Finding and filling the knowledge gaps in mechanisms of T cell-mediated TB immunity to inform vaccine design

Abstract

Mycobacterium tuberculosis, the bacterium that causes tuberculosis (TB), results in more human mortality than any other single pathogen, in part because of the lack of an effective vaccine. Although T cells are essential for immunity to TB, the mechanisms that provide protective immunity are poorly understood. In this Review, we describe current gaps in our knowledge about T cell-mediated immune responses to M. tuberculosis and discuss how recent technologies, including multiphoton intravital microscopy, spatial multiomics and high-resolution in vivo analyses of cell–cell interactions, may be used to gain insights that can inform the design of T cell-targeted TB vaccines.

Introduction

In 2024, Mycobacterium tuberculosis (Mtb), the bacterium that causes tuberculosis (TB), led to 10.8 million cases of TB and 1.25 million deaths (https://www.who.int/health-topics/tuberculosis). Unlike smallpox, which was successfully eradicated using vaccines, TB remains a major cause of human morbidity and mortality due to lack of an effective vaccine.

It is well established that T cells are required for immunity to Mtb, but the mechanisms of T cell priming and polarization, as well as the T cell antigen targets, effector functions and trafficking properties that provide optimal immunity against Mtb, are not yet defined. This is partially due to unique characteristics of Mtb, a genome that encodes ~4,000 proteins that are potential antigens. So far, 549 of these potential antigens, which contain T cell 2,293 epitopes (iedb.org, accessed 9 February 2025) have been reported to be recognized by human T cells. Moreover, a long coevolution with humans has enabled Mtb to develop numerous mechanisms to evade and exploit T cell responses, and there are obstacles that prevent direct interactions of T cells with infected cells. The long evolutionary relationship between humans and Mtb suggests that not all mechanisms of immunity and immune subversion are shared with other host species. Nevertheless, mechanisms that are conserved across species have established the value of model systems, including zebrafish and mice, in understanding host–mycobacterial interactions. Finally, the need for high level biocontainment for research on Mtb adds to the challenges of researching fundamental mechanisms of pathogenesis and immunity.

The landmark discovery that vaccine strategies based on intravenous¹ or intrabronchial² Mycobacterium bovis Bacillus Calmette–Guérin (BCG) or the administration of a rhesus cytomegalovirus (rhCMV) vectored subunit vaccine³ can elicit sterilizing immunity to Mtb in non-human primates (NHPs) has renewed hope that an effective vaccine can be developed for human use. There are currently eight phase 2b/3 clinical trials of TB vaccines. Of the vaccines in these trials, six are inactivated or attenuated whole bacterial vaccines and two comprise adjuvanted proteins (https://newtbvaccines.org/tb-vaccine-pipeline/). New technical approaches, including barcoded Mtb strains, single-cell RNA sequencing, spatial multiomics and multiplex confocal imaging, have revolutionized our ability to investigate TB in NHP and humans. Moreover, genetic engineering of mice enables in depth mechanistic studies using new technologies, including Mtb antigen-specific T cell receptor (TCR) transgenes, fluorescent protein reporters for marking specific cell types, activation states and lineage tracing (Table 1). Recent reports of a new subunit TB vaccine^4,5, as well as a system that establishes stable low-level chronic infection in a lymph node (LN) after intradermal injection of Mtb in mice, permits study of the impact of prior exposure and sensitization to Mtb⁶ and demonstrates that enhanced immunological control of Mtb after aerosol challenge in mice is possible. Multiphoton intravital microscopy, spatial multiomics and high-resolution in vivo analyses of cell–cell interactions can be used to investigate these and other emerging experimental systems, such as strains of mice that vary in susceptibility to Mtb. Together, these studies provide unprecedented insights into the mechanisms of effective immunity to Mtb, which will inform the development of effective TB vaccines for humans.

Table 1 |.

Complementary roles and trade-offs of experimental models in advancing TB research

Parameter	Mouse models	Non-human primate models	Human models
Immune similarity to humans	Limited; known and unknown differences in immune responses to Mtb compared with humans	High; closely mimics immune responses in humans	Exact
Mtb granulomas	Variable by mouse strain and immunologic competence	Well-formed and human-like	Highly variable in cell and bacterial content and overall architecture
Cavitation (necrotic lesions that communicate with airways)	Varies by mouse strain	Sometimes develops	Common in active tuberculosis
Latency modelling	Poor	Moderate to good	Accurate
Cost	Low	High	Very high
Ethical concerns	Moderate	Extensive	Strict
Disease progression	Rapid	Slow, similar to humans	Natural history of Mtb infection and tuberculosis disease is highly variable
Relevance for vaccine development	Suitable for initial screening	Considered important for preclinical efficacy testing	Required for final validation
Relevance for new discoveries	High, permitting genetic and mechanistic studies	Good for insights into immunity and pathology	Essential for confirming human relevance
Mechanistic understanding	Excellent for host-pathogen studies	Strong, with direct translational insights	Limited; observational studies dominate
Duration	Short	Moderate to long, limited by ethical considerations	Very long
Summary
Advantages	Cost-effective, rapid and excellent for mechanistic and genetic studies and guiding translational studies	High relevance for pathology and immunity, bridging preclinical to human studies	Critical for validating discoveries
Disadvantages	Some findings have limited translational relevance	Costly and ethically complex	Costly, limited in experimental flexibility

Here, we refer to the bacteria as Mtb, the active, pathologic disease caused by Mtb as tuberculosis (TB), and the state of infection with Mtb that may be active or ‘latent’ (‘a state of persistent immune response to stimulation by M. tuberculosis antigens with no evidence of clinically manifest TB’) as Mtb infection (MTI) (https://tbdictionary.org/)^7,8. We discuss the specific features of immunity to Mtb that remain poorly understood and how new technologies can be applied to studies of immunity to TB to address these knowledge gaps.

Current understanding of T cell immunity to Mtb

Mtb is an intracellular pathogen of phagocytes. Upon aerosol infection of the lungs, it is engulfed by alveolar macrophages⁹, replicates intracellularly, disperses including through death of the infected alveolar macrophages and then spreads to other cells including recruited monocytes that differentiate into subsets of monocyte-derived macrophages^10–12. These cell types differ in their control of Mtb in vivo, depending on the stage of infection. Before the initiation of T cell responses, alveolar macrophages support bacillary proliferation, whereas interstitial macrophages initially limit it^10,12,13. Once Mtb-specific T cells are recruited to the lungs, they efficiently limit replication of and even promote the killing of Mtb in alveolar macrophages but have less impact on infected monocyte-derived macrophages^12–14.

Both CD4⁺ and CD8⁺ T cells are essential to the control of the infection^15–17. The mechanisms of T cell differentiation and their acquisition of effector functions in response to MTI are poorly understood. Mtb predominantly induces interferon-γ (IFNγ)-producing CD4⁺ T cells (T_H1 cells), although there is increasing evidence that interleukin (IL)-17-producing (T_H17 cells) and cytolytic T cells are also induced^18,19. In NHPs, CD4⁺ T cells provide immunoregulatory roles that contribute to the control of Mtb after reinfection of animals that were initially infected and then treated with antimycobacterial drugs²⁰. After T cell priming and differentiation, antigen-specific T cells traffic to the lungs^21,22, where they need to bind their cognate antigens to be activated and to perform effector functions (Fig. 1). In the lungs, T cells contribute to organized structures termed granulomas (‘Tissue-localized T cells and TB granulomas’ section).

