The expression plasmid encoding the fusion protein of GM-CSF and ovalbumin (GM-OVA) was generated by cloning the murine GM-CSF cDNA, a flexible linker, the OVA cDNA and a HIS₆-tag as a single open reading frame into pcDNA3.1 (Fig. 1A). Three control plasmids expressing ovalbumin alone (OVA), ovalbumin fused to the leader peptide of GM-CSF (ΔGM-OVA), or ovalbumin fused to the open reading frame of rhesus monkey GM-CSF (GM^rh-OVA), which is biologically inactive in murine cells, were also constructed. Coexpression of murine GM-CSF and OVA from the same plasmid was achieved by a modified bidirectional promoter based on the immediate early gene of human cytomegalovirus (GM-DP-OVA). After transient transfection into 293T cells, expression of the recombinant proteins was detected in cell supernatants by Western Blot analysis. The OVA plasmid and the ΔGM-OVA plasmid expressed a protein of 50 kD, which corresponds to the size of ovalbumin, whereas both fusion proteins migrate at a size of about 75 kD (Fig 1B). The expression levels of the plasmids encoding the fusion proteins were slightly higher than those obtained with OVA and ΔGM-OVA, whereas the expression of ovalbumin from the GM-DP-OVA plasmid is reduced substantially, probably due to a lower activity of the bidirectional promoter. GM-DP-OVA lacks the C-terminal tag explaining the slightly lower size of the protein in the Western blot (Fig. 1B). For further in vitro characterization of GM-OVA, the protein was purified from transfection supernatants via Ni²⁺-NTA agarose beads. Using a GM-CSF-dependent cell line, comparable bioactivity was observed for GM-OVA and control supernatants from a GM-CSF producer cell line (data not shown). Adenoviral vectors containing the expression cassettes GM-OVA, GM-DP-OVA, GM^rh-OVA or ΔGM-OVA were constructed using the pAd-Easy system. Expression of the transgenes and GM-CSF bioactivity was confirmed for the adenoviral vectors after infection of 293 cells. The relative expression levels of OVA by the different adenoviral vector constructs mirrored those observed after transient transfections of the corresponding plasmid DNAs (data not shown). To compare the expression levels of GM-CSF, serial dilutions of the supernatants of 293 cells infected with the adenoviral vectors were tested in a proliferation assay with a GM-CSF-dependent cell line (Fig 1C.) In contrast to the OVA expression levels, the bicistronic vector expressed slightly higher levels of bioactive GM-CSF in comparison to the vector encoding the GM-OVA fusion protein. The supernatants of cells infected with adenoviral vectors enoding GM^rh-OVA or ΔGM-OVA did not induce proliferation of the GM-CSF-dependent murine cell line. Since GM^rh-OVA has GM-CSF activity on human cells (data not shown), rhesus monkey GM-CSF is not bioactive on murine cells.

GM-CSF plays a critical role in the differentiation of dendritic cells. Bone marrow derived monocytes (BMDM) can be differentiated into immature DCs by co-culture with GM-CSF and IL-4. The ability of purified GM-OVA to generate DCs was therefore compared to recombinant GM-CSF. Recombinant GM-CSF and GM-OVA were equally efficient in generating immature DC (low CD80, CD83, CD86) from BMDMs. To a similar extent, the immature DCs induced by GM-OVA or recombinant GM-CSF could be matured by LPS as evidenced by the upregulation of co-stimulatory molecules (Fig 2A). To test whether the fusion of GM-CSF to ovalbumin enhances the activation of antigen-specific CD8+ T cells by antigen-presenting cells (APC), splenocytes from T-cell receptor transgenic OT-I mice were labeled with CFSE and cultured in the presence of decreasing concentrations of GM-OVA or ovalbumin for four days. The proliferation-dependent decrease in CFSE fluorescence of CD8+ T-cells was analyzed by FACS. GM-OVA induced proliferation of OT-I cells at 10-fold lower concentrations than ovalbumin (Fig 2B). The addition of recombinant GM-CSF to ovalbumin-containing cultures did not enhance proliferation of OT-I cells (data not shown), suggesting that the fusion protein enhances presentation of ovalbumin peptides most likely by GM-CSF receptor positive APCs.

