A Broad Profile of Co-Dominant Epitopes Shapes the Peripheral Mycobacterium tuberculosis Specific CD8+ T-Cell Immune Response in South African Patients with Active Tuberculosis

We studied major histocompatibility complex (MHC) class I peptide-presentation and nature of the antigen-specific CD8+ T-cell response from South African tuberculosis (TB) patients with active TB. 361 MHC class I binding epitopes were identified from three immunogenic TB proteins (ESAT-6 [Rv3875], Ag85B [Rv1886c], and TB10.4 [Rv0288], including amino acid variations for Rv0288, i.e., A10T, G13D, S27N, and A71S for MHC allotypes common in a South African population (e.g., human leukocyte antigen [HLA]-A*30, B*58, and C*07). Inter-allelic differences were identified regarding the broadness of the peptide-binding capacity. Mapping of frequencies of Mycobacterium tuberculosis (M. tb) antigen-specific CD8+ T-cells using 48 different multimers, including the newly constructed recombinant MHC class I alleles HLA-B*58:01 and C*0701, revealed a low frequency of CD8+ T-cell responses directed against a broad panel of co-dominant M. tb epitopes in the peripheral circulation of most patients. The antigen-specific responses were dominated by CD8+ T-cells with a precursor-like phenotype (CD45RA+CCR7+). The data show that the CD8+ T-cell response from patients with pulmonary TB (prior to treatment) is directed against subdominant epitopes derived from secreted and non-secreted M. tb antigens and that variant, natural occurring M. tb Rv0288 ligands, have a profound impact on T-cell recognition.


Introduction
8.8 million incident tuberculosis (TB) cases and 1.45 million TB-related deaths were reported in 2010, with the highest TB rates occurring in sub-Saharan Africa [1].The clinical outcome and presentation of TB is shaped by the interaction of Mycobacterium tuberculosis (M.tb) genetics and the host's immunological response [2].
CD8+ T-cells contribute to the cellular immune defense against TB [3] early during M. tb infection [4].Activation and expansion of M. tb-specific and Major histocompatibility complex (MHC) class I restricted CD8+ T-cells requires the intracellular processing of M. tb antigens and subsequent presentation of epitopes by MHC class I cell-peptide complexes on the cell surface of antigen presenting cells.The immunogenicity of an M. tb epitope is influenced by several factors: i) the epitope must be generated and transported to the endoplasmic reticulum, by overcoming limitations in; cross-presentation [5] proteasome-dependent proteolysis [6] and the transporter associated with antigen processing [7]; ii) the epitope must bind to the MHC haplotype with appropriate affinity and dissociation rate; iii) the peptide-MHC (p-MHC) complex must then be recognized by an appropriate T-cell receptor (TCR) repertoire.
In order to determine immunogenicity of selected candidate M. tb epitopes, it is vital to correlate peptide MHC-binding to the actual detection of antigen-specific T-cells without in vitro manipulation (i.e. using in vitro expansion).This can be achieved in an unbiased way using MHC class I multimers loaded with molecularly defined pathogen-derived epitopes [8].
A marked heterogeneity characterizes the human immuneresponse against M. tb; the host-pathogen interplay dictates the outcome of infection.Different clustering of MHC alleles, human leukocyte antigens (HLA), are associated with geographic locations or ethnicity [9].This association may, in part, reflect local hostpathogen co-adaptation which has driven both MHC and pathogen diversification.Differences in ethnicity and MHC alleles have previously been shown to play a role in T-cell responses to a number of infectious diseases, e.g.Hepatitis B [10], Hepatitis C [11] and human immunodeficiency virus (HIV) [12].Associations between the host populations and defined lineages of M. tb have previously been identified [13,14], therefore we chose to study the impact of different host (MHC class I alleles) and pathogen (strainto-strain epitope) variations on M. tb peptide presentation on three levels: 1) peptide-MHC association, 2) TCR-recognition and 3) phenotype and cytotoxicity of CD8+ antigen-specific T-cells directed against different M. tb target antigens.
Up until now, only a few M. tb T-cell epitopes, restricted by a limited number of MHC class I alleles have been identified [15].Therefore, there is a great need to identify additional M. tb epitopes restricted by a more diverse MHC class I allelic repertoire that reflects ethnical populations in Africa.In many regions of sub-Saharan Africa including South Africa, the most common HLA-A molecules are A*02, A*23, A*30 and A*68, the most common HLA-B alleles are B*15, B*42, B*53 and B*58 and the most common HLA*C alleles are C*04, C*06, C*07 and C*17 [16,17], alleles for which only very few M. tb epitopes have previously been defined [15,17].
Not only the MHC class I alleles show diversity, also the pathogen shows variations.It was previously thought that the M. tb genome is relatively stable, however an increasing number of strain-to strain differences, i.e. insertions, deletions and point mutations in the M. tb genome have been observed [18].This genetic diversity might have both phenotypical and immunological impacts, e.g.variant T-cell epitopes may lead to different cellular immune response [19,20].Cellular immune responses directed to variant M. tb epitopes may also be responsible for the observation that mixed M. tb infections occur [21] and that TB patients are prone to re-infections [22].Only a limited number of M. tb strains have been sequenced and most genetic variations appear to reside in non-essential M. tb genes.This appears to be true for many M. tb proteins, with the exception of the small immunogenic protein Rv0288.This protein has the following identified naturally occurring single amino acid exchanges: A10T, G13D, S27N and A71S [23].All of these substitutions are part of previously reported CD8+ T-cell epitopes [17,24] and might therefore interfere either with antigen presentation or subsequent T-cell recognition.
We chose to include 6 different M. tb-derived antigens in this study, since M. tb proteins can be differentially expressed or secreted during the course of infection.The immunodominant secreted TB proteins Ag85B (Rv1886c) and ESAT-6 (Rv3875) are part of several new TB vaccine candidates, [25,26,27].They are early targets of cellular immune responses, and expressed at different time points of the infection, based on pre-clinical models [28,29].While Rv3875 (ESAT-6) is only expressed in M. tb and a few other mycobacterial species (e.g.M. marinum and M. kansasii) [30], Ag85B and TB10.4 (Rv0288), are, in addition to M. tb, expressed in the Mycobacterium bovis derived vaccine strain bacillus Calmette-Gue ´rin (BCG) as well as in mycobacteria other than tuberculosis.The non-secreted M. tb proteins, glycosyl transferase (1) (Rv2958c), glycosyl transferase (2) (Rv2957) and cyclopropanefatty-acid synthase (CFA synthase) (Rv0447c) are predominantly expressed in slow growing bacteria.They have previously been identified from our group as targets for humoral and cell-mediated immune responses [31,32,33].These M. tb antigens are involved in cellular metabolism [34,35,36] and are producing factors associated with resistance to macrophage mediated killing [37].
We report in this study the detailed mapping of M. tb peptides to some of the most frequent MHC alleles in South Africa, their impact on T-cell recognition and T-cell homing/differentiation markers using a broad panel of M. tb MHC class I-peptide multimers.