Fig. 1 |. T cell responses to Mtb.

a, After initial infection of resident alveolar macrophages, recruited monocyte-derived cells — inflammatory monocytes or monocyte-derived dendritic cells (DCs) — are infected with Mycobacterium tuberculosis (Mtb) bacteria in the lungs and then migrate to the lung-draining lymph node (LNs) via the lymphatics, transporting live Mtb with them. These monocyte-derived cells are ineffective at priming T cells. However, Mtb antigens (but not live bacteria) are transferred to LN-resident DCs, which present antigen to and prime naive CD4⁺ and CD8⁺ antigen-specific T cells with greater efficacy than the infected migratory cells. The primed T cells become polarized towards helper, cytolytic or regulatory phenotypes in the LN but the sources of the polarizing cytokines (such as interleukin (IL)-6, IL-12, IL-23 and TGFβ) have not been identified in the context of Mtb infection. b, The primed effector T cells (including CD4⁺ T helper 1 (T_H1) cells, T_H17 cells and regulatory T cells, as well as CD8⁺ cytotoxic T lymphocytes) traffic to the lungs, where they can be activated by binding to their cognate antigens presented by Mtb-infected cells. TCR, T cell receptor.

In humans, the majority of individuals who are exposed to Mtb are able to control the infection without becoming ill with active TB. A number of risk factors are known that increase the likelihood of progression to active TB. These include human immunodeficiency virus coinfection and tumour necrosis factor (TNF) blockade, where the mechanisms by which they enable progression is known^15,23, and other risk factors, such as malnutrition and diabetes mellitus, where the mechanisms are unknown. Mechanistic studies in animal models and translational studies in humans will be valuable in understanding the immunological basis of these and other risk factors that are associated with a higher frequency of progression to TB. These studies may also reveal mechanisms of effective and ineffective immunity to TB and inform vaccine development efforts.

In NHPs, the levels of Mtb-specific CD4⁺ and CD8⁺ T cells in the airways after intravenous¹ or intrabronchial² vaccination with M. bovis BCG correlate with protection from MTI. Although vaccination with rhCMV also induces Mtb antigen-specific CD4⁺ and CD8⁺ T cells, their abundance did not correlate with protection from MTI in an initial study³. These results demonstrate that protection against primary MTI can be achieved by vaccination, although there are practical barriers to implementing these strategies in humans due to concerns about the safety of intravenous-administered BCG (this is based on phase I studies in patients with cancer when used as investigational immunotherapy^24,25) and because of unresolved questions relating to the high frequency of preexisting immunity to CMV in most populations. It warrants consideration whether these barriers may be overcome by modifying the agents themselves and/or by altering their rate or route of administration.

Tissue-localized T cells and TB granulomas

TB granulomas are organized tissue lesions that are formed when macrophages aggregate in response to mycobacterial infection. Intravital microscopy of zebrafish embryos infected with M. marinum has provided insights into the dynamics of early granuloma development^26,27, though later events are less well understood. As granulomas mature and recruit other immune cells including T cells, they diversify in their cellular composition, organization and ability to restrict or support MTI^28,29.

Granulomas exhibit spatial segregation, with Mtb-infected macrophages at the core and T cells at the periphery, as shown for lesions in mice, rhesus macaques, and humans^19,29–35 (Fig. 2). The dynamics of this phenomenon in space and time are poorly understood, as this has only been studied in fixed tissue sections. Nevertheless, data exist that T cells with specific traits might be localized in specific areas within the granuloma, given localization of tissue-resident CD4⁺ and CD8⁺ memory T cells can vary in response to local signals³⁶.

Fig. 2 |. Spatial segregation of T cells and myeloid cells in Mtb granulomas.

a, Top: a granuloma from a C57BL/6 mouse infected with Mycobacterium tuberculosis (Mtb) 56 days earlier. The macrophages and neutrophils are stained for CD11b. The figure was adapted with permission from J. Infect. Dis. 218, 10 (15 November 2018) (cover photograph; refers to ref. 35). Middle: a granuloma from Mtb-infected rhesus macaque (image courtesy of Daniel Barber, NIAID). The macrophages and neutrophils are stained for CD11b. Laminin (extracellular matrix) staining indicates the boundary of the granuloma. Bottom: a human tuberculosis (TB) granuloma. The macrophages and neutrophils are stained for CD68⁺. The Mtb bacilli are not visualized in this image owing to incompatibility of staining methods for Mtb and for antibody targets (image courtesy of Carl Feng, University of Sydney). Although granuloma morphology can vary, particularly in human samples (probably related to chronicity of the infection), granulomas in all three species show evidence of spatial separation and infrequent contact between T cells and Mtb-infected macrophages in the granuloma core. b, A model of TB granulomas that permit (top) or restrict (bottom) Mtb replication. In this model, T cells in permissive granulomas are excluded from the central core region and are limited to the periphery of the granuloma, permitting necrosis of cells in the central region and accumulation of Mtb in the core. In restrictive granulomas, T cells access central regions of the granuloma, where they contact Mtb-infected myeloid cells, execute their effector functions, control the growth of Mtb and prevent central necrosis. High endothelial venules (HEVs) enable the efficient entry of memory T cells into the lung tissue and the formation of tertiary lymphoid structures adjacent to granulomas. Studies are not yet sufficient to determine whether HEV formation determines the outcomes of individual granulomas, although tertiary lymphoid aggregates have been associated with immune control of Mtb. Multinucleated giant cells, which are thought to be formed by fusion of multiple macrophages, are poorly understood, and their presence or abundance is not known to be associated with specific granuloma outcomes.

The segregation of different cell types may impair T cell efficacy against Mtb, as direct CD4⁺ T cell contact with infected cells in the lung is critical for control of intracellular Mtb in mice³⁷. This contrasts with infections with other pathogens such as Leishmania, where activation of CD4⁺ T cells within 80 μm of an infected macrophage controls the infection³⁸. The spatial segregation of T cells away from Mtb-infected macrophages also limits the efficacy of cytolytic CD8⁺ T cells that must contact target cells to eliminate them. In tumours, enhancing CD4⁺ and CD8⁺ T cell access to cancer cells can improve immune responses^39,40, highlighting the significance of overcoming spatial barriers to boost T cell effectiveness in TB.

Despite the spatial segregation, there is evidence that CD8⁺ cells play a role in responses to MTI^41–44. A study of MTI in cynomolgus macaques revealed that at an early timepoint after infection (~6 weeks), the impact of an anti-CD8α antibody, which depletes CD8α-expressing lymphocytes, including conventional CD8αβ⁺ T cells, CD8α⁺ donor-unrestricted T cells (including γδ cells), CD8α⁺ natural killer cells and CD8α⁺ CD4 T cells, was greater than the impact of an anti-CD8β antibody that only depletes class I MHC-restricted peptide-specific CD8αβ⁺ T cells⁴². In addition, protection conferred by vaccination with intravenous BCG in macaques also relies on CD4⁺ and CD8αα⁺ T cells⁴⁵. Although this supports a role for certain CD8⁺ cells in the control of early infection, further investigation is required to define the contributions of CD8⁺ T cells in immune control during later stages of infection. These cells could potentially be targeted with vaccines that prevent the progression from MTI to TB. Understanding how CD4⁺ and CD8⁺ T cells recognize and eliminate infected cells in vivo is crucial and depends on knowledge of antigen specificity, migratory properties, effector functions and the immune evasion mechanisms of Mtb. CD4⁺ T cell recognition of Mtb-infected cells is suboptimal, as reflected by their inability to reliably eliminate the pathogen and by the low frequency of their antigen-dependent activation in vivo in mice. When compared with responses of CD4⁺ T cells, CD8⁺ T cell recognition of infected cells is even poorer when tested in mice and in cultured human cells^46,47. An example of this is the Mtb TB10.4 antigen, which elicits an immunodominant CD8⁺ T cell response in mice after MTI or vaccination, but TB10.4-specific T cells do not recognize Mtb-infected cells for reasons that are not yet well understood^48–51.

