CD4⁺ T Cell Epitopes of FliC Conserved between Strains of Burkholderia: Implications for Vaccines against Melioidosis and Cepacia Complex in Cystic Fibrosis

Burkholderia pseudomallei is a saprophytic Gram-negative pathogen that causes melioidosis is endemic in northeast Thailand, northern Australia, and other tropical countries (1). B. pseudomallei is also classified as a category B biological agent (2). Although the majority of exposed individuals display asymptomatic seroconversion, sometimes followed by asymptomatic chronic persistence (3), a minority show local tissue infection, abscesses of the lung, spleen, CNS, and liver, pneumonia, and fatal septic shock. The pathogenicity of B. pseudomallei is multifactorial, depending on bacterial virulence (4). B. pseudomallei possesses flagella, which confer temperature independent motility and mutation of FliC, BPSL3319, the gene encoding flagellin, leads to attenuation of Bp in infected BALB/c mice (5); FliC immunization protects from B. pseudomallei challenge (6). A screen of melioidosis patients using a B. pseudomallei protein array identified flagellin as a major seroreactive Ag (7) and CD4 T cell responses to a number of these serodiagnostic Ags have been reported (8).

FliC proteins trigger innate immunity through TLR5 binding in several other bacterial infections, including Listeria, Pseudomonas, and Salmonella (9). Flagellin sequences are also stimulatory to adaptive immunity through recognition by B cell and T cell AgRs, and Salmonella FliC is recognized by T cells during murine and human Salmonella infection (10–12). Mapping of murine CD4 responses to Salmonella FliC identified multiple epitopes, including some within sequences conserved across multiple Gram-negative bacterial species and also encompassing the domain required for TLR5 binding (13).

Cystic fibrosis (CF) affects 80,000 individuals worldwide and is caused by the most common fatal autosomal recessive gene in Caucasians. Mutations in the CF transmembrane conductance regulator result in defective fluid transport in epithelial cells (14). CF affects the lungs, causing dehydration of the mucus layer, recurrent infections, and progressive fibrosis with airway remodeling. Treatment advances, including lung transplantation, have extended life expectancy to a mean of 40 y (15). In studies of predictors of mortality in CF, the key factors are infection with Pseudomonas or Burkholderia species (16). Burkholderia cepacia syndrome is the term used to describe infection with a group of related bacterial species that infect individuals with cystic fibrosis; Burkholderia cenocepacia and Burkholderia multivorans, account for the majority of these Cepacia complex infections. Despite evidence of poor immunity in CF, including the direct consequence of mutant CF transmembrane conductance regulator expression in T cells and dendritic cells (17, 18), little is known about recognition of immunodominant Ags from Burkholderia species or the strategies required to design effective vaccines for this at-risk patient group.

We report analysis of CD4 immunity to B. pseudomallei FliC through complementary approaches including protein and peptide immunization of HLA transgenic mice, measurement of peptide binding to HLA class II heterodimers, and screening of seropositive human donors for T cell immunity. We identify conserved epitopes across those Burkholderia species that cause melioidosis and that are associated with Cepacia syndrome in CF patients. This has implications for development of epitope-string vaccines that are broadly applicable across at-risk cohorts and offers potential for synergy in vaccine strategies and advocacy across both melioidosis and individuals with CF.

Recombinant FliC (rFliC) of B. pseudomallei was expressed in Escherichia coli and purified (Biomatik). Synthetic peptides of 20 aa overlapping by 10 aa from the FliC sequence of B. pseudomallei strain K96243 (BPSL3319, accession no. YP109915) (see Supplemental Table I), and analogs from B. multivorans (strain ATCC 17616, accession no. YP001947524.1), B. cepacia (strain E242, accession no. AAC38200.1), and B. cenocepacia (strain J2315, accession no. YP002229280.1) were synthesized by GL Biochem (see Supplemental Table II).

