Introduction

As one of the most prevalent communicable diseases, tuberculosis (TB) remains the leading cause of death from a single infectious agent, only being surpassed by COVID-19 during the 2019–2021 pandemic. There were an estimated 10.6 million new active TB cases and 1.3 million deaths in 2022, according to the latest World Health Organization (WHO) global tuberculosis report1. It is also estimated that one-quarter of the world’s population has been infected with Mycobacterium tuberculosis (Mtb), the causative agent of TB, and has a 5–10% lifetime risk of progressing into active TB disease2,3. Thus, the global control of TB remains challenging. The 1-century-old Mycobacterium bovis Bacille Calmette-Guérin (BCG) is the only licensed vaccine against TB. Although BCG offers protection against the aggressive forms of the disease, including meningeal and miliary TB in children, its protective efficacy ranges from 0% to 80% in adolescents and adults4. Considering that vaccines are one of the most effective ways of controlling infectious diseases, developing better vaccines is a crucial objective in TB research.

The development of a more effective TB vaccine faces obvious challenges, among which, the absence of validated correlates of protection (CoPs) is a crucial knowledge gap that significantly hampers progress5,6. CoPs are defined as biological parameters that predict the degree of protection against infectious disease; reliable CoPs could facilitate the development of new vaccines by assessing protective efficacy without exposing clinical trial participants to the infectious agent that the vaccine aims to protect against7,8. Currently, TB vaccine testing is entirely empirical. Since Mtb is a slow-growing bacterium and it takes months to years to progress into active TB diseases post-exposure in humans9,10, the lengthy TB vaccine clinical trials to measure differences in clinical endpoints require tremendous costs and resources. Hence, robust, generalizable, and vaccine-induced CoPs that serve as an intermediate outcome to infer efficacy are desperately needed.

In this review, we provide an overview of potential CoPs for TB vaccines and summarize the usage of different samples in TB vaccine CoPs research. The current major hurdles and potential opportunities for finding reliable CoPs are then discussed in some detail.

The identification of different immune cell populations in TB vaccine CoP research

Th1 immune responses

The most protective antigens and corresponding CoPs have often not been identified in complex diseases that depend heavily on T cell immune responses for protection. The genome of Mtb encodes ~4000 individual proteins that are dynamically produced during different infection stages, and this hinders the identification of the most protective antigens and the corresponding CoPs, which will be discussed in Section “The major hurdles of TB vaccine CoPs identification”. CD4+ T cells are crucial effector cells in the host’s response to Mtb infection, and the helper type 1 CD4+ T cells (Th1) that secrete cytokines IFN-γ, IL-2, and TNF-α are prominent in the immune response against Mtb11. Typically, in TB vaccine development, the effect of a vaccine on the number of bacteria in the lungs and spleens of mice was correlated with the numbers and proportions of multifunctional CD4+ T cells expressing IFN-γ, IL-2, and TNF-α (Fig. 1). Consequently, the Th1 cytokine profiles in antigen-specific T cell responses have been routinely used as biomarkers and assessed as predictors of protection in the evaluation of TB vaccines12,13. IFN-γ was found crucial for the expression of protective immunity, and IFN-γ deficiency causes a high susceptibility to mycobacterial infections14,15. However, excessive IFN-γ production can accelerate death in infected mice16, and current TB vaccine screening criteria that favor vaccines with robust antigen-specific Th1 memory immune responses may over-emphasize IFN-γ production. The MVA85A vaccine could induce high levels of Th1 immune responses in several animal models, such as mice17, guinea pigs18, cattle19, and non-human primates (NHPs)20, as well as in human clinical trials21,22, however, this vaccine failed to show protection in BCG-immunized infants in a phase IIb clinical trial23. Thus, although Th1 cells are considered central mediators of protection, accumulating evidence suggests that current Th1 cell measurements do not reliably indicate the establishment and recall of host resistance against Mtb infection.

Fig. 1: Potential correlates of vaccine-mediated protection against TB.
figure 1

Protective immunity is dependent on the crosstalk between innate and adaptive immune systems. The summaries of the multi-faceted protective immune responses are collected from several research models and reflect the main features overall. Some components are still understudied as endpoints. Abbreviations: AMs alveolar macrophages, Breg regulatory B cells, DCs dendritic cells, IFN interferon, Ig immunoglobulin, IL interleukin, ILCs innate lymphoid cells, Mø macrophage, MAIT mucosal-associated invariant T cells, Mtb Mycobacterium tuberculosis, NK natural killer cells, TNF tumor necrosis factor, TCM central memory T cells, TEM effector memory T cells, TRM tissue-resident memory T cells, Treg regulatory T cells. Created with bioRender.com.

