Personal View

Is a with a pulmonary portal?

Marcel A Behr, W Ray Waters

Tuberculosis most commonly presents as a pulmonary disease, in which infection, persistence, and induction of Published Online transmissible pathology all occur in the lungs. If viewed as a pulmonary disease, enlarged lymph nodes represent November 22, 2013 reactive , and extrapulmonary forms of tuberculosis (including lymphatic tuberculosis) are not transmissible, http://dx.doi.org/10.1016/ S1473-3099(13)70253-6 hence representing an evolutionary dead-end for the pathogen. In an alternative theory, tuberculosis McGill International passes asymptomatically through the lungs and rapidly establishes a chronic lymphatic infection. After a period of Tuberculosis Centre, McGill weeks to decades secondary lung pathology develops, ultimately allowing transmission to occur. Evidence that University, Montreal, QC, supports this lymphatic model includes historical descriptions of human tuberculosis from the preantibiotic era, Canada (Prof M A Behr MD); and analogy with other mycobacterial infections, observations of tuberculosis in non-human hosts, and experimental Bovine Tuberculosis Research Project, National Animal models of tuberculosis disease. At a fundamental level, a lymphocentric model proposes that spread of organisms Disease Center, Ames, IA, USA outside the lung parenchyma is essential to induce adaptive immunity, which is crucial for the generation of (W R Waters DVM) transmissible pathology. Furthermore, a lymphatic model could explain why the lesion associated with primary Correspondence to: infection (Ghon focus) is anatomically separated from the most common site of reactivation disease (the apex). More Prof Marcel A Behr, McGill practically, an alternative perspective that classes tuberculosis as a lymphatic disease might aff ect strategies for International Tuberculosis Centre, McGill University Health preclinical and clinical assessment of novel diagnostics, drugs, and vaccines. Centre, Montreal, QC H3G 1A4, Canada Introduction The pulmonary model of latent infection [email protected] According to police legend, when the recidivist burglar In the conventional, pulmonary model of tuberculosis Willie Sutton was asked why he robbed banks, he replied: (fi gure 1), several steps in infection and pathogenesis are “That’s where the money is”. This pragmatic view, known proposed. First, bacteria enter the terminal airways as Sutton’s Law, guides both medical care (selection of a where they are taken up by alveolar macrophages. Then, site to ) and medical research (selection of a model at or near the site of primary contact, successive waves of system to study a disease). In the case of tuberculosis, immune cells are recruited by the infected cells, either this reasoning extends to the observation that the aerobe directly (through paracrine processes), or by eliciting an Mycobacterium tuberculosis prefers the oxygen-rich immune response at the draining lymph nodes. These environment of the lung, potentially explaining the immune cells generate a that contains the reactivation of organisms in the oxygen-rich apexes of bacteria in this primary lesion, classically called the Ghon the lungs.1,2 Paradoxically, entire programmes of research focus. In this model, the granuloma is gradually broken are devoted to investigating how M tuberculosis survives down, which enables spread of bacteria into the airways hypoxia3 and whether the transcriptional response to and transmission to the next host via the expectoration of low-oxygen conditions is important for bacterial infectious aerosols. Thus infection, immune responses, persistence in the lungs.4 and disease occur in the lungs. On the basis of this At least 2 billion people are estimated to be latently theory, several questions about the pulmonary environ- infected with M tuberculosis, of whom a predicted 10% ment follow. What is the role of lung epithelium? What will progress to clinical tuberculosis.5 This large are the particularities of pulmonary macrophages and reservoir of infected individuals pays testament to the recruitment of into the lungs? What are M tuberculosis as a symbiont, able to infect and persist the metabolic conditions in the lung at the site of in human hosts, causing disease in only a few infected infection? people.6 Although animal models largely recapitulate In view of this pulmonary model, review of early the latter, disease stage, experimental models of publications on M tuberculosis infection is instructive. persistence