Enhanced Antigen Processing of Flagellin Fusion Proteins Promotes the Antigen-Specific CD8⁺ T Cell Response Independently of TLR5 and MyD88

Abstract

Flagellin is a highly effective adjuvant for CD4⁺ T cell and humoral immune responses. However, there is conflicting data in the literature regarding the ability of flagellin to promote a CD8⁺ T cell response. In this article, we report that immunization of wild-type, TLR5^−/−, and MyD88^−/− adoptive transfer recipient mice revealed the ability of flagellin fusion proteins to promote OVA-specific CD8⁺ T cell proliferation independent of TLR5 or MyD88 expression by the recipient animal. Wild-type and TLR5^−/− APCs were able to stimulate high levels of OVA-specific CD8⁺ T cell proliferation in vitro in response to a flagellin fusion protein containing full-length OVA or the SIINFEKL epitope and 10 flanking amino acids (OVAe), but not to OVA and flagellin added as separate proteins. This effect was independent of the conserved regions of flagellin and occurred in response to OVAe alone. Comparison of IFN-g production by CD8⁺ effector cells revealed higher levels of SIINFEKL peptide–MHC I complexes on the surface of APCs that had been pulsed with OVAe–flagellin fusion proteins than on cells pulsed with OVA. Inhibition of the proteasome significantly reduced Ag-specific proliferation in response to OVAe fusion proteins. In summary, our data are consistent with the conclusion that flagellin–OVA fusion proteins induce an epitope-specific CD8⁺ T cell response by facilitating Ag processing and not through stimulatory signaling via TLR5 and MyD88. Our findings raise the possibility that flagellin might be an efficient Ag carrier for Ags that are poorly processed in their native state.

Flagellin, the ligand for TLR5 (1-4), is a potent adjuvant in numerous model-system vaccination regimens (5-18). Over the past several years, a strong interest has emerged in developing flagellin as an adjuvant for use in human vaccines to stimulate humoral and cell-mediated immune responses. Currently, the clinicaltrials.gov database lists five ongoing or completed trials examining the safety of flagellin as a vaccine adjuvant for use in humans. In addition, we have made a vaccine designed to protect against Yersinia pestis (13) that will shortly begin a phase I clinical trial.

This enthusiasm for flagellin results from a combination of several factors. Early studies carried out by Arnon and colleagues (5-10) revealed the potential for flagellin as an adjuvant. Sub-sequently, numerous findings have been published that have examined the structure of flagellin (19-21) and the regions required for signaling via TLR5 (22-29), making flagellin an extremely well characterized molecule. The pathways that are engaged and ultimately result in NF-kB activation following binding of flagellin to TLR5 are well documented (reviewed in Refs. 30, 31). Key findings that explain the adjuvant mechanism of flagellin on the cellular level have also been published. Flagellin promotes a strong, Ag-specific CD4⁺ T cell response (32) through a mechanism that is dependent on direct stimulation of TLR5-expressing CD11c⁺ cells (33). Activation of CD4⁺ T cells, in turn, promotes a strong humoral immune response and results in high levels of protective Abs.

From a vaccine production standpoint, flagellin has several benefits over other adjuvants. Large quantities of recombinant flagellin can easily be produced in Escherichia coli. Use of flagellin-Ag fusion proteins further streamlines production and regulatory approval because the active components of the vaccine are contained in a single molecule. Pre-existing immunity to flagellin does not reduce its efficacy as an adjuvant (8, 11), so flagellin is well suited to serve as a vaccine platform for making vaccines designed to protect against a wide range of pathogens.

Despite the depth of our collective understanding of flagellin, the ability of flagellin to promote a CD8⁺ T cell response remains unsettled. Cuadros et al. (34) showed that cells from mice immunized with a flagellin-enhanced GFP (EGFP) fusion protein were far superior at lysing EGFP-loaded target cells than were cells from mice immunized with only EGFP. Similarly, Huleatt et al. (35) showed that the response to OVA_257-264, based on IFN-g production, was significantly greater in mice immunized with flagellin–OVA fusion protein than in mice immunized with only OVA. Immunization of mice with flagellin containing a Plasmodium yoelii circumsporozoite epitope resulted in increased IFN-g production following in vitro restimulation of CD8⁺ T cells (36). In contrast, Datta et al. (37) found that treatment of APCs with flagellin did not result in APC activation or increased OVA-specific CD8⁺ T cell response following incubation of APCs with flagellin and OVA. Similarly, Schwarz et al. (38) found that, compared with naive mice, mice immunized with flagellin mixed with virus-like particles containing the p33 epitope of lymphocytic choriomeningitis virus showed no significant increase in the percentage of circulating p33-specific CD8⁺ T cells or in the ability of immunized mice to control viral infection.

