A High Throughput Whole Blood Assay for Analysis of Multiple Antigen-Specific T Cell Responses in Human Mycobacterium tuberculosis Infection Free

The vast majority of individuals infected with Mycobacterium tuberculosis never develop signs or symptoms of active tuberculosis (TB) disease, thus providing compelling evidence that there are host immune responses that are capable of containing the infection. Both innate and adaptive immunity are essential components of the immune response to M. tuberculosis infection, with an important role for CD4 T cells demonstrated by animal models of M. tuberculosis infection (1–3) and human studies of individuals coinfected with HIV (4, 5). Induction and/or boosting of M. tuberculosis–specific CD4 T cell responses has been a central focus of novel TB vaccine candidates (6, 7). Despite the importance of CD4 T cells in M. tuberculosis infection, the Ag specificity, phenotype, and function of M. tuberculosis–specific CD4 T cells that correlate with immune control or risk of TB disease have not been well defined. Immunological tools that enable identification of individuals with latent M. tuberculosis infection (LTBI) who are at highest risk for development of active TB disease are currently lacking.

The M. tuberculosis genome contains over 4.4 million bp, encoding ∼4000 genes (8), thus imposing a significant challenge to conducting genome-wide characterization of M. tuberculosis–specific CD4 T cell responses. Several studies have measured IFN-γ secretion to evaluate T cell responses to panels of recombinant M. tuberculosis proteins in cross-sectional cohorts of individuals with LTBI and active TB disease, in an effort to identify novel vaccine candidates, as well as Ags associated with different states of M. tuberculosis infection (9–16). However, such studies have generally not determined CD4 and CD8 T cell reactivity or further defined specific T cell epitopes, and have also not been applied to longitudinal cohort studies to identify changes in M. tuberculosis Ag recognition during the course of M. tuberculosis infection or treatment.

Although the M. tuberculosis–specific CD4 T cell correlates of protection are not well defined, differential CD4 T cell responses that underlie or correlate with M. tuberculosis infection states may consist of different T cell phenotypes, functions, trafficking capabilities, activation states, and/or Ag specificities (17, 18). Tracking M. tuberculosis–specific CD4 T cells on an epitope-specific level requires additional, in-depth screening with peptides, including synthesis of overlapping peptides spanning the full sequence of an Ag, or synthesis of specific epitope peptides according to HLA binding predictions. Studies conducting comprehensive screening of M. tuberculosis–specific CD4 T cell responses in individuals with LTBI have been conducted recently using large panels of peptides corresponding to epitopes predicted to bind to a panel of HLA DR, DP, and DQ class II alleles (19, 20). These studies have further underscored the heterogeneity of CD4 T cell immunity in LTBI, and have identified antigenic islands of epitope specificity, which are largely focused on Ags related to bacterial secretion systems (20). However, it remains unknown whether distinct populations of M. tuberculosis–specific T cells have differential contributions to mediating control of M. tuberculosis infection. Moreover, it is currently unclear how the breadth and specificity of M. tuberculosis–specific T cell responses changes over time within the same individual in the setting of immunotherapeutic interventions, progression to TB disease, and antimicrobial treatment for either LTBI or active TB disease.

To further enhance the toolkit for evaluating M. tuberculosis–specific T cell responses in multisite and longitudinal studies of individuals with distinct states of M. tuberculosis infection, we sought to develop a high throughput and reproducible assay requiring low blood sample volumes to evaluate M. tuberculosis–specific T cell responses to multiple M. tuberculosis Ags simultaneously. We describe here the optimization of a microtiter plate-based, diluted whole blood stimulation assay utilizing overlapping peptide pools corresponding to a functionally diverse panel of M. tuberculosis Ags; furthermore, we describe implementation of the assay in the United States and a field site in Kenya, as well as application to longitudinal cohort studies of M. tuberculosis–infected individuals in a TB-endemic setting.