or latency are elusive, partly because First, in descriptive series published in the preantibiotic latency itself is probably more of a spectrum than a era, both Opie8 and Wallgren9 noted that the host is state.7 The potential value of an experimental model of typically asymptomatic during primary infection; in the mycobacterial persistence is at least two-fold: to few patients with symptomatic infection, these symptoms uncover fundamental processes associated with clinical were either constitutive (fever) or, less commonly, latency, and to guide new interventions, diagnostics, cutaneous (erythema nodosum).9 Unlike in infl uenza or , and vaccines, to detect, manage, and pneumococcal pneumonia, pulmonary symptoms are not prevent disease. Unfortunately, a key input to inform a manifestation of primary tuberculosis infection, such models is unknown: where does M tuberculosis perhaps because M tuberculosis does not produce classic reside? Many clinicians and tuberculosis researchers virulence factors that cause a locally invasive process. suggest that the organisms live in the lung, but they Second, M tuberculosis organisms enter the lungs in might admit uncertainty as to whether the organisms regions of greatest ventilation (where Ghon lesions have are intracellular, extracellular, or confi ned only to the been described),10 yet often exit from the ventilation-poor lung. apexes, where cavitary lesions are noted, both www.thelancet.com/infection Published online November 22, 2013 http://dx.doi.org/10.1016/S1473-3099(13)70253-6 1 Personal View

Entry Exit from a child with cervical lymphadenitis (scrofula), and Mycobacterium avium is frequently isolated from such cases in children.16 In livestock, M avium is detected soon after experimental Bacterial uptake by alveolar macrophage infection in the mesenteric lymph nodes at a time Spread of extracellular bacteria when it cannot be detected in the bowel wall.17 into the airways with transmission of M tuberculosis to next host fi shtank granuloma presents after a cutaneous inoculum as a sporothrichoid 18 Recruitment of cellulitis with a cord-like . Is the immune cells, resulting M tuberculosis complex an exception within the genus? in granuloma Lymph nodes are the primary site of M bovis in cattle, as established by more than a century of carcass inspection T T B for slaughterhouse surveillance. Procedural bias toward T N involvement in cattle does not exist, because N T Mφ N T Mφ N post-mortem inspection includes a thorough assessment B B of lymph nodes, lungs, serosal surfaces, mammary glands, B N Mφ Breakdown of granuloma, T leading to cavitation B N Mφ and abdominal organs for tuberculous lesions. In the T USA, during the past 7 years, more than 94% of M bovis T DC T T DC T isolates from naturally infected cattle were obtained from

T T T T lymph nodes, with fewer than 2% from lungs. In human beings, after the lung, the most common site of Figure 1: The pulmonary model of tuberculosis pathogenesis, in which granuloma formation and tuberculosis is lymphatic disease. Rich observed that “the transmissible pathology occur at or near the site of infection. total mass of the lymph node component of the primary Mϕ=macrophage. N=neutrophils. B=B . T=T lymphocyte. DC=dendritic cell. complex commonly exceeds that of the primary lesion itself”10 (fi gure 2). Although this fact might have been radiologically and in autopsy series.8 The site of familiar 60 years ago, the spread of bacteria to the lymph reactivation disease has been revisited in molecular nodes is mentioned inconsistently in contemporary epidemiological studies; in studies that stratifi ed reviews of tuberculosis pathogenesis.19–23 radiological fi ndings by restriction fragment length Experimental infections further emphasise the polymorphism patterns, most patients who were HIV importance of lymph nodes in the pathogenesis of negative with unique M tuberculosis isolates had upper tuberculosis. In experimental M bovis infections of lobe disease.11,12 Unless there is migration of the cattle or white-tailed deer, the highest bacterial burden granuloma across fi ssures, the entry granuloma must be per gram of tissue is in lymph nodes, irrespective of the distinct from the exit granuloma. Third, in children, the site of initial infection. Pulmonary exposure leads to principal feature of tuberculosis infection is enlarged infection of the pulmonary lymph nodes,24 ingestion hilar or mediastinal lymph nodes, or both. Although to leads to lymphatic spread