Seven-week-old C57BL/6 female mice were purchased from Harlan (Gannat, France). MyD88- (MyD88^−/−) (22) and mitochondrial antiviral-signaling (MAVS)- (MAVS^−/−) (23) deficient mice were bred in animal facilities of TAAM-UPS 44 (Orléans) and UMR 0892 (VIM, Jouy-en-Josas), respectively. All mice were conditioned for at least 1 week in our animal facilities before beginning of the experiments. All animal experiments were approved by Ethics Committee No. 26 (officially recognized by the French Ministry for Research) in accordance with the European Directive 2010/63 UE and its transposition into French Law.

AdWT [described as AE18 in Ref. (24)] is based on Ad5 and is deleted in E1 and E3 regions. The expression cassette cloned instead of the E1 region contains the promoter/enhancer from the immediate early gene of human cytomegalovirus, the Escherichia coli lacZ gene with a nuclear localization signal, and the SV40 late polyA signal (SV40pA). AdH-3OVA2 and AdF-3OVA2 derived from AdWT and displaying OVA_320–341 (3OVA2) epitope, respectively, in hexon or fiber protein were described previously (21). AdP*F-3OVA2, AdH*F-3OVA2, AdH*P*F-3OVA2, and AdS*F-3OVA2 disabled to a different extend in Ad interactions with their natural receptors, were derived from AdF-3OVA2 (Table 1). More precisely, AdP*F-3OVA2 was derived from AE74 (9) bearing a deletion of penton base RGD motif, impairing interaction with integrins. AdH*F-3OVA2 was derived from AdH[GA]24 (12) bearing a deletion of hypervariable region 5 (HVR5) of hexon protein, impairing FX binding. AdH*P*F-3OVA2 contains both deletions. Finally, AdS*F-3OVA2 was obtained from AdWT by replacing the fiber shaft with the shaft from Ad serotype 3 (25). All capsid-modified viruses (Table 1) were constructed by recombinational cloning in E. coli (26).

AdOVA (provided by Dr. D. Descamps, INRA, Jouy-en-Josas) has a wild-type capsid and encodes the complete amino-acid sequence of ovalbumin protein. AdControl encoding no transgene was described previously (27).

All viruses were obtained using previously described procedures (5), stored at −80°C in PBS-7% glycerol, and titrated by spectrophotometry [1 OD₂₆₀ = 1.1 × 10¹² viral particle (vp)/ml].

The 293A human embryonic kidney cell line (R705-07, Invitrogen) was maintained in modified Eagle medium supplemented with 10% FBS and 1% of non-essential amino acids. CHO-k1-hCAR and CHO-k1-pCDNA were kindly provided by J. M. Bergelson (School of Medecine, University of Pennsylvania). L929 murine fibrosarcoma cells were kindly provided by Dr. U. Greber (Institute of Molecular Biology, Zurich, Switzerland) and maintained in DMEM supplemented with 10% FCS.

Purified viruses (10¹⁰ vp) were resuspended in Laemmli lysis buffer, boiled for 10 min and loaded onto a 10% NuPage gel (Novex, Invitrogen, CA, USA). After electrophoresis, the gel was stained with a silver staining kit (Invitrogen, Carlsbad, CA, USA). Alternatively, the gel was transferred on nitrocellulose membrane and the membrane was incubated with a rabbit polyclonal antibody directed against the fiber protein.

In order to evaluate virus infectivity, cells were plated out in 12-well plates at 1 × 10⁵ cells/well 48 h prior to infection. On the day of infection, cells were counted and infected with the indicated multiplicity of infection (MOI) of different Ad in 400 µl of serum-free medium. After 24 h, βgalactosidase (βgal) activity was measured using a chemiluminescent assay (BD Biosciences, Clontech, CA, USA) and protein concentration was determined using the Bio-Rad Protein Assay (Bio-Rad Laboratories, Hercules, Marnes-la-Coquette, France). Results were presented as relative light units (RLU) per microgram of proteins.