Since the GM-OVA fusion protein maintained GM-CSF bioactivity and enhanced the proliferation of OVA-specific CD8+ T cells in cell culture, we expressed the GM-OVA fusion protein by DNA and adenoviral vectors in vivo and determined ovalbumin-specific CD8+ T cell responses. C57BL6 mice were either immunized by one or two (week 0 and 5) DNA injections (-/D and D/D), a single adenoviral vector injection (-/A) or a DNA prime (week 0) and adenoviral vector boost (week 5) regimen (D/A). The DNA and adenoviral vectors either encoded the GM-OVA fusion protein or ovalbumin. One week after the first or second immunization, OVA-specific CD8+ T cell responses were determined using SIINFEKL/H2-K^b tetramers. The functional activity of OVA-specific CD8+ T cells was assessed by stimulation with the SIINFEKL peptide followed by staining for interferon-γ and the degranulation marker CD107a 32. A single injection of DNA encoding either OVA or GM-OVA did not result in any detectable CD8+ T cell responses, while a single adenoviral vector immunization with either antigen resulted in similar levels of tetramer-positive and IFN-γ/CD107a double positive CD8+ T cells, with approximately 6% of the total CD8+ T cells being specific for the SIINFEKL peptide. After two injections of the OVA DNA vaccine, OVA-specific CD8+ T cells became detectable constituting approximately 0.4% of the CD8+ T cells. Two injections of GM-OVA DNA did not result in detectable levels (> 0.1%) of OVA-specific CD8+ T cells demonstrating reduced CD8+ T cell responses for the GM-CSF ovalbumin fusion protein. Significantly reduced CD8+ T cell responses could also be observed for the GM-OVA fusion protein in the DNA prime adenoviral vector boost immunization. The mean percentage of tetramer positive CD8+ T-cells decreased to 5,5% in the GM-OVA group compared to 13,8% of tetramer positive cells in animals immunized with the vaccines only expressing ovalbumin (Fig 3A). Results for the CD107a/IFN-γ staining paralleled the results obtained with the tetramers (Fig. 3B). Priming with the OVA DNA vaccine clearly enhanced CD8+ T cell responses after the adenoviral vector immunization. In contrast, CD8+ T cell responses after the adenoviral GM-OVA boost were not enhanced by priming with GM-OVA DNA (Fig 3A, B).

Since the fusion of GM-CSF to antigens has previously been shown to enhance humoral immune responses to the antigens, we also studied the effect of GM-OVA on OVA-specific IgG1 and IgG2a antibody levels. Since a single injection of the DNA and adenoviral vector vaccines encoding either GM-OVA or OVA did not result in detectable antibody levels, only the results for two repeated DNA injections and the DNA prime adenoviral vector boost regimen are shown (Fig. 4). Strikingly, mice immunized with GM-OVA by DNA or DNA plus adenoviral boost did not produce any detectable levels of ovalbumin-specific IgG2a antibodies. In contrast, immunization with vaccines encoding only ovalbumin resulted in readily detectable levels of antigen-specific IgG2a antibodies (Fig. 4B). Opposite effects were observed on ovalbumin-specific IgG1 antibody levels. These were significantly elevated for the GM-OVA groups in comparison to the groups immunized with the ovalbumin-expressing vaccines (Fig. 4A).