from the primary infection (in refer to these as reactive adenitis is tempting, knowledge the gut) to mediastinal or tracheobronchial lymph that these lymph nodes are culture positive, and that nodes,25 and intratonsilar inoculation leads to initial organisms from the nodes can transmit disease to rabbits, colonisation in the or medial retropharyngeal has existed since 1890.13 Finally, after experi mental lymph nodes with subsequent lymphatic spread to ingestion of in cattle, lymphatic pulmonary lymph nodes.26 The same process has been spread from the primary infection (in the gut) leads to observed in murine models of experimental mediastinal or tracheobronchial , with M tuberculosis. 2 weeks after aerosol infection of mice, or without accompanying pulmonary disease.14 In 1944, the number of bacteria isolated from the lymph nodes Arnold Rich summarised much of the early data. at day 14 is only one log lower than in the entire lungs, “A striking example of a diff erence in tissue susceptibility despite the lungs being several logs larger than the is provided by the freer multiplication of the bacillus in nodes. Furthermore, the results of careful temporal and the lymph nodes draining a site of primary infection than spatial tracking of immune responses during infection in the infected site itself”, he wrote.10 A primary lung have emphasised that lymphatic infection is not a infection is not a prerequisite for development of spillover process. Rather, induction of adaptive pulmonary tuberculosis; this fact was shown by immunity needs delivery of bacteria to the regional Sanderson in 1867 before Koch discovered the tubercle lymph nodes,22,27 and only after systemic immunity has bacillus.15 been generated are pathological responses recorded in the lungs.28 Spread of bacteria to the lymph nodes is not Evidence for tuberculosis as a lymphatic disease an incidental event that signifi es extrapulmonary Mycobacteria commonly cause lymphatic disease. The disease, but rather a core event during the pathogenesis type strain Mycobacterium intracellulare was isolated of tuberculosis infection. In an observational study of

2 www.thelancet.com/infection Published online November 22, 2013 http://dx.doi.org/10.1016/S1473-3099(13)70253-6 Personal View

eight cynomolgus macaques that developed asympto- matic infection after direct bronchoscopic delivery of bacteria into the right lower lobe, all eight had lesions in the hilar or mediastinal lymph nodes, or both, yet only three had lesions at the site of inoculation.29 These Lymph nodes fi ndings match autopsies of patients who died of conditions other than tuberculosis, in which tuberculous lymph nodes could be detected even in patients without 13,30 pulmonary pathology. Ghon lesion Pathogenesis as a lymphatic disease Guided by a lymphatic model of tuberculosis, obser- vational data from human series and natural or experimental animal models can be reconsidered. At a most straightforward level, images of lymphatic disease (fi gure 2) provide a compelling reason why children with M tuberculosis have enlarged lymph nodes on chest Figure 2: Ghon focus compared with the associated lymphadenopathy Reproduced from Rich.10 radiograph: these are not reactive nodes, they are sites of bacterial infection. Extended further, one can derive a model for M tuberculosis infection that places the nodes in Entry Exit Alveolus a central position and posits that the primary lesion and the most common site of transmissible pathology (apex) Alveolar macrophage are anatomically and temporally distinct processes (fi gure 3), bridged by the lymphatics. If persistent Epithelial barrier infection occurs in the nodes, primary infection in the Bacterial uptake Ghon lesion= Cavitary disease= right lung or the left lower lobe will spread to the right by dendritic cells entry granuloma exit granuloma Pulmonary paratracheal nodes, but not to the left paratracheal nodes disease (fi gure 4). M tuberculosis infection causes a lymphangitis soon after infection, as described in 1873 by Klein31 and Lymphatic Lymphocytes to Bacterial spread spread primary lesion via lymphatics reinforced more recently by Basaraba and colleagues.32 Such lymphatic disease can allow for migration of cells Lymphatic laden with bacteria in a retrograde manner to sections of Infection in hilar with or without disease the lungs with the lowest vascular perfusion (the apexes, mediastinal lymph nodes fi gure 4), with a right-side predominance, based on the asymmetric lymphatic drainage. Progression