The experimental murine TB model has elucidated host fate upon natural infection which is achieved by aerosol exposure. It has unveiled the complexity of the kinetics of the infectious process in great detail. This model has also enabled mutual integration of host and pathogen traits in experimental studies. However, mice do not fully recapitulate TB pathology. Granuloma liquefaction, cavitation and fibrosis remain undetected in Mtb-infected mice, and hence murine TB is an imperfect disease model. This model allows to comprehensively study the immune responses to Mtb. Pulmonary anatomy and immune mechanisms in mice have a great degree of similarity to humans (16) which make them ideally suited for studying immune dynamics within tissues and for testing vaccine efficacy. Although mice are not natural hosts for Mtb and are generally tolerant to TB (17), they have proven instrumental for understanding some disease mechanisms.

Susceptibility of laboratory mice to Mtb depends on both host and bacterial features. Among the host factors, mouse genetics, age, sex, and immune status control TB outcome (18–21). The route of infection (22), inoculum size and bacterial genetics, e.g. Mtb lineages and virulence factors, impact as well on the course of TB. For instance, aerogenic exposure to the East/Asian Beijing strain HN878, in contrast to infection with the reference Euro-American strain H37Rv, triggers heightened susceptibility in C57BL/6 mice (23) and granulomatous lesions resembling human-like pathology (24). Various Mtb strains differ in propensity to infect myeloid cells (25), disseminate (26) or trigger inflammation (27). Mtb attenuation by deletion of the PhoP regulon (28), or deletion of entire virulence coding genomic regions, e.g. region of difference 1 (RD1) (29), or pathogenicity factors such as the early secreted antigenic target 6-kDa protein (ESAT-6) (30, 31) cause reduced pathology. Such studies have also unveiled a dominant role of nitric oxide in antimycobacterial immunity in the murine host, unlike in humans (32). In addition, they confirmed the relevance of subcellular pathogenicity events, such as cytosolic translocation of the bacilli, for Mtb pathogenicity during lung infection.

Research on TB immunology and pathology heavily relies on inbred and knock-out (KO) mouse strains and has recently been enriched by the addition of the collaborative cross (CC) lines (33, 34), diversity outbred (DO) (35–38) as well as humanized mice (39, 40). Whereas inbred animals and respective KO lines have permitted targeted characterization of host factors essential for the susceptibility to disease, genetic diversity has contributed to unbiased identification of host susceptibility or resistance traits. Of note, mice can be infected by various MTBC bacteria, and usage of transgenic knock-in models has facilitated analysis of particular cell types or molecules during infection (e.g. fluorescent-tagged reporters), or enabled targeted cell deletions (e.g. Cre-lox system). The transgenic mice used in TB research offer opportunities to decipher disease pathogenesis. They do not confer essential cell entry host factors to mycobacteria, a situation common for COVID-19 murine models where infection is usually abortive in wild type animals (see section 2.1.2).