Suppression of CD8+ T cell responses by the GM-OVA vaccines could be directly due to the bioactivity of GM-CSF. If this were the case, the coexpression of GM-CSF and ovalbumin from bicistronic expression cassettes should also suppress CD8+ T cell responses. However, the opposite effect was observed. Despite substantially lower ovalbumin expression levels by the GM-DP-OVA vaccines coexpressing GM-CSF and ovalbumin, DNA prime adenoviral vector boost immunization with GM-DP-OVA vaccines enhanced the percentage of SIINFEKL-specific CD8+ T cells compared to vaccines only expressing ovalbumin (Fig 5, left column). This is in sharp contrast to the results obtained after GM-OVA vaccination, which reduced the percentage of tetramer positive CD8+ cells.

After immunization with the bicistronic vaccines more than a quarter of all CD8+ cells were specific for ovalbumin. Given this extraordinarily high CD8+ T cell response, the OVA-specific cells were further characterized functionally. Stimulation of the spleen cells of mice immunized with the bicistronic vaccine with the SIINFEKL peptide also resulted in more than 25% IFN-γ positive CD8+ cells. Most of IFN-γ positive cells also displayed cytotoxic activity as indicated by costaining with the CD107a degranulation marker (Fig 5, center column). In addition, the coexpression of GM-CSF enhanced the percentage of IFN-γ and interleukin 2 double-positive cells after peptide stimulation, confirming the polyfunctional nature of the CD8+ T cell responses induced by this potent immunization regimen (Fig 5, right column). About 90% of all tetramer-positive CD8 cells also showed the downregulation of CD62L expression (Fig 6), which further indicates an effector phenotype 3334. No qualitative differences among the various immunization groups were detectable, suggesting that GM-CSF only influences the quantity but not the functional properties of the vaccine-induced CD8+ T-cells.

In contrast to the GM-OVA vaccines, the coexpression of GM-CSF and ovalbumin did not suppress CD8 T cell responses. This indicates that the suppressive effects of GM-OVA vaccines on the CD8 T cell responses are not simply due to the biological activity of GM-CSF. To exclude the possibility that immune responses were affected by altered antigen expression levels, subcellular localization, and/or stability of the fusion protein, mice were also immunized with DNA and adenoviral vector vaccines expressing a fusion protein of rhesus macaque GM-CSF and ovalbumin. Although fully active on primate cells, rhesus monkey GM-CSF was inactive in rodent cells (Fig. 1C). In all parameters investigated, including Tetramer analyses and intracellular staining for IFN-γ, interleukin 2, and/or CD107a, the immune response induced by the rhesus monkey GM-OVA vaccines did not differ significantly from the response induced by the OVA vaccines (Fig. 5 and 6).

The GM-OVA vaccines not only suppressed cellular immune responses but also the antigen-specific IgG2a antibody response (Fig 4B). Therefore, ovalbumin-specific antibody levels of the IgG1 and IgG2a subtype were also determined for the vaccines coexpressing GM-CSF and ovalbumin or expressing the fusion protein of inactive rhesus GM-CSF and ovalbumin. Vaccinations with GM-OVA again suppressed OVA specific IgG2a antibody responses in comparison to the vaccine expressing only ovalbumin (Fig. 7). In contrast, the coexpression of GM-CSF or the expression of the fusion proteins of rhesus monkey GM-CSF and ovalbumin did not affect IgG1 and IgG2a antibody levels in the DNA prime adenoviral vector boost immunization regimen.

Whereas CD8 T cell responses upon injection of the Ad-OVA vaccine were significantly increased in DNA-OVA primed mice, such a priming effect was not observed with the DNA-GM-OVA prime and Ad-GM-OVA boost regimen (Fig. 3), suggesting that immune responses elicited by priming with DNA-GM-OVA might have suppressed the secondary response. Since neutralizing antibodies to GM-CSF have been observed previously after injection of recombinant GM-CSF proteins 3536, we analysed sera from immunized mice for the presence of GM-CSF neutralizing antibodies using the GM-CSF-dependent FDCP-1 cells. While the sera of mice immunized with the GM-OVA vaccines inhibited GM-CSF-dependent cell growth down to a 1:100 dilution, no inhibitory activity on FDCP-1 cells was observed for the sera of mice immunized with vaccines coexpressing GM-CSF and OVA or expressing the monkey GM-CSF ovalbumin fusion proteins (Fig. 8). Consistently, autoantibodies binding to GM-CSF in an ELISA were only detected in mice immunized with vaccines encoding the fusion proteins, but not in any of the other vaccine groups (data not shown).