In order to evaluate FX-dependent cell transduction, viruses were mixed with or without human FX (1 U/ml, Cryopep). Then, virus-FX solution was added to CHO-k1-pCDNA cells and cells were incubated for 24 h at 37°C. βgal activity was measured as described above.

To confirm the presence and accessibility of the epitopes on the capsid surface, viral particles were coated on 96-well plates (Nunc, Roskilde, Denmark). Viruses were inactivated by incubation at 56°C for 30 min, followed by addition of 0.1% SDS. Spectrophotometric measurements at 215 and 225 nm allowed to determine viral protein concentrations, and subsequently 100 ng was coated on 96-well plates. To analyze epitope presence and accessibility on virions, non-denaturated viruses (100 ng) were coated on the plates. After overnight incubation at 4°C, non-specific sites were blocked with 5% milk PBS-Tween for 2 h, then plates were washed and incubated for 1 h with an anti-ovalbumin rabbit polyclonal antibody (AB1225, Millipore, MA, USA). Upon washing, an anti-rabbit IgG peroxidase-linked Ab (NA934, Amersham Biosciences, Saclay, France) was added for 1 h and peroxidase activity was revealed by incubation with the substrate O-phenylenediamine dihydrochloride (Sigma-Aldrich, Lyon, France) for 30 min. The reaction was stopped by addition of 3 N HCl and spectrophotometric readings were performed at 490 nm. Each virus was assayed in sexdecaplicates and the experiments were repeated at least twice.

Capsid-modified viruses (10¹⁰ vp) in PBS (200 µl) were injected intraperitoneally. Repeated injections were performed at 2 weeks intervals with a total number of injections ranging between two and three. Blood samples were collected from the submandibular vein, before virus injection and at different time points thereafter. Mice sera were prepared and analyzed for the presence of anti-ovalbumin, anti-βgal, and anti-Ad Abs by ELISA as described below.

In some experiments, mice were depleted of coagulation factors by subcutaneous injection of 133 µg of warfarin in 100 µl of PBS at days 3 and 1 prior to virus administration as described previously (10). Ad-naive or Ad–immune mice were obtained after injection of PBS or AdControl (10¹⁰ vp), respectively.

Sera were analyzed for the presence of specific Abs by ELISA. After coating of 96-well plates (Nunc) with 1 µg of ovalbumin (Sigma), 100 ng of βgal (Sigma), or 100 ng of denaturated AdWT viral particles, serial dilutions of the sera in 5% milk PBS-Tween were added. Bound Abs were detected with peroxidase-conjugated anti-mouse IgG, IgG1, IgG2b, or IgG2c isotype goat Abs (Southern Biotechnology Associates, Birmingham, AL, USA). The peroxidase activity was revealed by incubation with the substrate O-phenylenediamine dihydrochloride (Sigma-Aldrich) for 30 min. The reaction was stopped by addition of 3 N HCl and spectrophotometric readings were performed at 490 nm. Titers were defined as the reciprocal of the highest dilution giving an OD₄₉₀ twofold above background values.

Data from in vivo experiments (titers) were log₂-transformed before analysis. Comparison between two groups was done using unpaired Mann–Whitney test. Comparison between multiple groups were done using Kruskal–Wallis test followed with Dunn’s post hoc test. Two-way repeated measures ANOVA was used for comparison of responses measured for different groups at different time points, then Bonferroni post hoc test was used to compare between groups at each time point. Differences were considered significant when P < 0.05. All graphs and statistical tests were obtained with the use of GraphPad Prism software.