HLA class II molecules were purified from B lymphoblastoid cells by affinity chromatography using mAb L243 for HLA-DR and SPVL3 for HLA-DQ (19). Binding of peptides to HLA class II heterodimers was assessed by competitive ELISA on an automated workstation (19). HLA-DR molecules were diluted in 10 mM phosphate, 150 mM NaCl, 1 mM n-dodecyl β-d-maltoside, 10 mM citrate buffer with appropriate biotinylated reference-indicator peptide, and serial dilutions of competitor peptides. Unlabeled biotinylated peptides were used as reference peptides to assess the validity of each experiment. The sequences and IC₅₀s of reference peptides chosen as high binders were HA 306–318 (PKYVKQNTLKLAT) for DRB1*01:01 (2 nM), DRB1*04:01 (14 nM), DRB1*11:01 (72 nM) and DRB5*01:02 (18 nM); YKL (AAYAAAKAAALAA) for DRB1*07:01 (5 nM); A3 152–166 (EAEQLRAYLDGTGVE) for DRB1*15:01 (37 nM), MT 2–16 (AKTIAYDEEARRGLE) for DRB1*03:01 (84 nM), B1 21–36 (TERVRLVTRHIYNREE) for DRB1*13:01 (46 nM), CTP 427–441 (VHGFYNPAVSRIVEA) for DRB1*09:01 (23 nM), TFR141–155 (TGTIKLLNENSYVPR) (360 nM) for DRB1*12:02, TFR607–620 (LNLDYERYNSQLLS) for DRB1*15:02 (4 nM); B7150–164 (LNEDLRSWTAADTAA) for DQ6 (DQA1*01:02/DQB1*06:02) (37 nM), and DQB45–57 (ADVEVYRAVTPLGPPD) for DQ8 (DQA1*03:01/DQB1*03:02) (98 nM). After 24- to 72-h incubation at 37°C, the samples were neutralized with 50 μl 450 mM Tris HCl pH 7.5, 0.3% BSA, 1 mM n-dodecyl β-d-maltoside buffer and applied to 96-well Maxisorp ELISA plates (Nunc) coated with 10 μg/ml Ab. Bound biotinylated peptide was detected by streptavidin-alkaline phosphatase conjugate (GE Healthcare). Emitted fluorescence was measured at 450nm upon excitation at 365 nm. Peptide concentration that prevented binding of 50% labeled peptide (IC₅₀) was evaluated, and relative binding affinity was expressed as a ratio of the IC₅₀ test/reference peptide. Mean + SEM were calculated from two to three independent experiments and relative affinities of 10 or fewer were considered high binders and 10–100 moderate binders.

Leukocytes from healthy donors were collected through the Blood Transfusion Center, Khon Kaen Hospital (Khon Kaen, Thailand). Ethical permission was obtained from Ethical KKU research no. HE470506. Healthy B. pseudomallei–seropositive samples were selected based on indirect hemagglutination assay titers of 1:40 or greater (20). PBMCs from were isolated by Ficoll-Hypaque (Sigma-Aldrich) density gradient centrifugation and stored at −80°C. HLA-DRB1 and -DQB1 genotypes were determined by PCR sequence–specific primer as described previously (21).

The frequency of human CD4 T cells secreting IFN-γ in response to Ag was evaluated by IFN-γ ELISPOT (Cellular Technology Limited). Precoated 96-well polyvinylidene difluoride plates (MultiScreen Immobilon-P; Millipore) were incubated overnight with 15 μg/ml IFN-γ Ab, and then, 100 μl 5 × 10⁶ cells/ml pulsed with 100 μl 5 × 10⁷ CFU/ml killed intact B. pseudomallei as positive control, with 50 μg/ml individual FliC peptides or medium alone as a negative control were added in triplicate, cultured for 48 h, and IFN-γ secretion detected by adding 1:250 of human IFN-γ detection solution at room temperature for 2 h; IFN-γ–producing cells were quantified by Immunospot analyzer (Cellular Technology Limited). Results are expressed as mean spot-forming cells (SFC)/10⁶ cells of triplicates in the presence of stimulus minus mean spots in the medium control (Δsfc), and considered positive if values were above the mean medium control + 2SD.