Peptide screening assay
96-well plates coated with biotinylated-bovine serum albumin (BSA) and avidin (from BeckmanCoulter, San Diego, USA) were used to immobilize 0.5 mg/ml biotinylated monomers.Overlapping nonameric peptides (JPT Peptide Technologies GmbH, Berlin, Germany) covering the three TB proteins TB10.4 (Rv0288), Ag85B (Rv1886c) and ESAT-6 (Rv3875) were used in binding, affinity and off-rate studies as well as variant peptides from Rv0288.The duplicated experiments were performed as previously described [39,40], shortly; the monomer-coated plates were stripped of the placeholder peptide leaving the heavy chain free to re-associate with an added candidate peptide after addition of b 2 -microglobulin.Positive binding was detected using a fluorescent labeled conformation-dependent antibody (A*30; IgM, clone 0273HA [One Lambda Inc, Canoga Park, USA], A*02, A*24, A*68, B*07, B*58 and C*07; fluorescein isothiocyanate (FITC)-conjugated anti-HLA-A, -B and -C [BeckmanCoulter, San Diego, USA]).The binding of each candidate peptide was compared to the binding of an appropriate control peptide (same peptides as used to produce the monomers, listed above).For selected peptides further characterization of the binding were made using affinity and off-rate assays.Affinity was measured by incubating titrated peptide concentrations (1024 to 1029 M) on the MHC coated 96-well plates in a manner similar to the binding assay.An ED 50 value, i.e. the peptide quantity needed to achieve 50% binding saturation was calculated.By incubating the candidate peptide for different time points (0, 0.5 h, 1 h, 1.5 h, 2 h, 4 h, 6 h and 8 h), the stability of the p-MHC complexes could be assessed as a t 1/2 value, which is defined as the time point when 50% of the initial peptide concentration has dissociated from the HLA-peptide molecule complexes.The binding, affinity, and offrate-values were calculated using the iTopia TM System Software (BeckmanCoulter, San Diego, USA).Sigmoidal dose response curves were generated using PrismH 4.0 (GraphPad).