Full size image

Th17 immune responses

Several studies have shown that IL-17-secreting T cells are involved in immune protection against Mtb infection at early stages of infection24,25,26. Th17 cells could recruit neutrophils, monocytes, and Th1 cells to the infection site and granuloma formation and inhibit Mtb replication and dissemination24,27,28,29. Mice lacking IL-17 receptors were also defective in exerting long-term control of Mtb infection30. We and others showed that Th17-inducing TB vaccines gave high levels of immune protection against Mtb challenge in animal models31,32,33,34,35,36. However, similar to IFN-γ-producing Th1 cells, the overexpression of Th17 cells could also enhance susceptibility to infection by exaggerating the influx of neutrophils and inflammatory monocytes to the site of infection, causing lung tissue damage37,38,39. In addition, repeated antigenic exposure led to increased IL-17 expression and immune-mediated pathology40. Thus, current findings suggest a critical role for IL-17-producing Th17 cells in protective immunity via an early but transient burst of IL-17 production, whereas long-lasting Th17 cell activation might contribute to tissue pathology.

CD8+ T cell immune responses

Antigenic peptides from intracellular pathogens such as Mtb can enter the cytosol and associate with major histocompatibility complex class I molecules, resulting in antigen presentation to CD8+ T cells. Upon antigen recognition, CD8+ T cells differentiate into immune effector cells, which can lyse infected target cells and kill the intracellular bacteria through a range of mechanisms41,42. Although less prominent than CD4+ T cells, CD8+ T cells also show a protective role in Mtb infection animal models such as mice43,44, cattle45, and NHPs46. Immune protection induced by several TB vaccines is mediated by robust Mtb-specific CD8+ T cell responses47,48,49,50,51. However, high-frequency vaccine-induced CD8+ T cell responses might contribute to the limited protection against Mtb infection52,53. Thus, further profiling of cytotoxic T cell sub-types that employ different mechanisms of cytotoxicity is likely to be rewarding. State-of-the-art technologies such as single-cell multi-omics might help define the heterogeneity among immune cells, including CD8+ T cells, and provide additional opportunities to find CoPs. This will be discussed in Section “The opportunities for identification of TB vaccine CoPs”.

Humoral responses/B-cell mediated responses

For many approved vaccines against extracellular pathogens, the titer of neutralizing antibodies has served as a reliable CoP and an admittance criterion. However, humoral responses fill a minor part of the literature on the host responses to Mtb, primarily due to the paradigm that humoral immune responses play little role in the protection against intracellular pathogens. We and others showed that T cell immune response-inducing TB vaccines could confer protection against Mtb infection without inducing detectable antigen-specific humoral immunity, and some humoral response-inducing vaccines failed to enhance protection54,55. Nevertheless, antibodies are a major arm of adaptive immunity and have the potential to modulate immune protection against Mtb. In 2019, a cohort study showed that “TB resisters”, who were defined as the Mtb-exposed individuals who kept healthy and were persistently negative in IFN-γ release assay (IGRA) and tuberculin skin tests (TST), possessed IgM and class-switched IgG antibody responses56. Evidently, exposure resulted in the presence of sufficient antigen to generate an immune response in at least the antibody arm of the immune system, either directly by antigen deposition or through initial bacterial proliferation. Thus, the definition of the “TB resisters” population suggests IFN-γ T cell-independent mechanisms exist, which provides a new avenue for identifying CoPs, which will be discussed in detail in Section “The opportunities for identification of TB vaccine CoPs”. Recently, distinct features of antibody responses in individuals progressing to active TB from non-progressing TB were profiled. A multivariate antibody score differentially associated with progression could identify progressors up to 2 years before TB diagnosis57. Furthermore, mucosal and systemic Mtb-specific antibody responses were correlated with protection in NHP models and human beings58,59. Thus, a series of studies has now elucidated the contribution of humoral responses/B cells to the modulation of immune responses to Mtb. For example, it was shown that Mtb-specific IgG purified from LTBI individuals enhanced macrophage activation and intracellular Mtb killing60. Besides producing antibodies, B cells are also competent antigen-presenting cells and produce a wide range of cytokines, which can further influence the function of other immune cells61,62. In addition, B cells are an important component of protective granulomas, which contribute to the control of Mtb infection and prevention of reactivation through various mechanisms63,64. The role of humoral responses/B-cell-mediated responses in TB infection was well reviewed elsewhere65,66,67,68. The uncovering of antibody-dependent protection against Mtb infection in the last decade might have opened a new avenue for finding CoPs in TB vaccine design.

Tissue-resident memory T cells (TRM)