Taken together, the above reports are consistent with the conclusion that for flagellin to effectively promote a CD8⁺ T cell response, it must be fused to the specific Ag. However, none of these studies tested the TLR5 dependency of the observed effect. Consequently, it is unclear whether the effects they observe result from stimulation of TLR5 by flagellin or from some other mechanism. In view of the robust effort to develop flagellin as an adjuvant for a broad range of humoral and cell-mediated vaccines, it is extremely important to determine if, indeed, flagellin can promote an Ag-specific CD8⁺ T cell response and associated memory. We have developed an experimental approach to address this issue and offer a mechanism that reconciles the contradictory results in the literature.

Materials and Methods

Mice

Female, 6-wk-old C57BL/6 mice and MyD88^−/− (39) mice were purchased from The Jackson Laboratory. OT-I transgenic mice (40) on a RAG^−/− background were kindly provided by Dr. Martha Alexander-Miller (Wake Forest University School of Medicine, Winston-Salem, NC) and crossed with CD90.1-expressing mice purchased from The Jackson Laboratory. F1 mice were used as donor mice for adoptive transfers and in vitro proliferation experiments. TLR5^−/− (41) mice have been previously described. CD11c-DTR/GFP mice (42) were purchased from The Jackson Laboratory and bred in our facility. All mice were housed in the Wake Forest University School of Medicine animal facility in accordance with institutional guidelines. All experiments were approved by the Institutional Animal Care and Use Committee.

Immunogens

Recombinant His-tagged Salmonella enteritidis flagellin was produced in E. coli and purified using a nickel affinity resin, as previously described (11, 43). Flagellin–OVA fusion protein was prepared as noted earlier (33). In this fusion protein, the hypervariable (HV) region has been replaced with full-length OVA. Recombinant OVA epitopes (OVAe)–flagellin was produced by cloning the OVAe recognized by the OT-I transgenic TCR (residues 257–264) and the OT-II transgenic TCR (residues 323–339), along with the 10 flanking amino acids on each side of both epitopes onto the N terminus of Salmonella and Pseudomonas flagellins. Flanking amino acids on either side of the epitope were included in the construct in case proximal residues are important for processing of the epitopes (44). We refer to the OVA portion of the fusion protein as OVAe throughout this article. For the HV fusion proteins, OVAe was cloned onto the N terminus of the HV region of Salmonella (aa 180–419) or Pseudomonas (aa 181–398) flagellin. The HV region of flagellin does not bind to TLR5 and thus does not trigger an innate immune response (22). Full-length OVA was purchased from Sigma-Aldrich. All proteins were passed through Acrodisc Mustang Q membranes (Pall Corporation) to remove contaminating endotoxin and nucleic acids. Endotoxin levels were verified to be,30 pg of endotoxin per microgram of protein by Limulus amebocyte lysate analysis (Associates of Cape Cod).

Adoptive transfers

Lymph nodes (LNs) and spleens were harvested from OT-I × CD90.1 mice, and cell suspensions were enriched for untouched CD8⁺ T cells using the Miltenyi negative selection protocol. This protocol typically resulted in a cell population consisting of 95% CD8⁺ T cells. CD8⁺-enriched cells were then labeled with CFSE (Invitrogen) for 10 min at room temperature, washed, and resuspended in PBS. A total of 1 × 10⁶ cells were injected via the tail vein into recipient mice in a volume of 300 μl PBS. Mice were i.m. immunized with 10⁻¹⁰ mol Ag 1 d following cell transfer. All mice were sacrificed 3 d following immunization. A single-cell suspension was generated from the draining popliteal LNs and stained to identify CD8⁺ (53–6.7) and CD90.1⁺ (OX-7) cells, using Abs from Becton-Dickinson. CFSE dilution by CD8⁺CD90.1⁺ cells was analyzed by flow cytometry. Samples were collected using a FACSCalibur flow cytometer (Becton-Dickinson), and data were analyzed using FlowJo software (Tree Star).