Selection of epitopes for multimer construction
Based on the data generated in the peptide binding assays, previously reported data [17,24,40,41,42], or high SYFPEITHI score [43], a set of 48 multimers covering epitopes from the TB proteins Rv0288, Rv1886c, Rv3875, Rv2957, Rv2958c and Rv0447c were obtained from BeckmanCoulter, San Diego, USA; from Immudex, Copenhagen, Denmark or produced in our lab as previously described [8].

Patient data
27 patients diagnosed with active pulmonary TB (prior to any treatment), diagnosed with at least two positive sputum smears for acid-fast bacilli or positive sputum culture for M. tb were enrolled at the University of Stellenbosch, Cape Town and at the Pretoria Academic Hospital in South Africa.Demographic data of the patients are provided in Table S1.Peripheral blood mononuclear cells (PBMCs) were separated at each location after informed consent.The participants provided their written consent to participate in this study and these documents are on file at the clinical sites with the PI.The written consent forms were also submitted to the respective ethical committees and were approved.The institutional review boards (reference no.45

Statistics
Statistical significance between different T-cell populations was evaluated using PrismH 4.0 (GraphPad Software, La Jolla, USA) using a Student's two-sided t-test.
For the much larger M. tb protein Rv1886c, 292 epitopes could be identified associating with the following alleles: A*02:01  S2).
In addition, we could identify 29 binding epitopes from Rv0288 associating with A*68:01 (n = 11), B*58:01 (n = 14) and C*07:01 (n = 4) (Table 2).We also wanted to elucidate the effect of naturally occurring single amino acid substitutions within these epitopes concerning the binding to different MHC class I molecules.The reported substitutions were part of 18 previously identified binding epitopes (from the sequence of H37Rv) and we compared the binding-rate of these M. tb epitopes to 24 novel epitopes which were constructed including one or two amino acid substitutions (Table 3).The variant epitope showed better MHC class I-binding as compared to the reference epitopes in only 13% of the tested peptide-MHC binding combinations.All of these were restricted by 'promiscuous alleles' like HLA-A*02:01, A*24:02 and A*30:02.Changing a binding epitope restricted by alleles with intermediate (A*68:01 and B*58:01) or a more restricted binding pattern (A*30:01, B*07:02 and C*07:01) resulted in a profound decrease of peptide binding or even total abrogated MHC class I-peptide binding.
Table 1.MHC class I binding, affinity and off-rate data for peptide-epitopes derived from Rv3875 (ESAT-6).Analysis of the nature of the p-MHC binding Next, we analyzed in greater detail the Rv3875 epitopes that bound with more than 50% to the respective MHC class I test molecules as compared to the positive control peptide (30 epitopes).The nature of peptide binding to an MHC class I molecule can be divided into affinity, i.e. the binding strength between the peptide and the MHC molecule, and the dissociation rate, i.e. the off-rate (t 1/2 ) of the peptide-MHC complex.The affinity ranged between 1 mM and 5 nM with the majority of the peptides exhibiting an affinity in the lower end of the interval.The dissociation rate ranged between 0.2 h and 4.9 h; both extremes applied for HLA-A*02:01 (Table 1).
Since we identified a considerable number of binding epitopes from the protein Rv1886c, we analyzed the binding characteristics of only a selected set of peptides.The peptides were chosen based on 1) not being previously reported, 2) high binding compared to the positive control and 3) promiscuity (i.e.binding to several MHC molecules).In total, we analyzed affinity and dissociation rate from 121 novel M. tb candidate epitopes.The affinity ranged between 400 mM and 5 nM with the majority between 9 mM and 10 nM.Some of the peptides showed a very rapid dissociation rate (,0.1 h), while some formed a stable MHC class I-peptide complex for up to 8 hours.We identified a high range of intraallelic differences (different peptides bound different to the same HLA allele) as well as inter-allelic differences (same peptides bound different to different alleles).The peptides bound generally in a more stable fashion to the HLA-A alleles as compared to the B and C alleles , with the exception of HLA-A*30:01 which exhibited an off-rate of less than 1 h.Also, peptides binding to alleles with a limited set of binding epitopes tended to dissociate more rapidly (Table S2).
We identified in total 29 novel epitopes from the protein Rv0288 binding to the alleles HLA-A*6801, B*58:01 and C*07:01 and analyzed the binding characteristics for the entire peptide set.The affinity ranged between 700 mM and 30 nM, and the off-rate between 0.1 h and 6 h.We could not see any specific inter-or intra-allelic differences regarding affinity but we could detect a very rapid dissociation rate of less than 20 minutes regarding all candidate epitopes binding to C*07:01 (Table 2).
Alterations in individual amino acids lead to an increased or decreased binding strength in an almost equal number of cases.However we were only able to measure affinity in 20 out of 53 cases due to that too low MHC-peptide binding makes it impossible to reliably study affinity and off-rate.This was the case for all the variant epitopes restricted by A*30:01, A*68:01, B*07:02 and C*07:01.In most cases, where the peptide variation resulted in increased affinity, it increased the total peptide binding.In cases where the MHC class I-peptide stability increased, the increase was only marginal.In general, the amino acid exchange lead to decreased MHC binding that could be explained by either a decrease in 1) binding strength and/or 2) stability of the MHC class I/peptide complex.However, a few outliers were identified for which we could see an increase in both parameters (binding strength and affinity) while the total peptide binding decreased, e.g.allele, bold -binding to .2 alleles).**Positive binding epitopes are marked in bold.Binding is reported as percent relative the binding of a positive control peptide, affinity is reported as an ED50 value (M), and off-rate is reported as a t 1/2 value (h), as described in material and methods.Multimer MHC-peptide complexes were constructed for the epitopes: A02-Rv3875  S1 and S2.

