Comparative immune responses to Mycobacterium tuberculosis in people with latent infection or sterilizing protection

Summary There is great need for vaccines against tuberculosis (TB) more efficacious than the licensed BCG. Our goal was to identify new vaccine benchmarks by identifying immune responses that distinguish individuals able to eradicate the infection (TB-resisters) from individuals with latent infection (LTBI-participants). TB-resisters had higher frequencies of circulating CD8+ glucose monomycolate (GMM)+ Granzyme-B+ T cells than LTBI-participants and higher proportions of polyfunctional conventional and nonconventional T cells expressing Granzyme-B and/or PD-1 after ex vivo M. tuberculosis stimulation of blood mononuclear cells. LTBI-participants had higher expression of activation markers and cytokines, including IL10, and IFNγ. An exploratory analysis of BCG-recipients with minimal exposure to TB showed absence of CD8+GMM+Granzyme-B+ T cells, lower or equal proportions of Granzyme-B+PD-1+ polyfunctional T cells than TB-resisters and higher or equal than LTBI-participants. In conclusion, high Granzyme-B+PD-1+ T cell responses to M. tuberculosis and, possibly, of CD8+GMM+Granzyme-B+ T cells may be desirable for new TB vaccines.

Highlights TB-resisters eradicate TB infection providing immunologic goals for new TB vaccines TB-resisters have higher CD8+GMM+GzB+ T cells than people with latent TB (LTB) TB-resistance is associated with higher GranzB+PD1+ polyfunctional T cells than LTB BCG generates less CD8+GMM+GranzB+ or polyfunctional T cells than TB-resistance INTRODUCTION Tuberculosis (TB) is a major global health problem. There are an estimated 2 billion people infected with M. tuberculosis (Mtb) and 1.5 million annual deaths. Vaccines are the most powerful tools for limiting the morbidity and mortality of many infectious diseases, but Bacillus Calmette-Gué rin (BCG)-the only licensed TB vaccine-confers limited protection. 1-7 BCG administered to infants at birth is 80% effective in preventing disseminated and central nervous system TB. In contrast, BCG has variable efficacy against pulmonary TB when administered to infants or adults, and some studies could not demonstrate any efficacy. [1][2][3][4][5][6][7][8] In addition, recent studies showed that BCG revaccination did not prevent acquisition of latent TB infection (LTBI) measured by IFNg release assays (IGRA) conversion in adolescents despite boosting immune responses, albeit more BCG than placebo recipients reverted IGRA 3 to 6 months after conversion 9-11 BCG generates TB-specific Th1 cell-mediated immunity (CMI) that has been considered necessary and sufficient for protection against TB, although this concept has been disputed by recent studies. [12][13][14] Moreover, recent vaccine candidates that generate more robust Th1 CMI than BCG against the Mtb antigen 85A, did not have greater efficacy against pulmonary TB than BCG. 12,[15][16][17][18] Additional evidence that Mtb Th1 responses may not predict protection against Mtb infection is that 10 to 20% of the individuals with latent TB infection, as defined by the presence of Mtb CMI measured by tuberculin skin test (TST) or IGRA, a measure of Th1 immunity, develop active TB disease over time. In contrast, 20% of individuals with household contacts with highly contagious active pulmonary TB infection do not develop clinical disease and have negative TST or IGRA. 19 These individuals, designated as TB-resisters, have antibodies and limited adaptive CMI responses to Mtb, confirming prior Mtb infection, but their T cells do not produce IFNg in response to Mtb ex vivo restimulation. 