Blood samples were collected in sodium heparin Vacuette tubes (Greiner Bio-One) from LTBI individuals and uninfected, healthy control adults enrolled at the DeKalb County Board of Health Refugee Clinic and the Emory Vaccine Center in metropolitan Atlanta, GA and the Kenya Medical Research Institute (KEMRI) Clinical Research Center in Kisumu, Kenya. Healthy control adults with a negative tuberculin skin test and/or a negative QuantiFERON-TB Gold (QFT) test and no history of exposure to active TB were enrolled only in Atlanta, whereas individuals with LTBI were enrolled both in Atlanta and Kenya. A total of 20 healthy control individuals were enrolled in Atlanta (65% Black, 20% White, 10% Asian, and 5% Hispanic); 44 individuals with LTBI were enrolled in Atlanta (66% Black, 2% White, and 32% Asian); and 18 individuals with LTBI were enrolled in Kenya (100% Black). All individuals with LTBI included in the study met the following inclusion criteria: asymptomatic adults ≥18 y of age; a positive QFT result; seronegative for HIV Abs; no previous history of diagnosis or treatment for active TB disease; no previous history of treatment for LTBI; and a normal chest X-ray. Individuals with LTBI who were enrolled in Kenya were household contacts of a sputum smear-positive active TB patient. All subjects provided written informed consent for participation in the study, which was approved by the Institutional Review Boards at Emory University, the Georgia Department of Public Health, and the KEMRI Scientific and Ethics Review Unit. The U.S. Centers for Disease Control and Prevention reviewed the protocol and chose to rely on the oversight of KEMRI Scientific and Ethics Review Unit.

Initial optimization experiments were conducted using pooled, overlapping 15- or 16-mer peptides corresponding to the sequences of CFP-10, ESAT-6, and TB10.4 (BEI Resources, National Institute of Allergy and Infectious Diseases, National Institutes of Health: Peptide Array, M. tuberculosis CFP-10 Protein, NR-34825; Peptide Array, M. tuberculosis ESAT-6 Protein, NR-34824; Peptide Array, M. tuberculosis TB10.4 Protein, NR-34826). 18-mer peptides (overlapping by 11 aa) corresponding to the 60 M. tuberculosis Ags listed in Table I were synthesized on a 10 mg scale using Fmoc chemistry (Synthetic Biomolecules, San Diego, CA) according to the H37Rv genome sequence. Peptides were pooled by Ag (10–121 peptides per pool, with a median of 39 peptides per pool); peptide pools were numbered consecutively according to Rv gene numbers (Table I). After pooling the peptides for each Ag, pools were lyophilized, followed by reconstitution in DMSO at a final concentration of 1 mg/ml per peptide. Peptide pools were further diluted in RPMI 1640 medium supplemented with 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin, to prepare working stocks for use in the stimulation assays described below. PHA-M (PHA; Sigma-Aldrich) was used as positive control.

Heparinized whole blood was transported to the respective laboratories at the Emory Vaccine Center and KEMRI, and processed within 2 h of collection. In initial studies, blood was diluted at varying dilutions up to 5-fold with response spectrum assay (RSA) media, consisting of RPMI 1640 medium supplemented with 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin. Diluted blood was added in a volume of 100 μl to each well of a sterile 96-well round-bottom, tissue culture–treated plate (Corning). Each well contained 100 μl of either RSA media alone (negative control), RSA media with M. tuberculosis peptide pools, or RSA media with PHA (positive control). The final volume in the assay was 200 μl per well, with the final blood dilutions ranging from 1:2 to 1:10. Two hundred microliters of sterile PBS was added to the empty wells on the outer edge of the 96-well plate (Fig. 1). Plates were incubated in a 37°C incubator with 5% CO₂ for 7 d. On day 7, plates were centrifuged at 2000 rpm for 5 min, and 150 μl of plasma supernatant was removed from each well and transferred to a V-bottom 96-well plate (Corning). Plasma supernatants were either used immediately or stored at −80°C until use in ELISAs.

IFN-γ ELISAs were conducted according to the manufacturer’s instructions (Human IFN-γ Uncoated Ready-SET-Go! ELISA kit; eBioscience). Fifty microliters of plasma supernatant from the whole blood stimulations was diluted with 50 μl of assay diluent for use in the IFN-γ ELISA. ELISA plates were read at 450 nm using a molecular diagnostics spectrophotometer.