Our previous results have shown that pre-existence of anti-Ad Abs strongly enhances anti-ovalbumin antibody responses elicited by AdF-3OVA2 bearing OVA_320–341 B cell epitope (21), suggesting that cell transduction is dispensable for the efficacy of this vector. To address specifically this point, we compared the ability of capsid-modified AdF-3OVA2 and AdH-3OVA2 (Table 1) to elicit anti-ovalbumin antibody responses relative to AdOVA, encoding the whole ovalbumin protein as a transgene, in both naive and immune mice. C57BL/6 mice were injected with PBS or with an Ad bearing a wild-type capsid (AdWT, 10¹⁰ vp) and 2 weeks later received one intraperitoneal injection of AdH-3OVA2, AdF-3OVA2, or AdOVA (10¹⁰ vp). In Ad-naive mice, AdOVA induced high levels of anti-ovalbumin Abs from day 14 p.i. up to 209 days p.i (Figure 1, upper panel). In accordance with our previous study (21), AdH-3OVA2 led to a significantly higher anti-ovalbumin Ab responses compared to AdF-3OVA2. However, these responses remain lower than the ones observed in AdOVA-injected mice (Figure 1, upper panel). In Ad-immune mice, AdF-3OVA2 triggered strong anti-ovalbumin antibody responses compared to AdH-3OVA2 as described previously (21) while AdOVA was unable to trigger any antibody response (Figure 1, middle panel). Interestingly, the immune responses elicited by AdF-3OVA2 were long-lasting with strong Ab titers detectable up to 7 months after the immunization. While AdF-3OVA2 and AdH-3OVA2 elicit strong anti-βgalactosidase (βgal) Ab responses in Ad-naive mice, none of them triggered significant Ab responses in Ad-immune mice (Figure S1 in Supplementary Material).

Altogether, our data indicate that, in contrast to AdOVA, AdF-3OVA2 was able to trigger Ab responses in both Ad-naive and Ad-immune mice, suggesting that virus transduction is not mandatory for the efficacy of this vector.

To examine more precisely whether virus transduction plays any role in controlling the efficacy of AdF-3OVA2, different vectors were produced displaying both 3OVA2 epitope inserted into the fiber protein and capsid modifications impairing binding to specific receptors. AdP*F-3OVA2 presents a deletion of the RGD motif in penton base protein in order to ablate interaction with integrins (9). AdH*F3-OVA2 contains a mutation in hexon HVR5 impairing binding to FX (12). AdH*P*F-3OVA2 possesses both hexon and penton base mutations. Finally, AdS*F-3OVA2 contains an Ad3 fiber shaft instead of Ad5 shaft (25) impairing binding to CAR receptor. All vectors were produced at titers comparable to AdF-3OVA2 (Table 1). SDS-PAGE analyses showed no difference in virus composition between the different Ad. As expected, all viruses but AdS*F-3OVA2 displayed a similar migration pattern for the modified fiber (MW = 63.3 kDa, Figure 2A). For AdS*F-3OVA2, the migration pattern of the fiber protein is consistent with its reduced size (MW = 33.8 kDa, Figure 2A). Using a polyclonal serum specific for the fiber protein, we confirmed the modification of fiber size (long fiber or short fiber) for all vectors displaying 3OVA2 epitope (Figure 2B). Additionally, 3OVA2 epitope was detected on native virions by ELISA using a polyclonal anti-ovalbumin antibody (Figure 2C). Interestingly, the detection of 3OVA2 epitope on AdS*F-3OVA2 was reduced compared to AdF-3OVA2 (p < 0.001).

With the exception of AdS*F-3OVA2, all vectors possess the same capacity to transduce CHO-k1-CAR cells, a cell line overexpressing Ad primary receptor. AdS*F-3OVA2 led to a reduced cell transduction (Figure 3A) consistent with previous results of our laboratory (Raddi et al., in revision). Next, we examined transduction efficiency using specific cell lines allowing to monitor the detargeting from native Ad receptors. AdP*F-3OVA2 and AdH*P*F-3OVA2 were tested on CAR-negative integrin-positive L929 cells. Both vectors led to a reduced βgal activity compared to AdF-3OVA2, confirming their reduced ability to bind integrins (Figure 3B). AdP*F-3OVA2 and AdH*P*F-3OVA2 were used to transduce CAR-negative CHO-pcDNA cells in the presence or absence of FX (Figure 3C). The results showed that both vectors displayed reduced ability to use FX in accordance with the role of hexon protein in FX binding (Figures 3C,D).