up lymph-node trunk to subclavian vein In addition to addressing the anatomical pattern of results in haematogenous dissemination tuberculosis disease, a lymphatic model might clarify Extrapulmonary certain epidemiological, clinical, and immunological disease observations associated with this disease. Silicosis is a Clinical presentations of organ-specific disease strong risk factor for development of active tuberculosis and silica exposure leads to lung damage, also with Figure 3: A proposed lymphocentric model of tuberculosis pathogenesis Lymphatics could be a bridge between the primary lesion (entry) and the site of transmissible pathology (exit). particles inhaled in lower lung zones producing an apical disease.33 A potential model for the silicosis– tuberculosis association is that silica taken up by worldwide vary genetically, genes coding for recognised macrophages causes a chronic lymphangitis or immunodominant antigens are quite conserved, that can be exploited by M tuberculosis, implying that ability to induce an immune response is or even the less virulent Mycobacterium kansasii, to cause crucial to the lifecycle of M tuberculosis.34 apical disease. A lymphatic model would also explain why surgical resection of tuberculosis still necessitates Implications of a lymphatic model for clinical treatment using antibiotics, because surgery for and fundamental tuberculosis research tuberculosis, unlike that for cancer, does not normally For both clinical and laboratory research, investigations are attempt to excise draining lymph nodes, but instead ideally based on proven observations about infection and aims only to alleviate local pulmonary damage. Finally, a disease. Unfortunately, clinical classifi cation schemes that lymphatic site of M tuberculosis provides an explanation are useful for the guiding of management decisions might for how M tuberculosis has coexisted with human beings not convey pertinent information for a research study. In a for millennia. Results of a whole-genome analysis of fi rst step, clinicians aim to diagnose a patient as having 21 M tuberculosis isolates showed that, although strains active tuberculosis disease, as opposed to latent infection. www.thelancet.com/infection Published online November 22, 2013 http://dx.doi.org/10.1016/S1473-3099(13)70253-6 3 Personal View

A such as contagion. However, such clinical categories might group together dissimilar patients while separating patients with similar disease processes into distinct groups—eg, using clinical criteria, tuberculosis lymphadenitis, tuberculosis , and tuberculosis meningitis would be called extrapulmonary tuberculosis, and pulmonary tuberculosis would be regarded as distinct from pleural tuberculosis.35 For research aimed at improvement in understanding of the progression of disease and the associated host response, to divide Macrophage tuberculosis into contained (intrathoracic) versus uncontrolled (dis seminated) disease might be more appropriate. In this way, disease of the lungs, the pleura, or the pericardium represents local spread, linked via Bacteria thoracic lymphatics. By contrast, after containment fails, bacteria will exit the nodes with the eff erent lymph and infection will progress to the subclavian vein, whereupon bacteraemia would occur, with seeding of distant sites, B such as the meninges. For biomarker studies and genetics of susceptibility, patients with contained intrathoracic disease might be the best control group for study of patients with uncontrolled, disseminated disease. Low Low Instead, diff erent disease processes, such as tuberculosis perfusion perfusion lymphadenitis and , could be grouped together, potentially obscuring important signals. In addition to the importance of classifi cations for aiding Bacteria in cavity clinical research, clinical categories could complicate the search for the perfect experimental animal model. In view of the prevailing idea of tuberculosis as a categorical construct (active vs latent), investigators using animal models of infection have found a latency model elusive, because experimental models generally show a dynamic infection process.36 Whereas one approach to this problem High perfusion High perfusion is to refi ne the models, using diff erent hosts and diff erent infective doses, an alternative method is to ask whether the Right lung Left lung experimental goal should be a clinically silent pulmonary Figure 4: Lymphocentric model infection. Is lymphatic infection dynamic in human (A) Primary infection in area of high ventilation with spread of infection to beings? What are the metabolic conditions encountered in lymph nodes. Filled spheres represent infected lymph nodes. (B) Lymphatic the lymph nodes, and how do these aff ect access to lipids spread of infection from infected nodes to region of lung with low vascular perfusion. Hollow red spheres represent uninfected lymph nodes. Normal and micronutrients that M tuberculosis depend on for direction of lymphatic fl ow from lungs towards subclavian vein is indicated by survival in the host? What are the key cellular players in red arrows. Dotted black arrows indicate direction of spread of infection. the pulmonary lymphatics and what is the role of newly described cellular hosts of M tuberculosis, such as However, as emphasised by others, tuberculosis might be adipocytes and mesenchymal stem cells?37,38 Furthermore, more of a spectrum than a set of distinct states separated by by investigating areas other than the lungs, we might clear checkpoints.7 Even if tuberculosis infection is a establish greater acceptance in novel model systems, such spectrum, this construct does little to explain why billions as the M marinum zebrafi sh model, in which a lymphatic of people have a positive skin test and a negative system is in the host, but the host does not have lungs. chest radiograph, unless all of these people have pulmonary That Rich wrote the following in 1944 is remarkable: “The lesions that are too small to be detected with this technology. frequency of involvement of certain tissues is determined A lymphatic model of tuberculosis proposes that many of by their position as sites at which bacilli entering the body these individuals have enlarged lymph nodes that might lodge fi rst. The lungs and the alimentary tract, respectively, have been noted merely as an ancillary observation, rather bear the initial brunt of the attack…the frequency of than as a key reservoir of tuberculosis infection. involvement of these tissues in natural infection in the After diagnosing active tuberculosis, a clinician will human being is largely a mechanical result of their classify a case as being pulmonary or extrapulmonary, for position and, in itself, throws no helpful light upon the established reasons linked to management decisions problem of tissue susceptibility.”10

4 www.thelancet.com/infection Published online November 22, 2013 http://dx.doi.org/10.1016/S1473-3099(13)70253-6 Personal View

Implications of a lymphatic model for applied portal of entry and exit. If M tuberculosis resides primarily tuberculosis research in our lymph nodes, might it be possible that the The aim of translational tuberculosis research is to organism has coexisted with humans for millennia by identify improved diagnostics, treatments, or vaccines residing within our immune system? to transform management of tuberculosis. For diagnosis Contributors of disease, the best tests are microbiological (culture or Both authors contributed to the conception, literature review, writing, nucleic acid amplifi cation), whereas for diagnosis and editing of this Personal View. of infection biomarkers such as the tuberculin skin test Confl icts of interest or interferon-γ release assays are used. Unfortunately, no We declare that we have no confl icts of interest. good predictors exist to identify the subset of patients Acknowledgments with infection who will progress to clinical disease. We thank Joel Ernst and Anne-Marie Demers for insightful discussions, and members of their laboratories for feedback on earlier drafts of this Serum-based tests (eg, transcriptomics or proteomics) to Personal View. MAB is funded by operating grant MOP-97813 of the predict disease progression have attracted active interest, Canadian Institutes of Health Research (CIHR). WRW is funded by the but their usefulness might be constrained by the United States Department of Agriculture (USDA). documented diff erences in systemic versus local host References responses.39,40 An alternative direction of enquiry would 1 Haas DW, Des Prez RM. Mycobacterium tuberculosis. In: Mandell GL, Bennett JE, Dolin R, eds. Principles and practice of be to better characterise the size and activity of infectious diseases. 4th edn. New York: Churchill Livingstone, 1995. intrathoracic lymph nodes—eg, via chest CT or PET 2 von Lichtenberg F. Infectious Disease. In: Cotran RS, Kumar V, scan. Although the signal might be too unspecifi c to use Robbins SL, eds. Robbins pathological basis of disease. 