Multimeric analysis of selected TB epitopes
The number of antigen-specific T-cell frequencies ranged between 0 and 3.9% in the CD8+ T-cell population, with an    2A and Table S3).Based on the construction of several different allele-specific multimers (e.g.11 for HLA-A*02, 10 for HLA-A*24 and 11 for HLA-A*30), we could identify differences in the allele-specific repertoires.Some epitopes were never recognized by T-cells in any individuals, while others were recognized in PBMCs from all patients, such as the A*02-restricted epitope KLVANNTRL from Rv1886c, indicating immunodominance (Figure S3).
We could also detect general allele-specific variations in the detection of antigen-specific T-cells.Rv3875 epitopes were, for example, generally recognized very strongly by A*24-restricted Tcells, due to the high frequent recognition of the immunodominant epitope Rv3875 ELNNALQNL .On the other hand, Rv2957 derived epitopes were not recognized in blood from HLA-A*02 and A*68 positive individuals at least the single epitope included in this study derived from this protein (Figure 2B).
Since we managed to produce 6 different multimers presenting the M. tb variant epitope (QI)MYNYPAM(LG), we were also able to examine the nature of antigen-specific T-cell recognition of 'MHC-promiscuous epitopes'.We could not detect significant differences in the detection of antigen-specific T-cells, neither inter-allelic, i.e. if the same epitope IMYNYPAML was presented via the different HLA alleles (e.g.A*02, A*24 and A*30:02), nor intra-allelic when the almost identical epitopes was restricted by the same allele, e.g.A*24 (IMYNYPAML), MYNYPAMLG [0.3-0.4%]), and A*30:02 (QIMYNYPAM and IMYNYPAML [0.3-0.4%]).These data may suggest non-allele-specific T-cell recognition of 'promiscuous epitopes' like (I)MYNYPAML(G) (Figure 2C and Table S3), at least in the study population examined in this report.

Phenotypic analysis of antigen-specific T-cells
Since we detected differences in epitope/antigen recognition based on the restricting MHC class I allele and in the numbers of antigen-specific T-cells, we proceeded to analyze the phenotype of M. tb antigen-specific cells.The phenotype of CD8+ T-cells can be defined by the cell-surface markers CD45RA and CCR7 [44,45].The phenotype frequency analysis revealed that precursor (CD45RA+CCR7+), effector memory (CD45RA2CCR72) and terminally differentiated (CD45RA+CCR72) are almost equally distributed (,30% in each population), while the average percentage of central memory cells (CD45RA2CCR7+) was around 6%. On the other hand, we detected a significantly higher number of antigen-specific CD8+ T-cells in the precursor compartment (45%) (p = 0.01) as well as a lower frequency of cells in the effector memory compartment (21%) (p = 0.002) (Figure 3A).
We were able to detect allele-specific differences concerning Tcell maturation and differentiation.The HLA-A*30-restricted (M.tb specific) T-cells were found, to a higher extent, in the terminally differentiated effector compartment (p = 0.02) as compared to the other HLA-A restricted antigen-specific T-cells.ˆAmino acid substitutions within an epitope are marked in bold italic text.ˆPositive binding epitopes are marked in bold.Binding is reported as percent relative the binding of a positive control peptide, affinity is reported as an ED50 value (M), and off-rate is reported as a t 1/2 value (h), as described in material and methods.MHC class I-peptide complexes were constructed for the epitopes: A02-Rv0288 AMLGHAGDM , A02-Rv0288 AMLDHAGDM , A02-Rv0288 MLGHAGDMA, A02-Rv0288 MLDHAGDMA , A24-Rv0288 MYNYPAMLG and A24-Rv0288 MYNYPTMLD .*Binding, affinity and off-rate have previously been reported in reference [24].**Binding, affinity and off-rate have previously been reported in reference [17].doi:10.1371/journal.pone.0058309.t003 A similar trend could be seen for the allele HLA-C*07 (Figure 3B).For effector memory antigen-specific CD8+ T-cells, we could detect an MHC class I allele family based difference with a significantly higher proportion of T-cells restricted by the allele HLA-B*07 and a similar trend for the other HLA-B alleles included in this study, i.e.B*58 (Figure 3C).For the other two phenotypic compartments (precursor and central memory T-cells), we could also identify a trend towards allele-specific differences, although none were significant (Figure S4) Comparison of differences within the phenotype of antigen-specific CD8+ T-cell between different M. tb proteins showed a profile independent of the nature of the M. tb antigens included in this study (Figure S4) and the same was found to be true for the different epitopes of the Rv0288 hot-spot (QI)MYNYPAM(LG) (Figure S5): differences in antigen-specific CD8+ T-cell maturation/differentiation were associated with the restricting MHC allele, yet not the presented M. tb epitope.