TRM represents a distinct subset of memory lymphocyte lineage that colonizes tissues independently of new recruitment of T cells from the blood and provides a rapid response against reencountered infections69,70. One of the major mechanisms by which Mtb escapes from adaptive immunity is by delaying the initiation of the cellular immune responses and the recruitment of activated circulating T cells to the lung mucosa, the site of primary infection9,71,72. Thus, the lung TRM cells participate in the first line of defense, helping to control the replication and dissemination of Mtb at the early stage of infection by immediately sensing infection and inducing recall immune memory responses73. In 2014, Daniel Barber et al. reported the protection afforded by lung parenchyma cells for the first time in Mtb-infected mice: they showed that the Mtb-specific memory CD4+ T cells in the lung could be divided into two subsets, namely, a population of protective cells colonizing the lung parenchyma, and another population which is less protective circulating in the blood vasculature, The populations were distinguished with surface markers as KLRG1−CXCR3+ and KLRG1+CX3CR1+cells, respectively74. Adoptive transfer experiments indicated the lung parenchyma cells (KLRG1−CXCR3+) to be highly protective against Mtb infection16,74,75. Based on these findings, we studied the properties of circulating CD4+ T cells in patients with active TB and found that high expression levels of KLRG1 were associated with disease progression and lower levels associated with protective responses in immunotherapeutics76. However, the definition of TRM cells in Mtb infection is complicated. In general, memory T cells do not develop when an antigen is still present, and actual memory only forms when the antigens disappear. Thus, it is still controversial whether actual memory can form in the murine models of TB since the host cannot eliminate the invading bacilli in this model. In addition, in Daniel Barber’s report, not all lung parenchyma cells are tissue-resident memory cells. Thus, the role of TRM in Mtb infection still needed to be further defined by using other animal models, such as low-dose NHP infection models, in which some of the animals could entirely clear the invading Mtb. Nevertheless, several studies indicated the protective roles of TB vaccine-induced TRM in Mtb infection, and a better understanding of vaccine-induced lung TRM could facilitate the identification of CoPs. In 2016, Stefan H. E. Kaufmann et al. reported that mucosal delivery of BCG-induced TRM mediated immune protection against Mtb infection in a murine model77. However, confirmation was not sought by intravascular staining for definitive identification of TRM. In 2017, we reported the use of a comprehensive intravascular staining technique and a surface markers staining assay to show that a novel recombinant virus-vectored TB vaccine induced high levels of lung CD8+ TRM49. Several TB vaccines were also reported to be able to induce lung TRM, determined by intravascular staining in murine models78,79,80,81. In an NHP model, intravenous (i.v.) immunization with BCG induced more lung parenchymal CD4+ T cells compared with intradermal inoculation and led to a sterilizing immunity against Mtb infection in 202082; this model provided a unique opportunity to define CoPs of TB vaccines, which will be discussed in detail in Section “The opportunities for identification of TB vaccine CoPs”.

Trained immunity/innate immune memory

Immune memory has been regarded as an exclusive characteristic of adaptive immunity. However, a growing body of evidence suggesting that innate immune cells also show memory responses has challenged this dogma. Innate immune memory also termed trained immunity, was defined as the long-term functional reprogramming of innate immune cells that is evoked by exogenous/endogenous stimulation, with the cells returning to the non-activated state yet showing altered inflammatory immune responses (such as IL-1β, IL-6, and TNF-α) and immune protection against the secondary homologous/heterologous infection83,84,85. The TB vaccine BCG was one of the first agonizts found to induce trained immunity in seminal studies of BCG vaccination of school children86,87. More recently, monocytes derived from BCG-immunized healthy volunteers showed enhanced release of inflammatory cytokines in response to heterologous bacterial and fungal pathogens infection; in addition, BCG could enhance lymphocyte-independent protection in immunodeficient SCID mice against disseminated candidiasis88. These non-specific protective effects remained for at least one year after vaccination89 and were mediated by glycolysis and epigenetic regulation90. Importantly, i.v. BCG vaccination-trained hematopoietic stem cells generated epigenetically modified macrophages/monocytes that provided protection against Mtb infection91, and a randomized clinical trial of BCG vaccination showed that BCG enhanced protection against heterologous infections and improved survival in the elderly92. We recently showed that intrinsic factors in different BCG strains contributed to different stimulating abilities in trained immunity and antimicrobial functions93. These studies suggest that TB vaccine-induced trained immunity might provide CoPs. However, a major challenge is the heterogeneity of individuals, such as in the history of heterologous infections, genetic polymorphisms, and microbiota, which might confound the determination of vaccine-induced trained immunity. A better understanding of trained immunity, especially Mtb protection-associated innate immune memory, is warranted to accompany the derivation of an accurate CoP for TB vaccination.

Tissue-resident alveolar macrophages (AMs)

AMs, resident macrophages of the airway lumen, provide innate immunity in the first line of defense against respiratory pathogens and infections. Mtb is transmitted from persons with active pulmonary TB via respiratory droplets, which are inhaled and deposited in the lung alveoli of close contacts. The deposited Mtb is actively engulfed by the resident AMs via phagocytosis. The early productive Mtb infection occurs almost exclusively within resident AMs for up to 10 days after infection94,95, suggesting that the immune responses induced by Mtb-infected AMs are delayed and/or weakened in the early phase of infection and specifically-acquired immune protection is slow to emerge96.