In vitro proliferation assays

CFSE-labeled OT-I cells were prepared as described above and combined with unlabeled C57BL/6 LN cells at a ratio of 1:9 and plated in a 96-well, round-bottom plate (Corning). A total of 5 × 10⁵ cells were plated per well and stimulated with the stated concentrations of immunogens in RPMI 1640 containing 10% FBS (Lonza). For experiments using chloroquine (Sigma-Aldrich) and lactacystin (Cayman), LN cells were incubated in 50 μM chloroquine or 20 μM lactacystin for 2 h prior to addition of Ag and for an additional 6 h following addition of Ag. Cells were then washed to remove Ag and incubated overnight in 5 μM chloroquine or 2 μM lactacystin. In total, cells were incubated in inhibitors for ~20 h (8 h at the higher concentration and 12 h at the lower concentration). The following day, LN cells were washed and plated with CD8⁺-enriched OT-I cells, using the same conditions as listed above. Three days later, cultures were harvested and analyzed by flow cytometry, as described above.

Effector cell assay

Production of clonal effector cells has been previously described (45). High-avidity OT-I effector cells were kindly provided by Dr. Martha Alexander-Miller. CD11c⁺ cells were enriched from the s.c. LNs of TLR5^−/− mice and pulsed overnight with Ag. The following day, APCs were washed and replated with OT-I effector cells. Cells were incubated for 6 h in the presence of brefeldin (Becton-Dickinson) and then harvested and stained to determine levels in intracellular IFN-γ (XMG1.2 from BioLegend).

Statistical analysis

Statistical analysis was performed with GraphPad Prism 5. Significance was based on the results of Student t tests (unpaired, one-tailed).

Results

Flagellin–OVA fusion protein promotes the in vivo proliferation of OT-I cells

We have previously shown that immunization with a flagellin–OVA fusion protein results in a significantly stronger OVA-specific CD4⁺ T cell response than does immunization with OVA and flagellin given as separate proteins and that this response is TLR5 and MyD88 dependent (33). To determine whether flagellin promotes Ag-specific CD8⁺ T cell responses, we employed an adoptive transfer system in which CFSE-labeled, CD90.1 + OT-I TCR transgenic T cells were transferred into sex-matched C57BL/6, congenic TLR5^−/−, or congenic MyD88^−/− mice. At 1 d following cell transfer, mice were immunized with 10⁻⁹ mol of OVA, OVA + flagellin, or flagellin–OVA fusion protein. TCR transgenic cells were recovered from the draining popliteal LN 3 d following immunization and analyzed for CFSE dilution by flow cytometry (Fig. 1). Immunization with OVA alone resulted in low levels of CFSE dilution by the Ag-specific cell population in each genotype of mouse. Inclusion of an equimolar dose of flagellin in the immunization mix had a small stimulatory effect on CFSE dilution by cells in wild-type mice and resulted in a significant increase (p < 0.05) in the number of OT-I cells in the draining LN. CFSE dilution and OT-I cell numbers in the draining nodes of TLR5^−/− and MyD88^−/− immunized with OVA + flagellin were similar to what was found in mice immunized with OVA only. Immunization with a flagellin fusion protein in which the HV region was replaced with full-length OVA resulted in extensive CFSE dilution and significant expansion of the OVA-specific cell population in wild-type, TLR5^−/−, and MyD88^−/− mice (p < 0.05 for all genotypes). These results are consistent with a mechanism through which fusion of flagellin to Ag facilitates a robust Ag-specific CD8⁺ T cell response (relative to OVA + flagellin) independently of TLR5 and MyD88.

FIGURE 1.

Flagellin–OVA fusion protein stimulates in vivo OVA-specific CD8⁺ T cell proliferation independent of a requirement for TLR5 or MyD88. C57BL/6 or TLR5^−/− mice received 1 × 10⁶ CFSE-labeled OT-I cells and were immunized the following day with 10⁻¹⁰ moles of OVA, OVA + flagellin, or OVAe-flagellin. Three days later, CFSE dilution (A) and OT-I cell number (B) in the draining popliteal LN was analyzed by flow cytometry. Plots are gated on CD8⁺CD90.1⁺ events. Data are representative of three mice per group. *p < 0.05 compared with the OVA group.

