Extremes in Timing of Mycobacterium Tuberculosis Infection: Implications For

Extremes in Timing of Mycobacterium tuberculosis Infection: Implications for

Managing Human Susceptibility to Tuberculosis

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy

in the Graduate School of The Ohio State University

Russell Ault

Biomedical Sciences Graduate Program

The Ohio State University

2020

Dissertation Committee

Joanne Turner, PhD, Advisor

Amy Lovett-Racke, PhD, Co-advisor

Larry Schlesinger, MD

Jeffrey Parvin, MD, PhD

Christopher Walker, PhD

Copyrighted by

Russell Ault

2020

Abstract

Tuberculosis (TB) is the leading killer due to a single infectious disease worldwide. It is caused by the bacterium Mycobacterium tuberculosis (M.tb) and is spread via the airborne route. Only a small proportion of people who become infected with M.tb develop disease and are capable of transmitting the bacteria to others. The capability to identify M.tb-infected people who will actually go on to develop disease is critical. With current methods, treating everyone who has ever been exposed to M.tb to prevent disease likely will cause more harm than benefit due to the potential adverse side effects of treatment. There is a crucial need to target preventive treatment to those who most need it.

With the aging of the global population, the case rate and deaths due to TB are highest in the elderly population. While general immunosenescence associated with old age is thought to contribute to the susceptibility of the elderly to develop active TB disease, very few studies of immune function in elderly individuals with M.tb infection or disease have been performed. Based on an analysis of T cell function and monocyte phenotypes in the blood, we observed no strong evidence that the peripheral blood immune response specific to M.tb in elderly individuals is altered relative to younger adults. This is consistent with the available strong epidemiologic data based on longitudinal birth

ii cohorts that TB disease incidence declines throughout the lifespan well into old age. Our work highlights the need for research identifying biomarkers that allow better assessment of an elderly person’s risk of developing TB disease after infection. It is critically important to know whether an elderly person has been infected recently, with a high risk of disease progression, or earlier in their life, with the lowest risk of progression during their entire lifetime.

Recent M.tb infection is the strongest clinical risk factor for progression to TB disease in immunocompetent individuals, who comprise the majority of TB disease cases. However, time since M.tb infection is challenging to determine in routine clinical practice. To define a biomarker for recent TB exposure, we determined whether gene expression patterns in blood RNA correlated with time since M.tb infection or exposure. First, we found RNA signatures that accurately discriminated early and late time periods after experimental infection in mice and cynomolgus macaques. Next, we found a blood RNA signature that identified recently exposed individuals in two independent human cohorts.

However, for M.tb infected adolescents and adult household contacts of TB cases, our

RNA signature of recent infection was unable to provide prognostic information for TB disease progression, possibly because of its brief duration. Our work supports the need for future longitudinal studies of recent TB contacts to identify biomarkers of recent infection that have sufficient duration to provide prognostic information of TB disease risk in individuals and to help map recent transmission in communities. Such a biomarker would be useful in all populations, including the elderly.

iii

Dedication

To the patients I have met who have suffered from tuberculosis. To those who will yet

suffer from this disease. I hope this work will help you in some small part.

To my devoted wife, Rachel, who has provided me immense support.

Thank you.

Acknowledgments

I would like to acknowledge my advisor, Dr. Joanne Turner. Thank you for accepting me into your research team and for your unparalleled support these past several years. Thank you for training me to be a scientist, with all the sacrifice and effort this has taken from you. I also thank you for the opportunity to come to Texas and to pursue my own ideas as reflected in the work of this dissertation.

I would like to acknowledge past and present members of the Turner laboratory, all of whom have contributed to the work described in this dissertation: Joshua Cyktor, Bridget

Carruthers, Elisha Koivisto, Jenna Nagy, Jacob Horvath, Cynthia Canan, Varun Dwivedi,

Colwyn Headley, Tucker Piergallini, Shalini Gautam and Julia Scordo. You have been wonderful to work with, and I greatly appreciate your help these past few years.

I would like to thank Dr. Jordi B. Torrelles and past and present lab members Michael

Duncan, Juan Moliva, Holden Kelley, Andreu Vilanova, Anna Guardia, Kizil Yusoof,

Anwari Akhter, Paula Tamayo and Angelica Olmo-Fontanez. My association with you and our discussions have been invaluable.

I acknowledge the BSL3 programs and former and present staff at both The Ohio State

University and Texas Biomedical Research Institute, including Lena Lynch, Austin

Hossfeld, Journey Cole, Ariana Duffey, Heather Guenther and Beata Clapp. Thank you to

Alison Whigham and the rest of the vivarium staff at Texas Biomed. I also especially thank Dr. Shannan Hall-Ursone for aid in mouse training and veterinary assistance. I thank the analytical cytometry shared resource at The Ohio State University and acknowledge the efforts of the Baylor Institute for Immunology Research Genomics core for assistance with sample preparation and microarray processing. I thank the molecular core facility at Texas Biomed, and especially Clinton Christensen and Jeremy Glenn. I thank Frederic Chevalier, Jack Kent, Sandra Smith and Richard Polich for helping me set up and for support with computing resources at Texas Biomed. I thank the Data Science

Club at OSU for inspiration in the beginning of my research.

I greatly thank my PhD dissertation committee members, Amy Lovett-Racke (co- advisor), PhD, Larry Schlesinger, MD, Jeffrey Parvin, MD-PhD, and Christopher

Walker, PhD, for providing me guidance and support throughout my PhD.

Thank you to all of our collaborators on this work: Karin Miller, MD, Kokila Nagendran,

MD, Indu Chalana, MD, Xueliang Pan, PhD, Shu-Hua Wang, MD, MPH&TM,

Alexander Hare, Esko Kautto, Quais Hassan, Asuncion Mejias, MD-PhD, Melanie

Carless, PhD and Blanca Restrepo, PhD. I would like to thank past and present members of the OSU Department of Microbial Infection and Immunity and former members of the

vi former OSU Center for Microbial Interface Biology, as well as scientists and staff at

Texas Biomed for feedback, advice, guidance and support in my path to become a scientist. I likewise thank the Medical Scientist Training Program at OSU, as well as the

Medical Scientist Training Program at UT-Health San Antonio, who have graciously allowed me to join in on programmatic events and training opportunities during my PhD.

I especially thank our human study participants as well as clinic staff at The Ohio State

University Wexner Medical Center and Columbus Public Health for assistance in subject recruitment.

Finally, I thank my wife, Rachel, who has supported me through the ups and downs of

PhD training and who has always believed in me.

The work described in this dissertation is contained in two manuscripts, both under

Creative Commons BY-4.0 International licenses

(https://creativecommons.org/licenses/by/4.0/), allowing free re-use and modification under terms of providing attribution and not restricting others’ use of the license.

Substantial text is taken verbatim, with some additions. These manuscripts are (1, 2):

R. Ault, V. Dwivedi, E. Koivisto, J. Nagy, K. Miller, K. Nagendran, I. Chalana, X. Pan, S.-H. Wang, J. Turner, Altered monocyte phenotypes but not impaired peripheral T cell immunity may explain susceptibility of the elderly to develop tuberculosis, Exp. Gerontol. 111, 35–44 (2018).

R. C. Ault, C. A. Headley, A. E. Hare, B. J. Carruthers, A. Mejias, J. Turner, Blood RNA Signatures Predict Recent Tuberculosis Exposure in Mice, Macaques and Humans, bioRxiv , 830794 (2019). vii

Vita

June 2008…………………………………………………Diploma, Fairfield High School

May 2014………..…………………………B.S. Molecular Biophysics and Biochemistry,

Yale University

2014-2016….………………………………………….Medical Student, Medical Scientist

Training Program, The Ohio State University

2016-2020…………………………………...…………………Graduate Research Fellow,

Biomedical Sciences Graduate Program, The Ohio State University

Publications

J. I. Moliva, M. A. Duncan, A. Olmo-Fontánez, A. Akhter, E. Arnett, J. M. Scordo, R. Ault, S. J. Sasindran, A. K. Azad, M. J. Montoya, N. Reinhold-Larsson, M. V. S. Rajaram, R. E. Merrit, W. P. Lafuse, L. Zhang, S.-H. Wang, G. Beamer, Y. Wang, K. Proud, D. J. Maselli, J. Peters, S. T. Weintraub, J. Turner, L. S. Schlesinger, J. B. Torrelles, The Lung Mucosa Environment in the Elderly Increases Host Susceptibility to Mycobacterium tuberculosis Infection, J. Infect. Dis. 220, 514–523 (2019).

viii

P. McGillivray, R. Ault, M. Pawashe, R. Kitchen, S. Balasubramanian, M. Gerstein, A comprehensive catalog of predicted functional upstream open reading frames in humans, Nucleic Acids Res. 46, 3326–3338 (2018).

R. Ault, A. Morales, R. Ault, A. Spitale, G. A. Martinez, Communication pitfalls in interpreted genetic counseling sessions, J. Genet. Couns. 28, 897–907 (2019).

Fields of Study

Major Field: Biomedical Sciences

Emphasis: Immunology

Table of Contents

Abstract ...... ii Dedication ...... iv Acknowledgments...... v Vita ...... viii List of Tables ...... xiii List of Figures ...... xiv List of Abbreviations ...... xvi Chapter 1 Introduction ...... 1 Tuberculosis, an Ancient and Persistent Disease ...... 1 Current tools in the fight against tuberculosis ...... 3 Diagnosis and treatment of tuberculosis disease ...... 3 Preventing tuberculosis via vaccination...... 5 Diagnosis and treatment of latent tuberculosis infection ...... 7 Consequences of untreated and treated tuberculosis disease ...... 14 New tools needed in the fight against tuberculosis ...... 15 Need for non-sputum tests for diagnosis of tuberculosis disease ...... 15 Need for non-sputum tests to predict risk of future tuberculosis disease ...... 17 Potential contribution of test for recent M.tb infection to preventing TB ...... 21 Recent infection is the strongest risk factor for TB disease in most people ...... 21 Known biological events of recent M.tb infection...... 28 Immune mechanisms of M.tb control ...... 35 Preliminary immune correlates of recent M.tb infection in humans ...... 44 Chapter 2 Impaired Peripheral T Cell Immunity does not Explain Susceptibility of the Elderly to Develop Tuberculosis ...... 48 Abstract ...... 48 x

Introduction ...... 49 Materials and Methods ...... 52 Subject Recruitment ...... 52 PBMC Isolation ...... 53 Measurement of ESAT-6 and CFP-10 specific IFN-+ T cells ...... 53 PBMC Culture and Stimulation ...... 54 Measurement of Cytokine Levels ...... 54 Flow Cytometry ...... 54 Statistical Analysis ...... 55 Results ...... 56 Cohort Description ...... 56 T cell cytokine responses to M.tb antigens are not altered by age in those with LTBI or active TB...... 57 Global T cell phenotypes are altered by M.tb infection and disease independent of age ...... 59 Old age and M.tb infection alter the monocyte/lymphocyte ratio and skew monocytes towards a nonclassical phenotype ...... 61 Discussion ...... 63 Chapter 3 Blood RNA Signatures Predict Recent Tuberculosis Exposure in Mice, Macaques and Humans ...... 82 Abstract ...... 82 Introduction ...... 83 Materials and Methods ...... 86 Study Design ...... 86 Mice ...... 86 Mouse aerosol infection and blood collection ...... 87 RNA processing and microarray hybridization ...... 87 Microarray data pre-processing ...... 88 RNA-seq data pre-processing ...... 89 Machine learning predictions ...... 89 Forward search to discover parsimonious 6-gene signature ...... 90 Cell type deconvolution, pathway and transcriptional module analysis ...... 91 Statistical analysis ...... 92

Data availability ...... 93 Code availability ...... 93 Results ...... 93 Blood genome-wide RNA expression accurately discriminates early vs. late M.tb infection time periods in C57BL/6 mice ...... 93 Blood RNA signature discriminates early vs. late M.tb infection time periods in cynomolgus macaques ...... 95 Blood RNA expression of 250 genes predicts time since active TB exposure in humans...... 99 Time since TB exposure in humans is associated with alteration in CD4+ T cell proportion and immune activation pathways...... 102 Application of reduced 6-gene expression signature of time since active TB exposure to adolescent M.tb infection acquisition cohort confirms its identification of recent infection in humans...... 105 Discussion ...... 107 Chapter 4 Discussion ...... 143 Implications for addressing TB in the elderly ...... 143 Progress towards developing a test for recent M.tb infection ...... 146 Biology to probe in future studies of recent M.tb infection ...... 153 Concluding Remarks ...... 159 References ...... 161

xii

List of Tables

Table 2.1. Study Population Characteristics...... 72

Table 2.2. Active TB Subject Characteristics...... 74

Table 2.3. Multivariate linear models of nonclassical monocyte proportion in adult