Studies in mice showed that highly transmissible Mtb strains induced rapid IL-1 receptor-dependent AMs migration and, subsequently, Th1 responses, whereas poorly transmissible strains delayed this process and developed a Th17 response, suggesting flexibility in the development of acquired immunity dependent upon the heterogeneity of innate immune AMs97. Prior exposure to mycobacteria, either by BCG vaccination or by a contained Mtb infection, impacted the initial response of AMs against infection through both AM-intrinsic changes and signals that depend on the altered lung environment98. Thus, the plasticity of AMs suggests that early innate responses could be targeted through vaccination to fine-tune the host response against Mtb infection. Recently, Zhou Xing et al. showed that respiratory adenovirus infection could induce immune memory in AMs, and this was associated with increased glycolytic metabolism and enhanced generation of neutrophil chemokines upon restimulation; the response protected against secondary bacterial infection, and the protection was independent of circulating monocytes99. The evidence suggests that AMs might be targeted by novel TB vaccines to impede the development of Mtb infection. Zhou Xing et al. further showed that a recombinant adenovirus-vectored vaccine, AdAg85A, induced an innate immune memory in AMs that contributed to anti-Mtb protection in mice100. Others have shown that the adenovirus-based TB vaccine and BCG vaccination also drove a long-term activation of AMs that restrained the growth of Mtb and disrupted early bacterial dissemination after the Mtb challenge101,102. Together, these studies indicate that differences in responsivity among lung resident AMs influence the outcome of Mtb infection and might contribute to the definition of TB vaccine CoPs.

Other immune cells

In addition to AMs, other innate and adaptive immune cells can also dictate the course of infection after encountering respiratory pathogens. Immune cells such as neutrophils, monocytes, dendritic cells (DCs), natural killer (NK) cells, γδ T cells, regulatory T cells (Treg), regulatory B cells (Breg), innate lymphoid cells (ILCs), and mucosal-associated invariant T cells (MAIT) have also been shown to tune innate immunity and activate antigen-specific adaptive immune responses (Fig. 1). The relevance of these cells to the efficacy of TB vaccines has been well-reviewed elsewhere103,104,105,106,107 and is not a focus of this review. A major hurdle for deriving innate immune cell-based vaccine CoPs is that innate immunity is non-specific. Thus, infection by non-tuberculous pathogens, allergy, or even sterile inflammation would confound the determination of Mtb-induced innate immune responses in naturally complicated conditions prevailing in clinical trials. Thus, although innate immune responses play important roles in host immunity against TB infection, our opinion is that an innate phenotype is unlikely to be used as a definitive CoP of TB vaccines in the coming years, primarily due to their complex, multilayered, and non-specific characteristics.

The available samples For Tb vaccine CoP discovery

The identification of CoPs relies on the knowledge of the immunological processes of the host after infection or vaccination. There are several limitations in seeking this knowledge, such as technical feasibility, financial costs, and ethical requirements. Given such restrictions, whole blood samples are always the main source of biological samples for studying CoPs. Whole blood is one of the most accessible samples in humans/animals. It can be separated into peripheral blood mononuclear cells (PBMCs), plasma, or serum to study circulating cells, soluble cytokines, and antibodies. Recently, the use of blood in mycobacterial growth inhibition assays (MGIA) has provided unbiased measures of ability to control mycobacterial growth in vitro108. These assays offer the opportunity to determine the contributions of multiple components of the immune responses, either in whole blood or cell samples. Thus, MGIA is used to down-select candidate TB vaccines at an early stage of their development109,110,111. In addition to whole blood, cells obtained by bronchoalveolar lavage (BAL) and lung and spleen cells from animal models are also widely used to determine adaptive antigen-specific T-cell immune responses. However, tissue samples are not readily accessible without invasive procedures in human clinical trials. Lymph nodes are another valuable and feasible source of samples to study CoPs, particularly in experimental animals. There is increasing interest in how Mtb is controlled in the lymph node in the NHP model82,112 and the contained Mtb infection murine model61,113 (Fig. 2).

Fig. 2: The available samples from different research models for immunological interrogation in TB vaccine CoP discovery.
figure 2

The samples that are typically obtained from mice, NHPs, and human beings in clinical trials are summarized. Abbreviations: BAL bronchoalveolar lavage, CoPs correlates of protection, TB tuberculosis, PBMCs peripheral blood mononuclear cells, PET-CT positron emission tomography-computed tomography, VOCs volatile organic compounds. Created with bioRender.com.

Full size image

The distribution of immune cells in the lung tissues is crucial, and new methods such as advanced imaging techniques, for example, positron emission tomography-computed tomography (PET-CT), can reflect the protective and pathological changes within the lung tissues and lesions after Mtb infection. Urine samples also offer some advantages over whole blood in terms of convenience of collection and safety. The potential of using urine samples as a source of CoPs in TB is inferred from the use of urine analysis to indicate TB disease progression. Among the potential CoPs, lipoarabinomannan, a cell wall component of Mtb detected in urine, shows promise in detecting patients with disseminated TB among human immunodeficiency virus-infected populations114,115. In addition, invasive sampling, for example, to obtain puncture biopsy samples116 or surgically resected human tissue samples117,118, offers valuable opportunities to study the local immune responses. Moreover, there have been studies using breath-detected volatile organic compounds (VOCs) in the diagnosis of TB119,120 (Fig. 2).

Considered together, analyses of the various types of biological samples could build a picture of the host immune responses against Mtb infection observed from different perspectives; a comprehensive analysis of such samples is needed to better characterize vaccine-induced protective responses.