HLA-DR04:01 transgenic (DR4) mice carrying HLA-DRA1*01:01 and HLADRB1*04:01 (22) were obtained from Taconic Farms. HLA-DR1 and HLA-DR15:01 and HLA-DQ8 transgenic mice carrying HLADRA1*01:01/HLA-DRB1*01:01, HLA-DRA1*01:01/ HLADRB1*15:01, or DQA1*03:01/DQB1*03:02 were as described previously (23, 24). All mice lacked expression of endogenous MHC class II. All animal experiments had passed institutional ethical review and were performed under the Animals (Scientific Procedures) Act 1986 and authorized by the Home Secretary, Home Office U.K.

Popliteal lymph nodes were removed 10 d after immunization in one hind footpad with 25 μg rFliC or peptide. Assays were performed in HL-1 serum–free medium (BioWhittaker, Lonza, Slough, U.K.), supplemented with l-glutamine and gentamicin (Life Technologies, Paisley, U.K.). The frequency of cells producing IFN-γ was quantified by ELISPOT (Diaclone; 2B Scientific, Oxon, U.K.). A total of 2 ×10⁵ cells and Ag were added to wells, and plates were incubated for 72 h at 37°C in 5% CO₂. Spots were counted using an automated ELISPOT reader (Autoimmun Diagnostika). A total of 50 ng/ml staphylococcal enterotoxin B was used as a positive control and culture medium only as a negative control. Results are expressed as for human ELISPOT data above. The presence of an epitope was confirmed when the majority of mice in a group responded, and frequencies as Δsfc/10⁶ cells categorized as low (<25), intermediate (<100), or high (>100).

FliC-specific T cell lines were generated from lymph node cells of FliC-immunized HLA-DR04:01 transgenic mice by periodic restimulation with Ag and irradiated spleen cells and expansion in rIL-2 as described previously (25) and then used to generate T cell hybridomas by polyethylene glycol fusion with BW5147 (TCRα⁻β⁻) cells (a gift from Dr. P. Marrack, Denver, CO). T cell hybridomas were shown to express CD4, CD3ε, and TCRαβ by flow cytometry and secreted IL-2 in response to rFliC- or peptide-pulsed APC. Bone marrow macrophages were generated as described previously (26), harvested after 6 d, and activated by treatment with 1 ng/ml IFN-γ (R&D Systems, Abingdon, U.K.) overnight when >90% expressed F4/80. Flat-bottom 96-well plates were seeded in triplicate wells with 4 × 10⁴/well macrophages, 5 × 10⁴/well T cell hybridoma cells, and a range of doses of rFliC, peptides or culture medium alone. In some experiments, macrophages were first infected for 30–360 min with B. cenocepacia J2315 or incubated overnight with heat-killed B. cenocepacia strain J2315 or with supernatants from overnight cultures of J2315 in Luria broth or Luria broth alone as a control. Plates were incubated for 24 h and frozen. Responses were determined as the amount of IL-2 released in a CTLL-2 bioassay (3 × 10⁴/well) in the presence of T cell hybridoma culture supernatants diluted 1:2. The bioassay was incubated for 24 h in triplicate wells of flat-bottom 96-well microtiter plates and cells labeled with [³H]thymidine. Results were plotted as mean cpm of triplicate wells ± SD. Experiments were repeated at least twice, and the data for representative experiments are shown.