(n=31) and elderly (n=35) controls based on clinical covariates...... 75

Table 3.1. Probes comprising 50-probe RNA signature of time since M.tb infection in cynomolgous macaques (regression of 1-6 months post-infection)...... 113

Table 3.2. Genes comprising 250-gene RNA signature of time since exposure to active

TB index case (6 months vs. baseline)...... 115

Table 3.3. Top significantly enriched canonical pathways in 250-gene RNA signature of time since exposure to active TB index case (6 months vs. baseline) by IPA...... 124

xiii

List of Figures

Figure 2.1. Study Subject Exclusion Flow Chart...... 76

Figure 2.2. T cell cytokine responses to M.tb antigens are not altered by age in those with

LTBI or active TB...... 77

Figure 2.3. Global T cell CCR7 phenotype is not altered by age in those with LTBI or active TB but PD-1 phenotype may be increased in elderly with active TB...... 79

Figure 3.1. Blood genome-wide RNA expression discriminates early vs. late M.tb infection time periods in C57BL/6 mice...... 128

Figure 3.2. Training and test set partition for cohort of cynomolgus macaques...... 130

Figure 3.3. Comparison of different machine algorithms to predict time period of M.tb infection in cynomolgus macaques...... 131

Figure 3.4. Blood RNA signature discriminates early vs. late M.tb infection time periods in cynomolgus macaques...... 133

Figure 3.5. Blood RNA expression of 250 genes predicts time since active TB exposure in humans...... 135

xiv

Figure 3.6. Time since TB exposure in humans is associated with alteration in CD4+ T cell proportion and immune activation pathways...... 137

Figure 3.8. Trajectory of 3-gene signature for recent M.tb infection before and after

IGRA conversion in adolescents who acquire M.tb infection...... 142

List of Abbreviations

ACS Adolescent Cohort Study

AUC area under the curve

BCG Bacillus Calmette-Guérin

CCR7 CC-chemokine receptor 7

CFP culture filtrate protein

CFP-10 10 kDa culture filtrate protein

CFU colony forming unit

ChIP-seq chromatin immunoprecipitation-sequencing

CNS central nervous system

COPD chronic obstructive pulmonary disease

COR correlate of risk

CORTIS Correlate of Risk Targeted Intervention Study

CT computed tomography

CyTOF cytometry by time of flight, mass cytometry

DosR dormancy survival regulon

DTH delayed type hypersensitivity

ELISA enzyme-linked immunosorbent assay xvi

ELISpot enzyme-linked immune absorbent spot assay

ESAT-6 6 kDa early secretory antigenic target

GC6-74 Grand Challenges 6-74 Study

HBHA heparin-binding hemagglutinin

HCT hematocrit

HHC healthy household contact

HIV human immunodeficiency virus

IFN- interferon gamma

IGRA interferon gamma release assay

IL Interleukin

ILC innate lymphoid cell

LAM lipoarabinomannan

LTBI latent tuberculosis infection

M.tb Mycobacterium tuberculosis

MAE median absolute error

MAIT mucosal-associated invariant T cell

MDSC myeloid derived suppressor cell

MHC major histocompatibility complex

MSMD Mendelian susceptibility to mycobacterial diseases

NK natural killer cell

xvii

NKT natural killer T cell

NTM nontuberculous mycobacteria

PBMCs peripheral blood mononuclear cells

PCA principal component analysis

PCR polymerase chain reaction

PD-1 programmed cell death protein 1

PET positron emission tomography

PPD purified protein derivative

QFT QuantiFERON test

SFU spot forming unit

SIADH syndrome of inappropriate antidiuretic hormone secretion

TB tuberculosis

TCR T cell receptor

Th1 Type 1 T helper cell

TNF/TNF tumor necrosis factor alpha

TPP target product profile

TST tuberculin skin test

VZV varicella zoster virus

WBC white blood cell count

WHO World Health Organization

xviii

Chapter 1 Introduction

Tuberculosis, an Ancient and Persistent Disease

Tuberculosis (TB) is an infectious disease that continues to have a devastating impact on human life and well-being. Evidence from art, literature and specimens of human remains suggest that the modern disease of tuberculosis afflicted people anciently in pre-

Columbian America and in Egyptian times (3, 4). According to the World Health

Organization (WHO), tuberculosis claimed the lives of more people than any other infectious disease during the past 200 years, with over a billion deaths (5, 6). Mortality from tuberculosis rose dramatically in the second half of the 1700s in Europe and North

America, correlating with overcrowded housing and poor socioeconomic conditions incident to the onset of the industrial revolution (7). Death in these regions due to tuberculosis then declined dramatically from 1800 to after the Second World war at approximately the same rate (7). Postulated explanations for this decline include improved living conditions and the application of basic public health measures, including the principle of isolation with the growing knowledge that tuberculosis was an infectious disease (7). A further, more rapid decline of tuberculosis mortality occurred following the invention of antituberculosis chemotherapy, inaugurated clinically in 1946 by the treatment of patients with various forms of tuberculosis with streptomycin (7, 8).

Well into the dramatic decline of tuberculosis in the Western world following its peak in the industrial revolution, Robert Koch demonstrated in 1882 the disease’s causative bacterium, Mycobacterium tuberculosis (M.tb) (9). In his elucidation of the microbiological cause of several diseases, including anthrax, Koch formulated a logical methodology to prove that microbes cause infectious disease, known as Koch’s postulates

(9, 10). The postulates include identifying the microbial organism, finding it associated with the host in all cases of the disease, growing the microbe in laboratory culture and causing the same disease in a new host by inoculating with the purified microbe (9, 10).

Despite the discovery of the causative organism of TB over 130 years ago and the demonstrated efficacy of combination chemotherapy with 3 or more antituberculosis drugs more than 50 years ago, TB is the leading cause of mortality due to a single infectious agent (since 2007) and one of the top 10 causes of death worldwide (9, 11, 12).

About 10 million people became sick with TB in 2018 and 1.5 million died (12). Most

TB cases in 2018 occurred in South-East Asia, Africa and the Western Pacific (12). There is a wide variance in TB incidence rates between countries, with under 10 incident cases per 100,000 population in most high-income countries, ~150-500 in most high TB burden countries and > 500 in the highest burden countries (12). Many high TB burden countries will not attain the 2020 goals of the World Health Organization’s End TB strategy (12).

The global case fatality ratio of TB was estimated at 15% in 2018 (12). Increased application of current methods to diagnose, treat and prevent TB, as well as improvement

2 in the socioeconomic determinants of TB epidemics, are needed to hasten progress in reducing the burden of TB on the world’s population (12). Improved technology to aid in all of these areas of TB control is also needed (12).

Current tools in the fight against tuberculosis

Diagnosis and treatment of tuberculosis disease

The mode of transmission of TB is the airborne route. M.tb is transmitted via droplet nuclei that are expelled from a contagious case’s lungs or larynx when that person coughs, sneezes, shouts or sings. Infection with M.tb is thought to confer a ~10% lifetime risk of developing symptomatic, transmissible active TB, with the remainder of individuals controlling the bacteria in a state known as latent TB infection (LTBI) (13).

TB disease is clinically characterized by its symptoms, which can include cough, fever, weight loss, night sweats, hemoptysis, chest pain, shortness of breath, fatigue and many others. While TB disease predominates in the lungs—the critical affected organ for allowing airborne transmission—M.tb can disseminate to virtually any organ in the body and cause symptomatic pathology thereby. Thus the constellation of symptoms can be different and atypical depending upon what area of the body is affected. The WHO recommends a 4-symptom screening rule, using a combination of current cough, weight loss, night sweats and fever, for exclusion of TB disease. At a 5% TB prevalence rate, this rule has a negative predictive value of ~98%, which is marginally improved by

3 including the presence of a chest X-ray abnormality (14, 15). Radiological abnormalities of pulmonary TB that can be visualized with chest X-ray or with higher sensitivity by computed tomography (CT) include lymphadenopathy, lung parenchymal disease such as lobe consolidation (pneumonia), cavitation, atelectasis, tree-in-bud opacities, diffuse miliary disease, pleural effusion, nodules representing the tuberculoma, which can be calcified, and others (16).

The definitive diagnosis of TB disease is established by microbiologic confirmation of

M.tb in any body tissue, most commonly sputum. This includes visualization of acid-fast bacilli on sputum smear via microscopy, growing live M.tb on culture and/or detecting

M.tb DNA or RNA by nucleic acid amplification testing. Because none of these methods of microbiologic confirmation are completely sensitive, a person can be treated empirically for TB disease in the absence of microbiologic confirmation based on the aggregation of clinical and epidemiologic history, physical exam, radiographic findings, tuberculin skin test and interferon gamma release assay results. The absence of the demonstration of live M.tb cannot rule out TB disease.

TB disease is treated with a combination of antibiotics. Based on studies of spontaneous

M.tb mutation rates and the knowledge that M.tb does not transfer drug resistance genes via horizontal gene transfer, the probability that drug susceptible M.tb can develop resistance to three or more drugs simultaneously is infinitesimally small (17, 18). The first application of three antituberculous drugs for treating chronic pulmonary

4 tuberculosis with cavitation was demonstrated with isoniazid, streptomycin and para- aminosalicylic acid (11). Patients treated for two years with this regimen had only a 2% relapse rate by year three post-initiation of treatment (11). The current first-line regimen for drug susceptible M.tb is isoniazid, rifampin, pyrazinamide and ethambutol (19). These four drugs are typically used for 2 months to rapidly reduce the bacillary load, eliminate transmission risk, clinically cure the patient and to avoid the risk of developing drug resistance (19). Thereafter, just isoniazid and rifampin are continued for an additional 4 months to reduce the numbers of persisting bacilli and thus to prevent relapse following therapy (19). Efforts are underway to systematically identify subpopulations where duration of treatment can be stratified by disease severity, with burgeoning evidence that lower amounts of bacteria in the sputum and the absence of cavitation may require less than 6 months of treatment, and their presence may require more than 6 months of treatment (20).

Preventing tuberculosis via vaccination

Prevention of TB transmission, as with most infectious diseases, is the key pillar of successful TB control using public health and medical measures. Prevention consists of preventing M.tb infection and/or TB disease in currently healthy people and treating those with TB disease to cure them and interrupt ongoing transmission. As with other infectious diseases, an ideal prevention for TB is preventing initial M.tb infection in healthy people via vaccination.

The Bacillus Calmette-Guérin (BCG) vaccine is the only licensed vaccine for tuberculosis. It was developed by Albert Calmette and Camille Guérin by passaging the causative organism of bovine tuberculosis, Mycobacterium bovis, on potato medium containing bile over a 14 year period (21, 22). A systematic meta-analysis of 14 prospective trials and 12 case-control studies estimated that over the course of the 20th century the average protective effect of BCG vaccination against risk of TB disease was

50%, with geographic latitude and study validity score explaining most of the variability between individual studies (23). A meta-analysis of 10 randomized trials showed that the average efficacy more than 10 years after BCG vaccination was 14% (95% confidence interval [CI] -9% to 32%) (24). While this suggests that in many settings BCG does not provide protection beyond 10 years after vaccination, there is one report that BCG vaccination provided durable protection in American Indians and Alaska Natives for up to 60 years (24, 25). There is also growing evidence that BCG vaccination can prevent initial M.tb infection, with the largest observational adult study showing an efficacy of

30% (95% CI, 13%–44%) (26). While BCG efficacy is estimated to be higher for severe forms of TB in children (73% against tuberculous meningitis and 77% against miliary tuberculosis) as compared to pulmonary TB in adults, efficacy in adults is the key for helping to reduce transmission (27). With the sustained and growing evidence that BCG provides partial efficacy against M.tb infection and TB disease in adults, and the potential for developing targeting revaccination strategies with subunit vaccines or BCG, vaccination continues to play an important role in TB prevention worldwide (28, 29).

However, the only partial efficacy of BCG vaccination is such that it is not sufficient as the only TB prevention strategy. It is also not yet clear whether this efficacy can be substantially improved upon in humans.

Diagnosis and treatment of latent tuberculosis infection

If M.tb infection cannot be prevented in an exposed individual via prior vaccination, then

TB disease can still be prevented by treating the infected individual with one or two antituberculous drugs prophylactically. Fewer drugs are required to prevent TB disease than to treat current TB disease likely because of the reduced bacillary load and because of the high likelihood that infected individuals will never develop disease even if untreated. In most people at risk of M.tb infection, confirming infection before deciding whether to initiate preventive treatment is desired (14).