The major hurdles of Tb vaccine CoPs identification

The quest for definitive CoPs for TB vaccines has encountered multiple obstacles. Firstly, the slow and intermittent growth characteristics of Mtb results in different disease states of TB infection, which hinders the definition of protective immunity against TB. According to the WHO’s annual TB reports, it is estimated that one-quarter of the world’s population has been infected with Mtb, and there is a 5–10% lifetime risk of developing active TB diseases1. Although individuals with asymptomatic latent TB infection (LTBI) fail to completely eradicate invading Mtb, an immune system that is persistently or intermittently responding against Mtb antigens might contribute to the absence of clinically manifest TB disease progression. Thus, immune profiles induced by latency-associated antigens might guide finding CoPs against LTBI. However, the recent recognition of the “TB resister” population has added a layer of complexity to defining protective immunity; this population might induce anti-TB immunity that is different from the typical CD4 Th1 responses. Thus, the search for TB CoPs should comprehensively consider the differences in protection against primary infection, persistent latent infection, reactivation, progression from latent infection to active disease, and reinfection. Recently, considering the lack of consistency in conceptual and diagnostic definitions of different TB states, with definitions that failed to reflect the spectrum of TB diseases, an International Consensus for Early TB classification framework was established to comprehensively consider macroscopic pathology, infectiousness, symptoms, and signs121. The new international consensus classification and novel diagnostics might solve this problem, improve cross-study data comparisons, and assist with future study design. Secondly, the protective recall antigen-specific immune responses induced by vaccines might not be identical to the protective immunity induced by natural infection. Therefore, a reliable CoP induced by natural Mtb infection might not be suitable to be used as a CoP induced by TB vaccination. Although cytokines are necessary for controlling intracellular infection with agents such as Mtb, excessive cytokine production induced by a vaccine might also cause cell necrosis and tissue damage. Thus, the optimal “cut-off” value for natural and vaccine-induced immune responses might differ. A phase III randomized controlled TB vaccine clinical trial is needed to solve this problem. Fortunately, the clinical trials of the M72/AS01E vaccine provide valuable guidance to finding CoPs by profiling vaccine-induced protective immunity in human beings, which will be discussed in the next section. Thirdly, vector-specific immune responses induced by different types of recombinant-vectored vaccines might confound the identification of CoPs. Subunit protein vaccines and nucleic acid vaccines, such as mRNA vaccines, could solve this problem. Fourthly, the biomarker dynamics and time of sampling will further influence the outcome of CoPs prediction. Thus, a rigorous and comprehensive study design is needed to precisely define the immune CoPs. Finally, the host genetics, such as species differences and polymorphisms in genes encoding proteins involved in intercellular cytokine signaling, are other confounding factors in identifying CoPs. Clinical trials that enroll large numbers of participants in different regions could solve this problem, but such a trial is expected to be prohibitively expensive. Alternatively, genetically diverse mice strains could represent an appropriate range of genotypes. This, too, will be discussed in the next section.

The opportunities for identification of Tb vaccine CoPs

In principle, key opportunities for finding vaccine CoPs arise when the vaccine induces protection in some but not all vaccines. Then, comparisons of the two groups, protected vs unprotected vaccines, could provide a solid basis for the definition of a robust CoP. Recently, several TB vaccines and vaccination strategies, such as the M72/AS01E protein subunit vaccine and BCG i.v. inoculation, have shown promise in providing substantial immune protection or even sterilizing immunity against Mtb infection in clinical trials and pre-clinical models. Moreover, new experimental models such as the controlled human infection model and the ultra-low-dose infection murine model, are available for clinical and pre-clinical vaccine evaluation, and the discovery of the TB resisters population in close contacts of active TB patients gives another opportunity. Conventional high-throughput technologies and novel state-of-the-art technological approaches such as omics and artificial intelligence are increasingly being adopted and will further facilitate the discovery of TB vaccine CoPs. These new opportunities suggest that the identification of reliable CoPs to determine the efficacy of TB vaccine candidates might be feasible shortly (Fig. 3).

Fig. 3: The opportunities of TB vaccine CoPs identification.
figure 3

The available experimental approach, research models, and resources that might facilitate TB vaccine CoPs determination are summarized based on current knowledge. Abbreviations: BCG Bacille Calmette–Guérin, CC collaborative cross, CoPs correlates of protection, DO diversity outbred, i.v. intravenous immunization, TB tuberculosis, Mtb Mycobacterium tuberculosis, ULD ultra-low-dose. Created with bioRender.com.

Full size image

Controlled human infection model

A controlled human infection model can substantially accelerate the development of new vaccines since the size of volunteer groups can be significantly smaller, the experiment duration can be shorter, and the expenses can be far less compared with traditional clinical trials. In a controlled human infection model study, a well-characterized strain of an infectious agent is administered at a controlled dose and by a specific route to carefully selected volunteers; volunteers are closely monitored for evidence of infection under medical supervision for diagnosis and management of early symptoms of diseases122. This model is a great tool for vaccine development and has contributed to the development of vaccines for other disease agents, including influenza, shigella, enteric fever, malaria, campylobacter, cholera, and respiratory syncytial virus122. It is particularly attractive for TB vaccine assessment since the low rates of natural exposure to infection, the slow growth characteristic of Mtb, and the slow speed of TB disease progression render the conventional assessments of protective efficacy extraordinarily time-consuming and costly. Thus, the controlled human infection model of mycobacterial challenge could substantially accelerate the development of TB vaccines towards clinical trials.