IFN-γ ELISA data were analyzed using SoftMax Pro v6.3 software (Molecular Devices). Background IFN-γ production for each individual was determined by calculating the average concentration of IFN-γ in the six negative control wells, and the mean background IFN-γ production was subtracted from the Ag-stimulated wells. A maximum concentration of quantifiable IFN-γ was set at 1000 pg/ml, corresponding to the concentration of the highest standard of recombinant human IFN-γ protein in the ELISA. IFN-γ concentrations below the level of detection by the ELISA standard curve were set to 0 pg/ml.

Statistical testing was done using GraphPad Prism v7.0b and R software programs. Paired comparisons were evaluated using the Wilcoxon matched-pairs signed rank test. Correlations were evaluated using the Pearson correlation, with statistical significance evaluated using the Kendall rank correlation coefficient. The p values <0.05 were considered significant.

To further enhance the evaluation of M. tuberculosis–specific T cell responses in cohorts of individuals across different states of M. tuberculosis infection, we sought to develop a high throughput, reproducible and transportable Ag-specific T cell RSA in which we could measure M. tuberculosis–specific T cell responses to multiple Ags within the same individual using small quantities of blood (Fig. 1). M. tuberculosis infection in humans generates a heterogeneous T cell response targeting highly conserved epitopes across a broad range of Ags (20–22). We selected a panel of 60 M. tuberculosis Ags for further evaluation that have been previously confirmed for recognition by T cells in humans with different states of M. tuberculosis infection (9, 20, 23, 24), and that represent a diverse range of functional categories, bacterial cell fractions, and gene families with specific characteristics (Table I). Overlapping 18-mer peptides were synthesized according to the full-length H37Rv sequence of each Ag; peptides were pooled by Ag.

Previous studies have measured cytokine production in 1:5 and 1:10 diluted whole blood cultures stimulated for 7 d with mitogens (25) and recombinant M. tuberculosis proteins expressed in Escherichia coli (10, 12, 14, 15). To determine the optimal blood dilutions and the kinetics of IFN-γ production for our panel of M. tuberculosis peptide pools, blood was collected from individuals with LTBI and stimulated with peptide pools corresponding to the immunodominant M. tuberculosis Ags CFP-10, ESAT-6, and TB10.4. Blood was used undiluted (200 μl blood per well), or diluted with RSA medium at 1:2, 1:5, and 1:10 (in a final volume of 200 μl of diluted blood per well). Peptide pools were stimulated in duplicate wells, with four replicate plates set up for harvesting supernatants on days 1, 3, 5, and 7. For all three M. tuberculosis peptide pools in both individuals tested, IFN-γ production, as measured by ELISA, was low to undetectable in 1:10 diluted cultures, compared with undiluted, 1:2, and 1:5 diluted blood (Fig. 2).

Harvesting of supernatants from each of the indicated blood dilutions at multiple time points indicated IFN-γ production was detectable as early as day 1, with peak detection by day 5 (Fig. 2). Given the similarities in concentrations of IFN-γ between days 5 and 7, combined with the logistical consideration of harvesting supernatants on weekdays and not weekends, we opted to proceed with harvesting of diluted blood supernatants on day 7.

The data above indicated diminished detection of M. tuberculosis–specific IFN-γ production in 1:10 diluted blood cultures, thus 1:10 diluted blood was excluded from further testing. In addition, we excluded undiluted blood, which would not be feasible given the large volumes of blood that would be required to test the full panel of 60 M. tuberculosis Ags and controls. Therefore, we proceeded with the next phase of optimization comparing 1:2 and 1:4 diluted blood.