The origins of 20th century and contemporary tests for M.tb infection lie in the claim of

Robert Koch in 1890 to have found a cure for tuberculosis (30). His process involved dissolving heat-killed, concentrated M.tb in glycerin and injecting it into guinea-pigs, and soon thereafter humans, and is termed “old tuberculin” today (30, 31). It was soon shown to be too toxic and of marginal efficacy for treatment of TB disease (30). Von Pirquet built upon this preparation by inoculating it into the skin, observing a local inflammatory reaction 24-48 hours later in children with TB disease but also in healthy children which increased in prevalence with age (32). Therefore the reaction to old tuberculin could be

7 used to diagnose LTBI (32). The intradermal injection technique used today was reported by Charles Mantoux in 1912 (33). In 1934 Seibert reported the method of isolating the main active ingredient of old tuberculin, which was protein antigens, using trichloroacetic acid precipitations, later refined with ammonium sulfate precipitations (34, 35). Her preparation was termed purified protein derivative (PPD) and soon became the reference standard upon which the tuberculin skin test (TST) is still based today (31, 34).

Administration of the TST in previously infected individuals results in a delayed type IV hypersensitivity response mediated through the infiltration of T cells into the injection site, vasodilation and edema, and the expansion of memory T cells specific to M.tb antigens in PPD. The test is read by measuring in mm the diameter of the induration caused. Previous BCG vaccination or exposure to environmental mycobacteria can cause a false positive TST result due to shared antigens between M.tb and these mycobacteria.

Immunosuppression, such as caused by HIV infection, can lead to false negative results.

In the United States, different thresholds in mm are used for calling a positive test result, based on the prior likelihood that a person has been exposed or is immunocompromised

(36). A threshold of 5 mm is used for immunosuppressed people or those with documented recent exposure to a TB case, 10 mm for those at a higher risk due to occupation, age or country of origin and 15 mm for most other persons (36). While BCG vaccination theoretically leads to reduced specificity of TST for detecting M.tb infection, in practice it does not due to vaccination occurring primarily in infancy and the waning of

BCG immunity over time (14, 37–40).

The TST relies on the in vivo recruitment and proliferation of T cells recognizing M.tb antigens. In the blood, most T cells recognizing M.tb peptides produce IFN-, which is a critical immune cytokine for controlling the intracellular growth of M.tb (41–43).

Beginning in the early 1990s, in vitro blood tests for M.tb infection were commercially developed, beginning in cattle, based on these principles (44–46). One advantage in the early test for humans was that it had improved sensitivity in an immunosuppressed population, specifically intravenous drug users (47). Two versions of this test are used in humans today, the QuantiFERON® (QFT) (Qiagen, USA) and T-SPOT®.TB (Oxford

Immunotec, UK) tests. In the QFT, whole blood is stimulated with M.tb antigens and

IFN- is measured in the supernatant with an enzyme-linked immunosorbent assay

(ELISA) using monoclonal antibodies to IFN-. In the T-SPOT.TB test, peripheral blood mononuclear cells (PBMCs) from whole blood are stimulated with M.tb antigens while resting on a membrane containing adhered monoclonal antibodies to IFN-, which is then developed to show spots where the original IFN--producing cells were. This is termed an enzyme-linked immune absorbent spot (ELISpot) assay. Both assays are called interferon gamma release assays (IGRAs).

Initially, IGRAs used PPD from M.tb as the stimulus, just like the TST (44, 47).

However, today IGRAs use a combination of antigens that are specific to M.tb and absent from BCG and most environmental mycobacteria. The first demonstration of an antigen partially specific to M.tb complex was by Harboe et al with the 24 kDa antigen MPB64

(also termed MPT64), and it was later shown via genetic methods to be absent only in some BCG substrains (44, 48, 49). The discovery of the ESAT-6 and CFP-10 proteins from M.tb culture filtrate, due to their immunodominance in mice, ushered in the first very specific antigens for LTBI diagnosis, as they are absent from all BCG strains and environmental mycobacteria except M. kansasii, M. marinum and M. szulgai (44, 50–52).

The current IGRAs use ESAT-6 and CFP-10, with the current QFT version also using

TB7.7 (53).

Neither the TST nor the IGRAs are a gold standard for diagnosing current M.tb infection.

They measure either an effector or memory T cell response that reflects exposure sometime in the past. The sensitivity of either method can only be calculated precisely in people with culture-positive active TB, who are known to be infected definitively.

Specificity can be approximated using healthy people with low likelihood of past exposure to M.tb. However, neither of these groups are the population of interest for

TST/IGRAs, who are healthy people with some likelihood of prior exposure to M.tb.

Thus, a common comparator study design for evaluating IGRAs and TST is determining their concordance in people with likely prior exposure, but by design such a study cannot prove which is superior for LTBI diagnosis (54). Given that therapy for LTBI to prevent

TB disease is available, the primary medical criteria for comparing performance of the

IGRAs vs. TST is their predictive value for risk of future TB disease. Several studies involving IGRA and TST testing coupled with follow-up for TB disease in untreated persons have shown that the positive and negative predictive values of both IGRA and

TST are not different as long as BCG vaccine status is accounted for with the TST size cut-off used (37, 55). Accordingly, the decision of which to use for LTBI diagnosis is primarily driven by availability and affordability (14). In those at high risk of LTBI in a low incidence setting, the positive predictive value for TB disease progression of all these tests was less than 5% (37). Therefore, tests for LTBI are of poor prognostic value, reflecting the truth that the attack rate for TB is low and these tests only indicate prior exposure, not disease susceptibility (55). Therefore, the decision of whether to test for

LTBI should already incorporate all known information of the person’s likelihood of progressing to TB disease (14). The decision of whether to test for LTBI with TST/IGRA should be accompanied with a commitment to treat the person if the result is positive for past exposure. TST/IGRAs are rule-out tests with negative predictive values for TB disease progression of >99% in both low and high incidence settings (37, 56).

There are individuals who persistently test negative on TST/IGRA despite being highly exposed to active TB cases (57). It has been shown that these individuals have had the same exposure history and epidemiologic risk factors as those who do convert to a positive TST/IGRA (58). Recently it was shown that these individuals are probably all infected at some time with M.tb, as they develop an adaptive immune response to ESAT-

6 and CFP-10 (59). Their adaptive immune response is characterized by a qualitatively different antibody response to M.tb than TST/IGRA+ persons and reduced T cell responses to M.tb that do not involve IFN- in response to ESAT-6 and CFP-10 (59). The reasons for this altered response have not yet been elucidated, but could involve

11 interaction with pre-existing immunity to BCG and environmental mycobacteria, earlier clearance of M.tb or possible host genetic factors (59). With the current absence of a well-powered study to answer the question, it is unknown whether people with persistently negative TST/IGRA are at an altered risk of TB disease compared to

TST/IGRA+ persons (57, 59). While this may appear to be a potential weakness of the

TST/IGRA tests to rule-out TB, their extremely high negative predictive value for TB disease (>99%) in both high and low incidence settings argues against this phenomena as having importance for current TB control efforts (37, 56). Therefore, such people either have a reduced risk of TB or they are so few in number that their contribution to the pool of active TB cases is negligible. Nevertheless, study of this group of people could have important implications for understanding protective immunity to M.tb that could be elicited by vaccines, if it is shown that their TB incidence rates are lower (59).

We have heretofore described treatment of current TB disease, prevention of TB via vaccination with BCG and the means for identifying people who have previously been infected with M.tb. We will now explain the development of the means currently used to treat people with LTBI to reduce their future risk of TB disease. Following its application in treating TB disease in humans and for preventing TB in animals in the 1950s, isoniazid was the first drug studied extensively to prevent TB disease in people with LTBI (60–63).

Its choice relative to other drugs at the time was due to its relatively low toxicity, low cost, ease of administration and efficacy in treating TB disease (60). In a placebo controlled trial, 1 year of isoniazid in asymptomatic children reduced TB incidence by

94% during treatment and 70% during the 8-year follow-up period (61). In adult contacts of active TB cases and in adults with inactive TB that was never previously treated, efficacy was 50-80%, being highest during treatment and closer to 50% during post- treatment follow-up (61). These and other studies led the American Thoracic Society to recommend preventive treatment for TB with isoniazid for high risk groups in 1965 (60).

Subsequent studies showed that the efficacy of isoniazid to prevent TB is not improved beyond 12 months of treatment (64). From 6 to 12 months of treatment, efficacy is improved, but the risk of hepatitis due to isoniazid and the risk of nonadherence likewise increase, resulting in 9 months of isoniazid being the standard of care in low incidence countries for many years (64, 65). The addition of rifamycins, including rifapentine and rifampin, to isoniazid preventive therapy, or rifampin alone, now allows for shortened therapy of 3-4 months with noninferior efficacy, higher safety and higher rates of treatment completion (14, 66–68). Efforts to shorten therapy even further are underway, with 1 month of rifapentine and isoniazid recently shown to be noninferior to 9 months of isoniazid alone in HIV-infected patients (69). However, serious adverse events from preventive treatment continue to occur, highlighting the need to test for LTBI and offer treatment only to those deemed at high risk of future TB disease based on their clinical and epidemiologic history (14, 68–70).

Consequences of untreated and treated tuberculosis disease

If M.tb infection and TB disease in a susceptible person have not been prevented via public health efforts, vaccination or preventive treatment, then the afflicted person must be diagnosed and treated for clinical TB, as previously described. Treatment for all persons with TB disease is critical as treatment has a very high likelihood of curing the person and it also interrupts further transmission of the bacteria from that infectious person. Untreated TB disease has a 10-year case fatality rate of approximately 70% if bacteria can be visualized microscopically in the person’s sputum, and approximately

20% if not (71).

The global treatment success rate for drug susceptible TB is 85% (12). With room for improvement, this high treatment success rate does not take into account the lifelong consequences of successfully treated TB disease. A recent systematic review estimated that those with tuberculosis have a 3-fold higher all-cause mortality than the general population, and this increased risk was not lower among those who completed treatment and/or were cured (72). Over half of the deaths in those who have had tuberculosis were accounted for by cardiovascular disease, cancer and respiratory disease, with the large majority of deaths not explained by tuberculosis recurrence (72). Health needs for TB patients following microbiologically successful treatment are very poorly characterized and understood, with excess morbidity likely influenced by tissue damage from TB

14 disease, comorbidities, social determinants and treatment toxicity (73, 74). Therefore, it is clear that prevention of TB disease is the best way to benefit both individuals and the community to reduce its burden on society. However, as previously described, current

LTBI diagnostics do not allow identification of individuals who do need treatment to prevent likely TB disease. In addition, new tools are needed to help identify all current

TB cases so that transmission can be stopped and excess death from untreated tuberculosis eliminated.

New tools needed in the fight against tuberculosis

Need for non-sputum tests for diagnosis of tuberculosis disease

The WHO has called for point-of-care non-sputum tests to triage patients for TB disease testing and to diagnose TB to direct treatment initiation (75). Such tests would ideally be able to be used in the field with minimal equipment and would be more sensitive than sputum smear microscopy or culture. If successfully developed, they could allow more identification of currently undiagnosed TB patients and allow treatment to be administered to currently untreated persons. The proposed triage test would prioritize sensitivity to detect TB disease so as to be able to rule out TB, and the diagnostic test would prioritize specificity relative to other respiratory diseases and conditions so that treatment can be confidently administered.

There are many published studies on identifying biomarkers that correlate with TB disease from blood, urine or breath (76). The majority have focused on the host immune response to M.tb, including antibodies, cytokines, host RNA and other host proteins (76).

Some have focused on M.tb molecules, with the most common being lipoarabinomannan

(LAM), a glycolipid on the M.tb cell wall (76). LAM-based tests have been commercialized for urine in select patients with HIV (76, 77). New technologies are being developed to increase the sensitivity of detection of pathogen molecules by enriching them in non-sputum samples, so as to be more widely applicable to the general population. These include detecting LAM in urine and CFP-10 and ESAT-6 peptides in blood (78, 79). The identification of blood host RNA signatures of TB disease was facilitated by the application of whole transcriptome microarrays to comprehensively profile gene expression in the blood (80). Type I and II interferon signaling genes and interferon-inducible transcripts have been the most widely reported genes associated with

TB disease (80). While machine learning methods have often been used to identify collections of multiple genes that together facilitate discrimination of TB disease from healthy controls and other diseases, a recent metanalysis of reported host RNA signatures showed that a higher number of genes than 3 did not increase the accuracy of signatures

(81). Currently there appear to be 2 RNA signatures that meet the minimal WHO Target

Product Profile (TPP) sensitivity and specificity criteria for a non-sputum triage test for

TB disease (75, 81).