20,21 They also have genetic traits that differentiate them from people with LTBI. [22][23][24][25] TB-resisters are deemed to have sterilizing immunity against Mtb. 26,27 However, the mechanisms of protection against Mtb infection in TB-resisters are incompletely understood. 21,28 Identifying these mechanisms would benefit the development of new vaccines that may cf. sterilizing immune protection against Mtb. 29 In addition to adaptive CMI, innate immunity is a critical mechanism of protection against infections. Innate immunity is rapidly deployed and constitutes the first line of immune-mediated defense. Moreover, innate immune responses critically contribute to the development of antigen-specific adaptive CMI and establish a feedback mechanism that allows them to be also boosted by adaptive CMI. [30][31][32][33][34][35][36] Persistent memory-like innate immune responses have been described against viruses, tumors, and other antigens, including Mtb. 32,[36][37][38][39][40][41][42][43][44][45][46][47] Natural killer (NK) cells, gd T cells, NKT cells, invariant NKT (iNKT) cells, macrophages, monocytes (mono), and dendritic cells (DC) have been shown to undergo clonal expansions and/or epigenetic modifications after exposure to immunogens that can cf. antigen specificity and/or allow them to activate transcription programs that improve their functionality. 47,48 Recent studies described a germline encoded, mycolyl-reactive iNKT cell subset (GEMT) that recognizes the lipid glucose monomycolate in the context of CD1b and participates in the elimination of Mtb from the host. 44,46,49 Compared to other iNKT, CD1brestricted GEMT cells have a more diverse T cell receptor repertoire, which permitted the identification of GEMT cell clonal expansions in patients recovering from TB and their persistence in the host for several years after Mtb elimination. 44,46 Another nonconventional T cell subset of interest is the mucosal-associated invariant T cells (MAIT) that developmentally share some characteristics both with iNKT and gd T cells. MAIT can recognize microbial riboflavin derivatives presented by the major histocompatibilityrelated receptor 1 (MR1) and play a critical role in antibacterial, including Mtb, pulmonary defenses via cytokine production and cytotoxicity. 50-52 MR1+ MAIT were shown to account for most CD8 + T cell production of IFNg in response to BCG. 53 MR1-MAIT lack the receptor for antigen recognition and are deemed to respond to secreted cytokines. 54,55 The role of innate immune responses in sterilizing immunity against Mtb infection has not been studied. However, in vitro, trained immunity controls mycobacterial outgrowth. 56 It is important to note that vaccines, including BCG, adjuvanted vaccines, and vectored vaccines, have been recently shown to generate trained immunity. [57][58][59][60] Thus, information on the association of trained immunity with sterile protection against Mtb infection may be used in the development of new and improved TB vaccines.
The overarching goal of our study was to identify innate and adaptive immune responses that differentiate TB-resisters from people with LTBI (LTBI-participants) among household contacts of active pulmonary TB cases that could be targeted by new TB vaccines. We leveraged a longitudinal cohort study of household contacts of individuals with active pulmonary TB disease [Cohort for TB research by The Indo-US Medical Partnership Multicentric Prospective Observational Study (C-TRIUMPH)] in Pune, India, by using advanced spectral flow cytometry technology and high-dimensional analytic tools on peripheral blood mononuclear cells (PBMC) archived from the study participants. 27,61, 62 To further understand how the responses to Mtb in LTBI-participants and TB-resisters may differ from those induced by BCG, we also included in our analysis a group with documented BCG vaccination and minimal exposure to TB.