We next evaluated the concentrations of M. tuberculosis peptide pools and PHA for use in the T cell RSA. Stimulation of 1:10 diluted blood for 7 d with peptide pools at a final concentration of 1 μg/ml per peptide has been previously described (16). Using 1:2 diluted blood from individuals with LTBI, we measured IFN-γ production by ELISA in cultures stimulated for 7 d with 2, 1, and 0.5 μg/ml of CFP-10, ESAT-6, and TB10.4 peptide pools (Fig. 3). For the three M. tuberculosis peptide pools tested, IFN-γ production in response to 1 μg/ml of peptide pools was not inferior to 2 μg/ml of peptide pools (Fig. 3A–C). However, IFN-γ production was lower following stimulation with 0.5 μg/ml ESAT-6 peptide pool, compared with 2 and 1 μg/ml (Fig. 3A–C). Therefore, to increase the potential of detecting low magnitude M. tuberculosis–specific T cell responses, and to economize use of peptides synthesized for large cohort studies, we opted to proceed with a final peptide pool concentration of 1 μg/ml for use in the T cell RSA.

To demonstrate that the IFN-γ production measured following stimulation of diluted whole blood is indeed Ag-specific, we stimulated 1:2 diluted blood from healthy QFT⁻ adults with CFP-10, ESAT-6, and TB10.4 peptide pools. In contrast to QFT⁺ individuals with LTBI, significant IFN-γ production was not detected in any of the three M. tuberculosis peptide pools, at any of the concentrations tested (Fig. 3A–C).

Lastly, we determined optimal concentrations of the positive control (PHA) in the T cell RSA. In both QFT⁺ and QFT⁻ individuals, stimulation of 1:2 diluted blood with 1 μg/ml PHA induced significantly lower levels of IFN-γ, compared with 5 μg/ml PHA (Fig. 3D). All individuals tested had a positive IFN-γ response to stimulation with 5 μg/ml PHA, thus this concentration was selected for use in subsequent assays.

The initial optimization experiments described above were conducted using a small panel of immunodominant M. tuberculosis peptide pools: CFP-10, ESAT-6, and TB10.4. A major goal in establishing the T cell RSA was to enable testing of a large panel of Ags while minimizing the amount of blood required. To address these issues, our next set of experiments incorporated stimulation of 1:2 and 1:4 diluted blood with 20 M. tuberculosis peptide pools. In order to select 20 Ags in an unbiased manner, we chose every third Ag from our consecutive list of M. tuberculosis Ags in Table I. Using heparinized blood samples collected from a cohort of 15 QFT⁺ individuals, we first evaluated the performance of 1:2 and 1:4 diluted blood in the negative (no Ag) and positive control (PHA) conditions. There was no difference in IFN-γ production in direct comparisons of 1:2 and 1:4 diluted blood for the negative and positive controls (Fig. 4A). Next, we correlated IFN-γ production in 1:2 versus 1:4 diluted blood stimulated with the selected panel of 20 M. tuberculosis peptide pools. There was a significant positive correlation between IFN-γ production to M. tuberculosis peptide pools following stimulation with 1:2 and 1:4 diluted blood from QFT⁺ individuals (Fig. 4B).

To evaluate intra-assay variability, each of the 20 M. tuberculosis peptide pools was tested in triplicate wells of the 96-well plate. Of the M. tuberculosis peptide pools eliciting a response >50 pg/ml of IFN-γ, the median coefficient of variation (CV) of IFN-γ across the triplicate wells was similar when comparing 1:2 and 1:4 diluted blood (25 and 29%, respectively, Fig. 4C). The intra-assay variability was significantly lower than interindividual variation in IFN-γ responses to the 20 M. tuberculosis peptide pools in this cohort of QFT⁺ individuals (Fig. 4C), thus indicating heterogeneity in the pattern of M. tuberculosis T cell Ag recognition in individuals with LTBI.

We next evaluated whether 96-well plates containing M. tuberculosis peptide pools and PHA could be prepared in batch and frozen at −80°C prior to use. There were no differences in IFN-γ production in the negative controls, PHA, or M. tuberculosis peptide pool stimulations when adding 1:2 or 1:4 diluted blood to freshly prepared Ag plates, or Ag plates that had been previously prepared in batches, frozen, and then thawed just prior to blood collection (Fig. 5 and data not shown). Taken together, these results indicate that 1:4 diluted blood correlates well with, and is not inferior to, 1:2 diluted blood when evaluating IFN-γ production following 7-d stimulation with a panel of 20 M. tuberculosis peptide pools. Moreover, large quantities of 96-well plates containing M. tuberculosis peptide pools and PHA can be preprepared in batch and frozen prior to addition of the diluted blood, thus further reducing intra-assay variability and facilitating on-site assay standardization for large cohort studies.