Mtb subversion of T cell responses

Mtb employs several different mechanisms to modulate T cell responses and evade elimination. These include interference with antigen presentation, inhibition of T cell signalling and the induction of regulatory responses (Table 2). For example, Mtb interferes with phagosome maturation in macrophages, thereby limiting the generation of peptide epitopes for MHC class II (MHC-II) presentation. Moreover, Mtb limits antigen presentation by blocking autophagy and apoptosis of infected cells and by ‘hijacking’ intracellular vesicular transport to divert secreted Mtb antigens away from the MHC-II presentation pathway. Mtb also downregulates MHC-I antigen presentation, and by inhibiting host cell death pathways, prevents the cross-presentation of Mtb antigens that would otherwise reach the extracellular space⁵². In addition, Mtb can directly affect T cells by disrupting TCR signalling and by activating inhibitory pathways to suppress immune responses^53–59 Whether TCR avidity and/or TCR signalling strength are affected by Mtb is currently unknown.

Table 2 |.

Mechanisms by which Mtb modulates T cell responses to promote persistent infection

Mtb activity	Host target	Mtb effector molecules	In vivo evidence	Refs.
Interactions of CD4+ T cells with APCs
Disrupts antigen processing	Phagosome maturation	Live (H37Ra) bacteria	No	224
Disrupts antigen processing and presentation	Endosomal sorting complex required for transport component HRS	EsxH	Yes	225,226
Decreases MHC-II, CD40 and CD86 expression by APCs and decreased production of IL-12 and IL-23. Decreased T cell activation and TH1 and TH17 differentiation	Signalling through TLRs and dendritic cell maturation	Hip1 serine protease	No	227
Decreases antigen processing	Phagosome acidification	PPE18	No	228
Decreases antigen presentation without decrease of surface MHC-II	Autophagy	PE_PGRS47	Yes	229
Decreases antigen presentation by inhibiting IFNγ-mediated induction of MHC-II expression	TLR2	Multiple Mtb lipoproteins	No	230–235
Decreases antigen presentation by inhibiting IFNγ-mediated MHC-II expression through induction of SOCS1 in host cells	Mincle (CLEC4E)	Trehalose-6,6-dimycolate	No	236
Facilitates export of Mtb antigens out of infected cells to divert them from MHC-II pathway and decrease antigen presentation	Kinesin 2-dependent vesicle trafficking	Live Mtb	Yes	237
Decreases antigen presentation without decrease in surface MHC-II expression	Antigen processing and/or trafficking	Esx1 locus dependent	Yes	238
Interactions of CD8+ T cells with APCs
Decreases trafficking of peptide-MHC class I complexes to cell surface by trapping β2-microglobulin in endoplasmic reticulum	β2-microglobulin	ESAT-6	No	239
Inhibits cross-presentation	Unknown	Viable bacteria	No	96
Inhibits apoptosis of Mtb-infected cells for antigen presentation by bystander cells	Superoxide anion	NuoG	Yes	240
Shifts cell death from apoptosis to necrosis to decrease cross-presentation	Plasma membrane repair and annexin 1 expression	Not determined	Yes	97,98,241
T cell intrinsic mechanisms
Attenuates TCR signalling in CD4+ T cells	Proximal TCR signal and decreased ZAP-70 phosphorylation	LAM, PIM	No	56
Attenuates TCR signalling in CD4+ T cells	Decreased ZAP-70, LCK and LAT phosphorylation	LAM	No	57
Decreases T cell activation in CD4+ T cells	Induction of GRAIL (gene related to anergy in lymphocytes)	LAM and Mtb membrane vesicles	No	53,59
Disrupts T cell-APC synapses in CD4+ T cells	Decrease LCK recruitment to synapses	LAM, whole Mtb infection of cultured cells	No	58
Decreases IFNγ secretion by both CD8+ and in CD4+ T cells (CD8 > CD4)	Decreases T-bet via block of MTORC1 activity	d-serine	Yes	54,55
T cell regulatory mechanisms
Decreases CD4+ and CD8+ T cell proliferation, skews CD4+ T cell differentiation to reduce TH17 cells and increase Treg cells	IDO	Not determined or induced by IFNγ	Yes	60–62,242
Decreases TH1 differentiation and IFNγ expression and promotes CD4 T cell apoptosis at site of Mtb infection	TGFβ	Not determined	Yes	125
Decreases accumulation of CD4+ T cells in regions of Mtb-infected cells in granulomas; decreases IL-2 production; promotes CD4+ T cell apoptosis	IL-27	Not determined	Yes	72
Restrains CD4+ T cell production of proinflammatory cytokines	PD-1	Not determined	Yes	109,243
Suppresses CD4+ and CD8+ T cell cytokine expression	TIM-3	Not determined	Yes	70

Host immunoregulatory mechanisms also affect the outcome of MTI. Indoleamine-2,3-dioxygenase (IDO), which is induced by IFNγ, is elevated in granulomas and affects T cells by depleting tryptophan and/or generating immunoregulatory kynurenine catabolites that inhibit T_H17 cells and promote regulatory T cell (T_reg) differentiation^60–62 T_reg cells, which are induced soon after infection but are later suppressed by IL-12 expressed by dendritic cells (DCs) and macrophages, can restrict effector T cell function²⁸. Recent single-cell RNA studies identified T_reg cells among CD4⁺ reg T cell populations, which potentially reduce inflammation during chronic infection⁶³. TGFβ, a negative regulatory cytokine that is produced by T_reg cells and other cells, was shown to inhibit T cell activation in lungs of Mtb-infected mice²⁹. Likewise, IL-10, which can be derived from T_reg and/or myeloid cells, also suppresses T cell responses. Although total genetic ablation of IL-10 had no effect on MTI in C57BL/6 mice, IL-10 expression is actively repressed by the transcription factor BHLHE40, and deletion of Bhlhe40 leads to enhanced IL-10 production and more severe TB^30–32. Finally, following infection, type I IFN is produced by plasmacytoid DCs and macrophages⁶⁴ and is associated with bacterial virulence and transition from latent to active TB in humans^64–66. The role of type I IFN during TB is complex, and it can have both beneficial and detrimental effects⁶⁷.

Coinhibitory receptors (checkpoint inhibitors) on T cells, such as PD-1, which are upregulated by TCR signalling, can reduce T cell proliferation and function and mediate T cell exhaustion. Although PD-1 blockade has shown promise in cancer by reversing exhaustion, in mouse models of MTI, it exacerbates disease by inducing excessive activation of CD4⁺ T cells⁶⁸. Similar results were observed in Mtb-infected macaques and in patients with cancer and preexisting MTI⁶⁹. By contrast, blocking of other checkpoint inhibitors on T cells, such as TIM-3, was shown to decrease lung bacterial loads and enhance T cell function in a mouse model of MTI⁷⁰. IL-27 secretion by activated macrophages and DC induces the transcription factor BLIMP1, which enhances IL-10 and TIM-3 expression in T cells⁷¹, and IL-27-knockout mice have an improved ability to contain Mtb but succumb prematurely from the disease due to increased inflammation^72–75. Although these data may discourage the use of checkpoint inhibitors in TB, they illustrate an important concept: TB pathogenesis is driven by immunopathology, given that tissue inflammation is out of proportion to the number of bacilli detected. Therefore, in the absence of antimicrobial therapy, the disruption of immunoregulatory responses may be detrimental to the host.

