The partial POMP91B recombinant protein was expressed in E. coli BL21(DE3) and purified using Ni-gel chromatography. The purified POMP91B protein fragment was analysed by SDS-PAGE as shown in Figure 1(a). For the expression of POMP91B into the flagellin of E. coli, the 5’ part of the pomp91B gene was cloned in frame with the fliC-trx fusion gene present on the plasmid pFliTrx. The gene was inserted in the unique RsrII site of the thioredoxin gene allowing directional insertion. Positive colonies were tested by PCR and confirmed by DNA sequencing. To control the expression of the new chimeric flagellin two tests were used. In the first test, cells were assayed by swarming test as described in the Experimental Section 4.2. Results in Figure 1(b) indicate that cells carrying the POMP91B hybrid flagellin were motile in semi-solid IMC medium supplemented with tryptophan. In the second test, tryptophan-induced cell lysates were fractionated on SDS gel, transferred to a membrane and probed with monoclonal anti- thioredoxin or anti-POMP91B antibodies, thus confirming the presence of POMP91B protein in transformed cells [Figure 1(c)].

The cellular responses were first evaluated by measuring the proliferation of splenic cells after in vitro stimulation with the recombinant POMP91B protein. Proliferation was determined with a classical [³H] thymidine incorporation assay. Groups of mice were divided into a subset that was antigen-boosted one week before the test and another subset that was not boosted. No significant differences in the cellular responses were observed between the two subsets (data not shown). Bacteria with non-modified flagellin (C‑B), plasmid with no insert (C-D) and PBS (C-P) were used as negative controls, respectively.

Unlike P27 antigen that was able to induce strong cellular proliferations (Figure 2 and [20]), especially after immunization with the recombinant bacteria having the p27 protein inserted in flagellin, the POMP91B antigen was unable to promote such an intense response, regardless of the immunization procedure used. Immunization with the bacteria (group B) gave the highest proliferation (93.517 cpm ± 19.254) for P27 antigen, whereas only 15.592 cpm ± 5.383 was obtained when the POMP91B antigen was used [Figure 2(a)]. Immunization by the recombinant protein emulsified first in CFA then by IFA (group P) induced a small increase in proliferation (20.544 cpm ± 3.850), but the levels were inferior when compared to those observed with the P27 antigen (38.023 ± 5.223). The weakest responses were obtained by DNA immunizations using pcDNA3 plasmid containing the pomp91B gene (group D), i.e., 3.795 cpm, or by the soluble protein without any adjuvant (group sP), i.e., 9.728 cpm.

Combined immunizations with bacteria followed by boosts with the POMP91B protein in IFA (group B/P) as shown in Figure 2(b), increased the intensity of the response as compared to that obtained by the modified bacteria alone (15.592 cpm ± 5.383 to 30.533 cpm ± 7.574), and gave the strongest proliferation for this antigen, a result that is contrary to that observed with the P27 antigen (decrease from 93.517 cpm ± 19.254 to 62.593 cpm ± 9.590). In contrast to POMP91B, immunization with protein in CFA followed by immunization with the flagellin modified bacteria (group P/B) led to a relatively decreased response (20.544 cpm ± 3.850 to 17.693 cpm ± 3.001). DNA priming followed by the POMP91B protein (group D/P) or modified bacteria (group D/B) boosts also led to an increase in response intensity (3.795 cpm to 16.703 cpm and 11.347 cpm respectively).