Since GM-CSF knock-out mice had an impaired CD8+ T cell response after immunization with peptides 37, we determined the relevance of neutralizing anti-GM-CSF antibodies induced by immunization with the GM-OVA vaccine, in the Friend-Virus (FV) infection model. Immunological control of FV infection was shown to be dependent on antibodies and CD4+, and CD8+ T-cell responses 3839. To induce neutralizing GM-CSF antibodies mice were primed with the DNA and boosted with the adenoviral vector vaccine, both expressing GM-OVA. As controls, mice were immunized against OVA with vaccines expressing either ovalbumin alone or GM-CSF and ovalbumin from a bicistronic expression cassette. One week after the adenoviral vector boost, neutralizing antibodies were detected in all animals immunized with the GM-OVA vaccines, but not in any of the other groups. The antibodies remained detectable during the whole observation period (data not shown). Induction of virus-specific CD8+ and CD4+ T cell responses was determined by MHC-I and MHC-II tetramer staining 11 days post challenge with FV. Similar levels of antigen-specific CD4+ and CD8+ T cell responses were observed in all immunized groups (Fig. 9A,C). In addition, FV loads in the blood and the spleen were not affected notably by the type of vaccine antigen (Fig. 9B,D). Thus vaccine-induced neutralizing antibodies against GM-CSF neither influenced T cell responses nor viral load after FV infection. Interestingly, all immunized groups had significantly reduced numbers of FV-specific CD8⁺-T-cells compared to the non-immunized control animals (Fig. 9A), suggesting some interference between vaccine-induced immune responses and antiviral CD8+ T cell responses after FV infection.

The DNA constructs used for the immunization studies are all based on the expression plasmid pcDNA3.1 (Invitrogen, Karlsruhe, Germany) or pShuttle-CMV 53. The coding sequence of murine GM-CSF was amplified by RT-PCR from RNA isolated from stimulated mouse splenocytes and cloned into the pCR-2.1-TOPO vector (Invitrogen). All other plasmids were constructed via standard cloning techniques including overlap extension PCR. Transgene expression is driven in all constructs by the immediate early promoter/enhancer region of human cytomegalovirus. The open reading frame of the fusion protein GM-OVA consisting of murine GM-CSF, a [Gly₄Ser]³ linker and OVA was cloned into the pcDNA3.1 vector. pOVA, also a pcDNA3.1 derivative, encodes ovalbumin itself. Both proteins have a C-terminal HIS₆-tag. The plasmids pGM^rh-OVA, pGM-DP-OVA, and pΔGM-OVA are based on the pShuttle-CMV vector. The GM^rh-OVA fusion protein is equivalent to the mouse one with the sequence of murine GM-CSF being replaced by the rhesus monkey homologue. In the plasmid pGM-DP-OVA, transgene expression is driven by a bidirectional version of the CMV/TetO2 promoter 54. In pΔGM-OVA the coding sequence of ovalbumin is preceeded by the 17 amino acid long signal peptide of murine GM-CSF. All DNA preparations for immunizations were carried out with the Endofree Plamid Mega or Giga Kit (Qiagen, Hilden, Germany).

All E1-deleted, replication-defective adenoviruses with the corresponding expression cassettes (Ad-GM-OVA, Ad-GM-DP-OVA, Ad-rhGM-OVA, Ad-ΔGM-OVA) were generated by the AdEasy-system 53. The pShuttle plasmids and pAdEasy1 were electroporated into BJ5183 bacteria as previously described 54. Correctly recombined plasmids were transfected into 293 cells. Viral vectors growing out were checked for transgene expression by Western Blot analyses and GM-CSF bioactivity, if applicable. Vector particles were purified by CsCl gradient centrifugation and quantified by optical density measurements. In addition, the TCID₅₀of the vectors were determined on 293 cells. The adenoviral vector preparations were also tested for endotoxin levels with the LAL quantification assay (Cambrex Bio Science, Verviers, Belgium), confirming that the dose used for immunization of mice contained less than 0,1 EU.