Knowledge gaps in T cell immunity

Do Mtb antigens presented in the lung correspond to those that prime T cells in the LN?

Thousands of Mtb-derived peptide epitopes that can be recognized by T cells have been identified in humans with prior or ongoing MTI. Mtb-specific T cells are probably primed by LN-resident DCs, after migratory DCs transport Mtb from the lungs to LNs, where LN-resident DCs take up exported Mtb antigens^4,22,76. Migratory DCs are poor primers of T cells, but the transfer of Mtb antigens to LN-resident DCs compensates for the limited capacity of Mtb-infected cells to activate T cells^11,76,77. Whether this mechanism also applies to non-secreted bacterial antigens and whether it permits optimal CD8⁺ T cell activation is unknown, and this is an issue that needs to be further investigated to achieve optimal vaccine design.

As discussed, optimal control of intracellular infection requires cognate recognition of Mtb-infected cells by CD4⁺ T cells in the lungs³⁷; contact is also essential for cytolytic T cell killing of Mtb-infected cells (Fig. 1). This requires that the Mtb antigens and epitopes presented to naive T cells by DCs in the LNs are the same as those presented by Mtb-infected lung macrophages. Differences in antigen expression and/or processing due to differential proteolytic activity^50,78 or MHC-II expression in DCs versus macrophages^79–81 may generate T cells that recognize epitopes that are not presented by Mtb-infected cells in the lung^50,78,82. Finally, immune evasion mechanisms may affect antigen presentation in Mtb-infected cells but not in uninfected cells that have acquired Mtb antigens. Mtb-infected cells in the lung include different types of DC (cDC1, cDC2 and monocyte-derived DCs), macrophages (alveolar macrophages, interstitial macrophages and monocyte-derived macrophages), monocytes and neutrophils. These differ markedly in their ability to restrict the proliferation of intracellular Mtb^{9,11–13,64,83}, and some of these differences can be attributed to differential recognition by T cells¹⁴.

The preferential acquisition of secreted Mtb antigens by uninfected DCs in LNs is likely to prime T cells that are biased in their antigen specificity towards antigens that are acquired and presented by uninfected DCs (Fig. 3). Activation of T cells by antigen-loaded uninfected cells might therefore promote inflammation without antimicrobial activity, potentially contributing to tissue damage and Mtb transmission.

Fig. 3 |. Interactions of CD4⁺ and CD8⁺ T cells with Mtb-infected myeloid cells in granulomas.

a, CD4⁺ T cells can interact with macrophages that are infected with low levels of Mycobacterium tuberculosis (Mtb), but this interaction may not affect intracellular bacterial replication. b, Progressive replication of intracellular Mtb enhances the recognition of infected cells by CD8⁺ and CD4⁺ T cells, probably because more antigen is produced and presented on MHC class II molecules, as well as MHC class I molecules (by cross-presentation). However, antigen presentation and T cell recognition can be countered by bacterial mechanisms that interfere with these processes (Table 2). c, CD4⁺ and CD8⁺ T cells can activate macrophage killing of Mtb through the secretion of tumour necrosis factor (TNF), interferon-γ (IFNγ) and possibly other cytokines that stimulate macrophage bactericidal activity and enhance macrophage survival. CD8⁺ T cells can also induce bacterial killing in macrophages via granulysin and other, less well characterized mechanisms. d, At later stages of infection of an individual myeloid cell, Mtb causes damage to the phagosome membrane, enabling the translocation of secreted Mtb antigens to the cytosol. Mtb-infected macrophages that evade recognition by CD4⁺ T cells can still be recognized and killed by cytolytic CD8⁺ T cells. e, Apoptotic macrophages release Mtb and Mtb antigens in apoptotic vesicles. The necrotic cells permit Mtb dispersal. Mtb released from dying cells can infect new cells, resulting in spread of the infection. f, Mtb in apoptotic cells can be engulfed (via efferocytosis) by macrophages and the bacterial antigens can be presented to CD4⁺ T cells and CD8⁺ T cells (by cross-presentation). g, Mtb-infected myeloid cells can also export Mtb antigens to the extracellular space by multiple mechanisms, including vesicular transport and exocytosis, release of exosomes and release from dying cells. The antigens can also be secreted by extracellular bacteria. h, Uninfected myeloid cells can take up (bacterial) cell-free antigens and process and present them to antigen-specific CD4⁺ and CD8⁺ T cells, resulting in their ‘distraction’ from recognizing Mtb-infected cells and in enhanced inflammation.

To develop robust vaccine strategies and understand Mtb immune evasion, deeper insight is required regarding the dynamics of T cell–APC interactions in MTI, both in LNs and lungs, particularly those involving different subtypes of DC⁸⁴. It is of note that in TB-endemic regions, BCG vaccination and exposure to environmental mycobacteria affect cross-reactive T cell immunity, probably influencing vaccine efficacy and geographic variability⁸⁵.

What is the role of direct versus cross-presentation?

The MHC-II pathway samples endosomal compartments for antigen presentation to CD4⁺ T cells (Fig. 3a–c), whereas MHC-I binds cytosolic antigens and presents these to CD8⁺ T cells (Fig. 3d,e). Mtb has been shown to cause phagosomal membrane damage in infected human and mouse DCs and macrophages via the ESX-I type VII secretion system, allowing antigens to enter the cytosol and to be presented by MHC-I^86,87. This mechanism explains why human and mouse CD8⁺ T cells mainly target secreted Mtb antigens^88–91. Another mechanism is cross-presentation, which allows endosomal antigens to be presented by MHC-I via a cytosolic or vacuolar pathway, and is a key feature of cDC1 cells. In the cytosolic pathway, antigens ‘cross’ from endosomes into the cytosol; in the vacuolar pathway, the MHC-I machinery is recruited to endosomes. Cultured human DCs cross-present Mtb antigens by both pathways^92–94, and Mtb-infected human and mouse macrophages retain their ability to cross-present exogenous antigens to CD8⁺ T cells^52,95–98. Macrophages also cross-present antigens from dying infected macrophages to CD8⁺ T cells^96,97 (Fig. 3e,f). Finally, exosomes, which are produced by living cells, are a source of Mtb antigens⁹⁹ (Fig. 3g,h). These pathways differ in whether they lead to antigen presentation by infected or uninfected bystander cells. A crucial issue for TB vaccine design is which of these pathways dominate during infection in vivo and whether the different pathways favour presentation of distinct Mtb antigens. Given that BCG, which lacks the ESX-1 type VII secretion system, is poor at inducing CD8⁺ T cell responses, and most prime-boost vaccine strategies bias T cell immunity towards CD4⁺ responses, it remains to be determined whether protective CD8⁺ T cell responses can be generated with vaccines.

How does infection with Mtb affect cellular interactions in lung-draining LNs?

DCs in LNs can provide signals that induce CD4⁺ T cell polarization. DC–T cell interactions in LNs are short range and facilitated by intranodal cell migration, which is determined by chemokine gradients and structural elements, particularly fibroblastic reticular cells and collagen-containing conduits^100,101. In studies of model antigens and a T_H1-promoting adjuvant, resident LN DCs undergo CCR7-dependent redistribution from peripheral to central locations in the T cell zone, where cDC1 and cDC2 cells occupy distinct locations and present antigens to CD8⁺ and CD4⁺ T cells, respectively¹⁰². Concurrent with intranodal migration of resident DCs is an influx of monocytes and migratory DCs. Although monocytes contribute little to antigen presentation, they can secrete IL-12, which promotes T_H1 polarization. Far less is known about the events following infection with Mtb, especially regarding intranodal cell trafficking or the sources of T cell polarizing signals^22,76. At later stages of MTI, robust expansion of B cells in LNs causes a redistribution of fibroblastic reticular cells and loss of paracortical T cell zones¹⁰³, interfering with CD4⁺ T cell priming. The drivers of B cell expansion and LN reorganization in MTI are unknown.

