Comprehensive definition of human immunodominant CD8 antigens in tuberculosis

Abstract

Despite widespread use of the Bacillus Calmette-Guerin vaccine, tuberculosis, caused by infection with Mycobacterium tuberculosis, remains a leading cause of morbidity and mortality worldwide. As CD8⁺ T cells are critical to tuberculosis host defense and a phase 2b vaccine trial of modified vaccinia Ankara expressing Ag85a that failed to demonstrate efficacy, also failed to induce a CD8⁺ T cell response, an effective tuberculosis vaccine may need to induce CD8⁺ T cells. However, little is known about CD8, as compared to CD4, antigens in tuberculosis. Herein, we report the results of the first ever HLA allele independent genome-wide CD8 antigen discovery program. Using CD8⁺ T cells derived from humans with latent tuberculosis infection or tuberculosis and an interferon-γ ELISPOT assay, we screened a synthetic peptide library representing 10% of the Mycobacterium tuberculosis proteome, selected to be enriched for Mycobacterium tuberculosis antigens. We defined a set of immunodominant CD8 antigens including part or all of 74 Mycobacterium tuberculosis proteins, only 16 of which are previously known CD8 antigens. Immunogenicity was associated with the degree of expression of mRNA and protein. Immunodominant antigens were enriched in cell wall proteins with preferential recognition of Esx protein family members, and within proteins comprising the Mycobacterium tuberculosis secretome. A validation study of immunodominant antigens demonstrated that these antigens were strongly recognized in Mycobacterium tuberculosis-infected individuals from a tuberculosis endemic region in Africa. The tuberculosis vaccine field will likely benefit from this greatly increased known repertoire of CD8 immunodominant antigens and definition of properties of Mycobacterium tuberculosis proteins important for CD8 antigenicity.

Fig. 1

Ex vivo CD8⁺ T cell peptide library screens. a The SFU from positive wells (as defined in methods) from the ex vivo CD8⁺ T cell peptide library screen for each individual well are depicted. The x-axis represents each of the 789 peptide pools arrayed by Tuberculist functional classification. b Compiled results from ex vivo screens (n = 20) are depicted. The top 5% of each subject’s screen are shown. Subjects with LTBI are shown in black whereas individuals with TB are shown in red

Fig. 2

Composition of the Mtb genome, peptide library and immunodominant peptide pools according to functional classifications. For the Mtb genome, peptide library, and set of immunodominant peptide pools, the proportion of proteins categorized by Tuberculist functional classification, or categorized as an Esx family member or not, or DosR/EHR family member or not is shown. Using hypergeometric distribution, the probability of observing at most 14 peptide pools representing PE/PPE proteins out of 49 immunodominant peptide pools by chance given the high (58.8%) prevalence of peptide pools representing PE/PPE proteins in the library is 0.00098%. The probability of observing ≥ 24 peptide pools representing cell wall proteins out of 49 by chance when the prevalence in the library is 23.5%, is 0.0053%. The probability of ≥8 peptide pools representing Esx proteins out of 49 by chance when the prevalence in the library is 2.4% is <0.0001. The probability of ≥16 peptide pools representing non-Esx cell wall proteins out of 49 by chance when the prevalence in the library is 21.2% is 0.0364

Fig. 3

Classification of secreted proteins within the Mtb genome, peptide library and immunodominant peptide pools. For the Mtb proteome, peptide library, and set of immunodominant peptide pools, the proportions of proteins categorized as secreted as defined in Tuberculist in 2005, as secreted by an experimental evidence definition, and/or as cell wall by Tuberculist functional classification, are shown. Using hypergeometric distribution, the probability of observing ≥ 22 peptide pools representing proteins with experimental evidence of secretion out of 49 immunodominant peptide pools by chance when the prevalence of peptide pools representing proteins with experimental evidence of secretion in the library is 28%, is 0.0053%. The probability of ≥15 peptide pools representing cell wall proteins with experimental evidence of secretion out of 49 by chance when the prevalence in the library is 13% is 0.0003

Fig. 4

Clinical Validation of Immunodominant Peptide Pools in Mtb-infected Individuals in Kampala, Uganda. a, b IFN-γ ELISPOT was performed on CD4/CD56 depleted PBMC from approximately 10 subjects with LTBI and 10 with TB. a Results are shown as percent positive assays. Black bars indicate participants with LTBI and green bars indicate participants with active TB. For the PPE51:PPE50 peptide pool, the asterisk denotes that the difference in proportion positive between those with LTBI and active TB was statistically significant (p = 0.004, Fisher’s exact test). b The associated magnitude of responses reported in SFU is shown. Black dots indicate participants with LTBI and green dots indicate participants with active TB. For the PPE51:PPE50 and PE_PGRS50 peptides pools, the asterisks denote that the difference in SFU between those with LTBI and active TB were statistically significant (p = 0.004 and p = 0.024, respectively, Mann Whitney). c, d Additional validation was performed for five selected antigens. IFN-γ ELISPOT was performed on CD4 depleted PBMC on 50 subjects with LTBI and 50 with TB. c Results are shown stratified by disease phenotype as a percent positive by ELISPOT assays. Black bars indicate participants with LTBI and green bars indicate participants with TB. d The associated magnitude of responses reported in SFU is shown. Black dots indicate participants with LTBI and green dots indicate participants with active TB