PBMCs from 46 healthy, B. pseudomallei–seropositive blood donors from Khon Kaen Hospital (Khon Kaen, Thailand) that were positive for a T cell response to heat-killed B. pseudomallei were tested against the FliC peptide panel, assessing each peptide individually (Fig. 1). Strong responses were seen to several epitopes, with clusters of responses centered on peptides 7/8, 16/17, 26/27/28, and 38 (Fig. 1). Consideration of the HLA-DR alleles carried by responders to each epitope confirms that these epitopes are broadly recognized by individuals carrying different HLA genotypes, including a proportion of donors with variants of HLA-DR5, -DR3, -DR12, and -DR15 (Supplemental Table III). HLA-DRB3, 4, and 5 loci could be inferred from the HLA-DRB1 typing, but without recourse to dissection of responses at the clonal level, we cannot resolve the extent to which individual peptides might be presented by each of these gene products on a given haplotype. Although it would be valuable in the interests of comprehensiveness to understand also the contribution of HLA-DP–restricted responses B. pseudomallei FliC as well as gaining a more complete picture of HLA-DQ restriction, we devoted the bulk of our resource to analysis of HLA-DR restriction, in light of the fact that, because of weaker expression of the HLA-DP and DQ isotypes, a far smaller component of the T cell repertoire tend in most cases to use these molecules and, in line with this, fewer reagents (such as HLA transgenic mice) have historically been generated to facilitate such analysis.

The 38 FliC peptides comprising the protein sequence were tested for binding to 13 common alleles of HLA-DR or DQ (Table I). The HLA panel included common HLA alleles found in Thailand and northern Australia, namely, HLA-DR12:02, HLA-DR15:02, and HLA-DR09:01, covering ∼35% of the Thai population (www.allelefrequencies.net). Peptide 38 elicited a response among the seropositive donors that is both common and of high frequency with respect to the magnitude of response in any given individual. Peptide 38 is also among those peptides with the property of binding a very wide array of HLA class II heterodimers with moderate to high affinity. This is also true of peptides 7, 8, 16, and 27, all of which show moderate to strong binding to HLA-DR12:02. It is noteworthy that the two HLA-DQ heterodimers assayed in this study are particularly strong binders of FliC peptides.

We investigated CD4 recognition of B. pseudomallei FliC in immunized HLA class II transgenics. HLA class II transgenics for the alleles HLA-DR1, -DR4, -DR1501, and -DQ8 (that is, expressing HLA-DQB1*0302) were immunized with rFliC and lymph node cells assayed 10 d later. The majority of HLA transgenics responded to FliC peptide 16 (Fig. 2A, 2C, 2D). Thus, the response in HLA transgenics appears highly focused on a single epitope, peptide 16, able to bind diverse HLA class II peptide–binding grooves, as demonstrated in the HLA class II binding assays shown in Table I. Note that peptide 16 also features strongly in the CD4 T cell responses of about half of the Thai human cohort of seropositive donors. The response to p38 that predominated in the human donors was not detected in any of these four transgenic lines. We take this to mean that this is a case where p38 can be presented by several of the HLA class II heterodimers that are common in that cohort (HLA-DR12:02, HLA-DR15:02, and HLA-DR09:01 haplotypes being particularly common) but not the four specific heterodimers screened in this study; several new transgenic lines would be needed to give the power to fully annotate responses in Southeast Asian populations.

HLA class II transgenics were next immunized with selected individual FliC peptides including p16. After priming with peptide, the majority of p16-immunized DR1 (five of five), DR1501 (three of five), and DQ8 (HLA-DQB1*0302) (five of five) transgenics responded to the immunizing peptide (Fig. 3A–C). Peptide 16–immunized DR1501 and DQ8 (HLA-DQB1*0302) transgenics did not respond to either flanking peptide, suggesting the presence of an epitope centered on p16 for each allele. All DR1 transgenics responded to p15, p16, and p17, most likely arguing for the presence of more than one CD4 epitope in the part of the FliC sequence from 131–170. Peptide 16 was not immunogenic in HLA-DR4 transgenics (Fig. 3D), even though this sequence binds DR4 with high affinity (Table I). Peptides 35 and 36, both of which bind HLA-DR4 with high affinity (Table I), elicit good T cell responses in HLA-DR4 transgenics (Fig. 3E, 3F).