Need for non-sputum tests to predict risk of future tuberculosis disease

These non-sputum diagnostics would help extend current abilities to diagnose TB disease. However, there is also a critical need for tests that predict the risk of TB disease development in asymptomatic persons. The IGRA/TST tests do not fulfill this need as they have very poor positive predictive value for TB disease progression, meaning that many (~37-64) people need to be treated to prevent just one case of TB disease (82). This subjects many people unnecessarily to the side effects of preventive treatment, and thus treatment is not universally recommended for asymptomatic LTBI+ individuals without other known clinical/epidemiological risk factors for TB progression (14). The TPP for this test for risk of TB progression is currently based around the concept of detecting incipient TB, which is the asymptomatic phase of early TB disease during which pathology progresses gradually before full-blown clinical TB (82–86). The test needs to be repeatable given that if someone is tested before the inciting biological event triggering progression occurs, the test result will be negative despite the fact that the person will progress (82). Therefore, the test needs to be based on an easily accessible tissue sample, such as blood. The test ideally needs to be positive during current TB disease detected via additional means (symptoms, radiology, microbiology) and should turn negative after a person is successfully treated, reflecting either elimination of the bacteria and/or resolution of the host inflammatory response (82).

Several host response blood signatures have been identified that meet several of these characteristics for a test for incipient TB. These include blood RNA signatures comprising 1-25 transcripts and mostly reflect the expression of interferon inducible genes that are virtually identical to those genes identified for TB disease diagnosis (81,

84–86). Host proteins, metabolites and circulating miRNAs have also been found to correlate with TB disease progression as well as current TB disease (87–89). Nearly all of these host signatures of TB disease progression have been derived from the Adolescent

Cohort Study (ACS) and/or the Grand Challenges 6-74 (GC6-74) Study (85, 86). The

ACS study enrolled adolescents who were IGRA+ with an unknown prior exposure history, and the GC6-74 study enrolled adults who were recent household contacts of pulmonary TB disease cases. Both followed untreated individuals for progression to TB disease and identified prognostic blood signatures of TB risk via a nested case-control design. A follow-up study of more host immune factors in the ACS cohort revealed that following the earlier detectable Type I/II interferon signaling and also complement cascade elevation, there were changes in myeloid cell inflammation, lymphoid, monocyte and neutrophil gene modules more proximal to the date of TB disease diagnosis among adolescents progressing from LTBI to clinical TB disease (90).

It is important to note that in these discovery and initial validation studies to date, these host-response based tests for TB progression have only marginally higher positive predictive value than that reported for TST/IGRAs (84–86). Therefore, it is currently uncertain whether treatment for LTBI based on the results of these tests, in particular the

RNA test that is furthest in development, will improve TB prevention relative to currently available tools. It is also unknown to what degree those with incipient TB are responsive to standard LTBI treatment versus LTBI+ persons without incipient TB, since it is possible that incipient TB reflects a higher bacillary load and thus may require a full active TB disease regimen. A randomized clinical trial, the Correlate of Risk Targeted

Intervention Study (CORTIS), is underway to answer whether preventive therapy with isoniazid and rifapentine reduces the rate of TB disease in those positive for the correlate of risk (COR) RNA test, vs. the standard of care for the general HIV- population, which is active surveillance without treatment (91). As a secondary outcome, CORTIS will prospectively compare the prognostic performance of the COR and IGRA for TB disease, which has never been performed (91).

A recent metanalysis of reported RNA signatures of TB disease risk showed that no published signature met the minimum TPP for biomarkers of incipient TB with the required high sensitivity and specificity for TB risk for the two years following sampling

(84). The eight best performing signatures showed equivalent performance and were predominately driven by IFN-, STAT1, IFN- and TNF, suggesting that these RNA signatures reflect the same host immune response in the incipient TB disease process

(84). No other profiled biomarkers in the ACS or GC6-74 studies have shown superior performance to these RNA signatures (87–89). There is some evidence that combining

RNA and metabolic signatures could improve predictive performance due to some statistical independence in their measurement, but this has not been shown in an

19 independent prospective cohort nor at the high specificity required for a test for incipient

TB (92). The metanalysis of RNA signatures showed that blood transcriptional biomarkers only exceed WHO TPP requirements if applied to 3-6 months after sampling, thus reflecting only short term TB disease risk (84). This suggests that incipient or subclinical TB measurable by a host-response based blood test on average develops less than 3 months before clinical diagnosis (84). Even among those within 3 months of clinical diagnosis, the sensitivity of the best performing signatures was 47.1-81.0%, suggesting either that these RNA signatures are imperfectly sensitive for the incipient phase of TB, or that very rapid disease progression occurs in some cases (84).

These studies of host response biomarkers for risk of TB disease have shown that there are measurable molecular aberrations that precede the presentation of clinical TB disease and are sustained during untreated disease. However, it is currently unclear to what degree these biomarkers may help with targeting preventive treatment to those most at risk of TB disease, due to their relatively brief duration in many people and/or their imperfect sensitivity at required specificity. Therefore, identifying other factors and/or biomarkers that inform a person’s TB risk, beyond indicators of the asymptomatic state of progressing early disease, is important and could be used to aid biomarkers of incipient

TB and vice-versa. Following the publication of the first RNA signature of TB disease risk, the first systematic uses of clinical, epidemiological and social data to predict TB disease risk in individuals and households have been reported (70, 93). In these studies, clinical, epidemiological and other social information are aggregated in a simple score

20 which has been shown to stratify high risk individuals/households from low risk individuals/households, in both discovery and validation cohorts (70, 93).

One limitation of these presumably easy to use clinical/epidemiological scores is that they are restricted to the setting of a household contact investigation where a known TB index case is identified (70, 93). However, the large majority of transmission in high incidence settings likely occurs outside of the home (94, 95). Accordingly, the CORTIS trial is designed for implementation in a mass “screen and treat” strategy, wherein virtually the whole healthy and immunocompetent population is eligible for screening

(91). Both of these approaches do not utilize the most common and strongest risk factor for TB disease in most persons, which is recent M.tb infection. In the next section we will discuss the abundant evidence that risk of TB progression is highest in the first 1-2 years after infection and how obtaining this information in an individual with a biomarker could aid and complement the aforementioned approaches to estimating TB disease risk and targeting preventive therapy to those at highest risk.

Potential contribution of test for recent M.tb infection to preventing TB

Recent infection is the strongest risk factor for TB disease in most people

It is already known that recent M.tb infection is the single strongest clinical risk factor for developing active TB disease in immunocompetent persons, who comprise the vast

21 majority of LTBI and active TB cases (96–101). While a recent study in the United States and Canada showed that recent contacts of active TB cases are at highest risk of TB disease in the first 1-3 months after the diagnosis of the TB index case, studies in other countries and other time periods show that the highest risk is in the first 1-2 years, with most cases accruing more than 2-3 months after a documented exposure (96, 99, 102–

105). It is highly likely that the specific timetable in persons and different cohorts is influenced by factors such as pre-existing immunity via BCG vaccination or host genetic, nutritional or other environmental factors, but the general period of 1-2 years is consistent, as will soon be described in more detail.

Unlike infections with an acute illness and high disease progression rate, or sexually transmitted infections, transmission events for tuberculosis cases are extremely difficult to pinpoint. Genotyping M.tb isolates from active TB cases coupled with comparative genomic analysis has permitted population-level identification of hotspots of localized transmission, but these data are mostly available retrospectively and thus do not allow real-time monitoring of TB transmission in a community, particularly in areas of high incidence (106). Much of our understanding of the initial period of M.tb infection in humans dates to the work of Poulsen in the 1930s and 1940s (102, 107). Working as a clinician in the isolated Faroe Islands before the advent of antituberculous drugs, he and colleagues observed patients who converted on a TST and who later developed TB disease after exposures that he could pinpoint to a two week period and often to a single day (102, 107). This was facilitated by administration of the TST to the majority of the

30,000 population at regular intervals (102). Most TST conversions occurred within 6 weeks of exposure (102). Among TST converters including both children and adults, approximately 70% had an accompanied fever, for most of whom lasted several weeks

(102). This fever was practically always accompanied by an elevated erythrocyte sedimentation rate and less frequently erythema nodosum (102). These clinical manifestations are known as primary tuberculosis. As revealed by X-ray fluoroscopy and/or plain film, hilar adenitis occurred in 64% of the converters, pleural effusion in

29% and pulmonary infiltrates in 26%, with ¾ of adults and nearly all children showing at least one X-ray finding (102). The vast majority of both pulmonary infiltrates and pleural effusions were observed within 6 months of conversion, with pleural effusions occurring later within that period (102). 19% of the adult contacts developed pulmonary

TB disease, with half diagnosed within two years after conversion and over 85% within 4 years (102).

Several other studies from currently developed nations in the pre-antibiotic era concur with the timeline reported by Poulsen (102). Wallgren reported that in Sweden in the pre- antibiotic period the clinical signs and symptoms of primary tuberculosis, when present, occurred by 6 weeks post-infection (108). In those who go on to develop pulmonary TB disease, disease occurred in ~80% by 2 years after the manifestations of primary tuberculosis (96, 108). He also described that generally the timing was different between the different organ system manifestations of TB disease: meningitis and miliary TB occurred earlier than disease with pleural involvement, and all three occurred before

23 musculoskeletal and pulmonary TB, with pulmonary TB having a slightly longer tail

(108). He argued that during the phase of primary tuberculosis, bacteria often spread throughout the body via the hematogenous route, with bacteria even sometimes observed in the urine (108). Gedde-Dahl around the same time also observed that 60% of those who developed TB disease did so within 3-9 months, with less than 10% doing so after 2 years (96, 103).

These studies occurred in the absence of chemotherapeutic intervention to prevent TB. A modern isoniazid preventive therapy trial among South African goldminers where TB transmission is extremely prevalent showed that while isoniazid treatment reduced risk of

TB during the 9 months of treatment by about 60%, it provided no protection once treatment was completed, with a convergence in TB rate in the untreated and treated groups (109). This suggests that reinfection could contribute to some of the tail observed after 2 years post-exposure in the pre-antibiotic era studies. A review of isoniazid preventive therapy trials in the United States from 1955-1966 summarized that isoniazid therapy abolished most of the early risk of TB disease in the 2 years following household contact with a case of TB disease, but the risk between treated and untreated individuals was similar thereafter, with both decreasing constantly until the end of 10-year follow-up

(110, 111). Therefore, the true risk of TB disease beyond 2 years post-exposure could be slightly overestimated due to the risk of reinfection in these historical studies. Today, in low incidence countries the risk of reinfection following an initial exposure is lower but not nonexistent. Among 158 close contacts of pulmonary TB patients in the United States

24 and Canada who themselves went on to develop TB disease, 51%, 75%, 81%, and 92% were diagnosed by 1, 3, 6, and 12 months, cumulatively (99). The timing, though not the risk of TB, was identical between those who did or did not receive preventive treatment

(99). A recent study from Amsterdam showed that over 75% of contacts developing TB disease did so within 6 months, regardless of their age, with over 90% doing so by 2 years (100). For culture positive cases bacterial genotyping was performed, and those who were diagnosed within 2 years had 93% genotype concordance with their index case while only 50% had concordance among those diagnosed greater than 2 years after contact with their index case (100). Other modern studies from the Netherlands and

Canada of recent household contacts with recent transmission established via bacterial genotyping confirm that the vast majority of TB cases accruing from such contact occur within 2 years of contact with the index TB case (105, 112).

The studies heretofore described had follow-up up to about 10 years after initial contact with an active TB case. They do not answer the question of whether there is some later point in life when TB risk from a remote infection may increase. This possibility would reduce the value of a test for recent infection when using it to target preventive therapy in adults. Hart and Sutherland observed 8838 TST+ or TST- adolescents in the control arm of a BCG and vole bacillus vaccine trial during a 20 year period from 1951 to 1970 in

Great Britain (113). There was a sharp increase in TB incidence in initially TST- adolescents, corresponding to their first inevitable exposure when TB incidence was highest in the first few years of the study (113). However, from study start in the initially

TST+ group and from the peak TB incidence in the TST- group by about 5 years after study enrollment, the incidence of TB disease decreased continuously in both groups through the 20 years of follow-up (113). There are also several cohort studies that consider the case rate of TB disease throughout the lifespan (114–116). These studies allow one to compare the case rate of TB within distinct decade-long birth cohorts over time, wherein a similar degree of M.tb exposure existed within each birth cohort. In these studies as well as the adolescent vaccine study, TB incidence in the general population was decreasing dramatically throughout the course of the study, leading to a continual decreased risk of reinfection. If waning immunity or other factors in old age caused an excess of TB disease from their initial remote infection, we should observe an increase in

TB disease rates at some point in advanced age. In almost every such cohort study, this was not observed (114–116). There was a decay of TB disease rates throughout the lifespan, well into old age (75+ years) (115, 116). Most of these cohort studies were of populations in currently developed countries, and thus represent longitudinal TB trends in a background of a dramatic reduction of transmission to very low levels over the course of the study.