Do all Mtb antigens behave similarly?

The Mtb proteome comprises ~4,000 potential antigens, but TB research typically focuses on a few immunodominant, secreted proteins (for example, Ag85A, Ag85B, ESAT-6 and TB10.4). Does recognition of other (non-secreted) antigens, particularly those expressed at different stages of MTI or associated with other bacterial compartments, generate T cells with distinct roles in immunity^79,104? Genome-wide studies of infected mouse lungs identified Mtb genes that are upregulated in vivo, and studies with human cells revealed that antigens encoded by some of these genes induce T cells that express cytokines other than IFNγ or TNF (such as IL-17 or GM-CSF). This suggests potential new targets for TB vaccine development¹⁰⁵. The relationship between Mtb and humans appears to differ from that of other pathogens that are characterized by antigenic variation to evade immune recognition. For example, most epitopes of Mtb antigens that are recognized by human T cells are highly conserved, suggesting unique evolutionary pressures¹⁰⁶. However, a subset of Mtb proteins, termed ‘Rare variable Mtb antigens (RVMA)’, undergo diversifying selection and induce CD4⁺ T cells that produce IL-17 and express RORγT, in contrast to the T_H1 responses to conventional Mtb antigens that are characterized by IFNγ-production¹⁰⁷. This indicates that TB vaccine research should investigate strategies to induce broader T cell responses beyond T_H1 responses.

Are there key Mtb antigens that induce protective immunity?

Identifying Mtb antigens that induce protective immunity is crucial for developing new vaccines. Traditional antigen discovery strategies identified Mtb proteins that are recognized by CD4⁺ and CD8⁺ T cells using IFNγ production as a readout¹⁰⁸. As IFNγ alone does not fully account for host resistance^41,109, and some Mtb antigens induce human T cells that produce effectors other than IFNγ¹¹⁰, it is likely that these screens did not reveal the full spectrum of antigens encoded by the Mtb genome. This may be important, as T cell activities, such as type 3 and cytotoxic responses, associate with protection in Mtb-infected mice, macaques and humans^{18,19,111,112}. Future antigen discovery for the development of vaccines should target antigens that are specific for different infection stages^80,113, as well as those under immune pressure and thus probably associated with protection^{107,108,114,115}. Antigen epitopes that are presented by infected cells in the lungs as opposed to those presented only in LNs may be useful vaccine targets to induce T cells that recognize Mtb-infected cells and thereby provide superior protective responses. These antigens might be discovered in mice and NHP, followed by targeted analyses using human samples. Small changes in TCR affinity can have a large impact on the fitness of antigen-specific T cells in both primary and memory T cell responses^47,116–118 and ultimately affect epitope abundance in complex ways⁵⁰. Heterogeneity in TCR affinity can lead to diversity in T cell differentiation and function during both initial T cell priming and memory formation. Novel concepts and technological approaches are necessary to identify the most relevant antigens associated with optimal T cell differentiation, localization and effector functions and protection against TB¹¹⁹.

Knowledge gaps regarding granuloma T cells

What mechanisms determine their location?

T cell homing to the lung, which involves crossing the pulmonary endothelia and navigating through the collagen-rich interstitium, is regulated by adhesion molecules such as the integrins αLβ2 (LFA-1), α1β1 (VLA-1) and α4β1 (VLA-4), as well as chemokines and their receptors¹²⁰. Effector CD4⁺ T cells are recruited to the lungs of Mtb-infected mice by CXCR3 and its ligands CXCL9 and CXCL10 (refs. 111,121). Chemokine signalling can be upregulated by injury sensing through the ATP/P2RX7 axis. In MTI and severe influenza, ATP release by damaged cells is sensed by CD4⁺ T cells that express P2RX7 and upregulate CXCR3, promoting their accumulation in the lungs¹²². The roles of CXCR3 and other chemokine/chemokine receptors or other determinants, such as neural guidance molecules, in determining the location of T cells within the granuloma is unknown.

In liver granulomas of M. bovis BCG-infected mice, CD4⁺ T cells were found to be highly motile and migrating throughout¹²³. By contrast, in Mtb granulomas in the lung, CD4⁺ T cells are often concentrated at the periphery^3,26,124, suggesting that CD4⁺ T cells do not access the core where Mtb-infected cells reside (Fig. 2). The mobility of T cells within these peripheral zones has yet to be understood, given that studies have only analysed fixed tissue samples^30,125. Considering that T cell efficacy against Mtb depends on direct contact with infected cells, it is critical to define the factors that segregate T cells from these targets within granulomas and to identify barriers hindering T cell access to core regions (Fig. 2 and Box 1). Strategies to improve T cell infiltration, function and survival could significantly enhance MTI outcomes^{120–122,126}. Factors that affect T cell positioning have been identified in tumours and during acute lung infection and include chemokines and their receptors, interaction with stromal cells and the extracellular matrix (ECM), and antigen recognition¹²⁷. The similarities and differences in the organization and activities of T cells in Mtb granulomas and in tumours are unclear and require further study. As an additional mechanism to explain the paucity of T cells in granuloma cores adjacent to Mtb-infected cells is that toxic metabolites produced by Mtb-infected cells may induce cell death in effector T cells. In other contexts, such as influenza virus infection and acetaminopen-induced liver injury, tissue-resident memory T (T_RM) cells are preferentially localized and maintained in specific niches that show signs of previous tissue injury^124,128. Low leukocyte recruitment into granulomas correlates with suboptimal Mtb control, suggesting that factors that determine T cell migration also affect MTI outcomes¹²⁷, though mechanisms are unknown (Box 1).

Box 1 |. Potential mechanisms that explain T cell segregation from Mycobacterium tuberculosis (Mtb)-infected cells in granulomas.

• Ineffective blood vessel trafficking

  • Dearth of high endothelial venules, which are known to facilitate lymphocyte entry in lymph node and tumours¹²⁹, correlates with absence of lymphocyte infiltration.
  • In recent studies, high endothelial venules detected in human and experimental tuberculosis (TB) granulomas were found to correlate with lymphocyte accumulation and improved immunity^131,132.
• Disrupted chemotactic signals

  • Impaired signalling via the CXCL9-10-11/CXCR3 axis may affect T cell migration^111,121,126.
  • Tissue injury and signalling via the ATP/P2RX7 axis that upregulates CXCR3 expression may regulate chemokine signalling and guide T cell positioning, as observed in TB granulomas¹²² and in viral infections, where tissue-resident memory T (T_RM) cells persist in damaged lung tissue^124,128.
• Physical barriers

  • The extracellular matrix may limit T cell access, as demonstrated in tumours where fibronectin fibres create a physical barrier, preventing T cells from entering the tumour core¹³⁶.
  • Dense clusters of infected and non-infected cells may restrict migration, as observed in Leishmania major infection¹⁶⁰. This is an active topic of investigation in TB.
• Failure to stably engage with infected cells

  • T cells may not stably interact with Mtb-infected cells, comparable with what was observed in Bacillus Calmette–Guérin-induced hepatic granulomas¹²³.
  • T cells may interact with uninfected cells that have acquired antigens. This is proposed to divert antigen-specific T cells away from Mtb-infected cells.
• T cell exhaustion and death

  • T cells in granulomas can become exhausted and show reduced proliferation and function^68–70.
  • Indoleamine-2,3-dioxygenase (IDO), an enzyme that is elevated in granulomas, can promote T cell death via the depletion of tryptophan and/or the generation of immunoregulatory kynurenine catabolites^60–62.

