On 20 January 2026, the Stop TB Partnership Working Group on New TB Vaccines (WGNV) convened a virtual, discussion-driven workshop focused on fit-for-purpose animal models for tuberculosis (TB). The aim of the workshop was to critically examine how well existing animal models reflect the human spectrum of Mycobacterium tuberculosis (Mtb) infection and TB disease, and how these models can be leveraged to generate human-relevant outcomes, including improved understanding of resistance to infection and containment of disease, and linking findings to human trials in support of TB vaccine development.
The workshop opened with welcoming and introductory remarks from David Lewinsohn, Chair of the WGNV, who emphasized the importance of carefully selecting preclinical models to answer clearly defined biological questions and translate findings from animal to human contexts. In this workshop summary, we highlight key points describing what each mammalian model is best suited to answer as a central theme to advancing human TB vaccine candidates.
Panel 1: Resistance to Mtb Infection
The first panel, chaired by Katrin Eichelberg (National Institute for Allergy and Infectious Diseases) and Monika Looney (South African Tuberculosis Vaccine Initiative), explored how different animal models are used to study resistance to Mtb infection and early disease containment, with a focus on immunological relevance, genetic diversity, and translatability to humans.
The first presentation by P. Ling Lin (University of Pittsburgh School of Medicine), highlighted the strengths of the non-human primate (NHP) model, discussing its unique ability to capture a wide spectrum of outcomes from Mtb infection that also appear in humans. Lin presented studies demonstrating that NHPs can exhibit pathology and immunological responses consistent with latency, asymptomatic infection, and active, symptomatic disease. Lin also described how the NHP model key features of human TB, including genetic variability, environmental exposure to non-tuberculous mycobacteria (NTM), heterogeneous granuloma formation, and clinically relevant signs such as weight loss, appetite changes, and cough. The use of PET-CT imaging to assess disease burden and dissemination was also discussed as a powerful surrogate for disease progression and outcomes in NHPs following infection. Signatures of Mtb infection in the blood and bronchoalveolar lavage can also be found across species with NHP, human, and even murine studies (PMID: 27837110, PMID: 31996462, PMID: 27509488). Lin emphasized how vaccine studies in NHPs, including work on intravenous BCG (ivBCG), have enabled the identification of potential correlates of protection (PMID: 31894150, PMID: 37267955, PMID: 37390827). Studies from the NHP model have also revealed immune factors that may be permissive to bacterial dissemination, which are unique to different physiological compartments. Furthermore, the breath and duration of antigen exposure may be important in vaccine-induced protection. These concepts used in the NHP model have the potential to inform strategies for next-generation TB vaccine development.
Shabaana Khader (The University of Chicago) discussed new approaches to the mouse model, outlining how different mouse strains provide complementary insights into Mtb infection and vaccine responses. Khader highlighted the utility of classical C57BL/6 mice for mechanistic and vaccine screening studies, although this model is generally resistant to Mtb infection and typically form non-necrotic granulomas, which is not fully representative of pathology observed in humans and NHPs. This is in contrast with more susceptible mouse strains, such as the C3HeB/FeJ mouse, which develops necrotic and cavitary lung pathology more consistent with human TB. However, one major caveat to these traditional mouse models is the inability to account for the interplay between genetic diversity of the host and immune response. To address this, Khader introduced the Collaborative Cross (CC) mouse resource, which incorporates extensive genetic diversity to better model variable host responses to infection and vaccination (PMID: 27651361, PMID: 31772048). The CC mouse model is generated from inbred and wild-derived mouse strains that undergo eight-way breeding to generate essentially limitless diversity in outbred mice. Khader also introduced the Diversity Outbred (DO) mouse resource (PMID: 22345611), which is a newly developed mouse population derived from progenitor lines of the CC mouse. As such, the DO mouse population segregates the same allelic variants as the CC mouse but embeds these in a distinct population architecture in which each animal has a high degree of heterozygosity and carries a unique combination of alleles. Phenotypic diversity in these mice is often divergent from phenotypes seen in the founder strains of the CC mice. Data from CC and DO studies demonstrated how these models generate a spectrum of phenotypes ranging from resistant to susceptible strains, which may more accurately recapitulate the influence of genetic diversity on disease progression and immune responses in humans. Khader also highlighted CC and DO mouse data demonstrating how host genetics fundamentally shape susceptibility, resistance, and vaccine-induced protection, and how shared immune pathways across mice, NHPs, and humans may point toward broadly relevant correlates of protection (PMID: 37200108, PMID: 31996462, PMID: 41556657).