OT-I proliferation in response to OVAe-flagellin is not dependent on the conserved regions of flagellin or OVA

In view of the TLR5 and MyD88 independence of the CD8⁺ T cell response to fusion protein, we evaluated which region or regions of flagellin and/or OVA were required for the Ag-specific CD8⁺ T cell response. Our hypothesis was that some portion of the conserved regions of flagellin would be required for this activity, although we also considered the possibility that fusion of OVA to flagellin might have resulted in a protein with novel stimulatory activity. The role of the conserved sequences of flagellin in signaling via TLR5 (43, 22, 24, 25) and IPAF (46) is well established. Similarly, in plants, residues 29–50 of flagellin, known as flg22, bind to leucine-rich repeats in FLS2 (47) and trigger an innate-like immune response (48). To test the possibility that the response to flagellin–OVA functions through the IPAF pathway or another unknown pathway that is dependent on the conserved regions of flagellin or OVA, we created a number of additional fusion proteins using flagellins from Pseudomonas aeruginosa and S. enteritidis in combination with residues 247–274 and 313–349 from OVA. These two portions of OVA, termed OVAe, are the epitopes recognized by the OT-I (257–267) and OT-II (323–339) transgenic TCR, along with the 10 flanking amino acids on each side of each epitope. OVAe was fused to the N terminus of full-length flagellin from S. enteritidis and P. aeruginosa, as well as the HV regions from each type of flagellin (Fig. 2).

FIGURE 2.

Design of flagellin full-length and HV region fusion proteins. OVA_247-274 and OVA_313-349 were fused to N terminus full-length flagellin or the HV region of flagellin from S. enteritidis (residues 180–419) or P. aeruginosa (residues 181–398).

If the CD8⁺ T cell response is the result of activation of APCs via IPAF or some other non-TLR5 pathway that is dependent on the conserved regions of flagellin, deletion of the conserved regions should dramatically reduce the stimulatory effect of flagellin. Using an in vitro proliferation system, we compared the ability of full-length flagellin and HV region fusion proteins to stimulate OT-I cell proliferation. CD8-enriched, CFSE-labeled OT-I cells were mixed with LN cell suspension from wild-type or TLR5^−/− mice and stimulated with 10⁻⁷ M OVAe, OVAe–flagellin, or OVAe–HV fusion proteins. Cultures were harvested 3 d following stimulation, and CFSE dilution by OT-I cells was measured by flow cytometry. All proteins containing the OVAe region stimulated significant levels of OT-I cell proliferation, regardless of fusion to full-length flagellin or the HV region. CFSE dilution between cells stimulated using wild-type or TLR5^−/− LN cells as APCs was similar (Fig. 3A). The greatest number of OT-I cells were recovered from wells stimulated with OVAe (Fig. 3B). Wells stimulated with OVAe also had the highest percentage of divided cells (Fig. 3C). This finding is not surprising, given the likely possibility that the limited processing required for OVAe relative to the flagellin fusion proteins may allow the OVAe to be more effective at equimolar levels. Comparison of the relative responses of cells stimulated with various Ags to cells stimulated with OVA or OVAe reveal that all fusion proteins stimulated significantly greater levels of proliferation than did OVA, but less than OVAe (Table I). These results clearly indicate that the OT-I response that occurs after stimulation with OVAe–flagellin fusion protein is not dependent on a specific region of flagellin and is thereby independent of TLR5, IPAF, or any other flagellin-responsive pathway. Importantly, the CD8⁺ T cell response to all versions of the fusion proteins we tested is also independent of any region of OVA, apart from the MHC I- and MHC II-restricted TCR epitopes and their immediate flanking residues.

FIGURE 3.

The enhanced proliferative response to the SIINFEKL epitope in fusion proteins is not dependent on the conserved regions of flagellin or on TLR5 expression by APCs. Total LN cell suspensions were generated from the s.c. LNs of wild-type and TLR5^−/− mice and mixed with CD8-enriched, CFSE-labeled OT-I cells at a ratio of 9:1. Cells were stimulated with OVA, OVAe, OVAe–flagellin_Se, OVAe–flagellin_Pa, OVAe–HV_Se, OVAe–HV_Pa, or flagellin–OVA and incubated for 3 d at 37°C before harvest. CFSE dilution (A), OT-I cell number (B), and the % divided of CFSE-labeled OT-I cells (C) were determined by flow cytometry. Data are representative of triplicate cultures from two independent experiments.

Table I.