To investigate the cytokine profile induced by the different immunization methods (plasmid DNA, recombinant protein, bacteria with modified flagellin), we measured the amount of antigen-specific IFN‑γ and IL-4 secreted by splenic cells isolated from immunized mice. The spleens were harvested one week after the last injection. Cell suspensions were stimulated in vitro with the POMP91B or P27 recombinant protein using different concentrations. The cytokine tests were performed on the cells supernatants after 48 hours and after one week of culture. At 48 hours a weak response was observed for both cytokines (data not shown). However, one week later high levels of IL-4 were secreted by splenic cells from the group of mice immunized by the recombinant protein emulsified in Freund’s complete and incomplete adjuvant [Figure 3(c) and Figure 3(d); indeed, one of the three groups that secreted the highest IL-4 concentration was that of mice immunized with the plasmid and then recombinant protein (333 pg/mL)]. The two other groups boosted with the recombinant protein in IFA, B/P and P, also induced a high release of IL-4 (212 and 243 pg/mL respectively). The group immunized with the protein alone had IL-4 concentration of 149 pg/mL. In the same conditions, P27 was unable to promote such a production of IL-4, and this level remains low for all other tested groups [Figure 3(c) and Figure 3(d)]. Cells from mice immunized by bacteria carrying the antigen POMP91B in their flagellin (B, D/B and P/B) produced lowest amounts of IL-4 (107, 39 and 52 pg/mL, respectively) whereas IL-4 released by cells from mice immunized exclusively with plasmid DNA (6.5 pg/mL) is not significantly different from the control.

IFN‑γ production [Figure 3(a) and Figure 3(b)] for the POMP91B antigen was low in respect to that obtained with the P27 antigen. The highest response (113 pg/mL) was obtained with cells from mice primed with flagellin-modified bacteria and boosted with recombinant protein in IFA (B/P group). All other groups (P, sP, D/B, D/P, P/B) showed comparable IFN-γ levels, i.e., between 26 and 72 pg/mL, with the exception of the groups B and D in which IFN-γ release was completely absent. This is opposite to the effect of P27 antigen since the B group induced the best response [Figure 3(a)].

To investigate the anti-POMP91B antibodies titer in the sera of mice immunized with different combinations of antigens and their controls, the levels of IgM, IgG1, IgG2a, IgG2b, IgG3 and IgA were determined by ELISA one week after the second and the third immunizations. The optical densities (ODs) obtained for the four IgG isotypes, IgA, and IgM, were measured with sera diluted at 1/500.

Similar to the cellular response, the antibody response depended on the PAMPs used during immunization. After the second (first boost) immunizations [Figure 4(a)], the group immunized with the protein in Freund’s adjuvant (P) gave the strongest antibody response; for IgG1 (O.D. = 1.130) and for IgG2a (O.D. = 0.702). Without Freund’s adjuvant, when the protein was in saline solution (sP), the response was also robust, but only for IgG1 (O.D. = 1.101) and not for IgG2a (O.D. = 0.156).

The group immunized with flagellin modified bacteria (B) only developed a weak IgM response (O.D. = 0.153), and the DNA immunized group did not show any significant antibody response. The response obtained by group (P/B) [Figure 4(b)], primed with recombinant protein in CFA and boosted by the flagellin-modified bacteria showed weaker IgG2a response as compared to that observed for the Freund’s adjuvant group (O.D. = 0.788 and O.D. = 0.748 for IgG1 and IgG2a respectively). No significant responses were noted for the other groups, except for a weak IgM response in the two groups primed with the modified bacteria and B/P (O.D. = 0.153 and O.D. = 0.187 respectively).

After the third immunization [Figure 4(c) and Figure 4(d)], stronger responses were observed in all immunized groups. The best response was obtained in mice primed with flagellin-modified bacteria and boosted with the protein emulsified in incomplete Freund’s adjuvant (B/P) [Figure 4(d)], i.e., O.D. = 2.554 for IgG2a, 2.018 for IgG2b and 1.903 for IgG1. Three other groups (P, D/P and P/B) present similarly weaker antibody isotype response with the highest response with IgG2a, followed by IgG2b and then IgG1. The order of intensity was different with the P27 antigen since the IgG2b was dominant, followed by IgG2a and then by IgG1. On the contrary to the P27 antigen, in these four groups POMP91B promoted IgG3 responses. In the group primed and boosted with the flagellin-modified bacteria (B), IgM was the only significant immunoglobulin (O.D. = 0.780). This is similar to the group boosted by the flagellin-modified bacteria after priming with DNA (D/B) (O.D. = 0.722). No significant augmentation of any isotype could be detected in sera from mice immunized with the DNA plasmid for the POMP91B antigen, in contrast to the effect of P27 where IgG2a could be observed. Mice in the sP group gave a strong IgG1 response (O.D. = 2.370) [Figure 4(c)] but has also a good IgG2a response (O.D. = 1.905) that is not seen with the P27 antigen. IgA responses were negligible regardless of the nature of the antigen used for immunization.