Analysis of degranulation and survival markers
We analyzed the degranulation marker CD107a (LAMP-1), which correlates with the cytotoxic capacity of CD8+ T-cells.We compared the frequency of total CD8+ T-cells and the M. tb antigen-specific cells concerning CD107a expression and could detect significantly higher frequencies of antigen-specific T-cells expressing the degranulation marker CD107a (4% vs. 19%, p = 0.0005) (Figure 3D).The increase in CD107a expression was independent of the restricting MHC allele and the nature of the M. tb antigen and ranged from 5% (B*07) to 50% (A*68).M. tb epitopes could be divided into 4 groups according to the frequency of the antigen-specific CD8+ T-cells expressing CD107a.To the first epitope group belonged T-cells with more than 70% expressing CD107a, i.e. predominantly A*68 restricted epitopes.The second group consisted of epitopes recognized by cells expressing CD107a on 40-70% of the cells, i.e. mostly A*30restricted epitopes.The third group consisted of epitopes recognized by cells expressing CD107a on 20-40% of the antigen-specific T-cells, i.e. mostly A*30 and A*24 restricted epitopes, and the rest belonged to the forth group.Generally, the HLA-A*30 and A*68 restricted antigen-specific T-cells expressed higher frequencies of CD107a (Figure 3E).
An important marker for T-cell survival is CD127 (IL-7 receptor a-chain).We were able to detect a significantly increased expression of this cell-surface marker on antigen-specific T-cells (81%) as compared to the total CD8+ T-cells (69%) (p = 0.001) (Figure 3F).This was also true for all MHC class I alleles presenting M. tb epitopes to CD8+ T-cells, except B*07 (Figure 3G).Based on variations between epitope-derived M. tb protein targets, we could neither detect difference regarding expression of CD107a nor of CD127 (Figure S6).This was also true when comparing the Rv0288 derived 'super-epitope' which binds to a higher number of MHC class I molecules (QI)MYNY-PAM(LG) (Figure S6) Also here, the expression of CD107a and CD127 was closely associated with the MHC class I restricting allele, and not the M. tb target protein.

Effect of amino acid substitutions on TCR recognition
An amino acid substitution within an epitope might abrogate or profoundly change the interaction to a specific MHC class I allele.However, the effect on TCR recognition is biologically equally relevant.To be able to study TCR-MHC-class I peptide interaction, we constructed 3 pairs of multimers consisting of a wild-type M. tb target epitope (from H37Rv) and epitopes containing one or two naturally occurring substitutions for the alleles A*02 and A*24.The two epitopes binding to HLA-A*02 (AML(G/D)HAGDM and ML(G/D)HAGDMA) had the mutations in the 3 rd or the 4 th amino acid potentially being part of the TCR recognition site, while the epitope associating with A*24 (MYNYP(A/ T)ML(G/D)) contained two mutations in the 6 th and 9 th position (of which the first position might affect TCR recognition).Although the association of these variant M. tb epitopes to its corresponding MHC molecule had either no effect, or showed decreased MHC binding, multimers containing these M. tb variant epitopes showed a tendency of detecting a higher frequency of antigen-specific T-cells as compared to multimers presenting the wild-type epitopes (A2 AML(G/D)HAGDM 0.12/0.20%,A2 ML(G/D)HAGDMA 0.13/0.25%and A24 MYNYP(A/T)ML(G/D) 0.44/0.51%)(Figure 4A).Next, we tested whether the wild type and the variant epitope were recognized by the same or different T-cell population(s).This was possible by costaining PBMCs with two separate multimers presenting the reference and variant epitope labeled with different fluorochromes.The results clearly showed that two different antigen-specific T-cell populations recognized the wild type and the variant M. tb epitope (Figure 4B).