TB outcome differs in various inbred mice which are classified as TB-resistant and TB-susceptible based on the time to death or bacterial outgrowth. TB-resistant mice, including C57BL/6 and Balb/c strains, control aerosol infection with relatively high doses (e.g. 500 colony forming units, CFU) of bacteria, do not develop typical granulomas and succumb rather due to aging. As such, they have been proposed as potential latency TB models (41). Studies in C57BL/6 mice have been critical for defining kinetics of the immune events post exposure, requirements for priming of adaptive immunity as well as kinetics and plasticity of T-cell responses in TB. Very early in infection alveolar macrophages support Mtb replication, as demonstrated in depletion studies (42), likely due to their metabolic imprinting towards oxidative phosphorylation (43) and anti-oxidant features (44). Elegant fate-mapping studies have highlighted that alveolar macrophages translocate into the lung interstitium (45) and Mtb gains access to less permissive glycolytic lung macrophages (43). Subcellular virulence factors, notably mycobacterial ESX1 secretion system (31), and host determinants of susceptibility, for instance phagosomal proteolysis (46) have been unveiled also in vivo in the context of macrophage plasticity in TB in C57BL/6 mice. Cell types conferring an Mtb-permissive environment, including lung monocyte-derived macrophages and dendritic cells (47), and neutrophils (48, 49), have been defined also in this murine model. Kinetics of T-cell responses (50) and the impact of their localization on disease outcome (51, 52) have been established in C57BL/6 mice receiving transgenic cells expressing an Mtb-specific T-cell receptor (TCR). T-cell depletion alone or combined with adoptive transfer of antigen-specific T-cells has indicated an essential role of CD4⁺ and CD8⁺ lymphocytes for TB control (53, 54) and highlighted Mtb escape strategies related to dominant epitopes and misplaced T-cells (55). Thus, C57BL/6 mice have substantially contributed to the delineation of immune events in primary TB. The major caveat of the C57BL/6 model lies in the lack of human-like pathology. Of note, application of ultra-low dose infection (ULD) (56) may render these mice amenable for pathology studies (57). Upon ULD, mice develop single, structured lesions upon inhalation of 1-3 Mtb CFU of the laboratory strain H37Rv. Organized granulomas have also been reported in C57BL/6 mice challenged with low-dose hypervirulent HN878 Mtb (24, 58). These murine models mirror, to some extent, human TB lesions, have organized granuloma-like lesions which contain foamy macrophages, develop central necrosis, yet still miss certain cellular components such as multinucleated giant cells and do not show fibrosis and calcification. A further utility of C57BL/6 mice has recently been reported. Intra-dermal Mtb infection resulted in localized spread of Mtb (59, 60), unlike systemic dissemination seen upon aerosol or intravenous challenge (22), and thus may represent a refined experimental model for latent TB infection (LTBI). Besides mechanistic understanding of immunity and pathology, C57BL/6 and Balb/c strains are also gold standards for chemotherapy studies and TB vaccine development.

TB-susceptible inbred mice encompass the C3HeB/FeJ, 129Sv (129S2/SvPas), I/St and DBA/2 mouse strains. C3HeB/FeJ mice are best suited for investigating pathology. Exposure of these mice to Mtb leads to the formation of well-formed, necrotic granulomas showing hypoxic regions (61), with liquefaction observable particularly upon i.v. challenge (62). Necrosis of Mtb-infected macrophages is controlled by the sst1 locus (63) and has been linked to the intracellular pathogen resistance 1 (Ipr1) gene (64). A role for neutrophils in susceptibly to TB in C3HeB/FeJ by controlling lesion progression (36, 62), likely via type I interferon (IFN-I)-driven NETosis (65), has been reported. The similarities to lesion progression in humans (66–68) make this model useful for pathology and immunopathogenesis studies. DBA/2 mice show a fast TB course with bronchogenic dissemination (69). Their susceptibility is driven by neutrophils (70) and limited accumulation of regulatory T-cells (Treg) within infected tissue (71). This phenotype is shared by the TB-susceptible inbred strain I/St (72, 73), which unlike their A/Sn counterparts cannot control TB. 129Sv (129S2) mice succumb early during TB (74) with extensive lung damage. Their susceptibility to TB is uncoupled from natural resistance-associated macrophage protein 1 (Nramp1) allele gene polymorphism (75). Mtb-triggered lethality is due to early neutrophil recruitment (74), heightened necrotic cell death (76), and likely Mtb-driven and acetyl-coenzyme A dependent foamy cell differentiation (77). TB-susceptible mice are suitable for deciphering host traits which favor a poor outcome in TB. They are also helpful for testing drugs and host-directed interventions given the development of lesions, notably well-structured granulomas containing transformed cell types, and environments, such as hypoxia and necrosis, characteristic of active TB.