We further investigated CD4 T cell epitopes restricted by HLA-DR4 by systematically immunizing mice with peptides predicted by in silico analysis (www.iedb.org) to bind to DR4. Peptides 2, 3, 7, 8, 10, 12, 15, 19, 22, 24, 26, 29, 30, 31, and 33 induced significant T cell responses in the majority of immunized mice (Fig. 4). The lack of responses to flanking peptides in every case except for p7 and 8 suggest separate epitopes in each peptide, and a single epitope shared by p7 and 8 (also seen as epitopes in the human T cell responses (Fig. 1). The majority of these peptides bind to HLA-DR4 with moderate to high affinity (Table I). Peptide immunization thus confirmed the presence of the epitope on p16 and identified at least 14 additional subdominant or cryptic HLA-DR4–restricted CD4 T cell epitopes.

We investigated whether T cells recognizing B. pseudomallei FliC epitopes would cross-reactively recognize epitopes from the CF-associated Burkholderia species, B. multivorans, B. cepacia, and B. cenocepacia. We generated cloned T cell hybridomas from Bp rFliC-immunized HLA-DR4 transgenics, mapping their recognition specificity to either p33 or p34 (data not shown). We showed above that five of five and two of five peptide-immunized HLA-DR4 transgenics responded to p33 and 34, respectively (Fig. 4W, 4X).

Comparison of the FliC sequences of B. multivorans, B. cepacia, and B. cenocepacia in the equivalent region FliC_321–350 show moderate sequence homology between species (Supplemental Table II). Bone marrow–derived macrophages from HLA-DR4 transgenics were used as APC in T cell hybridoma assays to test peptide analogs from the sequences of the above species. The results confirmed that p33-specific T cell hybridoma 1.20 recognized p33 but not p34 from B. pseudomallei, and in addition cross-reacted with the p33 analog from B multivorans, but not that from B. cepacia or B. cenocepacia (Fig. 5A, 5C). Similarly, the p34-specific T cell hybridoma 2.19 recognized p34 but not p33 of Bp and, in this case, cross-reacted with the p34 but not p33 sequences from B multivorans, B. cepacia, and B. cenocepacia (Fig. 5B, 5D). Finally, we investigated whether the T cell hybridomas recognized p33 and 34 analogs of B. cenocepacia presented by macrophages following infection with B. cenocepacia, J2315, to establish whether FliC is processed and presented during infection of macrophages. p33 and p34-specific T cell hybridomas both responded to macrophages infected with J2315 as well as to heat-killed J2315 and culture supernatants but not the control (Fig. 6A, 6B). We titrated the time of exposure of macrophages to viable or heat-killed J2315 for presentation of the p34 epitope, showing that processing and presentation began after 60 min and was delayed compared with presentation of synthetic peptide (Fig. 6C, 6D). The kinetics of presentation of the p34 epitope was also similar for viable and heat killed bacteria, and neither was dependent on IFN-γ activation of macrophages (Fig. 6C, 6D). The data suggest that the DR4-restricted CD4 epitopes within p33 and p34 are generated for presentation when macrophages are exposed to either live or heat-killed J2315 or to FliC released into the supernatant of cultured J2315 cells.

There are a number of beneficiaries from an improved understanding of immunity to Burkholderia species, which could inform new vaccination strategies. The case has most commonly been made either for melioidosis (1) or in the context of biodefense initiatives, since B. pseudomallei is a category B agent (2). Individuals with diabetes, chronic renal impairment, or alcoholic liver disease are considered an especially at-risk group for melioidosis, a cause of concern because of the fast-growing prevalence of diabetes in Asia (http://www.idf.org/diabetesatlas/data-visualisations). A recent study modeled the added value of implementation targeting dual application in biodefense and melioidosis (27). In this study, we consider the possibility that, in light of immunological cross-reactivity, vaccine strategies might aim to encompass also the need to protect CF patients (in whom infections with Pseudomonas and Burkholderia are the strongest indicators of poor survival) from the Burkholderia species associated with Cepacia syndrome (16, 28). Our goal was therefore to explore the rationale for such a cross-species approach to Burkholderia immunity, investigated at the level of T cell cross-reactivity.