In many developed countries, including the United States, the case rate of TB is highest in the elderly (> 65 years old) (12, 117). The preceding discussion argues that the cause for this is highly likely to be because the elderly lived through periods when TB transmission was much higher compared to currently living younger birth cohorts, rather than a biological susceptibility of the elderly to develop TB disease from a remote

26 infection. A test for recent infection that identifies the first 1-2 years post-infection has a high likelihood of increasing the positive predictive value of any approach to defining TB disease risk in individuals. The value of such a test would generally increase as a person ages and their initial infection becomes more and more remote. The elderly are a particularly vulnerable population as to TB disease as they are the most likely to die from an episode of TB disease and are more likely to experience side effects to preventive and curative TB treatment (117–121). A recent study from South Korea highlighted that the elderly are more likely to be IGRA+ than younger individuals and that both IGRA and

TST positivity do not differ between household contacts of TB cases vs. contacts of non-

TB pulmonary disease cases among elderly contacts (122). Therefore the limited predictive value of TST/IGRA for TB disease risk is even lower in this population, showing that a test for recent M.tb infection is especially needed for this vulnerable population.

We have so far reviewed the evidence that TB disease is most likely to manifest in the first 1-2 years post-infection, and thus a test for recent infection could be very useful in predicting TB disease risk in many populations. Given that this phenomenon of highest risk early on is common to many countries and time periods, representing many populations of diverse genetics, as well as a diversity of bacterial genetics, and has persisted despite dramatic changes in TB control efforts during the past century, it is reasonable to assume that it has a biological basis. The various components of the

27 biological basis of this phenomenon are important as these would be the targets of measurements to assess for a test(s) for recent infection.

A recent genetic study of early progression to TB disease (within 18 months) in

Peruvians demonstrated that the genetic architecture of early progression and disease from more remote infection are different (123). We will now outline several of the known biological events of early infection through which these genetic factors may mediate their effect in the host-pathogen interaction. This knowledge relies predominately on work in various animal models of M.tb infection, since there is no current test for recent M.tb infection that would allow sufficiently precise studies in humans.

Known biological events of recent M.tb infection

M.tb infects the host upon respiration when it overcomes the numerous physical barriers to particulates in the upper and lower airways and is able to reach the end of the lower airways, the alveoli where gas exchange occurs (124). The predominate cell types in healthy alveoli are epithelial cells and alveolar macrophages (124). A recent study in mice showed that within 48 hours of aerosol infection, both alveolar macrophages and epithelial cells were M.tb infected, but most infected cells were macrophages despite epithelial cells vastly outnumbering macrophages in the alveolar space (125). By day 8 post infection in this model, alveolar macrophages comprised all measurably infected cells, with no M.tb detected in epithelial cells (125). It was found that infected alveolar

28 macrophages traversed the airway epithelium into the lung interstitium in a manner dependent on secreted bacterial virulence factors and host inflammatory signaling (125).

In the interstitium these previously infected alveolar macrophages proliferated and comprised an initial cellular aggregate, being gradually replaced by macrophages of monocytic origin and neutrophils (125). Only following the migration into the interstitium did the bacteria disseminate to recruited neutrophils and macrophages of monocyte origin, which became the overwhelmingly predominate infected myeloid cell types by 21 days post-infection (125).

Using PET/CT imaging and individual bacteria DNA barcodes, Martin et al. provided key insights on the early lesion-level events of M.tb infection in cynomolgus macaques

(126). While most inoculated bacteria produced initial lesions of infection, with no evidence of host resistance to initial bacterial replication, only a very small proportion of inoculated bacteria disseminated locally in the lung to form new lesions or disseminated to thoracic lymph nodes, with the majority of lesions successfully containing the bacteria during the first 6 weeks of infection (126). Previous studies in mice had shown that dendritic cells of monocytic origin transfer live M.tb to the lung-draining, thoracic lymph nodes, where M.tb-specific T cell priming via uninfected dendritic cells occurs once a threshold number of bacteria are present in the lymph node (125, 127–130). The increased monocytic infiltration into the lung is accomplished by increased egress of monocytes from the bone marrow (127). Similarly in humans, it was historically observed that a tuberculous pulmonary lesion was almost always accompanied by a

29 tuberculous thoracic lymph node (131, 132). It was recently shown in both rhesus and cynomolgus macaques that thoracic lymph nodes remain infected with live M.tb during chronic M.tb infection with limited killing ability, are susceptible to complete destruction of their architecture from tuberculous lesions, and are less amenable to drug-mediated sterilization than lung lesions (131).

It has been proposed, but not yet proven, that a large quantity of pulmonary disease resulting in cavities and/or transmission is a function of M.tb that reseeds the lung via retrograde lymphatic transport from the infected thoracic lymph nodes (133). This could explain why the more highly transmissible TB pathology, such as cavities, more often occurs in the lung apex while the primary site of infection is more commonly near the lung base (133). It is clear from animal and human studies that M.tb often disseminates throughout the body and is possible to be found in the spleen, the gastrointestinal tract, the liver, the genitourinary tract, the kidneys, the meninges, the pericardium, the bone marrow, adipose tissue and other organs (43, 134–142). Apart from cases of clinical extrapulmonary tuberculosis, which comprise 15% of TB diseases cases globally, recent findings in individuals with latent tuberculosis and/or treated tuberculosis have focused mostly on mycobacterial DNA, sometimes RNA, in histologically normal tissue, which represents a limitation in that viability of the bacteria has not usually been fully demonstrated (12, 134, 136, 140–142). M.tb probably spreads from the lung to extrapulmonary tissues via the hematogenous route (136). This could occur by leakage into the pulmonary vasculature, or by efferent lymphatics or vasculature access once the

30 bacteria reach the thoracic lymph nodes (138). A detailed study in mice on the day-to-day timing of M.tb dissemination showed that in two strains of mice bacteria were culturable in the spleen and liver only after being culturable from the pulmonary lymph node, several days after the bacteria had already reached high levels in the lung (138). Adaptive immunity in the spleen only developed after dissemination to this organ, similar to what has been shown in more detail in the lymph node (130, 138). This study’s detailed, daily analysis argues against tissue damage and lung vasculature infiltration as a route of dissemination, with broad dissemination only occurring in a temporal sequence after bacteria reached the pulmonary lymph node (138).

It must be noted that most detailed mouse studies have been performed on inbred mice and thus do not capture the full genetic diversity of possibilities in outbred animals or humans, nor do they reflect the diversity in real-world circulating M.tb strains. Therefore, the actual events of M.tb dissemination, the likelihood of producing infection and immune responses in extrapulmonary sites and the true duration of infection at these sites

(and the lung) are virtually unknown in humans. In one series of cynomolgus macaques infected with M.tb, macaques who did not develop active TB disease had no extrapulmonary lesions (143). This uncertainty in humans is very important in considering possible host biomarkers to test for correlation with recent infection, as the search for a noninvasive, easily sampled test naturally begins in the blood, which is subject to influence from immune events all over the body.

Once adaptive immunity is initiated in the lymph nodes, T cells and B cells migrate to the lung and form a rim around the focus of infected and uninfected myeloid cells (144, 145).

The B cells can form a germinal center-like follicle (144). The total structure is called a granuloma and is a feature of other diseases, including diseases not thought to be infectious in origin such as sarcoidosis. The interior of the granuloma contains macrophages, which can include foamy and epithelioid macrophages and multinucleated giant cells, neutrophils and other immune cells (145). Central necrosis of the granuloma often occurs and can caseate (145). In cynomolgus macaques, adaptive immunity has a high capacity to sterilize granulomas, with 58% of granulomas in monkeys with LTBI being sterile in one study (146). Sterilization of granulomas also occurred in cynomolgus macaques who developed active TB, with nearly 40% of granulomas in monkeys with active TB being sterile (146). Efficacy in bacterial killing showed high lesion-to-lesion variability within individual monkeys. Moreover, bacterial loads were lower and sterilization percentages highest at time of necropsy in both LTBI and active TB animals compared to animals early on in infection at 4 weeks and 11 weeks post-infection (146).

Accordingly, in this study there was a correlation in granuloma histopathology type and killing capacity with time since infection in monkeys diagnosed with active TB, with caseous lesions predominating early (< 100 days) and exhibiting poorer M.tb control, with a switching to predominately fibrocalcific lesions at later times post-infection (> 200 days), associated with improved M.tb control (146).

The granuloma and its various forms do not represent the only important pathology of human or animal tuberculosis. In the aforementioned study of lesion-level dynamics in macaques, the most permissive pathology to M.tb growth was a tuberculous pneumonia observed in some monkeys with active disease (146). TB pneumonia showed absolutely no evidence of bacterial killing via comparison of bacteria colony forming units (CFUs) and DNA quantitation (146). Cavities were also observed in a minority of monkeys with active disease in a previous study in cynomolgus macaques (143). A review of historical reports of TB pathology from the pre-antibiotic era when such human specimens were plentiful together with contemporary cases of untreated TB has reminded the field that the primary pathology of pulmonary TB disease in those who have developed a competent adaptive immune response is an obstructive lobular pneumonia (147–149).

This pneumonia can then result in cavities in some persons. While cavities in M. bovis infection do result from erosion of caseating granulomas, this has never been observed in the case of M.tb infection in humans (147, 148). Surrounding granulomas and fibrocaseous disease can form as a reaction around TB pneumonia, and cavities can result as a dissolution of caseous pneumonia (147, 148). The mechanisms of TB pneumonia formation and the host and bacterial factors governing susceptibility to it are almost completely unknown. It is hypothesized that the pathologic process begins with a low level of bacteria infiltrating and surviving in alveolar macrophages that gradually accumulate mycobacterial antigens and host lipids in obstructed bronchioles (148).

To summarize this discussion of known and unknown biological events during the initial host response to M.tb infection, a test for recent infection could target aspects of the immune response that occur at different stages in the outlined sequence of events. It is conceivable that aspects of both the innate and adaptive immune response could be measured noninvasively in the blood. It is also possible that radiologic imaging could capture certain pathologic features, though narrowing the focus solely to the lung may not capture all the relevant events of early infection. While it has been shown, contrary to many historical assumptions, that there is no difference in general radiologic features of

TB disease at time of diagnosis between recent and remote infection identified by bacterial genotyping, the presence of abnormal X-ray findings during the first 2 months post-infection in cynomolgus macaques was strongly associated with the development of active disease and there may be observable radiologic features of incipient TB in humans

(143, 150–152). There is likely to be considerable heterogeneity between people in the various manifestations of the immune response and/or radiologically measurable pathology, but the events culminating in and directly resulting from M.tb dissemination to thoracic lymph nodes and initiation of an adaptive T cell response are likely to be common across many individuals given that these events are required for TST/IGRA conversion and these tests have an exceedingly high negative predictive value for TB disease risk.

Focusing on aspects of the host immune response that are measurable in blood, a test for recent infection could attempt to measure aspects of both innate and adaptive immunity

34 involved in mounting the initial immune response to M.tb. Several aspects of the initial innate and adaptive immune response are likely to be transient in duration and common to both those who develop TB disease and those who control the infection. By contrast, the IFN- T cell response measured by TST/IGRA is a stable component of this early immunity and thus not useful for determining time since infection. However, it is possible that several aspects of innate and adaptive immune perturbation in early M.tb infection begin to quantitatively decay as M.tb is controlled in individuals who do not develop TB disease. We will now describe several of these immune factors and their role in M.tb control in more detail. We will also highlight how some of these immune factors are altered in the aging process.

Immune mechanisms of M.tb control

M.tb is an intracellular pathogen, and as has been described, cell mediated immunity is critical for restricting its replication in phagocytes. Using M.tb containing plasmids that were degraded with bacterial replication in the mouse model, it was shown that innate immunity provides only modest bacterial killing before the onset of adaptive immunity at

3-4 weeks post-infection, with bacterial replication being similar as in vitro during the 1st month of infection (153). With the onset of adaptive immunity in the lungs, which corresponds to M.tb-specific CD4+ T cell infiltration, bacterial replication is slowed dramatically and killing increased (153). While in most observations of granulomatous lesions CD4+ T cells only infiltrate into the myeloid cell-dominated center infrequently,

35 it has been shown that direct cell-to-cell interaction through the Major Histocompatibility

Complex II (MHCII) is needed between CD4+ T cells and myeloid cells in the granuloma for M.tb control (154, 155). Direct cell-to-cell interaction enables cytokine-mediated activation and/or cytolysis of M.tb-infected macrophages and dendritic cells by CD4+ and CD8+ T cells (156–158).

In humans several monogenic disorders underlie the clinical syndrome Mendelian susceptibility to mycobacterial diseases (MSMDs). The genes causing this disorder are all involved in the IFN- pathway, which is the predominant cytokine produced by M.tb- specific T cells (41, 42, 159). These include both genes involved in influencing the production of IFN- by T cells (IL12B, IL12RB1, IRF8, ISG15, NEMO) and those involved in the response of infected cells (IFNGR1, IFNGR2, STAT1, IRF8, CYBB) (42,

159). IFN- is produced not only by αβ CD4+ and CD8+ T cells in response to IL-12, but it is also produced by γδ T, B, NK, ILC1, and ILC2 cells (160). By contrast, NKT, which are often CD1d-restricted, and MAIT cells, which are MR1 restricted, preferentially produce IFN- in response to IL-23 (160). Moreover, the development of IFN-- producing CD4+ T cells depends on both IL-12 and IL-23 (160). Consequently, it was recently shown that deficiency in several proteins resulting in unique impairment of IFN-

 production via IL-23 signaling, including TYK2 and IL-23R, can also result in MSMD or tuberculosis (160, 161).