Understanding of the immune control of TB has been nurtured by failed immunity in KO and immunodeficient mice which mirror catastrophic human genetic defects. Examples are mice with full or cell-type specific deletion of IFN-γ (78) or TNF-α (79, 80). Just as reported in humans with mendelian susceptibility to mycobacterial disease (81) or on suppressive anti-TNF-α immunotherapy (82) these cannot control Mtb infection or reactivate LTBI, respectively. Since immunity can be investigated within organs, KO mice also enriched knowledge about in situ roles of host factors. For instance, IFN-γ was shown to regulate neutrophil apoptosis (83) and TNF-α to regulate lesion stability by signaling in myeloid and lymphoid cells (80). Mice lacking lymphocytes (Rag2 KO and Rag2/γc KO) have demonstrated essential yet differential roles of T- and NK-cell derived IFN-γ in TB control (84). KO mice have supported reverse translation investigations in TB, as exemplified for miRNAs. For instance, miR-223 is enriched in human TB lesions and susceptibility of miR-223 KO has been linked to the regulation of IL-6, as well as of CCL3 and CXCL2, during acute disease (85). Thus, various KO mice have supported the understanding of TB pathogenesis at a molecular level. In contrast, susceptibility of certain KO lines has not translated to observations in human TB. Some examples are heightened mortality associated with mice lacking the adaptors MyD88 (86) and CARD9 (87). Failed models with observations distinct from human data are also exemplified by mice lacking NADPH oxidase subunits (88, 89) or indolamin-2,3-dioxygenase (90). These mice do not show a strong phenotype in TB despite susceptibility being linked to deficiency in these pathways in humans (68, 91). The inbred features of the most common murine models may contribute to such discrepancies. In this context, CC lines and DO mice may better mirror genetic diversity of the human host. CC lines have uncovered genetic loci associated with uncontrolled infection, including IFN-γ-independent phenotypes (33). DO models have confirmed that neutrophils are detrimental in progressive TB and highlighted roles of neutrophil chemoattractants in this process (36). Since mice and humans show variabilities in immune components, for instance cell abundance (humans belong to neutrophil-high species) or molecular constituents (mice lack granulysin and CD1-type-1 molecules), humanized murine models have been generated (39, 40). Their use is restricted due to financial and technological constraints as well as variability in immune reconstitution and persistence of a mixed human-mouse tissue environment. However, their usage could be critical for addressing co-infection of Mtb with viruses requiring human host factors for entry such as human immunodeficiency virus (HIV). Without doubt, KO models have substantially enriched the knowledge about immune cells and immune pathways in TB and provided causality proofs for disease pathogenesis. Embracing genetic diversity by usage of CC line and DO mice offers unique opportunities for mechanistic studies and may provide new ways to guide TB prophylaxis (92).

Collectively, murine models for TB are diverse and offer a spectrum of options to choose from (see Table 1 ). Experimental tools and feasibility of gene editing in mice, which permit cell fate mapping and tracing, will continue to support immunological research. There are yet several limitations related to the usage of mice in TB. Besides the drawback regarding TB pathology, mice are not suitable for transmission studies. They have been extensively used as models for primary TB. However, unlike humans, mice promptly allow Mtb dissemination to distal sites following aerogenic infection. Efforts to develop murine models for post-primary TB have been undertaken (41) and require additional evaluation.

Inbred laboratory mouse strains such as Balb/C and C57BL/6 are not susceptible to ancestral (B.1) SARS-CoV-2 infection. With the emergence of SARS-CoV-2 variants (e.g. Alpha (B.1.1.7), Beta (B.1.351), and Gamma (P.1)) with extensive mutations in the spike protein, particularly the N501Y mutation, laboratory mouse strains became susceptible to infection and virus replication, although without showing significant pathology (93). However, the inbred mouse strain 129S2 develops clinical disease and has been employed to assess the efficacy of monoclonal antibodies and vaccines (94, 95). Two major approaches have been pursued to amend the murine model for COVID-19 study: genetic engineering of mice for expression of the human ACE2 (hACE2) receptor protein, and adaptation of SARS-CoV-2 to enter murine cells via endogenously expressed receptors (96, 97). A comprehensive summary of the frequently used genetically manipulated murine models as well as adapted virus strategies that substantially contributed to reproduce key characteristics of SARS-CoV-2 infection has been provided recently (98). For comparative evaluation we integrate the murine COVID-19 models with TB models and highlight benefits and disadvantages for each model and infection ( Table 1 ).