Escherichia coli strains DH5a (Invitrogen, San Diego, CA, USA) and BL21 (DE3) (Novagen, San Diego, CA, USA) were used for cloning and overexpression of the recombinant protein, respectively. For the production of recombinant protein, the 5’part of the pomp91B gene was amplified by PCR from an original template plasmid provided by Dr. Olivier Grépinet (INRA Tours, Nouzilly, France) using oligonucleotide pair 5’-GGG AAT TCC ATA TGA AAC ATC CAG TCT ACT GGT T-3’ and 5’-CGC GGA TCC GGA CGG TGT TTG TAA CGT A-3’. The oligonucleotide primers (Eurogentec, Seraing, Belgium) contained a NdeI and a BamHI site (underlined) respectively. PCR products were purified, digested by the two restriction enzymes NdeI and BamHI and ligated during 16 h at 4 °C with T4 ligase (Eurogentec) into the expression vector pET15b (Novagen) pre-digested with NdeI and BamHI. This plasmid encodes six histidines tagged to the N-terminus of the expressed recombinant protein to be purified. Expression of the recombinant protein was induced by adding 1 mM isopropyl-b-_D-thiogalactopyranoside (IPTG) to BL21(DE3) bacteria (OD₆₀₀ = 0.6) carrying the recombinant pET-15b-pomp91b plasmid in 2 × TY medium at 28 °C for 3 to 4 hours. The bacteria were harvested by centrifugation, resuspended and lysed by incubation for 1 hour in lysis buffer (100 mM Na₂HPO₄, 10 mM Tris-HCl, 8 M urea, pH 8.0). After centrifugation (10,000 g, 30 min), the supernatant was added to a Ni-NTA resin (Qiagen, Hilden, Germany). The mixture was incubated for 1 h at room temperature. After two washes with the same buffer, recombinant proteins were eluted with urea 8 M at pH of 5.9, then 4.5. The purified recombinant proteins were renatured by dialysis against PBS pH 7.4 at 4 °C, quantified by a Bradford assay, filtered, and stored in the same buffer at ‑20 °C.

For DNA immunization, the gene encoding the N-part of the POMP91B protein was cloned in the eukaryotic expression vector pcDNA3 (Invitrogen). Plasmid DNA was amplified in E. coli DH5a, purified by Qiagen purification kit, dissolved in phosphate-buffered saline (PBS), and stored at ‑20 °C until use.

For immunizations with the bacteria carrying the POMP91B protein in their flagellin FliC protein, a plasmid (pFliTrx) coding for fliC gene and a specific E. coli flagellin-deficient strain were used as described [21]. The plasmid pFliTrx, carries the incomplete E. coli fliC coding sequence in which the coding region for amino-acid residues 244-351 of the E. coli flagellin has been replaced, in frame, with the entire coding sequence for the E. coli thioredoxin (trxA) gene. The partial pomp91b gene was cloned in the trxA gene at the unique RsrII restriction site. The two primers used (5’-GGG GGG GTC CGA AAC ATC CAG TCT ACT GGT-3’ and 5’-GGG GGG GAC CGG ACG GTG TTT GTA ACG TA-3’) have an AvaII restriction site (underlined) allowing a directional cloning. Plasmid constructs were then used to transform the non-motile E. coli GI809 strain [21], which carries a specific 512 bp deletion within the flagellin gene (fliC, GenBank accession #M14358). Transformed motile bacteria were assessed for motility by swarming assay and by western blotting using specific antibodies, before preparation of immunization solutions.