Relative proliferation of OT-I populations in response to antigenic stimulation

Mouse Strain	Immunogen	Response Relative to OVA	p Value	Response Relative to OVAe	p Value
Wild-type	OVA	1	NA	0.03	<0.001
	OVAe	31	<0.001	1	NA
	OVAe–flagellinSe	22	<0.001	0.73	<0.03
	OVAe–flagellinPa	26	<0.001	0.84	=0.1
	OVAe–HVSe	18	<0.001	0.58	<0.01
	OVAe–HVPa	21	<0.001	0.68	<0.03
	Flagellin–OVA	8.7	<0.001	0.29	<0.001
TLR5−/−	OVA	1	NA	0.02	<0.001
	OVAe	41	<0.001	1	NA
	OVAe–flagellinSe	23	<0.002	0.56	<0.004
	OVAe–flagellinPa	29	<0.001	0.70	<0.001
	OVAe–HVSe	20	<0.001	0.49	<0.001
	OVAe–HVPa	24	<0.001	0.57	<0.001
	Flagellin–OVA	6.9	<0.001	0.17	<0.001

Pulsing APCs with OVAe–flagellin results in higher levels of SIINFEKL peptide–MHC on the cell surface than does pulsing with OVA

Having ruled out a role for conserved regions of flagellin in the effect on CD8⁺ T cells, we considered the possibility that OVAe and OVAe fusion proteins are more efficiently processed and presented than is full-length OVA. In this scenario, the level of SIINFEKL peptide–MHC (pMHC) on the surface of the APC pulsed with OVAe fusion proteins would be of sufficient magnitude to be the driving force behind the greater OVA-specific response to just OVA (much like SIINFEKL peptide itself). To test this hypothesis, we used high-avidity OT-I effector cells (45), which do not require costimulation to produce IFN-g in response to cognate pMHC. CD11c⁺ cells were enriched from the LNs of TLR5^−/− mice and pulsed overnight with 10⁻⁷ M OVA, OVAe, OVAe–flagellin_Se, OVAe–HV_Se, SIINFEKL, or flagellin–OVA. The following day, CD11c-enriched cells were washed and plated with OT-I effector cells at a 1:1 ratio. At 6 h later, cells were harvested and IFN-g production by OT-I effector cells was determined by intracellular staining. APCs pulsed with OVAe stimulated 60% of effector cells to produce IFN-g, compared with 82% for the SIINFEKL-pulsed APCs (Fig. 4). Although not quite as robust, the response to APCs pulsed with OVAe–flagellin, OVAe–HV, and flagellin–OVA was quite strong (35% of T cells producing IFN-g). In contrast, only 1% of effector cells from cultures incubated with APCs from media control or OVA produced IFN-g. These data are consistent with the hypothesis that OVAe, OVAe fusion proteins, and full-length OVA inserted into flagellin are more readily processed for MHC I-restricted Ag presentation than is full-length OVA alone and that enhanced processing and presentation of fusion proteins is the mechanism underlying the stronger OVA-specific CD8⁺ T cell response to these immunogens.

FIGURE 4.

OT-I effector cells produce IFN-γ in response to incubation with TLR5^−/−CD11c⁺ cells pulsed with OVAe, OVAe fusion, or flagellin–OVA fusion proteins, but not with OVA. CD11c⁺ cells were enriched from the LNs of TLR5^−/− mice and pulsed overnight with 10⁻⁷ M immunogen. The next day, CD11c⁺ cells were washed and mixed with OT-I effector cells incubated for 5 h with brefeldin. Cells were then stained to determine intracellular production of IFN-γ by flow cytometry. Data are representative of triplicate cultures from two independent experiments.

The response to OVAe–flagellin fusion protein is dependent on CD11c⁺ cells and involves proteasome-mediated Ag processing

In view of our hypothesis that enhanced processing and presentation is responsible for the effectiveness of the flagellin fusion proteins (relative to just OVA), we sought to determine if the observed responses were dependent on CD11c⁺ cell and proteasome function.

In our previous work on the role of flagellin in stimulating an Ag-specific CD4⁺ T cell response, we demonstrated a direct requirement for tlr5^+/+CD11c⁺ cells for the adjuvant effect of flagellin (33). To determine whether the response to the OVAe–flagellin fusion protein was also dependent on CD11c⁺ cells, CD11c-DTR/GFP mice adoptively transferred with CFSE-labeled OT-I cells were depleted of CD11c⁺ cells by injection of diphtheria toxin and then evaluated for Ag-specific OT-I T cell proliferation based on CFSE dilution. Depletion of CD11c cells in adoptive transfer recipient mice dramatically reduced in vivo OT-I cell proliferation (data not shown).