From the experiments described above, we concluded that 1:4 diluted blood could be used to measure T cell IFN-γ production to M. tuberculosis peptide pools with an intra-assay variability that is substantially lower than the interindividual variability. One of the intended applications of the T cell RSA is in longitudinal cohort studies. Thus, to evaluate the reproducibility of the T cell RSA, we collected blood from 15 QFT⁺ individuals at three separate time points, spaced at weekly intervals. The T cell RSA was conducted using 1:4 diluted blood collected at each of the three study visits, and stimulated with the panel of 20 M. tuberculosis peptide pools depicted in Fig. 4. IFN-γ production to these 20 M. tuberculosis peptide pools was highly consistent across each of the three time points (Fig. 6).

We next initiated testing of the T cell RSA using the full panel of 60 M. tuberculosis peptide pools in cohorts recruited at study sites in the United States and in a TB-endemic setting in Kisumu, Kenya. A schematic overview of the fully optimized T cell RSA is shown in Fig. 1. In a cohort of M. tuberculosis–unexposed, QFT⁻ adults from the metropolitan Atlanta area, the median IFN-γ production to stimulation with our panel of 60 M. tuberculosis peptide pools ranged from 0 to 41 pg/ml, with 75% of the M. tuberculosis peptide pools eliciting a median IFN-γ response of 0 pg/ml (Fig. 7A). All M. tuberculosis–unexposed individuals had a maximal IFN-γ response to PHA (>1000 pg/ml) in this assay.

We next evaluated the performance of the T cell RSA in a cohort of refugees with LTBI who had recently arrived in the United States and were resettled in the metropolitan Atlanta area. In contrast with the healthy, uninfected individuals in Atlanta, multiple M. tuberculosis peptide pools elicited high concentrations of IFN-γ in the refugee LTBI cohort, with many individuals displaying maximal responses >1000 pg/ml to individual M. tuberculosis peptide pools (Fig. 7B). Ag85A (pool 57) elicited the highest median concentration of IFN-γ in this cohort, followed by ESAT-6 (pool 59).

We also evaluated the feasibility of establishing the T cell RSA in a TB-endemic region in Kisumu, Kenya. Multiple M. tuberculosis peptide pools elicited a positive IFN-γ response in the Kenyan LTBI cohort, although the responses were generally lower in magnitude than the refugee LTBI cohort in Atlanta. EspI (pool 60) elicited the highest median concentration of IFN-γ, followed closely by Ag85A (pool 57); all individuals in the Kenyan LTBI cohort elicited a maximal IFN-γ response (>1000 pg/ml) to stimulation with PHA (Fig. 7C).

Lastly, we evaluated the performance of the T cell RSA in a longitudinal cohort of 10 QFT⁺ household contacts of a smear-positive active TB patient in Kisumu, sampled every 6 mo over the course of a year (Fig. 8). As indicated in the heat map in Fig. 8, M. tuberculosis peptide pool stimulation induced IFN-γ production consistently across the three time points in these individuals. For example, PPE26 (pool 26) and PPE51 (pool 52) elicited a maximum IFN-γ response (>1000 pg/ml) in donor 12 at enrollment (baseline), at month 6, and again at month 12, whereas pool 30 (Rv1954c) elicited strong IFN-γ responses in this same donor at baseline and month 6, but decreased by month 12 (Fig. 8). Together these data indicate that the T cell RSA can be applied at several diverse study sites, and can be reliably implemented in longitudinal cohort studies to monitor the spectrum of M. tuberculosis–specific T cell responses over time; such tools are particularly useful in monitoring individuals during treatment for LTBI or active TB disease, and monitoring changes in T cell responses over time that are associated with or predict M. tuberculosis infection outcome.