Biochemical factors influencing antigen-specific T-cell recognition
Previous data suggested correlations between high MHC class Ipeptide affinity and dominant T-cell responses [46].We compared therefore the binding characteristics of the p-MHC complex with the actual frequency of antigen-specific T-cells recognizing a specific epitope.Based on our data concerning affinity, dissociation rate and M. tb epitope-specific recognition frequencies, it was possible to determine that among the 7 epitopes, which showed the highest average antigen-specific recognition (.0.4%), the affinity varied considerably (between 6 mM and 8 nM), as well as the offrate (0.5 h-27 h).In addition, we could detect allele-specific characteristics such as very high affinity regarding all the HLA-A*30 restricted epitopes and the opposite was found to be true regarding the B*07 restricted epitopes.However, this did not correlate with the frequency of antigen-specific CD8+ T-cells (in PBMCs from patients with active TB).The highest frequency of M. tb-specific T-cells for almost all epitopes was restricted by HLA-A*24 (Figure 5).

Discussion
Our study shows that, at all three levels (MHC binding, T-cellrecognition and the nature of the specific T-cells), we were able to detect allele-specific differences which potentially affect CD8+ Tcell immune-recognition.Regarding the p-MHC associations, we were able to confirm a broad peptide-binding repertoire for some alleles (e.g.HLA-A*02:01, A*24:02 and A*30:02) and a more restricted peptide-binding repertoire for other alleles (e.g.A*30:01 and B*07:02).This is in line with previously reported data (yet with different peptide ligands) [17,24,39,47].In addition, we included the less well studied alleles A*68:01 and B*58:01 as well as the (not yet reported) allele C*07:01.A*68:01 and B*58:01 showed a relatively broad binding repertoire, while C*07:01 showed a very restricted peptide binding.Structural constrains, electrostatic charge and hydrophobicity within the MHC binding pockets dictate the binding peptide repertoire [48] and are therefore part of the explanation why inter-allelic differences can be seen in the broadness of the peptide binding repertoires.In addition, these factors influence the stability of the p-MHC complex and might therefore explain some of the inter-allelic differences that could be seen in affinity and peptide dissociation rates.
The stability and off-rate are factors that are biologically important; they determine the time frame in which a certain epitope would be available for T-cell priming and for recognition in the effector phase of the immune response.The allele-specific differences in the p-MHC dissociation rate probably originate from intrinsic structural features of the allotype.This would imply that not only a limited number of epitopes would be available for presentation by these alleles but also that the time-frame in which the presentation could occur would be shorter.This scenario might entail a disadvantage for immune-recognition in individuals with certain allotypes, yet a fast peptide off-rate seems to be able to be counter-balanced by a high affinity, as shown for a number of M. tb epitopes included in this study.Affinities (ED50) between 100 mM and 10 nM have previously been identified using a similar experimental design and our findings in this study are in line with these data [24,39].
We identified highly promiscuous 'hot-spots' in M. tb proteins of inter-allelic binding such as Rv1886c [(KL)VANNTRL(WV) and (L)AAYHPQQF(I)] and Rv0288 [(QI)MYNYPAM(LG)], adding more alleles to the already identified hotspot in this protein [17,24].For some epitopes, promiscuous binding to up to 7 different HLA class I alleles including members of both HLA-A, B and C allele families have been identified in this study.In addition, some of these areas have previously been shown to be able to bind to the MHC class II alleles DR1, DR2 and DR4 [32].We have also been able to show allele-independent T-cell recognition for the Rv0288 derived epitope between five different HLA-A alleles.Since the individual T-cell responses were found to be similar, concerning T-cell frequency, T-cell phenotype and effector functions, these immunological 'hot-spots' might be potential candidates for peptide-based vaccine-strategies since the allelic coverage would include a large part of the world population, including the most common HLA-A allele in Caucasians (A*02), in Asians (A*24) and in Africans (A*30) [9].
Even though the importance of different MHC class I allotypes in shaping the possible peptide-repertoires is well established, we know in fact very little about how inter-allelic variations affect the generation, phenotype and functionality of a pathogen-specific Tcell repertoire.The immune-recognition by antigen-specific CD8+ T-cells in many diseases is often directed against a few of the many potential epitopes originating from a complex pathogen.This 'skewing' of the immune system gives rise to immune-dominant epitopes, which might be accompanied by less recognized subdominant epitopes [49].The driving force behind immunedominance and sub-dominance in different disease settings is not completely understood, but availability (time and receptor numbers) of a certain p-MHC complex on the cell-surface [50], as well as conformation of the epitope presented [51] and frequencies of T-cells precursors with the possibility of recognizing the epitope [52] seem to be crucial factors in this process.In this study, we were able to detect a dominant response in blood from some patients, for instance against the A*24 restricted Rv3875 epitope ELNNALQNL.This is interesting, since the majority of previously described dominant M. tb.epitopes seem to be HLA-B restricted [24,53].However, T-cell immune-dominance concerning a single epitope seems to be quite rare in active M. tb infection, since the mapping using 48 different multimers (some were previously described [17,24,40,41,54]) has not revealed a single dominant CD8+ T-cell epitope, yet a low frequency of a broad panel of co-dominant M. tb epitopes in the peripheral circulation of most patients infected with M. tb.The dominance and the immunological profile might however differ between the site of infection and the peripheral blood cells used in this study, as previously reported [42].
The generation of a broader T-cell repertoire may protect from the emergence of escape mutations in viral disease.A broader Tcell repertoire targeting 'subdominant' epitope may not be associated with clinical efficacy in TB, since these samples were taken from patients who failed to contain their disease (all individuals in the current study exhibited active pulmonary disease).This seems to be more in line with the low frequency of disease-reactive T-cells in the peripheral circulation that previously has been seen in the context of cancer [55] or the loss of certain immune-dominant epitopes in the peripheral circulation in chronic infection with simian immunodeficiency virus (SIV).Future studies will show, whether dominant T-cell responses arise after effective TB treatment, whether such immune responses are associated with latent TB, which represents a broad clinical spectrum, or even with continuous exposure to M. tb.