It is important to note that the syndrome of MSDM is characterized by early-onset, disseminated, life-threatening infections with BCG or environmental mycobacteria, not pulmonary TB (42). While studies of twins and families as well as animal models provide evidence that susceptibility to pulmonary TB has a strong genetic component, genome- wide association studies have not yielded associated common variants with replicability across different settings and ethnicities (159). Taken together with the truth that the majority of persons who develop pulmonary TB disease develop a IFN- T cell response, these results suggest that a deficiency in this critical arm of antimycobacterial immunity does not explain the pathogenesis of pulmonary TB disease. Indeed, a recent study in mice showed that IFN- accounted for ~30% of CD4+ T cell-dependent bacterial control in the lungs, but >80% in the spleen during the first 6 weeks of infection (43). Enhancing the IFN--producing capacity of CD4+ T cells exacerbated lung infection, and PD-1- mediated inhibition of T cell function was required to prevent mortality driven by IFN-

 from T cells (43). In macaques, only about 8% of T cells from granulomas were shown to respond with cytokine production after ex vivo stimulation, despite the general effectiveness of these granulomas at controlling M.tb growth (146, 162). In contrast to the pattern of IFN--producing CD4+ T cells to mediate extrapulmonary control of M.tb, it was recently shown that CD153, a member of the TNF superfamily, is a major mediator of CD4+ T cell-dependent control in the lung (163). CD153 expression by CD4+ T cells was independent of Th1 differentiation, despite its expression predominately in Th1- polarized cells, and the downstream mechanisms of its role in M.tb suppression remain to be elucidated (163). However, this finding highlights the fact that heretofore

37 uncharacterized aspects of CD4+ T cell-mediated immunity to M.tb may yet explain differential susceptibility of people to pulmonary TB disease.

While CD4+ and CD8+ T cells are among the most important cells for protective immunity to M.tb, CD4 and CD8 T cell epitopes of M.tb are hyperconserved, being even more conserved than essential genes (164–166). The vast majority show absolutely no amino acid variation (164, 165). Therefore, M.tb T cell epitopes reflect no immune evasion and are strong evidence that M.tb requires a robust T cell response for its ultimate survival and transmission. M.tb uses other mechanisms to evade T cell immunity in individuals, with inhibition of antigen presentation being chief among them (167–169).

In light of our incomplete knowledge on the role of T cells in the ultimate outcome of

M.tb infection in people, it is difficult to predict how known alterations of T cells with aging may affect the susceptibility of elderly individuals to M.tb infection. Naïve CD4+ and CD8+ T cells decrease with age, though being much more pronounced in CD8+ T cells (170, 171). This difference may not be the same in blood vs. spleen and lymphoid tissues, with an analysis of many human tissues from recently deceased individuals over the lifespan showing a similar decrease in naïve CD4+ and CD8+ T cells in lymphoid tissues (171, 172). Naïve T cells in humans are maintained in lymphoid tissues over the lifespan by clonal expansion within their resident lymphoid tissue (173). With homeostatic proliferation at low levels for naïve and memory T cells, it is thought that loss of T cell numbers or T cell receptor (TCR) diversity does not explain a decline in

38 adaptive immune responses in healthy elderly individuals, except for naïve CD8+ T cells

(171). For example, in a study of elderly people vaccinated with the live varicella zoster vaccine, many naïve T cells were recruited in response to the vaccine, with less boosting of dominate memory T cell clones (174).

While the pool of naïve and memory T cells may be adequate for adaptive immune function in old age, impairments in cell cycle control and DNA repair likely cause the observed more rapid decline of effector T cells after their peak proliferation and impaired new memory cell formation (171). Moreover, the heightened inflammatory state of old age, often termed inflammaging, likely contributes to reduced antigen presentation by myeloid cells as well as impairment in T follicular helper cell generation, thus decreasing the ability of T cells to coordinate antibody responses by germinal center B cells (171,

175–177). It is not yet clear how these impairments in T cell immunity with age may influence susceptibility to M.tb in humans who are exposed in old age. While previous work in an intravenous route M.tb infection model has shown that aged mice are more likely than young mice to die from a primary M.tb infection and delayed recruitment of

CD4+ T cells to the spleen contributes to this susceptibility, no study has demonstrated a clear defect in CD4+ or CD8+ T cell immunity to M.tb in aged mice using a pulmonary infection model (178–180). In one study of aerosol M.tb infection, old mice had higher bacterial burden at 2 months post infection than young mice but with equal numbers of

Ag85-specific T cells (180). Furthermore, boosting of CD4+ T cell responses to M.tb with rapamycin did not delay disease or alter survival in genetically diverse old mice

(181). Therefore, while old mice are more susceptibility to pulmonary TB disease when infected in old age, it is unclear to what degree alterations in T cell migration and function influence their poorer disease outcome.

In contrast to T cells, B cells and antibodies have been historically underrepresented in

TB immunologic research, likely due to the fact that early passive serum transfer experiments gave inconsistent results across studies and species (144, 182, 183).

Depletion of B cells with CD20-targeting rituximab does not have a significant impact on

TB risk in humans or macaques, with the important caveat that this reagent has only negligible impact on long-lived, antibody-secreting plasma cells, which do not express

CD20, in the bone marrow (182, 184). In contrast, recent work has shown that there are differences in antibody structure and function between different TB disease states in humans (59, 183, 185, 186). Differential antibody glycosylation in the Fc region and the ability to enhance human macrophage control of M.tb in vitro discriminated antibodies from individuals with LTBI from those with TB disease (183). Another recent study showed that antibodies from some highly exposed but healthy healthcare workers in a TB hospital but none from patients with TB disease conferred modest protection when transferred to mice subsequently infected with M.tb (185). Protection of these antibodies in a whole blood in vitro assay depended on CD4+ T cells and immune complex formation (185). Nearly half of the healthy individuals from whom these protective antibodies were isolated were IGRA-, and the observation that heavily exposed IGRA- persons can develop M.tb-specific antibodies, including class-switched antibodies, has

40 been reproduced in a larger study elsewhere (59, 185). These studies show that most, if not all, M.tb infected individuals possess antibodies specific to M.tb, regardless of their protective effect, and even many individuals who are LTBI- with current tests make

M.tb-specific antibodies, suggesting prior exposure.

Because neutralizing antibodies are strong correlates of immune protection for many infections for which we have vaccines, especially viruses, the functional impact of the decline of B and/or plasma cell function with aging is clearer than that for T cells. There is a decline in bone marrow plasma cells in humans and mice with age, and this correlates with reduced peripheral antibody levels to vaccine antigens for pathogens with unlikely continued natural exposure (187, 188). As for responses to new infections in old age, germinal centers are decreased in size and number, and the ability of B cells to undergo somatic hypermutation is dramatically compromised (187, 189). Given the unclear role for antibodies, B cells and plasma cells in determining susceptibility to TB disease in both animals and humans, it is difficult to extrapolate what the impact of aging for these compartments of adaptive immunity is on general TB susceptibility. Nevertheless, for those individuals for whom antibodies and/or B cells are likely to play a protective role, undergoing M.tb exposure in old age is likely to be negatively impacted by these aging defects.

As described in our discussion of the sequence of immunological events characterizing initial M.tb infection, it is clear that M.tb engages most arms of the innate and adaptive

41 immune systems. Beyond T and B cells, these include but are not limited to alveolar macrophages, recruited monocytes, neutrophils, dendritic cells, NK, ILC, NKT, MAIT and γδ T cells. However, just as with T and B cells, it is unknown to what degree differences in the function of these immune cell subsets underlies differential susceptibility of people to develop TB disease. In a recent multi-cohort study in humans,

NK cell proportion in peripheral blood was a more robust correlate of TB disease states than B or T cell proportions (190). NK cell proportion was increased in LTBI, decreased during the progression to active TB disease and returned to baseline levels upon clinical cure (190). While NK cells have a functional role in killing M.tb-infected cells and confer some protection in mice with T cell deficiency, depleting NK cells from immunocompetent mice did not alter infection outcome (190–193). While these mice studies do not reflect human genetic diversity and thus are to be interpreted with caution, they do reflect the possibility that the most robust correlates of TB disease state may not reflect causality. To briefly summarize the role of the other immune cell subsets, neutrophils may have an early protective role in M.tb infection but also correlate with later severe disease in animals (194). Alveolar macrophages are more permissive for M.tb infection than recruited monocyte-derived macrophages (195). ILC3s but not ILC1s or

ILC2s were recently shown to have an early protective role in M.tb infection in mice as well as correlate with TB disease and successful treatment (196). NKT cells can provide partial protection in some susceptible mouse strains, and NKT levels and function have been reported to be diminished in patients with TB disease (197, 198). γδ T cells have been recently shown to provide partial protection in nonhuman primates via adoptive

42 transfer or targeted vaccination (199, 200). By contrast, MAIT cell accumulation in the

M.tb-infected lungs of rhesus macaques are inconsistent and minimal, and show limited evidence of activation in granulomas (201).

Given the early stage of research with some of these immune cell subsets in TB pathogenesis, it is even less clear how aging could affect their potential role in the elderly. Macrophages isolated from the lungs of uninfected old mice have been shown to be at an increased inflammatory state compared to macrophages from young mice (202,

203). Old lung macrophages secreted more pro-inflammatory cytokines, and enhanced

M.tb uptake and phagosome-lysosome fusion upon in vitro M.tb infection (202).

Moreover, despite the increased susceptibility of old mice to M.tb infection, they manifest better M.tb control during the first 3 weeks of infection due to IFN- production by CD8+ T cells via an antigen-independent mechanism (204–207). In contrast, it has recently been shown that old mice have an increased population of pro-inflammatory

CD11b+ CD11c+ alveolar macrophages of monocytic origin that are more permissive to

M.tb infection and growth (208). The overall contribution of this M.tb-growth permissive alveolar macrophage subset to susceptibility of old mice to M.tb remains unclear. In

Chapter 2 we will examine in further detail inflammatory monocyte changes in old age that may affect susceptibility to M.tb infection in humans.

It is important to note that we expect that any of the above discussed immune components involved in the interaction of the host with M.tb could be discovered to be

43 effective correlates of recent infection for targeting treatment to those most at risk of disease. The success of such a correlate for this application doesn’t depend on its protective role during infection, but rather on its timing, duration and generalizability across different populations of people. For use in the elderly it would ideally be present in elderly people who are exposed in old age. We do not expect this to be an issue because while there is the general decline in immune function in old age, it is not concentrated dramatically in any one arm of immunity, being subtler in nature, as has been discussed.

Preliminary immune correlates of recent M.tb infection in humans

There are very few published studies on host immune components that discriminate recent from remotely acquired M.tb infection in humans. One of the few studies to attempt this focused on the hypothesis that bacterial replication is less restricted in both active TB disease and during early infection regardless of disease outcome and thus may affect effector T cell responses (209). They measured 3 T cell immune signatures assessing the memory and maturation phenotype of M.tb-specific T cells, all of which had previously and in their study discriminated active TB disease from remotely acquired

LTBI very accurately (209). Recent infection was defined epidemiologically by IGRA or

TST positivity and contact with a TB case in a low incidence country in the last 6 months, while remote infection was defined as IGRA+ or TST+, birth in a country of high TB burden and habitation for >2 years in a low incidence country with no known

TB contact during that time (209). Only 1 of the T cell immune signatures, the proportion

44 of PPD-specific CD4+ T cells secreting TNF- but not IFN- or IL-2 and having a differentiated effector memory phenotype (CD45RA-CCR7-CD127-), discriminated recent from remote LTBI per their case definition (209). However, this T cell signature was tested with a relatively small sample size, was not tested in an independent cohort, and there were relatively large confounders of sex, ethnicity, occupation and TST result between those defined as having recent vs. remote LTBI (209).

Another recent study showed that myeloid derived suppressor cells (MDSCs) are increased in both active TB diseases cases and household contacts of TB cases relative to both TST+ individuals with no recent exposure and TST- individuals, although they did not assess formally to what degree MDSC levels discriminated these populations (210).

A different study found an alternative pattern for recent vs. remote infection screening antigens in the M.tb DosR regulon (211). In this study the IFN- response to Rv2628 was uniquely higher in people with LTBI who were exposed over 3 years ago compared to recent contacts, active TB patients and uninfected controls (211).