Infection with M. tuberculosis induces a broad and heterogeneous Ag-specific T cell response, yet many questions remain to be addressed before comprehensive T cell signatures of successful immune control of M. tuberculosis infection can be defined. Among these knowledge gaps is an incomplete understanding of the relationship between the breadth and specificity of the M. tuberculosis–specific T cell response and infection outcome, and how this response changes with TB disease progression, anti-TB treatment, or vaccination. Previous studies on immunogenicity of multiple M. tuberculosis Ags in M. tuberculosis–infected individuals have used either diluted whole blood stimulation assays with IFN-γ secretion detected by ELISA (9, 10, 12, 15, 16), or PBMC-based IFN-γ ELISpot assays (19, 20). We have established a reproducible and high throughput diluted whole blood assay to further refine the immunological tools for monitoring multiple M. tuberculosis–specific T cell responses simultaneously that could be applicable to large cohort studies, multisite studies, and longitudinal studies of either natural history of infection, vaccination, or treatment interventions.

Important initial considerations in development of an assay for measurement of T cell responses to a spectrum of M. tuberculosis Ags were sample source and volume. In order to minimize the amount of sample processing and operator variability, we opted to use a diluted whole blood–based assay, initially described as a tool for measuring human cytokine production profiles following mitogen stimulation of 1:10 diluted blood cultures (25). Similar assays have subsequently been conducted for measurement of immunogenicity of recombinant M. tuberculosis Ags in 7-d, 1:10 diluted blood stimulation assays (10, 12, 14, 15). Although assays utilizing 1:10 diluted blood have the advantage of using minimal amounts of blood, the potential loss of detection of IFN-γ–secreting M. tuberculosis–specific T cells with increasing blood dilutions had not been systematically evaluated in these previous reports. By comparing levels of IFN-γ secretion to three M. tuberculosis peptide pools across multiple blood dilutions from the same individual, we determined that detection of IFN-γ in culture supernatants was substantially diminished following stimulation of 1:10 diluted blood with M. tuberculosis peptide pools, compared with 1:5, 1:2, and undiluted blood. Expanding our analysis to a broader panel of 20 different M. tuberculosis peptide pools, we determined that 1:4 diluted blood was not inferior to 1:2 diluted blood in our ability to detect M. tuberculosis–specific T cell IFN-γ secretion. In an assay volume of 200 μl per well of a 96-well plate, this corresponds to 50 μl of blood per test condition in 1:4 diluted blood assays, versus 100 μl of blood in 1:2 diluted blood assays. Using a final blood dilution of 1:4 enables evaluation of 60 M. tuberculosis Ags, as well as controls, with a total volume of 3.5 ml blood. Assays evaluating Ag-specific T cell responses to multiple Ags simultaneously, using small amounts of blood and no prior cell processing, are highly advantageous in pediatric studies in which blood volumes are limited, as well as resource-limited settings that lack the capacity to isolate and store cryopreserved PBMCs.

In addition to sample type and volume, another key consideration in establishment of this assay was the nature of the M. tuberculosis Ag used to stimulate T cells. Our goal is to define the Ag specificity of M. tuberculosis–specific T cell responses down to the epitope level; thus we opted to pursue assay optimization using 18-mer peptides overlapping by 11 aa, and pooled by Ag. Most previous studies using 1:10 diluted blood assays to identify immunogenic M. tuberculosis Ags have stimulated with recombinant M. tuberculosis proteins, expressed in E. coli (9, 12, 15, 26). However, this approach does not directly facilitate further downstream identification of specific T cell epitopes targeted; moreover, bacterial products from the protein expression process may also elicit responses by cells other than M. tuberculosis–specific T cells. Previous studies utilizing synthetic peptides to screen M. tuberculosis–specific CD4 T cell responses have used HLA binding prediction algorithms to synthesize predicted epitopes (19, 20), although this approach primarily detects epitopes that bind to multiple HLA class II alleles. To facilitate use of the assay in diverse international settings and cohorts for which HLA types may not already be well defined, we synthesized 18-mer overlapping peptides spanning the length of each Ag to increase the likelihood of detecting any M. tuberculosis–specific T cell response within a given Ag. Large peptide libraries consisting of 15-mer peptides overlapping by 9 aa have been used successfully to comprehensively define Ags and epitopes targeted by human M. tuberculosis–specific CD8 T cells (27–29).