Methods of using either BCG or a conditionally replicating Mtb strain are being developed in studies of controlled human mycobacterial infection, and intradermal or respiratory/mucosal routes of infection have been used. Initially, because deliberately infecting healthy volunteers with virulent Mtb is not ethical in the absence of a reliable method of anti-TB treatment that could completely eradicate infection, the group of Dr. Helen McShane, in 2012, used an attenuated M. bovis BCG vaccine strain that does not cause disease or latency in healthy humans; intradermal “infection” allowed straightforward quantification of mycobacteria from punch biopsies by PCR and bacterial culture123. This model was then used to evaluate TB vaccines in human beings124,125. Subsequently, the intradermal route was changed into the aerosol route to more closely mimic the natural route of Mtb infection and was used to evaluate TB vaccines in several clinical trials (NCT03912207 and NCT04777721)126. Live BCG has also been directly instilled into the lung lobes via bronchoscopy and was well-tolerated in healthy volunteers127. In addition to optimization of the vaccination route, other modifications, such as the incorporation of an auto-luminescent fluorophore reporter gene or detection of exhaled volatile compounds by mass spectrometry, can be proposed to avoid invasive sampling for mycobacterial recovery and quantification128,129. Currently, a conditionally replicating Mtb strain with a genetically inserted suicide mechanism is being developed for controlled infection in man. In this approach, the survival of Mtb is dependent on two exogenous compounds so that the invading bacteria are killed if the compounds are withdrawn130. The proof-of-concept of using the conditionally replicating Mtb strains has been obtained in mice and NHPs models; next, the strains will be reconstructed to meet, at a minimum, compliance with good manufacturing practice before usage in a human challenge study131.

These controlled human infection models could not only be used to provide early proof-of-concept of the protective efficacy of TB vaccine candidates but could also sharply facilitate the interrogation of immunity for the identification and validation of CoPs at defined time points after infection.

Ultra-low-dose aerosol infection and genetically diverse mouse model

In general, in the pre-clinical evaluation of TB vaccine protective efficacy, mice are aerosol-infected with a “conventional” (50–200 colony-forming units [CFU]) or a “high” (>250 CFU) dose of Mtb; however, in natural infection, most people are infected by one or two bacilli that successfully implant after inhalation of low doses132. This divergence between laboratory models and clinical scenarios hinders the discovery of CoPs. To solve this problem, a model of ultra-low-dose (ULD) aerosol Mtb infection in mice was established in 2021, in which infection was initiated by only 1–3 CFU133 and induced well-circumscribed granulomas that shared features with human granulomas and more closely replicated human TB disease. The model revealed a blood transcriptional signature that correlated with the wide range of lung bacterial burdens and might predict Mtb infection outcomes in both NHPs and human infections. It has been used to evaluate the protective efficacy of several TB vaccines134,135,136 and showed that the elevated serum levels of IFN-γ, IL-17A, IL-6, CCL2, CXCL1, and CXCL2 were associated with vaccine-induced protection134. The observation that vaccine-induced Th1/Th17 cytokines secretion enhancement correlates with protection is consistent with murine models that used conventional infection doses; however, whether an excessive production of these cytokines leads to exacerbated lung lesions, as suggested by conventional infection doses16,38, remains to be defined in ULD models. The model provides a unique experimental system for studying the host immunology and pathogenesis of Mtb infection, which will accelerate the discovery of a reliable TB vaccine CoP.

Genetic diversity plays a significant role in individual variation in vaccine-induced immune responses. However, classical inbred mouse strains, in which all individuals are genetically identical, were routinely used in vaccine studies, including those for TB. Collaborative cross (CC) and diversity outbred (DO) mouse strains are new mammalian animal resources that more accurately represent the genetic and phenotypic heterogeneity observed in naturally outbred populations. They were recently used to evaluate TB vaccines. It was shown that CC strains varied dramatically in primary susceptibility to Mtb infection and produced qualitatively distinct immune states137. BCG protected only half of the CC strains against TB, and this was correlated with heterogeneity in several inflammatory mediators in the lung microenvironment138. Similarly, DO mice were used to study immune correlates of TB disease and risk translation across species139; BCG-vaccinated DO mouse populations exhibited a wide range of disease outcomes. Sixteen chromosomal loci were found to be associated with complex infection traits and cytokine production140,141. Thus, these genetically diverse murine models can be used to facilitate precise definitions of vaccine-induced CoPs during the development of vaccines that protect a range of genotypes that mimic the heterogeneity in human populations.