Does T cell entry require specialized blood vessels?

High endothelial venules (HEVs) are specialized blood vessel endothelial cell structures that mediate the extravasation of leukocytes, including naive and central memory T cells, into LNs. At sites of chronic inflammation and in tumours, HEV-like vessels are induced in non-lymphoid tissues, where they promote lymphocyte entry into tissues. HEV abundance in tumours correlates with immune cell infiltration, therapeutic responses and patient survival^129,130. HEVs are detected in granulomas in human and animal models of TB and correlate with lymphocyte accumulation and improved immunity following vaccination^131,132 (personal communication, Rasmus Mortensen, Statens Serum Institute).

How does tissue structure affect T cell migration?

Characteristics of the ECM, including its density, stiffness and spatial organization, affect T cell trafficking through tissue¹³³. Intravital imaging shows that T cells often follow fibrillar collagen structures, guiding their movement^134,135. In sections of human lung tumours, T cells in perivascular regions were found to migrate along fibronectin fibres, which blocks their entry into the tumours. ECM degradation increased the ability of T cells to contact tumour cells in the sections, as determined by intravital microscopy¹³⁶. As TB granulomas are also surrounded by dense ECM that is composed of fibronectin, collagen and laminin^30,137, these structures might limit T cell access to the granuloma core. Although a dense matrix encasing the granuloma is readily identified by light microscopy, little is known about how it affects T cell entry and localization in TB granulomas.

Does antigen presentation promote T cell arrest?

The recruitment of T cells to sites of inflammation requires chemotactic signals and adhesive interactions, whereas TCR-dependent recognition of antigen leads to stable T cell–APC interactions that can halt T cell migration¹³⁸. Intravital microscopy imaging of liver granulomas in mice infected with attenuated M. bovis BCG revealed highly motile mycobacteria-specific T cells, but only a minor fraction of these stably interacted with APCs¹²³. It will be important to determine whether such interactions are also rare in infection with virulent Mtb. Nevertheless, samples from the respiratory tract from patients with TB reveal activated T cells, as determined by their expression of CD38, HLA-DR, PD-1 (refs. 139,140; see also ref. 141 and references therein) and bronchial fluids from infected macaques contain TNF-producing and/or IFNγ-producing CD4⁺ T cells³⁰. The antigen specificity of these T cells and the triggers of their activation remain unclear. As Mtb-specific T cells infrequently contact infected cells in granulomas, there is a need for more research on how this impacts MTI persistence and progression.

Does Mtb evade or subvert T cell recognition of infected macrophages?

Several mechanisms by which Mtb evades T cell recognition of infected macrophages have been described (Table 2). Studies in mice show that only a small fraction of Mtb-specific CD4⁺ and CD8⁺ T cells are activated by their cognate antigens^46,49, indicating that immune subversion mechanisms operate in vivo. Intravenous administration of Mtb epitope peptides to Mtb-infected mice markedly increased the frequency of Mtb antigen-specific T cell activation in vivo³⁵, indicating that T cells maintain the ability to be activated if they have encountered their cognate antigens. However, despite increased T cell activation, only minor reductions in lung bacterial loads were observed, probably due to the spatial separation of T cells from Mtb-infected cells³⁵. These results support the hypothesis that Mtb-specific T cells do not optimally perform effector functions in tissues, and further studies are required to determine the significance of distinct mechanisms that contribute to suboptimal T cell functions in MTI.

How do T_RM cells contribute to immunity in TB?

Considerable evidence suggests that preexisting immunity to Mtb, from prior exposure or vaccination, leads to distinct immune responses in both humans and animals, highlighting how immunological memory affects MTI outcomes^28,142. Memory T cells, which are compartmentalized across different tissues, play a significant role in these responses. Central memory T cells (CD62L^hiCCR7^hi) reside in secondary lymphoid tissues, effector memory T cells (CX3CR1^hi) circulate through peripheral tissues, and peripheral memory T cells (CX3CR1^int) predominantly remain in peripheral tissues¹⁴³. By contrast, T_RM cells localize in peripheral tissues without circulating^144,145, expressing surface proteins such as CD103 (αE), CD69 and VLA-1 (α1β1) that anchor them in specific locations. CD4⁺ T_RM cells in the lungs can be derived from T_H17 cells — so called ex-T_H17 cells — and have a role in the clearance of Klebsiella bacteria¹⁴⁶. Mucosal BCG vaccination generates lung T_RM cells, providing better protection than subcutaneous vaccination^147–150. Antigen persistence is critical for shaping lung CD4⁺ T_RM cell responses after mucosal BCG vaccination, but distinguishing the roles of antigen persistence and T_RM cells in protection remains challenging and requires tools for selective, time-dependent depletion of T_RM cells^{124,151–154}. Similarly, mucosal vaccines containing Mtb antigens, using various vector platforms or BCG, also increase the presence of T_RM cells^147,155, potentially contributing to their protective efficacy. T_RM cells are present at different infection sites in patients with TB¹⁴⁰ and T_RM cells are enriched in the airways of individuals who are considered resistant to TB⁴⁴. Whether T_RM cells reside in distinct locations in granulomas and have unique access to infected macrophages is unknown^{36,128,156,157}.

New technologies to fill knowledge gaps

Spatially resolved ‘omic’ analyses to characterize T cell localization and activation

Insights into the significance of T cell localization relative to Mtb-infected cells and the mechanisms that control it has been limited by low resolution methods. Spatial ‘omics’ technologies now reveal details about cell states and interactions at a molecular level while preserving spatial context. These technologies can analyse DNA, RNA, chromatin, proteins, metabolites, glycans and drugs, using advanced microscopy, mass spectrometry or sequencing techniques. Together with quantitative image analysis and bioinformatics, spatial omics methods offer new insights into cell subpopulations and their interactions with the ECM and other tissue components¹⁵⁸. For example, spatial transcriptomic analysis of granulomas in Mtb-infected C3HeB/FeJ mice revealed the necrotic centre as immunosuppressive, whereas the periphery contained activated T cells and macrophages³¹. Studies in both mice and humans indicate that T and B cell infiltration into granulomas increases over time, driving the formation of tertiary lymphoid structures that enhance local immune responses against Mtb³². A current limitation of spatial transcriptomics technologies is the inability to perform transcriptome-wide characterization at the single-cell level.

High-parameter microscopy and spatial analyses using metal-labelled antibodies and multiplexed ion beam imaging by time of flight have allowed the characterization of a large number of proteins and cell types within human TB granulomas. This detailed mapping revealed 37 different surface and intracellular proteins that allowed the identification of 19 different immune cell subsets and eight regional cellular microenvironments, indicating that the regulation of immune processes is highly localized within the granuloma. It also revealed that, as also observed in the studies of Mtb-infected C3HeB/FeJ mice, myeloid rich regions are associated with immunosuppressive signals (such as IDO-1, PD-L1 and TGFβ), T_reg cell infiltration and the absence of T cell activation³¹. These spatial ‘omic’ approaches now offer the possibility to define APC subsets that present Mtb antigens in LNs and lung and the consequences of APC diversity on T cell activation and polarization. They can also be used to assess myeloid cell trafficking to LNs and evaluate LN reorganization and tertiary lymphoid follicle formation (Table 3).

Table 3 |.