HEK293 and 293T cells were cultured in DMEM containing 10% FCS and 1% penicillin/streptomycin. RPMI 1640 supplemented with 10% FCS, 2 mM L-Glutamine, 10 mM HEPES, 50 μM β-Mercaptoethanol and 1% antibiotic/antimycotic (all Gibco, Karlsruhe, Germany) was used for the lymphocyte cultures (R10-medium). Bone-marrow cultures were grown in RPMI 1640, supplemented with 10% FCS, 1% penicillin/streptomycin, 4 mM L-Glutamine, 1 mM sodium pyruvate and recombinant mouse IL-4 (1 ng/ml) and GM-CSF (5 ng/ml) (Biomol, Hamburg, Germany).

293T cells were transfected using the calcium/phosphate precipitation method. Supernatants were collected 48 h after transfection and tested for protein expression by Western Blot analysis. A combination of rabbit-α-OVA (Chemicon International, LTD, Hampshire, UK) and goat-α-rabbit-HRP (Sigma, Munich, Germany) antibodies was used for detection. To confirm the bioactivity of GM-OVA, it was purified from supernatants of transfected cells by Ni-NTA-affinity chromatography as described by the manufacturer (Qiagen, Hilden, Germany). The purified protein migrated as a single band of the expected size in Coomassie stained polyacrylamide gels. The GM-CSF-dependent FDCP-1 cells were incubated with the purified GM-OVA or supernatants from NIH 3T3 cells stably expressing GM-CSF. The GM-CSF content of the supernatants had been determined by ELISA. After 48 h, GM-CSF-dependent cell growth was monitored by MTT-assay as described elsewhere 5556. Additionally, the ability to generate DCs from bone marrow cultures was tested as previously described 57. Briefly, bone marrow derived monocytes were cultured in the prescence of IL-4 and recombinant GM-CSF or GM-OVA for 8 days, with or without LPS maturation for 24 h. Their phenotype was characterized by surface staining with the antibodies α-CD11c-APC, α-CD80-FITC, α-CD86-FITC (all BD Bioscience, Heidelberg, Germany) and α-CD83-PE (eBioscience, San Diego, USA) in FACS analyses.

Splenocytes of transgenic OT-I mice were incubated with CFSE (3μM; Molecular Probes, Eugene, Oregon, USA) at a density of 8 × 10⁷ cells/ml for 6 min at room temperature with gentle mixing. The labelling reaction was stopped by adding one volume of FCS after which the cells were washed twice with PBS. Thereafter, cells were plated into 96-well plates at a density of 1 × 10⁶ cells/well and incubated with different concentrations of either GM-OVA or OVA (500 to 0,5 ng/ml) for 4 days. After washing the cells twice with PBS containing 0,5% BSA and 1 mM sodium azide (PBS/BSA/Azid), cell proliferation was measured in FACS analysis. Cells, which were incubated with OT-I peptide, were used as positive controls whereas non-stimulated cells served as negative ones.

6–8 week old female C57BL/6N mice were purchased from Janvier (Le Genest-ST-Isle, France) and housed in singly-ventilated cages in accordance with the national law and institutional guidelines.