Deciphering the phenotype of antigen-specific CD8+ T-cells is important in the context of memory formation as well as for immune effector functions [44,45].Previous studies have identified the majority of antigen-specific T-cells residing in memory (CD45RA2) [40,56] or terminally differentiated (CD45RA+CCR72) [41] T-cells.Yet, an increasing number of studies identified a majority of M. tb-specific or BCG-reactive Tcells in both adults and children belonging to 'naı ¨ve'/precursor (CD45RA+CCR7+)/(CD45RA+CD28+) [17,57,58,59] compartment.In this study, we were once again able to detect that the majority of M. tb antigen-specific T-cells resides in the (CD45RA+CCR7+) precursor population -in contrast to the phenotype of total CD8+ T-cells which showed equal number of naı ¨ve, effector memory and terminally differentiated cells.It has previously been speculated for HIV infection, where this phenomenon also has been observed, that the antigen-specific T-cells with cell-surface expression of both CD45RA and CCR7 might represent effector memory cells that reverted back to a 'naı ¨ve' like phenotype without losing this subset's ability of proliferation upon antigen-stimulation [45], yet it may also be due to increased lymphopoiesis in patients with active TB.
We detected allele-specific variations, rather than antigenspecific, regarding frequencies and the differentiation/maturation profile (based on CD45RA/CCR7) of antigen-specific T-cells.First, we hypothesized that this finding was due to differences in the presented M. tb epitope and thus due to the nature of the target protein, since the promiscuous epitopes were similarly recognized by inter-allelic restricted T-cells.However, when we compared several different epitopes, restricted by the same allele, allelespecific patterns of immune-recognition emerged while the antigen-specific patterns were remarkably constant.This was in contrast to the levels of vaccine-induced antigen-specific T-cells which were found to be higher targeting epitopes originating from Rv1886c (compared to Rv0288) in one previous study [25], it was also in contrast to the different phenotype of antigen-specific cells found in another study [41].However, the immune-profile of active TB disease and the vaccine-induced profile might not correlate, and in the second study the phenotypic profile of only a single epitope/protein was compared.
To our knowledge, no correlations between the T-cell profile and the restricting allele have been shown before in TB.One example has been reported in HIV infection where some alleles like B*57 and B*58, that have been associated with 'elite controllers', associated with higher numbers of specific cytokine producing cells [60].Interestingly, the prevalent African alleles HLA-A*30 and C*07 restricted epitopes give rise to higher proportions of terminally differentiated effector CD8+ T-cells in African TB patients and this T-cell phenotype has been shown to be important in controlling the disease [3].One possible explanation is that the long-standing co-evolution between mycobacteria and humans has taken part in the regional selection of the most prevalent MHC class I alleles based on an advantage in CD8+ T-cell responses, i.e. to expand immune effector T-cells.In line with this observation is also the fact that these alleles seem to restrict M. tb antigen specific T-cells with high CD107a expression ex vivo (indicating cytotoxic function).
We were able to detect biologically relevant differences concerning the naturally occurring amino-acid substitutions in the protein Rv0288.Since M. tb previously has been shown to express low sequence variability in most of its epitopes, it was speculated that some immune-recognition might be beneficial for the bacteria, for instance in the establishment of a latent infection or by contributing to transmission [23].We have studied the effect of the epitope-based polymorphism on both the MHC level as well as on the TCR recognition level for the protein Rv0288.An explanation for the higher frequency of amino acid substitutions in this protein [23] (indicating immune-evasion) might be that most of the substitutions are part of immunological hot-spots with broad inter-allelic immune recognition [17,24].We could show in the current study that most of the substitutions affected MHC class I binding negatively, indicating a potential loss of peptide presentation.Some of the substitutions reside within an anchor residue of the peptides and might therefore directly interfere with allelespecific peptide-binding.On the other hand, explanations for the negative effect of other substitutions might be that they cause steric hindrance, or changes in intra-peptide constrains changing the conformation of upward pointing residues [61].The outcome of individual p-MHC-TCRs interactions is determined by affinity and avidity dictating positive and negative downstream signaling events such as CD3 zeta-chain phosphorylation [62].In this study, we could still detect an epitope-specific antigen-specific T-cell population after introducing the amino acid substitution, although these T-cells were found to belong to a complete separate cell population as compared to the T-cells recognizing the original epitope.This change in TCR-repertoire usage upon amino acid substitution in an epitope has previously been seen in SIV infections [63] as well as within an influenza epitope restricted by H2-D b [61].It would be of interest to elucidate if these different Tcell populations exhibit similar functionality showing a full or partial agonistic, or even antagonistic, effect on the TCR repertoire recognizing wild-type or variant M. tb epitopes.
In conclusion, we constructed a broad panel of novel tools to be able to study M. tb-specific responses in African with TB infection.We could detect a low frequency of a broad repertoire of codominant M. tb epitopes in the peripheral circulation of patients with active TB.The study of host-pathogen effects on immunerecognition in M. tb derived CD8+ T-cell epitopes revealed 1) MHC class I allele specific differences in the broadness of the peptide-binding repertoire, 2) associations of the frequency and phenotype of M. tb antigen-specific T-cells with the MHC class I restricting alleles and 3) differences between variant M. tb epitopes concerning peptide binding and T-cell recognition.This study provides a first view of the nature of the anti-M.tb specific CD8+ response in patients with an African background.It is our hope that these findings, as well as these novel tools will aid future research on M. tb in Africa.