The clinical trials of the M72/AS01E vaccine

The novel TB vaccine M72/AS01E is composed of the M72 antigen, which is a recombinant fusion protein based on two immunodominant antigens Mtb32A (Rv0125/PepA) and Mtb39A (Rv1196/PPE18), and a liposome-based adjuvant AS01E, which contains monophosphoryl lipid A and the saponin QS21142. In 2019, a proof-of-concept Phase IIb trial showed that the M72/AS01E vaccine provided 49.7% (95% confidence interval [CI], 2.1-74.2) protection against microbiologically-confirmed pulmonary TB disease in Mtb-infected IGRA-positive adults who did not have active TB143. The T cell and antibody responses were evaluated in according-to-protocol subgroups totaling 244 participants (120 in the vaccine group and 124 in the placebo group) to test the immunogenicity. Among participants in the M72/AS01E group, the frequencies of multifunctional M72-specific CD4+ T cells (IFN-γ/IL-2/TNF-α-secreting cells) were substantially increased after immunization, and increases persisted for at least 36 months, with no evidence of waning. Robust M72-specific IgG responses were also observed. In light of this finding, this cohort was believed to offer a unique opportunity to identify CoPs against TB disease progression.

A “TB Immune Correlates Program” was launched for TB vaccine M72/AS01E-induced CoPs in 2020144, which aims to provide the strategy and governance framework to facilitate finding CoPs through an international consortium of investigators. These projects were planned to (1) define candidate biomarkers and assays to be performed for the discovery of TB vaccine CoPs; (2) develop the statistical framework to evaluate the pre-specified hypotheses; (3) pursue additional hypothesis-generating exploratory efforts; (4) ensure rigorous, efficient and streamlined usage of the cryopreserved samples from these two trials144. The priority is informed by existing knowledge and recent findings from animal studies and clinical trials, including antigen-specific T cell responses (functional/activated/memory profiles, recognized epitopes, TCR repertoire/gene expression, and secretion of immunomodulatory factors), humoral responses (the titers/sub-classes/avidity/specificity of antibodies, Fc-mediated function, and mucosal antibody responses), donor-unrestricted T cells, innate immunity, etc.144.

The sterilizing immunity induced by RhCMV/TB and BCG i.v. in the NHP model

Recently, two promising pre-clinical studies showed that TB vaccines could induce sterilizing immunity against Mtb infection in the NHPs model. Firstly, in 2018, RhCMV/TB, a cytomegalovirus-vectored vaccine encoding 9 Mtb antigens (ESAT-6, Ag85A, Ag85B, Rv3407, Rv1733, Rv2626, Rpf A, Rpf C, and Rpf D), reduced the overall extent of pulmonary and extra-pulmonary Mtb infection and disease by 68% compared to unvaccinated controls in rhesus macaques145. Notably, 14 out of 34 RhCMV/TB-vaccinated rhesus macaques showed no detectable TB disease by computed tomography (CT) scan and necropsy, compared with 0 out of 17 unvaccinated controls. This is the first demonstration of sterilizing immunity against highly pathogenic Mtb infection in the NHP model. Analysis showed that multifunctional CD4+ and CD8+ T cell immune responses were induced by RhCMV/TB145.

In 2020, a study compared different vaccination routes of BCG in the NHP model and found that BCG i.v. immunization protected 9 out of 10 rhesus macaques against Mtb challenge, which contrasted to limited protection afforded by other routes at the same dose; notably, 6 out of 9 protected macaques had no detectable infection, as determined by PET-CT, mycobacterial growth, pathology, and granuloma formation82. This encouraging result was confirmed by another study in 2023, which showed i.v. BCG protected 9 out of 12 SIV-infected animals; no culturable bacteria were detectable in tissues146. Ironically, the high level of protection made it challenging to detect cellular or humoral CoPs since 90% of animals were protected. In several studies, the i.v. BCG vaccination was “optimized” to achieve a lower (~50%) protective efficacy and facilitate correlates analysis. Multivariate analysis incorporating longitudinal cellular and humoral immune parameters of cells in BAL showed that, at 2–12 weeks post-i.v. BCG vaccination, the numbers of NK cells and CD4+ T cells producing TNF-α/IFN-γ or TNF-α/IL-17 (Th1/Th17 cells) were correlated with protection147. In another study, single-cell RNA sequencing was used to transcriptionally profile the responses to PPD among BAL cells that were collected 13/25 weeks, or 14–15 weeks (2 batches) after vaccination. High-dose i.v. BCG vaccination drove a local influx of AMs into the airways, and the cells exhibited a basal activation phenotype up to 25 weeks post-immunization; in addition, after PPD stimulation, intercellular immune signaling pathways between myeloid and T cells were enhanced, with robust Th1/Th17 transcriptional phenotypes in T cells, and augmented IFN-γ response in AMs148. The humoral immune responses in the BAL 4 weeks after immunization were correlated with the Mtb burden post-challenge149. Recently, it was shown that depletion of CD4+ T cells and CD8α-expressing lymphocytes (both innate and adaptive), but not CD8αβ+ T cells, before Mtb challenge of BCG i.v.-vaccinated NHPs resulted in the loss of protection150. These studies indicate that the cross-talk between myeloid cells and lymphocytes in the local infection sites is crucial and long-lasting during Mtb infection, and corresponding parameters might be used as CoPs.