Characteristics of the study population
The study used PBMC from 13 TB-resisters and 11 LTBI-participants in C-TRIUMPH, who had household contact with adults recently diagnosed with active pulmonary TB (index TB case) and lived in an area of high TB endemicity. TB-resisters were defined by TST and IGRA negative results at entry and during the 2-year C-TRIUMPH follow up. LTBI-participants had IGRA and/or TST positive results at entry in C-TRIUMPH. There were no appreciable differences in the baseline demographic characteristics and TB exposure scores between TB-resisters and LTBI-participants (Table 1). TB exposure scores have been designed to predict development of active and/or latent TB infection in household contacts of TB index cases. In C-TRIUMPH, the exposure score was comprised of 11 items, including presence of cough, pulmonary TB, smear positivity of the index case, if the index case was the household contact's primary caregiver or mother, sleep location of the household contact, and whether the household contact lived in the same house as the index TB case. High exposure was defined as a score R6 for adults and R5 for children. Although TB exposure scores were found to be associated with TB-resister versus LTBI status primarily in people 5-15 years of age, 63 we collected the exposure scores of all the participants in this study as an additional measure of uniformity between groups. Also recruited were 14 BCG-recipients with limited exposure to TB by virtue of having spent most of their lives in the United States or other countries with low incidence of TB. In addition to these differences in upbringing and environmental exposures, we noted that BCG-recipients significantly differed from C-TRIUMPH participants in race/ethnicity, but not in age or BMI.  (Table S1). In the T cell panel, we characterized CD4 + and CD8 + conventional T cells (Tconv), NKT, gd T, GEMT, iNKT, MR1+ and MR1-MAIT, and NK cells. Functionality was assessed by expression of CD25 and CD69-activation markers; PD1 immunologic checkpoint receptor; CD107a (lysosome degranulation) and granzyme B (GranzB) cytotoxicity markers; GMCSF, IL2, IL10, IL17, IFNg, and TNFa cytokines; and Ki67 proliferation marker. In the APC panel, we characterized functionality of CD14 + monocytes (Mono), CD123+ plasmacytoid DC (pDC), CD141+ cDC1, and CD1c+ cDC2 subsets using the following functional readouts: CD40, CD80, and CD83 activation markers; PDL1 immunologic checkpoint ligand; and GMCSF, IL1b, IL8, IL10, IL12p40, IL27, and TNFa cytokines. Gating trees are shown in Figures S1-S12.
Mtb-memory responses were characterized by subtracting frequencies of unstimulated from Mtb-stimulated corresponding PBMC. Using this definition, we identified 6 differentially expressed functional subsets with significantly higher frequencies in LTBI-participants compared with TB-resisters: CD25 + and CD69 + gd T cells; CD4 + CD25 + Tconv; CD8+IFNg+ iNKT; and CD8 + CD107a+ NKT ( Figure 3, and Table S4). There were no Mtb-memory subsets with higher frequency in TB-resisters than in LTBI-participants. Of note, cytokine-producing conventional T cell subsets, including CD4 + T cells expressing GMCSF, IFNg, IL2, IL10, IL17, and TNFa and CD8 + T cells expressing IFNg, IL2, IL10, and TNFa were excluded from the analyses based on the pre-specified criteria of a difference <0.1% in R33% of participants in each group (see statistical methods).

Polyfunctionality of the immune cell subsets in TB-resisters and LTBI-participants
Tconv polyfunctionality has been previously shown to predict protection against TB and other infections. [64][65][66][67][68] Here, we expanded the investigation of polyfunctionality to nonconventional T cells and NK subsets. Polyfunctionality was analyzed using Boolean gating of the six markers most abundantly expressed by NK and T cells (CD25, CD107a, GranzB, IFNg, IL10, and PD1) in Mtb-stimulated conditions. Higher marker co-expression was observed in TB-resisters than in LTBI-participants on CD4 + and CD8 + Tconv and also on

Coordination of the immune responses in TB-resisters and LTBI-participants
We complemented the analysis of manually gated PBMC subsets with Phenograph unbiased analysis, which identified 8 T cell and 1 APC clusters that significantly differed between TB-resisters and LTBI-participants in unstimulated PBMC; 3 T cell, and 1 APC clusters in Mtb-stimulated conditions; and 5 T cell and 3 APC Mtbmemory clusters ( Figure S15). Most clusters confirmed findings already observed in the manually gated analysis, with the following exceptions: (1) Two T cell and one APC Mtb-memory clusters ( Figure S15) with higher iScience Article relative frequencies in TB-resisters than LTBI-participants, in contrast to the manual gating analysis that did not identify excess Mtb-memory subsets in TB-resisters compared with LTBI-participants; and (2) A CD8+glucose monomycolate (GMM)+GranzB+ Tconv subset that had not been included in the prespecified manual gating analysis ( Figures 5, and S15). This subset had higher frequency in unstimulated PBMC from TB-resisters compared with LTBI-participants (FDR-corrected p = 0.008).