To determine whether the Ag-specific CD8⁺ response we were studying was proteasome dependent, we tested the effect of inhibition of proteasome function by lactacystin and inhibition of lysosomal acidification by chloroquine on the ability of LN cell suspension to stimulate OT-I proliferation. Processing of proteins for MHC I-restricted presentation is normally dependent on processing by the proteasome (49); thus, inhibition proteasome function in APCs should result in a significantly reduced OT-I response to OVAe–flagellin. A single-cell suspension was generated from the s.c. LN of TLR5^−/− mice and incubated for 2 h with lactacystin (20 μM) or chloroquine (50 μM) prior to the addition of Ag for 6 h. Cells were then washed to remove Ag and incubated overnight with a lower dose of lactacystin (2 μM) or chloroquine (5 μM). The following day, cells were washed to remove lactacystin or chloroquine and mixed with CD8-enriched OT-I cells. Cultures were harvested 3 d later, and the OT-I cell number was measured by bead-based flow cytometric counting (50). Ag-pulsed LN cells in the media and DMSO control groups stimulated approximately an 11-fold increase in the number of OT-I cells recovered per well, compared with unstimulated controls. Treatment of APCs with chloroquine did not result in a significant decrease in the number of OT-I cells recovered from culture, compared with the control conditions. However, treatment of APCs with lactacystin resulted in a significant decrease in the number of OT-I cells recovered from culture, compared with controls (Fig. 5). Viability of LN cells treated overnight with lactacystin and chloroquine was similar to that of media and DMSO control groups. Overnight treatment of LN cells with Z-FA-FMK to inhibit cathepsin activity or Z-VAD-FMK to inhibit caspase activity had no significant effect on the number of OT-I cells recovered per well (data not shown). These results confirm the hypothesis that processing of exogenous OVAe fusion protein for presentation to CD8⁺ T cells is dependent on the proteasome.

FIGURE 5.

Inhibition of proteasome function significantly reduces CD8⁺ T cell proliferation in response to stimulation with OVAe–flagellin fusion protein. Total LN cell suspension was generated from the s.c. LNs of TLR5^−/− mice and treated with chloroquine (50 μM) or lactacystin (20 μM) for 2 h before the addition of Ag for 6 h. Cells were then washed and incubated overnight in chloroquine (5 μM) or lactacystin (2 μM). The following day, cells were washed and mixed with CD8-enriched OT-I cells. Cultures were harvested 3 d later, and the number of OT-I cells recovered from each condition was determined by bead-based flow cytometric counting. Data are representative of triplicate cultures from two independent experiments. *p < 0.0001.

Discussion

Our results demonstrate that fusion of full-length OVA or OVAe to flagellin promotes an Ag-specific CD8⁺ response as a result of increased levels of Ag presentation (Fig. 4) rather than from activation of APCs via TLR5 (Figs. 1, 3), MyD88 (Fig. 1), IPAF (Fig. 3), or any other pathway dependent on the conserved regions of flagellin (Fig. 3). The hypothesis that flagellin fusion proteins are easier to process and present than OVA is consistent with our observations. The OVAe constructs used in our studies consist of OVA_247–274 fused to OVA_313–349. OT-I cells recognize OVA_257–264. We included the 10 flanking amino acids on each side of the epitope to ensure that processing of the epitope out of the linearized fusion protein would be the same as in linearized OVA. Replacement of the HV region of flagellin with full-length OVA also resulted in stronger Ag-specific CD8⁺ T cell responses (Figs. 1, 3, and 4) and provided further confirmation that observed effect is not dependent on amino acids proximal to the OVA_257–264 epitope. Thus, the step or steps in processing fusion proteins likely occur upstream of final processing by aminopeptidases and probably before proteasomal processing as well.