Non-motile E. coli GI809 cells were transformed by a flagellin containing pFliTrx plasmid and placed in a Petri dish containing IMC semi-gelosed medium with an agar specific for motility (Difco, Detroit, MI, USA) at 0.3%. The incubation was realized at 28 °C and the motility was observed at least 15 h after the start of the culture.

Purified proteins and bacterial extracts were heated 5 min at 100 °C in the lysis buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 4% SDS, 5% b-mercaptoethanol) and electrophoresed through a 12.5% polyacrylamide gel, transferred to nitrocellulose membranes, then blotted with anti-POMP91B or anti-thioredoxin antibodies as described in Le Moigne et al. [19].

Adult female BALB/c mice (Janvier, Le Genest, France) were divided in groups of six mice and immunized either with 50 mg of plasmid DNA carrying the antigen coding sequence, in 100 mL PBS, three times with one month interval in the quadriceps muscle (DNA immunization), with 25 mg of antigen emulsified in complete Freund's adjuvant (CFA, containing the H37Ra, ATCC 25177 M. tuberculosis strain) for the first immunization or incomplete Freund's adjuvant (IFA) for the two following immunizations, by a subcutaneous route (Classical immunization), with 50 mg of antigen in PBS (Soluble antigen) or with 10⁸ modified E. coli bacteria which express the antigen into the flagellin at the surface of the bacteria given intraperitonneally (Flagellin immunization). Control mice for DNA immunization were immunized with empty pcDNA3 plasmid, whereas control mice for classical immunization and soluble antigen immunization were immunized with PBS incorporated in Freund's adjuvant or PBS alone, respectively. For the flagellin immunization, control mice were immunized with non modified E. coli (bacteria which express normal flagellin). Combined immunizations between the different methods were applied as in Table 1.

Blood samples were collected one week after the last immunization from the retro‑orbital plexus and sera were stored at −20 °C until use. All animal experiments were done in accordance with institutional and national ethical guidelines.

Plates coated overnight at 4 °C with 1 µg/mL of recombinant proteins in 100 µL of carbonate-bicarbonate buffer (0.1 M, pH 9.6), were washed with phosphate-buffered saline-Tween 20 (PBS-T) (0.05%; v/v) and blocked for one hour at 37 °C with PBS-T containing 0.5% gelatin (PBS-T-G). Then serum dilutions (1/500) in the same buffer were added. After 1 h of incubation at 37 °C and four washes, goat anti-mouse IgM, IgG1, IgG2a, IgG2b, IgG3 and IgA alkaline phosphatase-conjugate (Southern biotechnology, Birmingham, AL, USA) were added and the plates were incubated for another 1 h at 37 °C. After four washes, 100 µL of 1 mg/mL of p-nitrophenylphosphate (Sigma, Saint Quentin Fallavier, France) in diethanolamine buffer (pH 9.8) was added, and then plates were read at 405 nm with a Titertek Multiscan instrument (Skatron, Oslo, Norway).

To check the effect of antigen boosting in our system, only half of the mice were boosted intraperitonneally with 25 mg of soluble antigen in PBS six months after the first immunization. One week later, splenic cells (5 × 10⁵ cells/well) from antigen-boosted or non-boosted mice were suspended in RPMI-10% (FCS) and plated in 96-well flat bottomed plates containing serial dilution of antigen. Two days later, 0.5 µCi of [³H] thymidine was added to each well and incubated for an additional 18 hours, then cells were harvested and thymidine incorporation was measured using a liquid scintillation spectrophotometer Packard 1600 Tricarb (Packard, Downers Grove, IL, USA).

Splenic cells (5 × 10⁵ cells) from boosted, non-boosted and control mice were cultured in RPMI-10% fetal calf serum (FCS) at 37 °C in 5% CO2 in 96-well plates containing different concentrations: 0, 1.1, 3.3 and 10 µg of recombinant proteins per mL. Supernatants were harvested after 48 h and one week of incubation and stored at ‑20 °C until use. IL-4 and IFN‑γ contents were measured with a commercial ELISA kit (Bender MedSystems, Vienna, Austria) according to the manufacturer’s protocol.