi⌬402-481, Fi⌬EF and Fi⌬485-486, did not change significantly the DC phenotype, suggesting that the double-stranded region EF, and more specifically the short ␤-strand F, was responsible for the maturation effect.
Knob-stimulated DC Induced a Specific CD8 ϩ T Cell Response in Vivo-To compare the in vivo functioning of knob-stimulated DC with that of DC stimulated with other Ad capsid components, we tested their ability to generate a cellular immune response. We immunized mice with DC pulsed with a model antigen and determined whether this could lead to more rapid elimination of syngeneic donor lymphocytes pulsed with the same antigen. The D b -restricted immunodominant peptide GP33 from lymphocytic choriomeningitis virus was chosen as model antigen, with the influenza virus D b -restricted NP366 peptide as negative control. Syngeneic donor splenocytes were labeled with the fluorescent vital dye CFSE to trace adoptively transferred splenocytes by flow cytometry and to monitor their elimination by specific CD8 ϩ T cells induced by DC immunization.
The population of CFSE ϩ spleen cells loaded with GP33 decreased within 24 h in all mice immunized with GP33-pulsed DC, regardless of how the DC were stimulated or unstimulated (Fig. 7a). However, a significantly faster spleen cell rejection was observed in mice treated with knob-stimulated DC, at a rate similar to that observed in mice immunized with GP33 emulsified in incomplete Freund's adjuvant. The curve of rejection in mice immunized with knob-stimulated DC loaded with the irrelevant NP366 peptide was similar to the curve observed in mice immunized with GP33-loaded DC stimulated by Pb or Hx or non-stimulated (Fig. 7a). In the latter mice, the CFSE ϩ cell diminution in blood was likely due to a non-specific redistribution of splenocytes in the secondary lymphoid organs. Consistent with this hypothesis, analysis of CFSE-labeled cells in spleens of recipient mice 10 days after adoptive transfer showed that only 5% of CFSE ϩ cells persisted in mice immunized with 3 ϫ 10 4 knob-stimulated DC, whereas they still represented 72 and 60% in mice immunized with 2.5 ϫ 10 5 Pbor Hx-treated DC, respectively (Fig. 7a). The effect obtained using 3 ϫ 10 4 knob-stimulated DC, as compared with that obtained using 2.5 ϫ 10 5 Pb-or Hx-treated DC, implied that the knob-treated DC were at least 10 times more efficient in inducing splenocyte rejection than DC treated by other Ad capsid proteins.
At day 20 after DC immunization, which corresponded to 10 days after the adoptive transfer of CFSE-labeled splenocytes, the number of IFN␥-secreting CD8 ϩ T cells was determined by ex vivo ELISPOT assays. GP33-specific T cells were not significantly induced in mice immunized with Hx-or Pb-stimulated and GP33-loaded DC or in mice immunized with knob-stimulated DC loaded with control peptide NP366 (Fig. 7b). In contrast, significant numbers of IFN␥-secreting T cells were detected in both spleen and blood of mice immunized with knobstimulated and GP33-loaded DC, and the induction of the GP33-specific T cell response seemed to correlate with the number of knob-stimulated DC injected (Fig. 7b). However, no T cell response was detectable after immunization with 3 ϫ 10 4 DC despite the rejection of CFSE ϩ cells shown in Fig. 7a. The cytolytic activity of in vitro restimulated spleen cells was then tested. No specific lytic activity was observed in effector cells from mice immunized with 3 ϫ 10 4 NS-DC, but effector cells from mice immunized with knob-stimulated DC were found to lyse GP33-loaded EL4 cells with a high efficiency (Fig. 7c). DISCUSSION DC are at the interface between innate and adaptive immunity and thus play a crucial role in interpreting the stimuli delivered by their environment. Their resulting maturation status is critical for the development of the immune response. In the present study, we provide evidence showing that fiber knob protein, the 180-residue C-terminal domain of Ad5 fiber capsomer, interacts with DC and activates signals that render DC phenotypically and functionally mature. Using deletion mutants in the knob domain of recombinant Fi protein, we found that short deletions that altered the double ␤-strand EF, as in mutants Fi⌬EF and Fi⌬485-486, resulted in the loss of their capacity to induce DC maturation. It is noteworthy that Fi⌬HI, a Fi protein deletion mutant that occurs in a monomeric form (24), retained its full DC-maturation activity. This suggested that the effect of the knob on DC maturation depended upon the integrity of the ␤-strand F (residues 485-486), but not upon the HI loop, and did not require the trimeric structure of the knob.
Ad5 uptake by permissive cells is a two-step process consisting of (i) the binding of the Fi terminal knob to the attachment receptor CAR, a high affinity receptor for most Ad subgroups (e.g. subgroups C Ad2 and Ad5 and subgroup A Ad12), except for subgroup B serotypes such as Ad3 and Ad7 (36). This step is followed by (ii) the endocytosis of cell-bound virions, mediated by the Pb RGD motifs and ␣ v ␤ 3 / 5 integrins (32). In the initial step of virus-cell binding, the ␤-stranded region EF represents one of the critical regions involved in the interaction of the knob with CAR (Fig. 1c), and mutant fibers Fi⌬EF and Fi⌬485-486 have been shown to bind to CAR with a significantly lower affinity than WT fiber (24,30,31,34). Since the DC do not express CAR (Fig. 5a) (17,33), our data suggested that Ad5-and Fi-mediated DC maturation resulted from a direct interaction between knob and DC via another receptor (Fig. 4b). However, the partial overlapping of the DC-binding site(s) on the knob with those specific for CAR is not coincidental, if one considers that the regions of the knob involved in the interaction with cell components must be very exposed on the protein surface and thus accessible to many other receptors and/or ligands. For example, a CAR-blocking anti-knob monoclonal antibody has been found to inhibit the binding of Ad5 virions to alternative receptors, heparan sulfate proteoglycans and sialoglycoproteins (37).