To focus the comparisons of BCG-recipients with TB-resisters and LTBI-participants, we analyzed the polyfunctionality of Mtb-stimulated T cell subsets with significantly different polyfunctionality in TB-resisters compared with LTBI-participants, including CD4 + and CD8 + Tconv, CD8 + iNKT, CD8 + NKT, and iScience Article MR1-MAIT cells ( Figure 6). The results showed significantly higher polyfunctionality in TB-resisters compared with BCG-recipients for all subsets except for CD8 + iNKT, which did not differ between the two groups. Notably, polyfunctionality of CD8 + iNKT cells was significantly higher in BCG-recipients than in LTBI-participants. Conversely, polyfunctionality of CD4 + Tconv was significantly higher in LTBI-participants than in BCG-recipients. CD8 + Tconv, CD8 + NKT, and MR1-MAIT subsets showed similar polyfunctionality in BCG-recipients and LTBI-participants. These data indicate that BCG and, possibly, other vaccines may generate higher polyfunctional responses than LTBI and reach levels observed in TB-resisters.
One of the two recent descriptions of GMM+ Tconv showed their presence in response to BCG administration in infants, but not in adults. 69 On comparing the frequency of CD8+GMM+GranzB+ T cells in BCGrecipients with TB-resisters and LTBI-participants in our study, we also found very small frequencies of this cell subset (%0.005% out of total lymphocytes) in BCG-recipients, significantly lower compared with either of the Mtb-exposed groups (FDR-corrected p % 0.00013; Figure S22).

DISCUSSION
The goal of this study was to characterize responses unique to TB-resisters compared to LTBI-participants that may be targeted by new TB vaccines. A consistent finding was increased GranzB-expressing conventional and nonconventional T cells in Mtb-stimulated and unstimulated PBMC of TB-resisters. This hallmark was observed in the prespecified manual gating and in the unsupervized cluster analyses and included expression of GranzB by Tconv, NKT, and iNKT cells. The analysis of Mtb-memory responses revealed higher cytotoxic responses in TB-resisters using the unbiased cluster analysis but not the manual gating. Collectively, these findings suggest that TB-resisters may mount higher and/or faster cytotoxic responses upon exposure to Mtb than LTBI-participants, which may contribute to the clearance of the infectious agent before establishing latency.
Of special interest was a population of CD8+GMM+GranzB+ Tconv that was not a part of the prespecified manual gating but was identified by the Phenograph cluster analysis. This cell subset, which has only iScience Article recently been described, was found in people with active TB, in whom CD8+GMM+ T cells displayed upregulated cytotoxic transcription profiles. 70 The authors concluded that this cell subset may contribute to the clearance of Mtb-infected host cells. Notably, we found higher proportions of CD8+GMM+GranzB+ T cells in TB-resisters compared with LTBI-participants, suggesting that this cell subset may contribute to the sterilizing immune protection of TB-resisters against Mtb infection.
TB-resisters also had higher proportions of PD1-expressing Tconv, iNKT, and MR1-MAIT than LTBI-participants after ex vivo Mtb stimulation. PD1 is an immunologic checkpoint inhibitor commonly expressed on activated T cells during acute or chronic infection. 71 PD1 expression contributes to quenching the immune response in tumors and may play a similar role in infections after removal of the stimulating agent, limiting the inflammatory response and tissue destruction. Recent reports showed an association of increased T cell activation with TB morbidity and with increased risk of active Mtb infection in BCG-recipients. [72][73][74] Moreover, PD1 deficient mice as well as people treated with PD1/PDL1-blocking agents have increased risk of Mtb infection, morbidity, and reactivation. [75][76][77] Collectively, these observations suggest that the increased expression of PD1 in TB-resisters may limit their susceptibility to Mtb tissue destruction and, perhaps, propagation of infection.