The most commonly used K18-hACE2 model, where hACE2 is expressed under control of the human keratin 18 promotor, in addition to the murine ACE2, appears to be the most susceptible COVID-19 model reported to date using human SARS-CoV-2 isolates (99). This model has contributed to the clarification of disease pathophysiology. For instance, it provided evidence for SARS-CoV-2 invasion of sustentacular cells as the cause of subsequent anosmia (100). As inflammation drives severity of COVID-19 in humans, details on dynamics of inflammatory responses obtained in this mouse model could be valuable for the design of therapies. The K18-hACE2 mouse resembles severe COVID-19 disease (101), developing cytokine storm (102), prompt accumulation of immune cells within infected lung (103), loss of plasmacytoid dendritic cells (104) and alveolar macrophages paralleled by accumulation of monocyte-derived macrophages (105). The contribution of host genetics to inflammation control has been further evaluated using collaborative cross (CC) x K18-hACE2 F1 progeny mice (106). In this model survival was associated with early IFN-I expression and production of proinflammatory factors. Similarly, disease severity was driven by CXCR6 and CCR9 in a comparable mouse model approach (107). The K18-hACE2 model has been also useful to demonstrate the relevance of lymphoid cell depletion, which together with the impaired antigen presenting cells/T-cell axis, is a specific feature of severe SARS-CoV-2 infection (108). Furthermore, evidence for the protective roles of T-cells was demonstrated by the fact that vaccination with immunodominant T-cell epitopes provided partial or even full protection in K18-hACE2 mice in the absence of neutralizing antibodies (109, 110). Comparative RNAseq analysis (human vs. mouse) has revealed that at the broad level of immune responses and inflammation pathways, highly overlapping patterns between the two species exist suggesting that the K18-hACE2 mouse model emerges as a representative and relevant animal model of COVID-19 (111). It remains yet unclear whether innate immunity alone could under particular circumstances, for instance low inoculum, eliminate the virus in these transgenic mice. A disadvantage of this model is that it does not mirror mild disease, and interference of signaling from both murine and human ACE2 adds an additional layer of complexity when investigating SARS-COV-2 variants which bind the murine receptor (104). Further, the K18-hACE2 mouse model also has the disadvantage of hACE2 expression in the brain of the transgenic animals. The severity of the disease and the reason for humane endpoints are therefore usually the artificial occurrence of a severe infection of the brain with encephalitis (101, 112). Brain invasion has been demonstrated in humans (113), it is though not a common manifestation of COVID-19 (114). Of note, aerosol delivery in contrast to intranasal challenge bypasses brain involvement (115), suggesting that the route of infection may be relevant for the phenotype of the K18-hACE2 murine model. The K18-hACE mouse model has been essential for vaccine research and its preclinical value is impressive. The critical role of the murine models is highlighted by the fact that mRNA vaccine preparations were extensively tested in the mouse model before licensing in the U.S. under Emergency Use Authorization. Additionally, next generation SARS-CoV-2 vaccines covering multivalency or mucosal application have been similarly evaluated in mouse models [i.a (116–119)].

Adenovirus-, lentivirus- or adeno-associated virus-driven transient hACE2 expression in the murine lung has also been established multiple times in different laboratories [e.g (97, 120, 121)]. However, the virus-induced expression comes with the disadvantage of potential induction of unspecific inflammatory responses, non-uniform expression of hACE2 in the lung epithelium and interference with vector-based vaccines (121). Nevertheless, it has been utilized to study COVID-19 pathology and for preclinical vaccine investigations, including mechanism of action studies. The pathology in this model is restricted to the respiratory tract, with milder disease and in most cases self-resolving inflammation (121, 122). Using this model, it has been shown that IFN-I responses are associated with inflammation and myeloid cell infiltration, but not with SARS-CoV-2 control (122, 123). The mild and localized pathology, along with the possibility to induce hACE2 expression in KO strains, have enabled to study mechanisms of SARS-CoV-2 clearance in naïve and vaccinated animals with different genetic backgrounds. These studies have confirmed the essential role of the adaptive immunity for resolution of inflammation and viral clearance (124). Furthermore, protection of neutralizing antibodies has been confirmed in this animal model (120).