A few cross-sectional studies have been published suggesting differences in antibodies to

M.tb between recent and remote infection. Two studies have shown that TST conversion within the past year is associated with uniquely higher IgM antibodies to M.tb HspX protein, with no difference among those with active disease, TST- individuals and TST+ individuals without a documented recent conversion (212, 213). In another cross- sectional study, IgA and IgM to M.tb heparin-binding hemagglutinin (HBHA) was

45 distinct in controls without documented recent exposure (53% IGRA+), as compared to

TB patients and recent contacts (72% IGRA+) who had similar levels (214). This is the same pattern as that described for the IGRA response to Rv2628 (211).

None of these studies reported follow-up sampling on recently exposed individuals to determine when their immune signatures returned to levels of remotely or unexposed individuals (or the opposite for Goletti et al or Belay et al). Likewise, they did not attempt nor were powered to test an association between these potential biomarkers of recent infection and TB disease risk. There is a clear need for formal and directed investigation to identify biomarkers of recent M.tb infection in humans and to test whether these biomarkers can be used to help estimate TB disease risk and thereby aid targeting preventive treatment to those truly at risk. Longitudinal analysis of individuals would provide critical benefits as it would mitigate the risk for many confounding factors inherent to these cross-sectional observation studies; longitudinal analysis also importantly allows for testing of whether biomarkers of recent infection change as the exposure becomes more remote and potentially allows for correlation with TB disease risk.

The following chapters of this dissertation focus on the two extremes of recent and remote M.tb infection, contributing to the search for useful biomarkers for predicting TB disease risk in humans. Chapter 2 evaluates the hypothesis that there are changes in peripheral immunity to M.tb in old age that may explain the higher case rate of TB in

46 elderly persons. We did not observe any strong evidence for such a phenomenon, which concords with the cohort data that TB risk does not increase as a person ages. In Chapter

3 we perform a multi-species analysis showing that RNA expression in the blood changes soon after recent M.tb infection, thereby being a potential biomarker of recent infection.

However, these RNA expression changes are likely too brief in duration to help inform

TB disease risk in programmatic conditions for targeting TB preventive therapy. In conclusion, we propose the need to find additional biomarkers of recent M.tb infection that have sufficient duration to capture the majority of people who are at increased risk during the first 1-2 years post-exposure. Such a biomarker would have important practical implications in treating those at highest risk to prevent TB and in tracking TB transmission in communities to help inform public health efforts.

Chapter 2 Impaired Peripheral T Cell Immunity does not Explain Susceptibility of the Elderly to Develop Tuberculosis

Abstract

Tuberculosis (TB) is the leading killer due to a single infectious disease worldwide. With the aging of the global population, the case rate and deaths due to TB are highest in the elderly population. While general immunosenescence associated with old age is thought to contribute to the susceptibility of the elderly to develop active TB disease, very few studies of immune function in elderly individuals with Mycobacterium tuberculosis

(M.tb) infection or disease have been performed. In particular, impaired adaptive T cell immunity to M.tb is one proposed mechanism for the elderly’s increased susceptibility primarily on the basis of the decreased delayed type hypersensitivity response to tuberculin-purified protein derivative in the skin of elderly individuals. To investigate immunological reasons why the elderly may be susceptible to develop active TB disease, we performed a cross-sectional observational study over a five year period (2012-2016) enrolling participants from 2 age groups (adults: 25-44 years; elderly: 65 and older) and 3

M.tb infection statuses (active TB, latent TB infection, and healthy controls without history of M.tb infection). We hypothesized that impaired peripheral T cell immunity plays a role in the biological susceptibility of the elderly to TB. Contrary to our

48 hypothesis, we observed no evidence of impaired M.tb-specific T cell frequency or altered production of cytokines implicated in M.tb control (IFN-, IL-10) in peripheral blood in the elderly. Instead, we observed alterations in monocyte proportion and phenotype with age and M.tb infection that suggest their potential role in the susceptibility of the elderly to develop active TB. Moreover, our results highlight the need for further research into the biological reasons why the elderly may be more susceptible to disease and death from TB, as well as research into biomarkers that allow better assessment of an elderly person’s risk of developing active TB after infection.

Introduction

The global population is aging. It is estimated that by the year 2050, 17% of the global population will be 65 years of age or older, compared to 8.5% in 2015 (215). Increasing age is the major risk factor for a host of diseases including diabetes, cardiovascular disease, cancer and neurodegenerative disease (216, 217). Increasing age is also a major risk factor for infectious diseases (218, 219). Tuberculosis (TB), caused by the airborne pathogen Mycobacterium tuberculosis (M.tb), is currently the leading killer due to an infectious disease, and increasing age is a significant risk factor for disease and death due to TB. For example, in the United States, the case rate of symptomatic, active TB is highest in those ≥ 65 years of age, and this age group is the most likely to die from TB

(117). Worldwide, the case rate is similarly highest among those ≥ 55 years of age (220).

Moreover, according to the 2016 Global Burden of Disease estimate, 62% of deaths due

49 to TB globally occurred among people older than 50, with more than half of these deaths occurring in those ≥ 65 years of age (221). We can expect that as the global population continues to age, the global burden of TB in the elderly will also increase (222).

Infection with M.tb is thought to confer a ~10% lifetime risk of developing symptomatic, transmissible active TB, with the remainder of individuals controlling the bacteria in a state known as latent TB infection (LTBI) (13). While the prevalence of active TB has decreased in the last 50 years, the risk that the elderly convert from LTBI to active TB poses an additional challenge to TB elimination efforts in the foreseeable future (13, 223–

225). Research into the clinical and biological reasons for why the elderly may be susceptible to develop active TB can help inform strategies to combat this present and future public health problem.

General immunosenescence associated with old age has been cited as one biological reason for the increased rate of active TB in the elderly (226–230). However, very few studies of immune function in elderly individuals with M.tb infection (LTBI or active

TB) have been performed. A historical hypothesis in the field is that impaired adaptive T cell immunity to M.tb plays a role in this biological susceptibility. This hypothesis stems from observations that reactivity to tuberculin-purified protein derivative (PPD) in the tuberculin skin test (TST), a measurement of a delayed type hypersensitivity (DTH) response for LTBI diagnosis, is reduced with advancing age past 65 years despite epidemiological evidence of higher exposure earlier in life (231, 232). A few articles

50 have shown that elderly patients with active TB had increased T-suppressor cell activity, or suppression of lymphocyte proliferation in vitro (233, 234). More recently, T cell responses to specific latency-associated antigens of M.tb have been investigated in the elderly, though without specific comparison to younger cohorts (235). These studies, taken together, do not provide definitive proof that age-related changes in T cell function contribute to risk of active TB in the elderly.

To provide additional investigation of immunological reasons why the elderly may be susceptible to develop active TB, we performed a cross-sectional observational study over a five year period (2012-2016) enrolling participants from 2 age groups (adults: 25-

44 years; elderly: 65 and older) and 3 M.tb infection statuses (active TB, LTBI, and healthy controls without history of M.tb infection). We hypothesized that impaired peripheral T cell immunity plays a role in the biological susceptibility of the elderly to

TB. Contrary to our hypothesis, we observed no evidence of impaired M.tb-specific T cell frequency or altered production of cytokines important for M.tb control in peripheral blood in the elderly. Instead, we observed alterations in monocyte proportion and phenotype with age and M.tb infection that suggest their potential role in the susceptibility of the elderly to develop active TB. Our results highlight the need for further research into the biological reasons why the elderly may be more susceptible to disease and death from TB. Additionally, research into biomarkers that allow better assessment of an elderly person’s risk of developing active TB after infection could be critical to addressing the needs of elderly persons with M.tb infection.

Materials and Methods

Subject Recruitment

Study participants were from two age groups, 25-44 years and 65 years and older.

Subjects were recruited from The Ohio State University Wexner Medical Center and

Columbus Public Health. Eligible subjects were enrolled in the inpatient and outpatient setting and consented according to an IRB approved protocol. Subjects’ medical histories were obtained by a questionnaire at study enrollment, and this information was verified and additional information obtained by review of their medical records. Clinical data collected for active TB/LTBI diagnosis included signs and symptoms of active TB, physical examination, QuantiFERON®TB-Gold Test (QFT) (Qiagen, USA), acid fast bacilli smears, M.tb PCR, cultures and imaging studies if available. Most subjects with active TB/LTBI were recruited at or before the day they started receiving treatment for active TB/LTBI. Upon study enrollment, peripheral blood was collected in lithium heparin coated tubes (Becton, Dickinson, NJ). Time from blood collection to peripheral blood mononuclear cell (PBMC) isolation was ≤ 4 hours.

PBMC Isolation

PBMCs were isolated by Ficoll-Paque (GE Healthcare Life Sciences) density centrifugation and washed twice before use in subsequent assays. Briefly, whole blood was gently overlaid onto Ficoll-Paque and centrifuged at room temperature for 22 minutes at 1000 x g, with no break so as not to disturb the separated layers. The upper plasma layer containing platelets was carefully removed before isolating the white PBMC layer into a new tube. Washes were performed with RPMI-1640 with L-glutamine

(Gibco), pelleting cells via centrifugation at room temperature for 7 minutes at 600 x g.

Measurement of ESAT-6 and CFP-10 specific IFN-+ T cells

The T-SPOT®.TB (Oxford Immunotec, UK) ELISPOT kit was used to measure the frequency of ESAT-6 and CFP-10 specific IFN-+ T cells in PBMCs, per the manufacturer’s protocol. Spots were counted manually by two lab members independently. The results gave identical trends between counters (results of one count are shown). In accordance with the manufacturer’s protocol, the data shown are from counts with the count of the negative control well subtracted.

PBMC Culture and Stimulation

2.5 x 105 PBMCs were cultured in triplicate in AIM-V media (with Albumax, 50 g/mL

Streptomycin, 10 g/mL Gentamycin) (GIBCO) with 10 g/mL H37Rv culture filtrate protein (CFP) (BEI Resources) or either media only or 10 g/mL concanavalin A

(Sigma-Aldrich) as negative/positive controls, respectively. Cell-free supernatants were removed after 48 hours culture at 37°C, 5% CO2 and stored at -20 °C until further analysis.

Measurement of Cytokine Levels

Human IFN-, IL-10 and IL-4 were measured in PBMC culture supernatants using

ELISA kits (BD Biosciences), according to manufacturer’s instructions. Samples outside the linear range were serially diluted and re-assayed. Some samples were thawed twice.

IFN- data from the first 72 recruited subjects were excluded from analysis due to the assay falling outside the linear range of the standards and samples being insufficient for further testing.

Flow Cytometry

5 x 105 PBMCs were washed in flow buffer (RPMI-1640, 10.4 mM HEPES, 0.1% sodium azide), stained for 20 minutes at 4°C with antibody and washed twice more.

Antibodies used were: CD4-APC-Cy7 (Clone: RPA-T4), CD8-PE (Clone: HIT8a),

CCR7-PerCP-Cy5.5 (Clone: 150503), PD-1-APC (Clone: MIH4), CD16-FITC (Clone:

3G8) and CD14-PerCP-Cy5.5 (Clone: M9). All antibodies were purchased from BD

Pharmingen. Data were acquired on a BD FACS Canto II with FACS Diva software (BD

Biosciences). 50,000 events per sample were acquired. The data were analyzed using

FlowJo software (Version 10.3.0). Lymphocytes were identified by FSC/SSC before gating on CD4 and CD8. The CD4/CD8 ratio was calculated from these gates. CCR7 and

PD-1 positivity were determined using isotype controls. Monocytes were identified by gating on PBMCs using FSC/SSC before gating on CD14+ PBMCs. Monocyte subsets were subsequently identified by CD14 and CD16: classical (CD14++CD16-), intermediate (CD14++CD16+) and nonclassical (CD14dimCD16+).

Statistical Analysis

Data were analyzed using R v. 3.4.1 (R Foundation for Statistical Computing, Vienna,

AUT). The chi-squared test or ANOVA were used to assess univariate differences in clinical covariates between our main study groups, for categorical and continuous variables, respectively. The non-parametric Mann-Whitney U-test was used in all pairwise comparisons for T cell function assay data. For our T cell and monocyte flow cytometry data, the non-parametric Kruskal-Wallis test was used, after which Dunn’s post hoc test was used for pair-wise comparisons if the Kruskal-Wallis test gave p < 0.05.

To assess the contribution of confounding clinical covariates to these comparisons,

55 multivariate linear models were fit using the “lm” function in R and analysis of variance performed using the “Anova” function from the car R package (236). IFN- ELISA data were log-transformed before performing multivariate analysis.