Peripheral blood samples are far more convenient than BAL samples for COP analysis. In analyses of blood immune features, the early innate transcriptional response on day 2 after BCG i.v. immunization, especially an innate response module enriched for type I interferon and RIG-I-like receptor signaling pathways, was highly correlated with antigen-responsive lung CD4+ T cells at week eight and with Mtb burden following challenge in the NHPs model151. In addition, the antibody responses induced by i.v. BCG was also correlated with protection against Mtb infection149. These studies suggest that analysis of peripheral blood signatures early post-vaccination holds promise for predicting vaccine-induced anti-TB protection.

TB resisters and latent infection

Resisters (people) and persisters (bacteria) deserve further consideration. These appear at first to be relatively defined states compared to the spectrum of immunological and pathological conditions found in tuberculosis disease. After initial Mtb exposure in human beings, 30% will develop a primary infection, among which 85–90% of immune-competent people develop asymptomatic LTBI, in which there is a state of persistent immune response against Mtb stimulation (IGRA or TST positive) but the bacillary replication is controlled without clinical evidence of active TB disease152,153. Once asymptomatic LTBI is established after initial exposure, there is only a 5–10% lifetime risk of developing into active TB disease2,3. Thus, the majority of the LTBI population is “resistant” to active TB diseases. However, the bacteria are not fully dormant and undergo periods of growth, so the immune response to the antigens may fluctuate between suppressive and facilitatory states throughout the lengthy latency period, rendering the identification of CoPs a challenge. We and others have profiled several latency-associated antigens and designed vaccines containing corresponding antigens54,154,155,156,157,158, but analysis of response to these antigens has not yet led to the identification of CoPs.

The term “TB resisters” was defined as individuals who were heavily and repeatedly exposed to Mtb but resistant to active TB disease and consistently tested negative for reactivity in the IGRA and TST assays159. The basis of IGRA and TST negativity is not established but may be due to minimal bacterial multiplication that the antigen dose does not reach a threshold for those responses. A cohort study in 2019 showed that TB resisters possessed IgM and class-switched IgG antibody responses and non-IFN-γ T cell responses to Mtb-specific antigens56. Thus, TB resisters, at least some of them, mounted those host immune responses after Mtb exposure. In 2020, we also found that the T cell activation status marker CD69 was associated with Mtb infection and may have the potential to distinguish LTBI and TB resisters160. More recently, several adaptive immune states were reported to be elevated in TB resisters, such as higher levels of circulating CD8+ glucose monomycolate+ granzyme-B+ T cells161, antigen-specific antibody Fc galactosylation/sialylation162, anti-LpqH human monoclonal antibody163, and MR1-restricted T cells164, as well as innate immune responses, such as enrichment of the TNF signaling gene set165, higher free fatty acid transcriptional responses166, lipid accumulation in monocytes167, and histone deacetylase 6 in monocyte-derived macrophages168. Moreover, pantothenate (vitamin B5) was significantly elevated in blood cells169, and some non-coding RNAs were enriched in PBMCs from TB resisters170. Whether any of these features, separately or together, will correlate with the elimination of the infection or yield a useful CoP remains to be seen.

Omics and artificial intelligence

Systems vaccinology leverages high-throughput omics technologies, such as transcriptomics, metabolomics, epigenomics, and mass cytometry, coupled with computational approaches to construct a global map of the complex processes that mediate the protective immune response to vaccination. These technologies provide the opportunity to interrogate biological models at a comprehensive systems-level instead of empirical approaches that characterize a few genes or proteins in an experiment. With multilayered omics data generated in the case-cohort studies, an integrative multivariate framework that draws together and analyzes data from the different technologies should provide key insights. The cross-talk between innate and adaptive immune responses before and after immunization and/or infection might conceal new candidate biomarkers or verify the hypothesized CoPs. The advent of sophisticated artificial intelligence based on deep learning and machine learning algorithms marks a seminal point in multimodal omics data integration. It will create powerful tools for determining novel correlates of vaccine protection171.

Conclusions

Despite Mtb co-evolving with human beings for centuries and being intensely studied for decades, the CoPs, especially vaccine-induced CoPs, against Mtb infection and TB disease have yet to be fully defined. Their discovery has been hindered by an incomplete understanding of host protective immunity against Mtb and uncertainty as to the predictive value of animal models. The finding of CoPs is particularly important in TB vaccine development since it is a slow-growing bacterium, and it takes months to years for a Mtb-exposed person to progress into active TB disease and reveal the effect of vaccination. This incurs tremendous costs in clinical trials to measure differences in clinical endpoints. However, the positive results of the M72/AS01E vaccine recently evaluated in clinical trials not only strengthen our confidence in designing an effective novel TB vaccine but also provide new opportunities to find CoPs. Moreover, new models, such as the controlled human Mtb infection model and new state-of-the-art technologies, will further facilitate the findings of CoPs. The sterilizing immunity induced by i.v. BCG and CMV-vectored TB vaccines in the NHPs model may provide a crucial benchmark to understand the immune correlates and mechanisms of sterilizing protection against TB. Thus, there is renewed optimism that a reliable CoP or a tailored combination of independent biomarkers might be realizable. When realized, a reliable CoP would enable the down-selection of vaccines that enter the clinical pipelines, accelerate the process of clinical trials, and become the game-changer of TB vaccine development.