New technologies to fill knowledge gaps in TB immunity

Knowledge gap	Hypothesis	Technology	Models
Ag presentation/T cell activation
Which cells present Mtb antigens in LNs and in lungs? Is their efficiency altered by infection with Mtb?	One or more specific subsets of DCs, including both infected and bystander cells, prime T cells in LNs; multiple subsets of macrophages present antigens in lungs, with uninfected/bystander cells superior to Mtb-infected cells	Proximity labelling methods (quantify/characterize Mtb antigen-specific T cell interactions with target cells)	Mouse
		Spatially resolved high throughput analyses (quantify/characterize/locate)	Mouse NHP Human
What are the epitopes presented by DCs in LNs, and by macrophages in the lungs? What epitopes are presented by DCs and not macrophages?	Differential proteolysis and expression of MHC pathway proteins result in differential antigen presentation	High-resolution mass spectrometry; differential activation of epitope-specific T cells (identify epitopes)	Mouse NHP Human
What are the mechanisms of T cell differentiation and polarization in LNs?	Infection with Mtb perturbs lymph node organization; CD4+ and CD8+ T cells are primed by uninfected DCs in distinct LN regions. T cell polarizing cytokines are produced by uninfected cells	Spatially resolved high throughput analyses (quantify/characterize/locate)	Mouse NHP Human
T cell trafficking and positioning in the infected lung
Which signals control and limit T cell entry into granulomas?	High endothelial venules and vascular inflammation combine with T cell intrinsic properties to promote T cell egress from blood vessels in or adjacent to granulomas (facilitated by specific chemokines and adhesion molecules)	Spatially resolved high throughput analysis (quantify/characterize/locate)	Mouse NHP Human
Which signals determine macrophage and T cell positioning in granulomas?	Chemokines, chemokine receptors, danger signals, extracellular matrix and neural guidance molecules establish granuloma architecture and cell-cell interactions
Which signals control and limit T cell arrest and recognition of Mtb-infected cells in granulomas?	Physical barriers (matrix, NETs), suboptimal antigen presentation by Mtb-infected cells; arrest on uninfected bystander cells; poor T cell receptor signalling and T cell exhaustion	Multiphoton intravital microscopy (characterize cell behaviours to test hypotheses and candidate mediators and mechanisms)	Mouse
What are the determinants of T cell activation and killing of Mtb-infected cells?	Time of arrest, number of T cells per infected cell

Multiphoton intravital microscopy to study T cell migration and interactions with infected cells

The reasons why T cells are spatially segregated from Mtb-infected macrophages in granulomas remain unclear. Potential causes include ineffective blood vessel trafficking, disrupted chemotactic signalling, physical barriers such as the ECM, failure to recognize and arrest on infected cells, diversion by antigens that are presented by uninfected APCs or death of activated T cells in the granuloma core. Intravital microscopy offers a way to explore these dynamics by providing real-time visualization of dynamic processes within natural environments.

Intravital imaging has been instrumental in understanding T cell behaviour in tumours. It revealed that T cell infiltration varies by tumour region, suggesting that variation in T cell infiltration rather than a defect in the cytotoxic activity contributes to inefficient antitumor responses¹⁵⁹. In tumours, low T cell infiltration is correlated with a low frequency of HEVs, and the importance of these specialized vessels as the main sites of lymphocyte arrest and extravasation in tumours was revealed by microscopy¹²⁹. Real-time observation of T cell migration in granulomas, similar to studies in tumours, can elucidate the roles of chemokines, neural guidance molecules and responses to danger signals from dying cells in directing T cell movements. In situ imaging may help assess how the ECM and physical barriers influence T cell access to the core of granulomas, informed by studies in Toxoplasma gondii and Leishmania major infections that highlight how structural barriers affect immune cell dynamics^{38,135,136,160} (Table 3).

Intravital imaging is the only method that allows to determine the efficiency of T cell interactions with and responses to Mtb-infected cells and to provide insights into T cell activation and cytotoxicity. Various tools have been developed to monitor T cell activation and the death of infected cells in vivo. These include TCR signalling reporters^161,162, pH-based parasite death reporters¹⁶³, pathogen-encoded reporters of calcium flux that are selective for infected cells¹⁶⁴ and a FRET-based dual fluorescent protein that detects cells undergoing apoptosis¹⁵⁹. These and other reporters can be used to correlate T cell function with bacterial death during infection. Yet, in vivo imaging of Mtb-infected lungs faces novel challenges. These include challenges specific to living lungs such as the need to maintain respiratory functions and movements that disrupt imaging. These issues have been addressed with techniques such as mechanical ventilation and stabilizing lung windows¹⁶⁵. Moreover, performing these procedures in biosafety level 3 containment adds complexity^166,167 and is currently only possible in very few centres.

New tools to investigate immune cell activation and Mtb in vivo

The study of immunity to Mtb and the response of Mtb to the immune system in vivo can benefit from specific tools for flow cytometry and intravital imaging, including reporter mice and engineered mycobacterial strains. Tools, such as mice expressing fluorescent proteins, can help to identify immune cell subsets including neutrophils, macrophages, DCs, natural killer and T cells^168–170. Techniques such as transferring cells or injecting fluorescent antibodies or Fab fragments are employed to stain cells for in vivo and ex vivo studies^171–173. These methods allow the tracking of T cell activities, such as T_H1/T_H2 polarization and cytotoxic functions, and macrophage heterogeneity and death^174–177. Advanced imaging techniques, including MRI, CT, PET and SPECT, along with second harmonic generation imaging, provide non-invasive, real-time insights into tissue structure and cell dynamics^178,179.

Reporter systems for Mtb are also available, including strains that express constitutive or inducible fluorescent proteins. Pathway-specific transcriptional reporter strains of Mtb can probe host-imposed stresses, such as acidic pH or hypoxia encountered by Mtb during infection^180–182. Although the response of Mtb promoters to multiple stresses is a potential limitation, these innovative tools allow observation of the immune landscape and pathogen behaviour within the granuloma microenvironment, aiding the study of Mtb responses to host defences.

High-resolution dynamics of APC–T cell interactions using proximity labelling

Cell–cell interactions are a crucial component of all immune responses and can be studied by proximity labelling techniques, where one cell type provides a donor molecule that tags target cells with which they interact. Fuco-ID, which employs fucosyltransferase-meditated fucosyl-biotinylation, and quinone methide-assisted identification of cell spatial organization are tools to chemically modify donor cells to identify cell–cell interactions, but they have limited utility in vivo^183,184. The recent development of genetically encoded systems such as ‘Labelling Immune Partnerships by SorTagging Intercellular Contacts’ (LIPSTIC) provides new opportunities to study cell–cell interactions in vivo, including in MTI. The key to LIPSTIC is the Staphylococcus aureus sortase A (SrtA), an enzyme that transfers peptides with the amino acid motif ‘LPXTG’ to a polyglycine (G5) acceptor. The initial version was designed to study CD4⁺ T cell interactions with APCs, so SrtA was expressed fused to CD40L, and CD40 was tagged with five N-terminal glycine residues (G5)¹⁸⁵. When CD40L-SrtA-expressing T cells and LPETG-biotin peptide were administered to mice expressing G5 on CD40, the APCs that interacted with the transferred T cells were labelled with the biotinylated peptide, allowing them to be quantified and characterized. These studies revealed that T cell–APC interactions in vivo are more complex than previously appreciated.

Originally based on the CD40L-SrtA/CD40-G5 interaction¹⁸⁵, the system was made more universal (‘uLIPSTIC’) by incorporating SrtA and G5 sequences into donor or acceptor cell membranes specified by Cre recombinase activity. This setup allows Cre-dependent expression of SrtA in designated (‘donor’) cells, enabling detailed study of T cell interactions, including the specificity and kinetics of their responses in Mtb-infected mice. Biotin-labelled APCs isolated by cell sorting can then undergo further analysis like single-cell transcriptomics, providing insights into the cellular outcomes of these interactions¹⁸⁶. Proximity labelling can readily complement results of cell–cell interactions observed by intravital imaging and can enable downstream analyses by isolating the biotin-labelled target cells (Table 3).