4th edn. as a stand-alone diagnostic test,41 in conjunction with Philadelphia: Philadelphia, 1989, pp 307–434. 3 Sherman DR, Voskuil M, Schnappinger D, Liao R, Harrell MI, other evidence of tuberculosis infection, these results Schoolnik GK. Regulation of the Mycobacterium tuberculosis hypoxic could conceivably stratify risk, or help to monitor response gene encoding alpha-crystallin. Proc Natl Acad Sci USA treatment responses. For new treatments, most 2001; 98: 7534–39. 4 Voskuil MI, Schnappinger D, Visconti KC, et al. Inhibition of tuberculosis drugs probably enter the lymph nodes, but respiration by nitric oxide induces a Mycobacterium tuberculosis do we know as much about the metabolic conditions of a dormancy program. J Exp Med 2003; 198: 705–13. granuloma in a lymph node as about conditions within 5 Lonnroth K, Castro KG, Chakaya JM, et al. Tuberculosis control and 42 elimination 2010–50: cure, care, and social development. Lancet the lung? Finally, in the case of vaccine development, if 2010; 375: 1814–29. M tuberculosis enters the lung and replicates in the 6 Veyrier FJ, Dufort A, Behr MA. The rise and fall of the parenchyma until such time as primed lymphocytes Mycobacterium tuberculosis genome. Trends Microbiol 2011; migrate to the site of infection, should a conceptual goal 19: 156–61. 7 Barry CE III, Boshoff HI, Dartois V, et al. The spectrum of latent be to pregenerate a population of mediastinal tuberculosis: rethinking the biology and intervention strategies. antimycobacterial lymphocytes? As has been emphasised Nat Rev Microbiol 2009 7: 845–55. before, we might not be screening for protective immune 8 Opie EL. Pathology of the tuberculosis of childhood and its bearing on clinical work. BMJ 1927; 2: 1130–35. responses when providing vaccines through the skin and 9 Wallgren A. The time-table of tuberculosis. Tubercle 1948; 29: 245–51. 43 measuring systemic responses. Might aerosol BCG 10 Rich AR. The pathogenesis of tuberculosis. 1st edn. Springfi eld, IL: function better than parental BCG,44 not because of Charles C. Thomas, 1944. antigen delivery to the lungs, but by establishing a BCG 11 Jones BE, Ryu R, Yang Z, et al. Chest radiographic fi ndings in patients with tuberculosis with recent or remote infection. infection in the mediatinal lymph nodes (something that Am J Respir Crit Care Med 1997; 156: 1270–73. might be achievable with an oral vaccine)? 12 Geng E, Kreiswirth B, Burzynski J, Schluger NW. Clinical and radiographic correlates of primary and reactivation tuberculosis: a molecular epidemiology study. JAMA 2005; 293: 2740–45. Conclusion 13 Loomis HP. Some facts in the etiology of tuberculosis, evidenced by When a scientifi c challenge is not being addressed thirty autopsies and experiments upon animals. Medical Record completely by research models, two broad options exist: 1890; 689–98. to look harder within our existing theory, using larger 14 Calmette A, Guerin C. Origine intestinal de la tuberculose pulmonaire et mechanisme de l’infection tuberculeuse. sample sizes and more refi ned technologies; or to ask Ann Inst Pasteur 1906; 20: 609–24. whether primary assumptions should be revisited. As 15 Sanderson, J B On the communicability of tubercle by innoculation. Stephen Jay Gould wrote in his classic essay on the On the communicability of tubercle by innoculation. In: Tenth report of the Medical Offi cer of the Privy Council, vol 6. London: 45 validation of continental drift, “New facts, collected in George E Eyre and William Spottiswoode, 1867: 11–151. old ways under the guidance of old theories, rarely lead 16 Engbaek HC. Lymph gland processes caused by atypical to any substantial revision of thought. Facts do not ‘speak mycobacteria and Mycobacterium avium. Bacteriologically verifi ed cases in Denmark from 1935 to 1961. Acta Tuberc Pneumol Scand for themselves’; they are read in the light of theory.” 1964; 44: 108–37. Although tuberculosis clearly involves pulmonary 17 Wu CW, Livesey M, Schmoller SK, et al. Invasion and persistence pathology, it is also a systemic disease, and, as such, of Mycobacterium avium subsp. paratuberculosis during early stages of Johne’s disease in calves. Infect Immun 2007; would be an unusual chronic infectious disease if it 75: 2110–19. involved only one organ from infection to transmission. 18 Wolinsky E, Gomez F, Zimpfer F. Sporotrichoid Mycobacterium Perhaps we should consider the possibility that marinum infection treated with rifampin-ethambutol. tuberculosis is a lymphatic disease, with the lungs the Am Rev Respir Dis 1972; 105: 964–67. www.thelancet.com/infection Published online November 22, 2013 http://dx.doi.org/10.1016/S1473-3099(13)70253-6 5 Personal View