Another approach that allows the use of standard laboratory mice and, more importantly, genetically modified mice, is to adapt SARS-CoV-2 to the mouse (96). These viral strains are therefore particularly suitable for studies in specific KO mouse lines. Thus, the age and sex dependency of human disease severity could be shown with an adapted ancestral SARS-CoV-2 strain (125, 126). However, such adaptations must be carried out separately for different virus variants which do not naturally infect mice and thus are disadvantageous due to the extensive time required for adaptation.

All mouse models come with a substantial drawback related to viral transmission. Even humanized and genetically modified mice are unable to transmit the virus to contact animals (127). Of note, recent investigations in a neonatal K18-hACE2 mouse model have reported virus transmission in a SARS-CoV-2 variant specific manner (128) and such promising observations require validation.

The murine models used for TB and COVID-19 differ substantially, primarily due to the distinct natural susceptibility of mice to Mtb and SARS-CoV-2 ( Table 1 ). For both infections, mouse models are not amenable for investigating transmission and generally have limitations due to dissemination of infection at distal sites as well as at recapitulating human pulmonary pathology. Nonetheless, they are suitable for the mechanistical understanding of immune responses and thus have been extensively employed for vaccine studies. The diversification of the mouse models in TB during the last decade is remarkable ( Figure 1 ), and attempts to employ systems approaches for vaccine discovery (129) further emphasize their value in pre-clinical research. Whereas transgenic knock-in mice have been essential for the progress of COVID-19 vaccines, such strains have rather targeted utility in TB. Irrespective of the peculiarities of the murine models, in both infections experimentation in mice has permitted evaluation of biological processes at subtissular and molecular scale and have advanced interventions.

In the 1960s and 1970s RM were used for the first time in TB research for vaccine and drug efficacy testing. For two decades RM and CM have offered substantial novel insights into pathology, immunology, vaccine and therapies for TB. Nowadays the NHP model is considered the most relevant for translational human TB research (203).

Depending on the dose (10¹-10⁵ CFU), Mtb strain (e.g. Erdman, H37Rv, CDC1551), and route of infection (intravenous, intratracheal or aerosol) RM and CM reflect the full TB spectrum (acute, LTBI and re-activation of LTBI) including all stages of human-like granuloma (204). NHP and human mature, adaptive granulomas, are structured into necrotic cores surrounded by layers of macrophages and lymphocyte zones (205), including immunocompromised microenvironments (68). Akin human lesions (206), NHP granulomas contain tertiary lymphoid structures with key roles in anti-mycobacterial immunity (207). Progression to active TB can be monitored in NHP by longitudinal MRI or PET-CT scans which correlate with bacterial burden and inflammation (208). Such clinical measurements revealed that even under non-clinical disease (e.g. LTBI) NHP lungs contain active, necrotic lesions and sterilized healing lesions at the same time (209), an observation which has been confirmed with similar methods in humans (210). RM are more susceptible to Mtb infection than CM, with RM showing increased pathology and progression to disease compared to CM (211). Evidence for variable baseline of anti- and pro-inflammatory status of the myeloid compartment resulting in increased anti-inflammatory responses in RM after Mtb infection compared to CM pro-inflammatory responses has been provided (212) and likely additional factors underlying diverging susceptibility exist.

Since the whole spectrum of human TB can be observed in NHP models, correlates of protection or susceptibility have been singled out by comparing progressor versus non-progressor animals and by comparing individual progressing versus sterile granulomas from the same animal (213). Treatment of NHP with an antibody against TNF-α leads to increased disease progression in line with observations in humans (214, 215). Similarly, co-infection of NHP with Mtb and Simian immunodeficiency virus (SIV) leads to active disease with pathological features comparable to HIV-1 co-infection in humans (216, 217). Consistent with human TB and many mammalian models, CD4+ T-cells play an essential role in protecting NHP against development of active TB (218, 219). Single-cell transcriptomic signatures of different granulomas from the same individual lung sample revealed that healing or sterile granuloma were associated with IFN-γ/IL17 producing Th1 CD4+ T-cells (213). However, T-cells alone do not seem to be sufficient to control Mtb infection in NHP and humans. Tertiary lymphoid structures (e.g. inducible Bronchus-Associated Lymphoid Tissues (iBALT) or granuloma-associated lymphoid tissue (GrALT)) are significantly associated with non-progressors for active TB (220). A recent study of these GrALT structures has revealed that Mtb-specific B-cells induce T follicular helper cells (Tfh cells) to promote such protection, while depletion of B-cells or impairment of Tfh cells would lead to reduction of GrALT and bacterial growth (207). These findings now require further investigations in human TB and highlight the power of the NHP model to advance knowledge about human TB.