A distinguishing characteristic of the immune response in TB-resisters from LTBI-participants was polyfunctionality. TB-resisters had higher proportions of polyfunctional Tconv and nonconventional T cells in Mtbstimulated conditions. Polyfunctional CD4 + Tconv, characterized by expression of IL2, IFNg, and/or TNFa have been previously associated with protection against active TB infection in animal models. 64,65 Although the role of polyfunctional CD4 + T cell responses in protection against Mtb infection in humans remains iScience Article uncertain after a recent study showed that IL2, IFNg, and/or TNFa polyfunctional CD4 + T cell responses generated by BCG did not correlate with protection against TB disease in vaccinated infants, 14 there is evidence from other human infections supporting the protective role of polyfunctional T cells. [66][67][68]78,79 In our study, cytokine production made a minor contribution to the polyfunctional T cell responses, which predominantly expressed GranzB and PD1, and involved both conventional and nonconventional T cells. Polyfunctionality of nonconventional T cells has not been previously studied. Here, we showed increased polyfunctionality of NKT, iNKT, and MR1-MAIT cells in TB-resisters compared to LTBI-participants, which may contribute to enhanced protection against TB infection in TB-resisters through cytotoxicity.
Mtb-memory T cell responses identified by the expression of single activation markers were higher in LTBIparticipants than in TB-resisters, although two Mtb-memory CD4 + Tconv clusters revealed by the Phenograph analysis had higher frequencies in TB-resisters than in LTBI-participants. These data confirm that TB-resisters develop Mtb-specific T cell memory, but in general Mtb-memory responses are higher in LTBI-participants. This finding may be related to the observation that 25% of people with LTBI have evidence of active Mtb replication, which continuously exposes the immune system to Mtb antigens and maintains high levels of Mtb-specific T cells. [80][81][82][83] APC displayed few differentiating features between TB-resisters and LTBI-participants. However, the Phenograph analysis identified an activated Mono cluster in response to Mtb stimulation with higher frequency in TB-resisters than LTBI-participants. Additional studies are needed to further characterize the potential role of APC-trained immunity in protection against TB.
The analysis of Mtb-memory responses in BCG-recipients using single markers of activation showed both higher and lower responses in multiple T cell and APC subsets compared with TB-resisters or LTBI-participants. These differences may have been related to the difference in environmental exposures, including Mtb, in addition to genetic backgrounds of BCG-recipients compared with TB-resisters or LTBI-participants. The Boolean analysis of polyfunctional responses helped focus the differences between BCG-recipients and LTBI-participants or TBresisters. Overall, BCG-recipients had lower frequencies of Mtb-stimulated polyfunctional T cells than TB-resisters, but similar to LTBI-participants. Notably, CD8 + iNKT cells had higher polyfunctionality in BCG-recipients compared to LTBI-participants clearly demonstrating the ability of the vaccine to generate polyfunctional responses higher than the infection. The corollary of this observation is that new TB vaccines may generate immune responses that protect against TB disease better than the immune responses of people with LTBI.
We conclude that both innate and adaptive responses distinguish TB-resisters from LTBI-participants. Unique elements of the cell-mediated immune responses to Mtb in TB-resisters are high-cytotoxic potential, expression of immunologic checkpoint inhibitors, and polyfunctionality of Tconv and nonconventional T cells, suggesting that some or all may be key factors in an effective immune response against Mtb.
Although TB-resisters displayed these characteristics even without ex vivo Mtb stimulation in many cases, it is reasonable to propose that a vaccine that increases some or all these responses in the form of Mtbmemory may be highly effective in protection against TB.