Björn Corleis (Friedrich-Loeffler-Institut) provided an overview of other mammalian animal models, including guinea pigs, rabbits, pigs, goats, sheep, cattle, ferrets, and hamsters. Corleis discussed how these models reflect different aspects of human TB, such as caseous granuloma necrosis, cavitary disease, latent infection, and transmission dynamics. While acknowledging that many of these models have limited experimental toolboxes (barring pigs), Corleis highlighted their value as systems for studying interspecies similarities and universal protective immune mechanisms. In particular, pigs, ferrets and guinea pigs were noted for their relevance to transmission studies, while cattle were underscored as an interesting model of resistance to Mtb infection (the mechanisms for which remain undescribed, representing an important knowledge gap for the field). Corleis also presented data describing multi-interspecies assays that revealed translatability of Th1-mediated effector functions across models (PMID: 41648316), which can also serve as tools to investigate universal protective mechanisms.
Summary. Across panel 1, speakers emphasized that resistance to Mtb infection is shaped by complex host-pathogen interactions and immune responses operating across multiple anatomical compartments, including the airway, lung tissue, and granulomas. A key takeaway was that no single model can fully recapitulate human TB; instead, strategically combining models, and selecting models for a specific biological or translational questions, offers the most robust path forward for understanding pathology, identifying correlates of protection, and advancing TB vaccine development.
Panel 2: Containment and/or Elimination of Mtb
The second panel, chaired by Alison Kraigsley (Gates Foundation) and Elly van Riet (TuBerculosis Vaccine Initiative, TBVI), focused on animal models used to study the containment and elimination of Mtb, with particular emphasis on how vaccine-induced immunity can prevent establishment, dissemination, or persistence of infection.
Frank Verreck (Biomedical Primate Research Centre, BPRC) spoke to the bottlenecks (ethics, regulation, costs, and capacity) and knowledge gaps (biomarkers and mechanisms of protection) that underscore the continued relevance and importance of animal models in TB vaccine development. Verreck focuses on NHP Macaca spp. as a model that captures a wide range of human-relevant pathologies and immune responses. Verreck highlighted that intrinsic pathogen factors (strain, virulence, drug resistance), intrinsic host factors (genetic susceptibility), and extrinsic host factors (pre-exposure, coinfection, comorbidity, malnutrition, ageing, immunosuppression) influence TB susceptibility and vaccine responses and therefore, should be accounted for as variables in a study. Notably, the validity of NHP models in vaccine studies is illustrated by a 50% vaccine efficacy where highly susceptible Rhesus macaques showed higher percent survival as well as reduced pathology and more stable clinical chemistry and haematology indicators over a 1-year follow up in animals vaccinated with BCG (unpublished). Disease burden, dissemination, immunological and metabolic changes are also affected by route of administration.
Verreck postulated that subunit vaccine candidates may, to date, fail to outperform BCG not only due to the makeup of the candidate itself, but also because of how infection is modeled in NHPs. Verreck described the repeated limiting dose (RLD) model, in which animals are exposed to limiting amounts of Mtb multiple times, rather than a single high- or low-dose challenge. This approach allows for more stochastic Mtb exposure and infection and more accurately reflect the real-world context of human infection. Importantly, and while pulmonary mucosal BCG administration shows superior suppression of TB pathology in either of the exposure conditions, the RLD model may enable vaccine-mediated prevention of infection in NHPs (PMID: 32286384), an outcome that is difficult to assess using traditional challenge models.