After the characterization of produced vectors, we examined their capacity to mount anti-ovalbumin humoral responses. C57BL/6 mice were injected intraperitoneally twice 2 weeks apart with 10¹⁰ vp. Sera were collected 2 weeks after each administration and anti-ovalbumin, anti-βgal, and anti-Ad Abs were quantified by ELISA. Remarkably, no difference in ability to trigger anti-ovalbumin Abs was observed between detargeted vectors (AdH*F-3OVA2, AdP*F-3OVA2, AdH*P*F-3OVA2, and AdS*F-3OVA2) and AdF-3OVA2, neither after priming nor after boosting (Figure 4A). In contrast, anti-βgal antibody titers were significantly reduced after priming in AdH*P*F-3OVA2-injected mice (Figure 4B, p < 0.05) compared to AdF-3OVA2-injected mice. After boosting AdH*P*F-3OVA2- and AdS*F-3OVA2 were significantly impaired in their ability to trigger anti-βgal Ab (Figure 4B, p < 0.05 and p < 0.01 respectively). Of note, all vectors induced comparable anti-Ad Ab responses at all time points (data not shown). Altogether, these results underline that while detargeting Ad from its natural receptors reduces its ability to trigger humoral responses against the transgene, it has no impact on its ability to trigger humoral responses against an epitope displayed on the capsid.

In parallel to FX-detargeting through hexon modification, we also used a pharmacological approach based on administration of warfarin, a drug able to deplete all vitamin K-dependent blood factors. To do so, mice were pre-treated with warfarin or PBS before each AdF-3OVA2 administration. Blood factor depletion was confirmed by the measurement of FX activity in mice sera harvested prior to virus delivery (data not shown). Measurement of anti-ovalbumin titers showed no significant difference between warfarin- and PBS-pre-treated mice neither after priming nor after boosting (Figure S2A in Supplementary Material). Additionally, levels of anti-βgal and anti-Ad Abs also remained unmodified in warfarin-pre-treated mice (Figures S2B,C in Supplementary Material). Thus, both genetic (Figure 4A) and pharmacological (Figure 2A in Supplementary Material) approaches used to detarget Ad from FX were unable to decrease humoral responses toward 3OVA2 epitope.

Collectively, these results underline that in contrast to humoral responses against βgal transgene, humoral responses against 3OVA2 epitope displayed on the fiber protein do not rely on Ad interaction with their cellular receptors.

To get further insight into molecular bases controlling the efficacy of vaccination with Ad displaying epitopes, we investigated the role of innate immune pathways. First, since TLR and MyD88 were shown to participate in Ad innate immunity, we examined their role in shaping humoral responses by using MyD88-deficient mice. No significant difference was found in anti-ovalbumin IgG Ab responses in wild-type and MyD88-deficient mice, neither after priming nor after boosting (Figure 5A). In addition, both strains elicited comparable levels of anti-βgal (Figure 5B) and anti-Ad Abs (Figure 5C). Interestingly, compared to AdF3-OVA2-injected wild-type mice, AdF3-OVA2-injected MyD88^−/− mice displayed a strong increase in IgG1 anti-ovalbumin Abs (Figure 5D, p < 0.05) and a trend toward reduced levels of IgG2b and IgG2c anti-ovalbumin Abs. In contrast, MyD88^−/− mice displayed very low titers of anti-βgal IgG1 and IgG2b and a strong reduction in IgG2c (Figure 5E, p < 0.05). A reduction in IgG1 anti-Ad Abs was also found in MyD88^−/− mice (Figure 5F, p < 0.05). Then, we examined the role of RIG-I-induced innate immune pathway using mice deficient in MAVS protein. The levels of total IgG (Figure 6A) as well as IgG1, IgG2b, and IgG2c anti-ovalbumin Abs (Figure 6B) were comparable in wild-type and in MAVS^−/− mice, ruling out a major role of RIG-I/MAVS pathway in controlling anti-epitope humoral responses. In addition, no difference was found in anti-βgal (Figure 6C) or anti-Ad (Figure 6D) Ab responses between both strains.

Altogether, these results underline that TLR/MyD88 and RIG-I pathways are both dispensable in mounting antibody responses against 3OVA2 epitope. However, they also reveal that TLR/MyD88 pathway influences the isotype nature of antibody responses against the displayed epitope.