Protection of NHP models against disease progression provided by BCG depends on the route of vaccination, the NHP model, the Mtb challenge strain and dose. Overall, intradermal BCG vaccination of NHP provides variable protection against pulmonary TB which might reflect BCG efficacy in humans (221–224). BCG appears to be more efficient when delivered via aerosol in low dose (225, 226) or when administered intravenously (224, 227). NHP have also been used extensively to test safety and efficacy of preclinical and clinical TB vaccine candidates (228). The vaccine candidate M72/AS01E, which showed 54% efficacy in a human clinical phase 2 trial (229) also showed efficacy in the CM model (230).

Thus, the NHP model greatly contributes to the understanding of TB pathology, correlates of protection and vaccine efficacy. NHP recapitulate active, latent TB, and TB reactivation, and are amenable to longitudinal studies with serial sampling, including imaging, as well as study of TB comorbidities. Limitations of this model are the high housing costs, ethical concerns and shortage of RM and CM for experimental studies. In addition, the variability in route of infection, inoculum, Mtb strain and NHP model lead to heterogenous outcomes, making it challenging to select the most appropriate experimental setup for translational studies.

The ACE2 receptor for SARS-CoV-2 in NHP is identical to hACE2 (231), which is an advantage over other mammalian models. Pathogenesis, vaccine and therapeutic studies have been primarily performed in RM (232), CM (233) and AGM (234) almost simultaneously and immediately after the start of the pandemic. In general, experimental infection resembles mild and/or moderate COVID-19 in humans. It reflects a mild to moderate disease course (235) including lung pathology, viral replication in the upper respiratory tract, vascular involvement including thrombosis (232) and more severe clinical symptoms in aged NHP (236). A direct comparison of RM and CM after SARS-CoV-2 challenge has demonstrated that both models are comparable in the clinical course of infection, viral replication, as well as humoral and cellular immune response (237).

The moderate clinical course in the NHP model allows investigations regarding the correlates of protection. The acute phase and viral replication peak at around 2-4 days post infection and virus genomic RNA and clinical signs decline rapidly afterwards (238). The dynamics of the viral burden are mirrored by influx of neutrophils, dendritic cells and monocyte/macrophage populations into the lung which peak around day 3 and resolve one week later (239). The inflammatory response in the lung of NHP seems dominated by infiltrated monocyte-derived macrophages and is required for clearance of infected pneumocytes and inflammation afterwards (240). This indicates that in NHP the innate immune system likely contributes to the control of virus replication and resolution of inflammation. In line with this, the decline of virus replication and inflammation was associated with IFN-I activated myeloid cells before the induction of adaptive immunity (241). The established immunity protects RM against re-infection, which is similar to observations in humans (242). However, in this case it is most likely mediated by humoral and cellular responses in the upper respiratory tract (243, 244).

NHP have been extensively used as a preclinical model for all currently licensed vaccines against COVID-19 (245). In this context NHP proved to be relevant to investigate correlates of protection and mechanisms of action of COVID-19 vaccines. Systemic neutralizing antibody titers have been found to provide protection induced by the mRNA-1273 vaccine in non-human primates and humans (246, 247). In case of declining antibody titers over time SARS-CoV-2-specific CD8+ T-cell responses provide additional protective immunity and T-cell responses correlate with protection (level of SARS-CoV-2 sgRNA) in RM vaccinated with mRNA-1273 (246). In summary, NHP serve as an excellent model for moderate human COVID-19 cases as well as for investigations of correlates of protection and vaccine efficacy. However, cost restraints and ethical concerns along with a shortage of RM for experimental studies (248) require complementation by other models for studying COVID-19 pathology and for vaccine development.