3 *
Promiscuous epitopes are marked in bold or italic (italic -binding .1

Figure 1 .
Figure 1.Alignment of the major histocompatibility complex (MHC) class I-binding epitopes with the amino acid sequence of Rv1886c.Epitopes identified for HLA-A*02:01 are shown in light blue, for A*24:02 in dark blue, for A*30:01 in dark green, peptides for A*30:02 in light green, for A*68:01 in yellow, for B*07:02 in orange, for B*58:01 in brown and for C*07:01 in red.Alignments of MHC class II restricted epitopes (HLA-DR1 -dark grey, DR2 -medium grey and DR4 -light grey) that have previously been reported in reference [32] are included as well.The signal sequence of Rv1886c is shown in red and the rest of the protein in black.doi:10.1371/journal.pone.0058309.g001
*Promiscuous epitopes are marked in italic.**Positive binding epitopes are marked in bold.Binding is reported as percent relative the binding of a positive control peptide, affinity is reported as an ED50 value (M), and off-rate is reported as a t 1/2 value (h), as described in material and methods.MHC class I-peptide complexes were constructed for the epitopes: A68-Rv0288 HAMSSTHEA and A68-Rv0288 ANTMAMMAR .doi:10.1371/journal.pone.0058309.t002

Table 3 .
MHC class I binding, affinity and off-rate data for the variant peptide-epitopes derived from Rv0288 (TB10.4).Bind Aff O-rate Bind Aff O-rate Bind Aff O-rate Bind Aff O-rate Bind Aff O-rate Bind Aff O-rate Bind Aff O-rate