A critical component of any assay for measurement of Ag-specific T cell responses is intra- and interassay variability. Testing our panel of 60 M. tuberculosis Ags and controls in a 96-well plate format necessitated that each M. tuberculosis peptide pool be tested in a single well. During the optimization phase, we measured IFN-γ concentrations for a subset of 20 M. tuberculosis peptide pools tested in triplicate wells, which indicated an intra-assay CV of 29%. Using the same subset of 20 M. tuberculosis peptide pools tested in triplicate wells for intra-assay variability, we also conducted repeat assays at three separate time points, spaced at weekly intervals, which also indicated a high level of reproducibility and low interassay variability. Previous studies screening T cell responses to broad panels of M. tuberculosis Ags have compared groups of individuals with different states of M. tuberculosis infection in a cross-sectional manner (9, 10, 12, 15, 16, 19), and thus intra-assay variability and assay reproducibility had not been previously well established.

In contrast with commercially available IFNγ release assays, which contain a pool of CFP-10 and ESAT-6 peptides (30), or the megapool approach, which pools hundreds of peptides from dozens of different mycobacterial Ags in a single peptide pool (31), the T cell RSA provides information on responses to individual M. tuberculosis Ags, which can be tracked and monitored on a per Ag basis over time. We therefore anticipate that the T cell RSA will be advantageous as a tool to test new hypotheses, including the hypothesis that T cell responses to individual M. tuberculosis Ags vary in distinct and discordant ways that may depend on the clinical or microbiological state of infection. In support of this hypothesis, a recent study reported that CD4 T cells targeting ESAT-6 and Ag85B have distinct phenotypic and functional profiles, and differential capacity to mediate protection against M. tuberculosis (32). Moreover, evaluating T cell responses to M. tuberculosis Ags associated with different phases of infection, such as latency-associated Ags or resuscitation-promoting factor-like proteins, may enable differentiation of individuals across a range of M. tuberculosis infection states (12–14), and may facilitate identification of individual M. tuberculosis Ags that have prognostic value in predicting progression to TB disease in prospective longitudinal cohort studies.

Using cohorts of M. tuberculosis–unexposed United States adults, foreign-born refugees with LTBI who were resettled in the United States, and a population of Kenyan adults with LTBI who are household contacts of an active TB case, we have applied this T cell RSA to begin to identify differences in the pattern of Ag recognition by M. tuberculosis–specific T cell responses across these diverse cohorts. Our preliminary findings indicate that differences in the breadth and magnitude of M. tuberculosis–specific T cell responses are evident in populations of individuals with LTBI from diverse geographical regions. Our preliminary findings are consistent with previous studies screening T cell responses to multiple M. tuberculosis Ags in geographically diverse populations worldwide (10, 16, 19). Differences in the spectrum of T cell Ag recognition in M. tuberculosis infection are likely driven by multiple factors, including host genetics, environmental mycobacteria exposure, microbiome composition, nutritional status, and coinfection with other pathogens. Further studies incorporating HLA typing and TCR sequencing will be important for defining host factors influencing the patterns of M. tuberculosis T cell Ag recognition and immunodominance in human M. tuberculosis infection (33, 34).