Immature DC are instructed of danger through TLR-mediated signaling. TLR belong to a family of receptors that recognize molecular motifs conserved among large groups of microorganisms. Due to conserved domains in their cytoplasmic tail, TLR share a common signaling pathway leading to the activation of MAPK and to the translocation of NF-B in the nucleus, where it directs the expression of a number of genes involved in inflammatory and immune responses (38). Interestingly, previous data have shown that Ad induced maturation of DC via early activation of NF-B independently of viral gene expression (16). However, with the exception of TLR4, which mediates RSV protein F signaling and activation of B cells by murine retroviruses, the role of TLR in immune responses to viruses in mammals remains poorly understood (39). Réa et al. (40) recently reported that chimeric Ad5 virions carrying fibers from subgroup B virus induced DC maturation. We also observed DC maturation upon stimulation with an Ad vector carrying chimeric fibers composed of tail and shaft from serotype 5 and knob from serotype 3 (data not shown). These data suggest that DC surface molecules could recognize phylogenetically distant Ad knobs and that the DC-knob interaction probably involves conserved secondary structures, rather than conserved individual amino acid residues in the knob domain.
Many studies have focused on recombinant Ad as vaccine vectors for viral and tumor antigens. One of the major reasons of their vaccine efficacy has been attributed to their natural ability to infect DC in vivo (41). If the Ad Fi knob carries peptide motif(s) and/or conformational structure(s) that medi-FIG. 7. Knob-stimulated DC primed GP33-specific CD8 ؉ T cells in vivo. a, rejection of donor cells by recipient mice vaccinated with DC. CFSE-labeled splenocytes from C57BL/6 mice were loaded with GP33 and transfused into C57BL/6 recipient mice that had been immunized with (i) GP33-loaded DC (2.5 ϫ 10 5 , 8 ϫ 10 4 ,and 3 ϫ 10 4 cells, respectively) stimulated with knob; (ii) GP33loaded DC (2.5 ϫ 10 5 ) stimulated with Pb or Hx, or NS; (iii) NP366-loaded DC (2.5 ϫ 10 5 ) stimulated with knob; or (iv) 50 g of GP33 emulsified in incomplete Freund's adjuvant (IFA). The rejection level (expressed as the percentage of CFSE-positive cell population in naive mice) was determined by FACS analysis in blood (left part of panel a) and spleen (right part of panel a) at different time points after adoptive transfer. b, ex vivo ELISPOT-IFN␥ response of the same mice. Freshly isolated spleen cells or peripheral blood lymphocytes were incubated with GP33, and results were expressed as spot-forming cells/10 5 CD8 ϩ T cells (mean from triplicate tests Ϯ S.D.). c, cytolytic activity of effector cells derived from mice vaccinated with GP33-loaded and NS or knob-stimulated DC (3 ϫ 10 4 cells). After in vitro restimulation with GP33, spleen cells were incubated with 51 Cr-labeled EL4 target cells loaded with GP33 or NP366 at different effector-to-target (E:T) ratios and assayed for 51 Cr release. Data shown are mean of duplicate experiments Ϯ S.D. ate DC maturation, our finding that the double-stranded region EF carries major DC ligands might provide a clue for the generation of synthetic adjuvants that would directly target DC in vivo. This has recently been exemplified with the bacterial molecule OmpA (42). In line with this observation, we showed that knob-stimulated DC secreted TNF␣ and IL12 and stimulated allogeneic CD4 ϩ T cells. Immunization of C57BL/6 mice with knob-stimulated DC loaded with GP33 induced a potent GP33-specific T cell response.
The results presented here should also have implications in gene therapy as prolonged transgene expression requires minimal vector immunogenicity. In the last generations of Ad vectors, most of the regulatory viral genes, or even the whole viral genome, have been deleted, in part to prevent expression of immunogenic viral proteins (43,44). This has led to some prolonged transgene expression in mouse models, although cellular and humoral responses to the vector and the transgenic protein have still been detected (43,44). Thus, persistent immunostimulatory properties have been assigned to capsid proteins of the input virions (8). Deletion of the entire knob domain, or replacement of knob regions by cellular ligands with new cell-binding specificities, might confer upon the vectors both retargeting properties and a decrease in their immunogenicity, as suggested by the low level, if any, of the DC maturation effect of our knobless virions AdGFP⌬knob (Fig. 4c). However, it has to be kept in mind that fiber knob constructs must respect several biological functions that are carried by the knob domain and are essential to the viability and growth of Ad (24,45,46). In this case, the use of Ad5 vectors carrying cleavable knob domains by means of the insertion of a restriction protease site in the fiber shaft (23) might represent a possible alternative to reconcile contradictory requirements such as a low immunostimulation and a high gene transfer efficiency.