Limitations of the study
Our study has several limitations, including the relatively small number of participants, which may have restricted our ability to detect differences with small effect sizes and increased the risk of outliers to bias the results. Cytokine-producing conventional T cell subsets, including CD4 + T cells expressing GMCSF, IFNg, IL2, IL10, IL17, and TNFa and CD8 + T cells expressing IFNg, IL2, IL10, and TNFa were excluded from the analyses of Mtb memory responses because the low variability of the results and the tendency to recapitulate results in unstimulated conditions. This may have decreased our ability to identify Mtb memory responses in TB-resisters. In addition, BCG-recipients had different demographic characteristics compared with C-TRIUMPH participants (e.g., race/ethnicity). Although our study did not measure antibody responses, which have been recently shown to differentiate TB-resisters from LTBI-participants, 21,84 the CMI analysis was comprehensive and revealed findings unique in TB-resisters.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

DECLARATION OF INTERESTS
The authors declare that they have no competing interests related to this manuscript.

Boolean analysis
Boolean combinatorial gates were created using the 6 cytokines/activation markers, generating 64 distinct activation phenotypes. Graphical representation was performed using SPICE 6.0 software. 85 The data were analyzed using permutations (10,000 iterations) included in the software.

UMAP analysis
For T cell clustering, we used 67,000 down-sampled events from 13 TB-resisters and 9 LTBI-participants for a total of 2,948,000 events in the concatenated file. For APC clustering, we used data from 11 TB-resisters and 8 LTBI-participants for a total of 502,479 events in the concatenated file. Parameters were rescaled to ArcSinh prior to running UMAP with phenograph in FlowJo on the concatenated file using parameters, Euclidean distance, 15  To compare the average cell lineage frequencies obtained from manual and data-driven gating between groups, we used beta regression (betareg R package, v3.1-3) with a logit-link to model cell lineage percent (regression outcome; between 0 and 100% of events) on group (primary explanatory variable of interest; LTBI-participant versus TB-resisters), adjusting for participant age in years as a covariate. 87 Similarly, to test for functional differences among unstimulated conditions, we modeled percent positivity (outcome; between 0 and 100% of cells in given lineage expressing functional marker) on group, correcting for age. ''Rare'' outcomes (i.e., outcomes of 0% in over 1/3 rd of participants) were excluded from testing due to their low variability among our participants. For each set of statistical tests, we defined significant between-group differences as a non-zero group regression coefficient (Wald test, multiple testing adjustment using the Benjamini-Hochberg false discovery rate < 0.05). 88 We confirmed the statistical approach following exploratory data analyses (e.g., raw marker frequencies by group) and regression goodness-of-fit metrics (e.g., residual plots). We report effect sizes in terms of odds ratios (OR, G 95% confidence intervals) and the median G interquartile range for percent positivity.
To examine differences in functional markers under Mtb-stimulated conditions, we repeated this modeling procedure but with the stimulated percent positivity values as the regression outcome variable, adjusting for age. We also examined differences in Mtb-memory by additionally adjusting for each participant's corresponding baseline values (unstimulated percent positivity values in their paired control sample) as a covariate. The Mtb-memory analysis thus emphasizes markers that were not already differentially abundant between groups prior to stimulation, or stimulation effects potentially masked by unstimulated differences.
Here, adjusting for baseline levels is conceptually similar to ''subtracting'' the unstimulated values from the Mtb-stimulated values; as a consequence, we visualize Mtb-memory with D-values, defined by stimulated frequencies minus unstimulated frequencies. Functional markers with limited changes following Mtb-stimulation (i.e., D-values of magnitude % 0.1% in over 1/3 rd of participants) were excluded from testing due to their low variability among our participants and tendency to recapitulate unstimulated tests. Significant between-group differences in Mtb-memory were based on FDR < 0.05.
These modeling procedures was also repeated for the TB-resister versus BCG recipient comparisons and the LTBI-participants versus BCG recipient comparisons. All analyses were performed in the R statistical package v4.0.2 unless otherwise specified. Additional methodical details and analysis code is publicly available at https://github.com/chooliu/Sterile_vs_Latent_TB_Immunity.