Identification of Mtb antigens that are targets of protective immunity

Identifying the relevant antigens in Mtb is challenging given the large number of proteins expressed. However, it is important because it has been observed that in infection with other pathogens, some antigens induce protective T cell responses whereas others do not^187,188. Various approaches, including in silico prediction and biochemical and immunological characterization, have allowed the identification of targets of T cell recognition¹⁸⁹; however, they have not yet led to the identification of antigens that preferentially induce protective immunity. Challenges include variations in antigen presentation during infection^{46,49–51,190}, unpredictability of peptide splicing¹⁹¹, variability in the expression level of native Mtb proteins and their ability to be presented to T cells, which depends on the APC and the stage of infection, as well as limitations of current algorithms in predicting non-peptide antigens. Nevertheless, studies of humans with distinct MTI outcomes (progression or control) may reveal antigens and epitopes that are associated with protection^46,49–51. Another approach is the direct identification of naturally processed Mtb epitopes on infected cells using mass spectrometry¹⁹². Improved tools for mass spectrometry and bioinformatics make the characterization of naturally presented MHC ligands feasible^193,194. This approach allowed the identification of epitopes presented on THP-1 cells infected with BCG, human Mtb-infected monocyte-derived macrophages and splenocytes from Mtb-infected mice^195–197. Although technical improvements have reduced the number of cells required to ~10⁸, only the most abundant antigens remain accessible, and these methods have not been able to determine whether there are differences in the peptides presented by Mtb-infected DCs versus macrophages⁷⁸. Advances in iPS cells that are differentiated to closely resemble distinct APC phenotypes may help to overcome cell number limitations and provide valuable insights^198–200. Deeper analysis will require more starting material or enhanced methodologies including machine learning²⁰¹. Advances in lipidomics are facilitating the identification of CD1-presented lipid and glycolipid antigens²⁰², which could simplify the process of identifying mycobacterial lipid antigens recognized by T cells^203,204. A remaining challenge for optimal vaccine design is to determine the relationships between epitopes presented by cultured cells and those presented in vivo by Mtb-infected cells, as Mtb-containing myeloid cells in vivo vary in their antigen presenting characteristics.

Another approach to discover Mtb antigens that induce protective immunity leverages TCR repertoire analysis in humans with distinct MTI states. For example, certain TCRs that cluster according to the structures of their epitope recognition domains are more prevalent in controllers than in progressors, and these can be further analysed to identify their epitope specificity^104,205. However, identifying the targets of the TCRs in the clusters of interest presently involves slow and labour-intensive efforts and may benefit from recent progress in predicting the molecular structures of protein–protein interactions using artificial intelligence²⁰⁶. Another new technology that can be applied to human samples uses DNA-barcoded epitope probes, single-cell sequencing and analyses of TCRs, as well as transcriptomic technologies to characterize T cells in individuals with distinct MTI states. With highly multiplexed (hundreds to thousands) epitope probes and focus on HLA allotypes that are prevalent in regions with a high burden of TB, this approach offers the potential to markedly accelerate the discovery of T cell characteristics that are associated with protective immunity and inform vaccine design²⁰⁷. These efforts may further benefit from studying tissue-derived T cell populations from human bronchoalveolar lavages or lung tissues or animal models^{19,44,157,208}. These strategies aim to differentiate between antigens that trigger effective immunity and those that do not, improving our ability to design effective Mtb vaccines.

New animal models to better mimic human pathophysiology

Mechanistic studies of immunity in mice have usually been carried out in C57BL/6 strains because of the availability of genetically engineered strains on this background. Other strains can be informative as they have different phenotypes following vaccination and/or infection. For example, C3HeB/FeJ mice develop lung lesions that morphologically resemble human granulomas²⁰⁹. A genetic locus (sst1) that confers this property was identified, and Sp140 was identified as the responsible gene. Sp140 gene-targeted C57BL/6 mice are highly susceptible to Mtb and enable studies using genetic tools that were previously only available for the more resistant C57BL/6 mice⁶⁴. To overcome the limitation that inbred mouse strains have a limited gene pool, collaborative cross (CC) and diversity outbred (DO) mice were generated from eight parental mouse lines, which included representatives from all three Mus musculus subspecies including three strains derived from the wild. Ultimately, ~70 recombinant inbred mouse strains were generated from the CC funnel breeding scheme, whereas the DO mice continue to be randomly intercrossed, generating mice that resemble an outbred population^210,211. Both CC and DO mice are now being used for studies of host responses to infection with Mtb and, as predicted, reveal diverse phenotypes. These resources can be used to identify the genetic basis and immunological mechanisms associated with host resistance and responses to vaccines^212–214.

Additional emerging technologies to study T cell immunity in MTI

Beyond the tools described above, other emerging technologies offer opportunities for enhanced discovery and determination of mechanisms of protective immunity to Mtb.

Mass spectrometry imaging (MSI) does not depend on antibodies or other probes and exploits the sensitivity of mass spectrometry to localize and visualize diverse molecules in tissues at the cellular and subcellular levels²¹⁵. MSI can localize specific proteins, lipids, signalling molecules, metabolites and drugs^216–219. Studies of Mtb-infected samples have already revealed insights into differential tissue distribution of drugs that are used to treat TB²²⁰. MSI studies of Mtb-infected tissues require inactivation of the bacteria using methods that do not destroy the analytes of interest.

Engineered mouse models that express specific elements involved in immune responses in humans are advancing rapidly and should be adapted for specific studies of MTI. For example, VelociT mice are engineered to generate T cell responses that closely mimic those in humans by substituting the human TCRα and β variable regions and CD4⁺ and CD8⁺ extracellular domains for their mouse counterparts²²¹. Although the antigen processing and T cell signalling machinery are from mice, when combined with human HLA class I or class II alleles, vaccination (and presumably infection) of these mice can generate T cell responses with human specificity and may be useful for characterizing the immune responses to specific TB vaccines.

In other fields, the generation and characterization of three-dimensional organoids has provided unique insights into cell–cell interactions and determinants of tissue architecture that are otherwise not possible to study, especially in humans^222,223. An adaptation of these technologies, including variants at advanced stages of development that incorporate microvasculature, offers opportunities to discover and characterize mechanisms of cell trafficking, positioning and cell–cell interactions that may differ between mice and humans.

As high throughput and high-resolution characterization of immune responses to infection with Mtb and other immunological states generate large datasets, efforts to share access to these datasets are also advancing and have already generated several examples. These include The Human Immunome Project (https://www.humanimmunomeproject.org/), ImmPORT (https://www.immport.org/home) and ImmuneSpace (https://immunespace.org/), as well as the Immune Epitope Database (www.iedb.org). A TB-specific data repository is also available (https://tbportals.niaid.nih.gov/), although it does not currently include immunological data. The current strengths of these public data repositories lie principally in their genomic and cytometry data; additional efforts will be needed to provide similar access to imaging data.

Conclusion

The development of effective TB vaccines has proven challenging because of unique characteristics of the pathogen, Mtb and the requirement for performing research work in high level biocontainment. Consequently, approaches that reveal mechanisms of immunity and inform development of vaccines against other pathogens have not been applied or have been insufficient to inform TB vaccine development. In other fields of biology and immunology research, new technological platforms have enabled unprecedented discoveries and insights and have opened new avenues of research. These are available to be applied to fill the knowledge gaps that currently hinder the development of efficacious TB vaccines, with the expectation that they will build on existing knowledge to accelerate design, development, and evaluation of effective TB vaccines.