Prior to proteolytic processing, exogenous Ags must be unfolded (51). Reduction of disulfide bonds within a protein is a key step in the unfolding process (52), and cross-presentation of disulfide bond-containing Ags is dependent on expression of IFN-γ-inducible thiolreductase by dendritic cells (53). Thus flagellin, which lacks disulfide bonds (20), should be easier to unfold and process than OVA, which contains one disulfide bond between residues 74 and 121 (54, 55). Importantly, the OVAe fusion proteins we used for this study do not contain the residues involved in disulfide bond formation in OVA. Studies of other proteins support this line of reasoning. Removal of an intramolecular disulfide bond from hen egg lysozyme resulted in increased levels of MHC II-restricted epitope generation, whereas the introduction of intramolecular cross-linkers reduced T cell epitope generation (56). Similar results have been obtained using unmodified influenza hemagglutinin protein, which has four intrachain disulfide bonds. The S3 epitope of hemagglutinin is located in the stalk region and can be presented without reduction of the protein. By contrast, the S1 epitope, located in the constrained globular domain, is not presented without reduction of the protein (57). Depression of intracellular glutathione, an important reducing thiol, significantly decreased presentation of disulfide bond-containing OVA and hen egg lysozyme, but had no effect on presentation of Ags that were reduced (58). For CD4⁺ T cell responses, disruption of disulfide bonds in an Ag also resulted in increased in vitro responses (59). For MHC I-restricted epitopes, increased presentation of reduced Ags, compared with nonreduced Ags, may also result from processing through additional pathways. Using a P. falciparum Ag that contains two disulfide bridges, Prato and colleagues (60) found that TAP-deficient APCs were unable to present the nonreduced form of Ag but were able to present reduced Ag to a CTL line. In view of these findings, we propose that the absence of a disulfide bond or bonds in OVAe–flagellin makes it more easily processed and presented than OVA. This conclusion is supported by the observation of increased IFN-γ production by OT-I effector cells, which indicates increased levels of pMHC on the surface of the APC following treatment with OVAe–flagellin compared with OVA (Fig. 4).

The ability of flagellin to facilitate a CD8⁺ T cell response independent of TLR5 and MyD88 is in sharp contrast to other actions of flagellin. For example, the Ag-specific CD4⁺ T cell response to flagellin fusion proteins is highly dependent on direct stimulation of TLR5⁺/CD11c⁺ cells by flagellin (33). Similarly, the Ab response to flagellin fusion proteins also requires expression of TLR5 (61). The lack of a requirement for TLR5 in the CD8⁺ T cell response suggests that Ag rather than TLR signaling is the rate-limiting step in this response.

Our findings resolve the controversy in the literature regarding the ability of flagellin to function as an adjuvant for Ag-specific CD8⁺ T cell responses. We have demonstrated that fusion of MHC I-restricted immunogenic epitopes to flagellin can create a pseudo-adjuvant effect that functions through increased Ag presentation on the surface of the APC and is independent of TLR5, MyD88, and the conserved regions of flagellin. This effect results from more efficient processing of the flagellin–Ag fusion protein than native-state Ag. This mechanism is in striking contrast to the situation with CD4⁺ T cells, in which flagellin functions as a true adjuvant via its stimulatory effect on CD11c⁺ cells (33).

Replacement of the HV region of flagellin with poorly immunogenic, disulfide bond-containing Ags may result in essentially unlocking the inherent antigenicity of these proteins by preventing normal disulfide bond formation and thus facilitating activation of Ag-specific CD8⁺ T cells. For example, we have found that mice immunized with the flagellin–OVA fusion protein in which the HV region was replaced with full-length OVA do not produce IgG that recognizes native OVA (K.N. Delaney, E.T. Weimer, J.T. Bates, and S.B. Mizel, unpublished observations), a finding consistent with the conclusion that the OVA in the flagellin fusion was not in its native configuration. However, the nonnative state of the OVA in the fusion protein clearly facilitated the activation of OVA-specific CD8⁺ T cells (Figs. 1, 3, and 4). Thus our findings raise the interesting possibility that flagellin may be an excellent platform for vaccines that incorporate poorly processed protein Ags in which specific CD8⁺ T cell epitopes have been identified.

In summary, the ability of flagellin to promote CD8⁺ T cell activation is independent of its ability to signal via TLR5 and operates through an unconventional pseudo-adjuvant mechanism, namely, delivery of an Ag in a form that facilitates Ag processing.

Abbreviations used in this article

EGFP

enhanced GFP

HV

hypervariable

LN

lymph node

OVAe

OVA epitopes

pMHC

SIINFEKL peptide-MHC