Kevin Urdahl (Seattle Children’s Research Institute) introduced the concept of prevention of detectable infection as an important and distinct metric of vaccine-induced protection against TB disease. Urdahl contrasted conventional mouse challenge models (which rely on relatively high-dose aerosol infection, usually 50-100 CFU, and result in progressive disease despite vaccination), with an ultra-low dose (ULD) infection model using 1-3 CFU of Mtb. This ULD model more closely reflects physiological exposure and results in relatively well-contained infection with transient bacteremia (PMID: 33142108). Using the ULD model, Urdahl described three complementary measures of vaccine-mediated protection: reduction in overall lung bacterial burden, prevention of dissemination to the contralateral lung, and prevention of detectable infection (defined as the absence of culturable bacteria after plating the entire lung) (PMID: 38011264). Importantly, these outcomes were shown to be mechanistically distinct: depletion studies revealed that CD8 T cells played a critical role in preventing detectable infection, a protective effect that was only evident under ULD challenge conditions, while CD4 T cells were more important for controlling bacterial burden and limiting dissemination. These findings highlight complementary, yet distinct, roles for CD4 and CD8 T cell responses in TB immunity. The discussion further emphasized that the ULD model can measure prevention of detectable infection, which does not necessarily correlate with reduced lung bacterial burden. Certain vaccine candidates were able to prevent detectable infection without significantly lowering bacterial load compared with BCG, underscoring the idea that different immune mechanisms may underlie these outcomes. While prevention of infection cannot be directly measured in humans, the panel noted that most TB disease in endemic settings likely arises from new infection, making this a potentially important and informative metric for preclinical vaccine evaluation.
Summary: Finally, considerations around model selection were discussed, including the trade-offs between genetic diversity and experimental feasibility. While genetically diverse models can better reflect variation seen in humans, they are often more complex and costly to use. One practical approach is to first assess immunogenicity across multiple mouse strains before conducting protection studies in defined F1 crosses. Additionally, redesigning how animals are infected with Mtb to better represent the stochastic, lower dose of real-world exposure in humans reinforces the importance of thoughtful model design aligned with specific research questions.
Concluding Reflections
Across both panels, speakers underscored that animal models remain essential tools in TB research and pre-clinical vaccine development. They provide controlled systems in which mechanisms and correlates of protection, containment, and disease progression can be investigated in ways that are not possible in humans. However, a clear consensus emerged: no single animal model can fully capture the heterogeneity of human TB. Immune responses, granuloma architecture, clinical manifestations, disease outcomes and vaccine responses vary widely in humans, and each model reflects only part of this spectrum. A recurring theme throughout the workshop was the importance of selecting models that are fit-for-purpose to provide valuable insights and links to human trials. Researchers must carefully consider both intrinsic and extrinsic factors, including host genetics, immune architecture, pathogen strain, infectious dose, route of exposure, and environmental context. Experimental design should be guided by a clearly defined biological question, rather than relying on conventional or historical precedents. Aligning model selection with the specific (human-relevant) outcome of interest, whether prevention of infection, containment, dissemination, or disease control, is critical for meaningful, translational progress. Speakers also emphasized that ethical considerations must remain central to all animal research. Responsible use of animal models requires careful justification, refinement of experimental approaches, and ongoing efforts to maximize scientific value while minimizing harm.
The discussion also highlighted several important gaps in the field. Besides a single model that fully recapitulates human TB disease and immune responses, we also have limited experimental “toolboxes” for several less conventional but biologically informative animal models. There is also a need for broader and more diverse pipelines of pre-clinical TB vaccine candidates, which is essential for maintaining and developing the current pipeline of new TB vaccines in clinical development. On that note, participants also emphasized the importance of sustained, strategic funding and investment in pre-clinical TB vaccine research involving fit-for-purpose animal models. Robust pre-clinical testing and screening are essential to ensure a strong and diverse vaccine pipeline, enabling identification of candidates with the greatest likelihood of success in clinical development.
NIAID does not endorse nor should endorsement be implied of the Working Group on New TB Vaccines, its products or services.