19 Ehlers S, Schaible UE. The granuloma in tuberculosis: dynamics of 33 Goodwin RA, Des Prez RM. Apical localization of pulmonary a host-pathogen collusion. Front Immunol 2012; 3: 411. tuberculosis, chronic pulmonary , and progressive 20 Walzl G, Ronacher K, Hanekom W, Scriba TJ, Zumla A. massive fi brosis of the lung. Chest 1983; 83: 801–05. Immunological biomarkers of tuberculosis. Nat Rev Immunol 2011; 34 Comas I, Chakravartti J, Small PM, et al. Human T cell epitopes of 11: 343–54. Mycobacterium tuberculosis are evolutionarily hyperconserved. 21 Russell DG, Barry CE, III, Flynn JL. Tuberculosis: what we don’t Nat Genet 2010; 42: 498–503. know can, and does, hurt us. Science 2010; 328: 852–56. 35 American Thoracic Society/Centers for Disease Control and 22 Ramakrishnan L. Revisiting the role of the granuloma in Prevention/Infectious Diseases Society of America: controlling tuberculosis. Nat Rev Immunol 2012; 12: 352–66. tuberculosis in the United States. Am J Respir Crit Care Med 2005; 23 Ernst JD. The immunological life cycle of tuberculosis. 172: 1169–227. Nat Rev Immunol 2012; 12: 581–91. 36 Lin PL, Flynn JL. Understanding : a moving 24 Waters WR, Palmer MV, Nonnecke BJ, et al. Effi cacy and target. J Immunol 2010; 185: 15–22. immunogenicity of Mycobacterium bovis DeltaRD1 against aerosol 37 Neyrolles O, Hernandez-Pando R, Pietri-Rouxel F, et al. Is adipose M bovis infection in neonatal calves. Vaccine 2009; 27: 1201–09. tissue a place for Mycobacterium tuberculosis persistence? PLoS One 25 Palmer MV, Waters WR, Whipple DL. Milk containing 2006; 1: e43. Mycobacterium bovis as a source of infection for white-tailed deer 38 Das B, Kashino SS, Pulu I, et al. CD271(+) bone marrow fawns (Odocoileus virginianus). Tuberculosis (Edinb) 2002; 82: 161–65. mesenchymal stem cells may provide a niche for dormant 26 Palmer MV, Waters WR, Thacker TC. Lesion development and Mycobacterium tuberculosis. Sci Transl Med 2013; 5: 170ra13. immunohistochemical changes in from cattle 39 Nemeth J, Rumetshofer R, Winkler HM, Burghuber OC, Muller C, experimentally infected with Mycobacterium bovis. Vet Pathol 2007; Winkler S. Active tuberculosis is characterized by an antigen 44: 863–74. specifi c and strictly localized expansion of eff ector T cells at the site 27 Wolf AJ, Desvignes L, Linas B, et al. Initiation of the adaptive of infection. Eur J Immunol 2012; 42: 2844–50. immune response to Mycobacterium tuberculosis depends on antigen 40 Brighenti S, Andersson J. Local immune responses in human production in the local lymph node, not the lungs. J Exp Med 2008; tuberculosis: learning from the site of infection. J Infect Dis 2012; 205: 105–15. 205 (suppl 2): S316–24. 28 Chackerian AA, Alt JM, Perera TV, Dascher CC, Behar SM. 41 Kuo CH, Chen HC, Chung FT, et al. Diagnostic value of EBUS-TBNA Dissemination of Mycobacterium tuberculosis is infl uenced by host for lung cancer with non-enlarged lymph nodes: a study in a factors and precedes the initiation of T-cell immunity. Infect Immun tuberculosis-endemic country. PLoS One 2011; 6: e16877. 2002; 70: 4501–09. 42 Via LE, Lin PL, Ray SM, et al. Tuberculous granulomas are hypoxic 29 Lin PL, Rodgers M, Smith L, et al. Quantitative comparison of in guinea pigs, rabbits, and nonhuman primates. Infect Immun active and latent tuberculosis in the cynomolgus macaque model. 2008; 76: 2333–40. Infect Immun 2009; 77: 4631–42. 43 Ottenhoff TH, Kaufmann SH. Vaccines against tuberculosis: where 30 Yik Wang C. An experimental study of latent tuberculosis. are we and where do we need to go? PLoS Pathog 2012; 8: e1002607. Lancet 1916; 188: 417–19. 44 Morello M, Krone CL, Dickerson S, et al. Dry-powder pulmonary 31 Klein EE. Contributions to the normal and pathological anatomy of insuffl ation in the mouse for application to vaccine or drug studies. the of the lungs. Proc R Soc Lond 1874; 22: 133–45. Tuberculosis (Edinb) 2009; 89: 371–77. 32 Basaraba RJ, Smith EE, Shanley CA, Orme IM. Pulmonary 45 Gould SJ. The validation of continental drift. Ever since Darwin: lymphatics are primary sites of Mycobacterium tuberculosis infection refl ections in natural history. London: Norton, 1992. in guinea pigs infected by aerosol. Infect Immun 2006; 74: 5397–401.

6 www.thelancet.com/infection Published online November 22, 2013 http://dx.doi.org/10.1016/S1473-3099(13)70253-6