All vaccines were diluted in PBS and injected subcutaneously in both hind foot pads. For single dose experiments, mice were immunized on day 1 with either 50 μg of DNA (pOVA or pGM-OVA) or 5 × 10⁹ adenoviral particles (Ad-ΔGM-OVA resp. Ad-GM-OVA) corresponding to 2 × 10⁸ 50% tissue culture infectious doses (TCID₅₀). After one week, serum samples were collected and on day 8 animals were sacrificed to analyze the CTL responses. In the prime-boost experiments, all animals received an additional DNA injection (pOVA or pGM-OVA) 5 weeks prior to the above mentioned protocol. Serum samples were collected 28 days after the first and 7 days after the second immunization, whereas the T-cell assays were carried out 8 days after the second immunization.

Blood was taken retro-orbitally and serum was collected after centrifugation for 5 min at 5.000 g in a table top centrifuge. Ovalbumin protein (Sigma) was coated on 96-well plates (MaxiSorb, Nunc, Wiesbaden, Germany) at a final concentration of 5 μg/ml. After blocking with 5% milk powder, serum samples were added at appropriate dilutions and incubated for 1 h, followed by intensive washing. Alkaline phosphatase-coupled antibodies against mouse IgG1 or IgG2a antibodies (BD Bioscience) were added and incubated for 1 h. The enzymatic reaction was developed with the pNPP substrate (Sigma) for 30 min. Reaction was stopped by sodium hydroxide solution (1 M) and the optical densities were measured at a wavelength of 405 nm.

Splenocytes were collected at indicated time points. After red blood cell lysis, 1 × 10⁶ cells were plated in 96-well round-bottom plates (Nunc) for each staining.

For the tetramer staining, cells were washed once and incubated with 2 μl of SIINFEKL/H-2K^b-APC tetramers (Sanquin, Amsterdam, NL) in total volume of 100 μl PBS/BSA/Azid for 40 min at room temperature. After surface staining with α-CD8-FITC, cells were incubated with 7-amino-actinomycin D (7-AAD) for 5 min to exclude dead cells from subsequent FACS analyses. In some experiments, α-CD62L-PE antibodies were included in the tetramer analysis to characterize the CD8⁺ activation status.

For ICS, samples were stimulated for 6 h in the presence of 2 μM Monensin, which inhibits the cytokine secretion, and 1 μl α-CD107a-FITC, which is a marker for lymphocyte degranulation 32. Cells were either stimulated by α-CD3 and α-CD28 antibodies (2 μg/ml and 1 μg/ml, respectively) or the OT-I/SIINFEKL peptide (2 μg/ml, Genaxxon, Biberbach, Germany) and compared to non-stimulated cultures. After stimulation, surface staining was carried out with αCD8-PerCP or αCD4-FITC (BD Bioscience). Cells were fixed in 2% paraformaldehyde, followed by permeabilisation with 0,5% Saponin in PBS/BSA/Azid buffer. Cytokines were detected with αIFN-γ-PE and αIL-2-AlexaFluor647.

Mice were injected intravenously with 0.5 ml phosphate buffered saline (PBS) containing 3,000 spleen focus-forming units (SFFU) of the Friend virus complex (FV). The B-tropic, polycythemia-inducing FV complex used in all experiments was from uncloned virus stocks obtained from 10% spleen cell homogenates as described previously 58.

For infectious center assays, single-cell suspensions from infected mouse spleens were cocultivated with Mus dunnis cells at 10-fold dilutions. Cultures were incubated for 5 days, fixed with ethanol, stained with F-MuLV envelope-specific monoclonal antibody 720, and developed with peroxidase-conjugated goat antimouse antibody and aminoethylcarbazol to detect foci. Tetramer analyses were done by flow cytometry as described previously 59. For the quantification of Friend virus-infected blood cells, single-cell suspensions of nucleated, live cells were analyzed by flow cytometry. To detect Friend virus infection cells were stained as described previously with tissue culture supernatant containing Friend murine leukemia virus glycosylated Gag-specific monoclonal antibody 34 (AB34) 60.

Results are expressed as the means ± standard errors of the means (SEM). Statistical comparisons were performed by one-way ANOVA test, followed by a Bonferroni post test using the Prism 4.0, GraphPad Software. P < 0,05 was considered as statistically significant.