Several limitations of our diluted whole blood T cell RSA should be considered. First, by utilizing 50 μl of whole blood per stimulation condition, it is possible that the frequency of a given Ag-specific T cell population may be below the threshold of detection using this small volume of blood. However, the stimulation period of 7 d allows for proliferation of Ag-specific T cells, thus increasing the likelihood that sufficient IFN-γ will be secreted during the course of the assay to be detectable above background in the ELISA. Second, as with previous M. tuberculosis Ag screening assays, the functional readout is IFN-γ secretion, thus M. tuberculosis–specific T cells that produce cytokines other than IFN-γ will not be detected. Sufficient supernatants are harvested from each well to store for future analyses using multiplex bead-based arrays to detect a much broader panel of cytokines and chemokines secreted by Ag-specific T cell populations. Third, the ability to precisely quantify IFN-γ concentrations in supernatants from individuals with very strong M. tuberculosis–specific T cell responses is blunted by the dynamic range of the standard curve in the ELISA. The IFN-γ ELISAs in this study were conducted using 2-fold diluted RSA supernatants to perform a qualitative evaluation of IFN-γ production. However, individuals with high M. tuberculosis–specific T cell IFN-γ production (>1000 pg/ml) will require the ELISA to be repeated at higher dilutions of supernatants in order for these IFN-γ concentrations to be within the dynamic range of the standard curve for accurate and precise quantification. Additionally, there is a high degree of homology between M. tuberculosis and other mycobacterial species, including M. bovis bacillus Calmette-Guérin (BCG) vaccine strains, and a high level of sequence conservation in T cell epitopes in M. tuberculosis and BCG (21, 35). Given that 57 of the 60 M. tuberculosis Ags we tested are expressed by both M. tuberculosis and BCG, and that universal childhood BCG vaccination is recommended in over 150 countries worldwide (36), it is possible that this assay will detect T cell responses primed by BCG vaccination, as well as M. tuberculosis infection. However, previous studies evaluating mycobacteria-specific T cell responses indicate significantly higher frequency and magnitude of Ag-specific T cell responses in individuals with LTBI, compared with M. tuberculosis–uninfected individuals who have been vaccinated with BCG (19), thus suggesting T cell responses primed by BCG do not predominate over Ag-specific T cell responses detectable in the setting of natural infection with M. tuberculosis.

In summary, we describe here the development of a standardized diluted whole blood–based assay platform for evaluation of multiple Ag-specific T cell responses, with conditions optimized to incorporate additional Ags, and the ability to be conducted in resource-limited settings, using small blood volumes. The intra-assay variation is lower than the interassay variation, thus allowing conclusions to be drawn in terms of biological variability that may reveal novel information on the breadth and spectrum of M. tuberculosis Ag recognition that is associated with protection from progression to TB disease. We anticipate that this assay will be applicable in multiple settings, including cross-sectional studies across multiple study sites and longitudinal cohort studies to monitor Ag-specific T cell responses associated with M. tuberculosis infection outcomes, treatment interventions, vaccination, and possibly assessment of M. tuberculosis exposure and/or reinfection.

T. Adekambi (Emory Vaccine Center, Emory University School of Medicine), S.C. Alcantara (Emory Vaccine Center, Emory University School of Medicine), R. Ahmed (Emory Vaccine Center, Emory University School of Medicine), L. Sharling (Division of Infectious Diseases, Emory University School of Medicine), L. Elon (Department of Biostatistics, Rollins School of Public Health, Emory University), S. Jabbarzadeh (Department of Biostatistics, Rollins School of Public Health, Emory University), A. Rao (Department of Epidemiology, Rollins School of Public Health, Emory University), R.S. Goldstein (Department of Epidemiology, Rollins School of Public Health, Emory University), J. Gharbin (DeKalb County Board of Health), F. Jameel (DeKalb County Board of Health), S. Shah (Global Tuberculosis Branch, Center for Global Health, U.S. Centers for Disease Control and Prevention), A. Sette (Division of Vaccine Discovery, La Jolla Institute for Allergy and Immunology), and H. Siefers (Aeras).

J.A. Agaya (Center for Global Health Research, KEMRI), D. Awilly (Center for Global Health Research, KEMRI), F.H. Odhiambo (Center for Global Health Research, KEMRI), and L.O. Okayo (Center for Global Health Research, KEMRI).

We thank many additional members of the Kisumu-based KEMRI team who helped with enrollment and evaluation of participants, along with the participants themselves.

The authors have no financial conflicts of interest.