Compared with other life-threatening human bacterial pathogens, relatively little is known about T cell specificity in immunity to Burkholderia infection. Faced with a 7.2-Mb bacterial genome of virtually uncharted territory for T cell immunology, it has been unclear where the search for key immunogens should start. Clues have come from immunomic serological screening, classifying 109 Ags as immunodominant and 31 as serodiagnostic for melioidosis (8). Among the strongly serodominant Ags is FliC, a target both of innate and adaptive immune recognition in infections by diverse bacterial species. As would be expected of a pathogen that inhabits a wide range of environments including soil, air, water, plants, and diverse mammalian tissues including human APC, B. pseudomallei has a highly variable, condition-dependent transcriptome (29) and FliC is among the transcripts preferentially up-regulated during lung infection. We have reported that CD4 T cells from exposed individuals respond to FliC, although responses were not different between healthy seropositive donors and those recovering from clinical melioidosis (7). Multiple CD4 T cell epitopes have been identified in FliC from Salmonella in the setting of murine and human infection (30). The dual properties of Pseudomonas flagellin as an immunodominant Ag and a strong TLR5 agonist make it in some respects a candidate vaccine immunogen in the context of Pseudomonas aeruginosa infections (31). However, an important caveat comes from the immunogenetics of melioidosis: the 1174T mutation of TLR5, present in 5–8% of the population and which causes loss of the TM and CY signaling domains to yield a hypofunctional receptor, is associated in the Thai population with significant protection from organ failure and death in the context of clinical melioidosis. This suggests either a proinflammatory, immunopathogenic component to disease ensuing from innate TLR5 signaling or, conversely, a deleterious dampening of immunity through TLR5-dependent IL-10 (32). Earlier studies demonstrated the ability of FliC constructs to protect mice from experimental B. pseudomallei infection (6) and for this reason, such studies were not repeated in this study; the clinical immunogenetic findings argue the need to use mouse models to now fully dissect the contributions of pro- or anti-inflammatory consequences of WT and mutant TLR5 to B. pseudomallei–mediated disease.

Preliminary earlier findings on B. pseudomallei FliC immunity have, in the study, been extended by identification of immunodominant CD4 T cell epitopes, HLA class II restriction and conservation of epitopes across Burkholderia species in different clinical settings. We have previously reported the use of panels of HLA class II transgenics in conjunction with HLA class II binding studies for deciphering patterns of HLA–peptide presentation to CD4 T-cells in the context of bacterial infection (23, 24, 33). We show in this study that epitopes of B. pseudomallei FliC can be presented by alleles of both HLA-DR and HLA-DQ heterodimers. Although HLA-DQ is considered a weakly expressed HLA isotype with poorer representation in the restriction patterns of the CD4 T cell repertoire, the finding in this study of restriction through both HLA-DR and HLA-DQ is of interest because susceptibility to melioidosis has been associated with HLA-DQ polymorphisms (34). At first sight, there may appear to be a disparity between the complex array of CD4 T-cell epitopes identified in the human cohort (with response to p38 especially common) and the high degree of focus we show on a single, different, epitope, p16, in the HLA class II transgenics. We conclude that, although p38 must likely be presented by several different HLA class II heterodimers among our human donors, these do not include those covered in our HLA transgenic mouse panel. The Thai cohort of 46 donors used in this study encompasses some 17 major DRB alleles that are variably present in heterozygous combinations, and the majority of these HLA-DR haplotypes encode for two expressed heterodimers (e.g., the DRB1 and DRB5 encoded products from the DRB1*15:01 and DRB1*15:02 haplotypes). Thus, the pattern of presented epitopes in the human donor panel is complex and the use of single HLA heterodimer transgenics would be expected to simplify the range of epitopes presented by each allele.

Life-threatening Burkholderia infections that affect CF patients on the one hand and susceptible individuals in SE Asia on the other are normally considered as distinct biomedical challenges, with distinct agendas for achieving sufficient momentum for successful vaccine trials. Considering the extensive conservation of Ag sequences between Burkholderia species, datasets such as the one that we present in this study argue strongly that there may be clear operational advantages in seeking to develop vaccines that can serve both patient groups.

The authors have no financial conflicts of interest.