Results

Cohort Description

We recruited a total of 205 subjects to our study, and a total of 169 subjects were included in the final analysis. Subjects excluded from analysis mostly included those whose samples were used for assay standardization or who had a previous history of treatment for active TB or LTBI (Figure 2.1). 6 study groups were defined for our primary analysis, with 2 age groups (adults: 25-44 years; elderly: 65 and older) and 3

M.tb infection statuses (active TB, LTBI, healthy controls without history of M.tb infection). All but 3 subjects with active TB were culture confirmed to have M.tb complex, with the other three diagnoses being clinical TB uveitis disease which improved on TB treatment and two subjects with pleural TB diagnosed based on abnormal chest x- ray and chest-computed tomography (CT) and improvement following TB treatment.

Subjects with LTBI were either clinically diagnosed by a positive QFT (n=39), or had a positive TST (n=1). All were deemed absent of active TB based on symptoms, chest x- ray, CT and/or negative sputum samples. All control subjects had a negative QFT result

(n=18) or no known QFT result with no prior history of a positive TST (n=98). No

56 control subjects had a clinical history of active TB. Moreover, most control subjects originated from a location of very low TB incidence.

Table 2.1 shows clinical covariates for our study subjects, and Table 2.2 shows culture confirmation and tissue site affected in our subjects with active TB. Due to our complex study design and available patient populations, not all covariates are balanced across all 6 study groups. For several of our immunological assays showing significant differences between age groups, we used multivariate analysis to assess the effects of our known clinical confounding variables.

T cell cytokine responses to M.tb antigens are not altered by age in those with LTBI or active TB

We first sought to determine whether among M.tb infected individuals, the elderly had decreased peripheral M.tb-specific T cell numbers and function compared to adults.

ESAT-6 is a virulence factor for M.tb and is released as a heterodimer with CFP-10 through a type VII secretion system (237, 238). ESAT-6 and CFP-10 are both immunodominant M.tb antigens and are not present in BCG or most nontuberculous mycobacteria (NTM), and thus T cell responses to these two antigens represent a specific response to infection with M.tb complex. We measured the frequency of ESAT-6 and

CFP-10 specific T cells using the T-SPOT.TB ELISPOT kit. Of note, peptide lengths

(15-mer) were designed to contain MHC II epitopes (239, 240). We observed no

57 significant differences in numbers of both ESAT-6 and CFP-10 specific IFN-+ T cells in the elderly relative to adults, in both latent and active TB groups (p>0.05, Figure 2.2A-

B). In fact, we observed a trend for increased ESAT-6 and CFP-10 specific IFN-+ T cells in elderly subjects with LTBI compared to adults.

It is possible that while the elderly have increased CFP-10 and ESAT-6 specific T cells in the blood, the amount of IFN- that these and other cells produce is decreased. Indeed, the ability of T cells to produce IFN- in response to the large collection of M.tb antigens contained in M.tb culture filtrate protein (CFP) is decreased in adult active TB cases with higher clinical disease severity (241). To test whether M.tb-specific IFN- production was reduced in the elderly, we stimulated PBMCs with M.tb CFP and measured supernatant

IFN- by ELISA. Contrary to our hypothesis, there was no significant difference in CFP- specific IFN- production with age, either among subjects with LTBI or active TB

(p>0.05, Figure 2.2C). We also measured IL-10 and IL-4 production by CFP-stimulated

PBMCs. We observed a trend for an increase in IL-10 production by CFP-stimulated

PBMCs from subjects with active TB relative to those with latent TB, consistent with what is known in the TB literature (241–243) (Figure 2.2D). However, we saw no significant differences in IL-10 production between age groups among those with LTBI or active TB (p>0.05). Moreover, we observed no change in IL-4 production between age groups (Figure 2.2E). Taken together, our data on M.tb-specific T cell frequency and cytokine production provide no evidence of M.tb-specific T cell dysfunction in the periphery in elderly people ≥ 65 years of age. While we did not measure many other 58 aspects of T cell function, including production of other cytokines, proliferation or cytotoxic function, IFN-, IL-10 and IL-4 are critical mediators of antimycobacterial immunity. IFN- is absolutely required, though not sufficient, for antimycobacterial immunity in humans, particularly in controlling extrapulmonary disease (42, 43). Both

IL-10 and IL-4 inhibit IFN--dependent immunity to mycobacteria through a variety of mechanisms, including inhibiting T-helper 1 cell differentiation and function (242, 244–

247).

Interestingly, we did observe a statistically significant decrease in IFN- production in response to CFP in the elderly who had no history of M.tb infection, relative to adults

(Figure 2.2C). This suggests potential hypo-responsiveness of the innate immune system to mycobacterial antigens in the elderly. This observation was not confounded by diabetes, cancer, immunosuppressive treatment or the presence of another comorbidity, using graphical analysis (data not shown). Accordingly, the effect of age alone remained significant using a multivariate linear model including these variables (p=0.004) or all categorical clinical covariates from Table 2.1 (p=0.001). In addition, BCG vaccination status had no effect (p>0.05).

Global T cell phenotypes are altered by M.tb infection and disease independent of age

We also studied global T cell phenotypes by flow cytometry in our cohort. Consistent with what is known in the clinical literature, the CD4/CD8 ratio was decreased in

59 subjects with LTBI and was further decreased in subjects with active TB relative to controls (data not shown). There was no significant difference in the CD4/CD8 ratio with age in any group (data not shown). We further sought to determine whether global T cell memory phenotype or inhibitory receptor status was altered by M.tb infection and age, by measuring CCR7 and PD-1 on CD4+ and CD8+ lymphocytes. CCR7 is expressed on naïve and central memory T cells, endowing them with the ability to home to lymphoid tissues, whereas effector memory T cells do not (248, 249). PD-1 negatively regulates T cell function and activation and can be associated with T cell exhaustion in chronic viral infections and cancer (250). However, the absence of PD-1 appears to promote TB disease in mice and humans through excessive immunopathology mediated through CD4

T cells (251–254).

We observed a trend for a decrease in both CD4+CCR7+ and CD8+CCR7+ lymphocytes in adults with LTBI and a trend for a decrease in adults with active TB, with significance reached for CD8+CCR7+ lymphocytes (p<0.05), relative to adult controls (Figure 2.3A-

B). However, we observed no significant difference in these parameters between the elderly and adults among those with LTBI or active TB (Figure 2.3A-B). Among our controls without history of M.tb infection, we did observe a significant decrease in

CD8+CCR7+ lymphocytes and a trend for decreased CD4+CCR7+ lymphocytes with age, consistent with the aging literature (255, 256) (Figure 2.3A-B). As for inhibitory receptor status, in a majority of subjects we measured no PD-1+ CD4+ or CD8+ lymphocytes (Figure 2.3C-D). There was a trend for increased PD1+ cells in both CD4+

60 and CD8+ lymphocytes in elderly with active TB relative to all other groups, though our low sample size in this group makes this conclusion tentative. Taken together with our cytokine production data, our data do not suggest a clear mechanism for impaired peripheral T cell immunity as a dominant cause of potential biological susceptibility of the elderly to active TB.

Old age and M.tb infection alter the monocyte/lymphocyte ratio and skew monocytes towards a nonclassical phenotype

Human monocytes are divided into three main subsets based on their phenotype: classical

(CD14++CD16-), intermediate (CD14++CD16+) and nonclassical (CD14dimCD16+)

(257). These subsets in humans are thought to exist on a developmental continuum, with classical monocytes egressing from the bone marrow before differentiating into intermediate (gain in CD16) and then nonclassical monocytes (reduction in CD14) (258).

Classical monocytes have increased phagocytic capacity, intermediate monocytes secrete more proinflammatory cytokines and nonclassical monocytes patrol the vascular endothelium (258–260). It is increasingly appreciated that monocyte phenotypes correlate with susceptibility to a variety of diseases (80, 257, 261–263). Thus, in addition to our examination of T cell phenotype and function, we sought to determine whether monocyte proportions and phenotypes were altered by old age and M.tb infection.

In children and adults, the monocyte/lymphocyte ratio is associated with risk of active TB and disease severity, with a higher ratio being associated with highest risk and a very low ratio also associated with increased risk (264, 265). Interestingly, we observed an increased monocyte/lymphocyte ratio in the elderly with active TB, relative to all groups

(p<0.05) (Figure 2.4A). We also observed a decreased monocyte/lymphocyte ratio in the elderly with LTBI, relative to both control groups and adults with LTBI (p<0.05 for controls and p=0.062 for adults with LTBI). In our examination of monocyte subset proportions defined by CD14 and CD16 expression, we observed a significant increase in the proportion of nonclassical (CD14dimCD16+) monocytes in elderly controls relative to adults, which has been observed by others (266–268) (Figure 2.4B). This increase remained significant when taking into account diabetes, cancer, immunosuppressive treatment and the presence of another comorbidity (p<0.05), although not when taking into account all categorical clinical covariates from Table 2.1 (p>0.05) (Table 2.3). Even more interesting, we observed a marginally significant increase in nonclassical monocytes in the elderly with LTBI, compared to elderly controls and adults with LTBI

(p=0.106 and p<0.05, respectively). This suggests the hypothesis of an age-dependent increase in nonclassical monocytes in LTBI. Mirroring the results in nonclassical monocytes, we observed a trend for a decreased classical monocyte (CD14++CD16-) proportion in elderly controls relative to adults and a trend for a decreased proportion in elderly with LTBI relative to elderly controls and adults with LTBI (data not shown). We observed no significant differences in intermediate (CD14++CD16+) monocyte proportions with age or M.tb infection status (data not shown). Contrary to our results

62 examining M.tb-specific T cell function in the elderly, our monocyte data suggest the hypothesis that changes in monocyte proportion and phenotype may contribute to susceptibility to develop active TB in the elderly.

Discussion

The global burden of TB in the elderly will increase as the global population continues to age (222). From available epidemiologic data (CDC, 117, WHO, 220, 222), it is clear that changes incident to age contribute to the likelihood of the elderly to die from an episode of active TB. In areas of low TB incidence, the increased likelihood of the elderly to develop active TB could be due to two effects. Epidemiologically, the elderly had higher exposure to TB in earlier years of life, when TB was more prevalent, compared to younger birth cohorts. Thus, a higher case rate in the elderly would simply reflect a larger reservoir of LTBI that could reactivate to active TB (13, 116, 269). Conversely, the increased case rate of TB in the elderly could be due to an age-related biological susceptibility to transition from a latent TB infection to active TB (115, 224, 225). In

South-East Asia, where new and relapse TB case notification rates are in general on par with case rates in Africa, those older than 55 years of age still have the highest incidence rate of TB, suggesting that the elderly may be biologically susceptible to developing active TB (115, WHO, 220, 270). Our cross-sectional observational study fundamentally cannot provide information to discriminate these two possibilities. Accordingly, our study assumes that the elderly are biologically more susceptible to developing active TB.

We hypothesized that impaired T cell immunity to M.tb plays a role in the biological susceptibility of the elderly to develop active TB. This has been a historical hypothesis in the field, stemming mainly from observations that reactivity to the PPD skin test, a measurement of a DTH response for LTBI diagnosis, is reduced with advancing age past

65 years despite epidemiological evidence of higher exposure earlier in life (231, 232).

Contrary to our hypothesis, we observed no difference between adults and elderly people with LTBI or active TB when we measured the frequency of IFN- producing T cells specific to ESAT-6 or CFP-10 or the amount of IFN- or IL-10 produced in response to

M.tb culture filtrate protein in PBMCs. Similarly, we saw no statistically significant effect of age on CCR7+ subsets of CD4 or CD8 T cells among those with LTBI or active

TB. Therefore, among these measurements of M.tb-specific and global T cell phenotype and function, we saw no evidence of impaired peripheral T cell immunity as a dominant cause of biological susceptibility of the elderly to develop active TB. The discordance of our functional T cell data with these historical PPD results in the elderly is consistent with the work of Mori and Kobashi which demonstrated that though PPD positive results are dramatically reduced in patients over the age of 80 who have culture-positive TB, their reactivity to a CFP-10 and ESAT-6 IFN- release assay using peripheral blood is not diminished relative to younger adults (271, 272). It is important to note that ESAT-6 and

CFP-10 are immunodominant M.tb antigens and are not present in BCG or most nontuberculous mycobacteria (NTM), including M. avium. We chose M.tb culture filtrate protein, which does contain many epitopes in NTM and BCG, to measure total cytokine

64 production in subject’s PBMCs to assess the wider antigenic response to M.tb.

Importantly, the median LTBI and active TB subject IFN- response to culture filtrate protein was more than 10-fold higher than that of controls (Figure 2.2C). This suggests that most of the response to culture filtrate protein was specific to M.tb complex.

Interestingly, we observed decreased IFN- production in response to CFP in the elderly who had no history of M.tb infection, relative to adults (Figure 2.2C). Because BCG status had no effect on this observation and we do not expect significantly different prior

NTM exposure between the adult and elderly controls, we believe this suggests hypo- responsiveness of the innate immune system of the elderly to the mycobacterial antigens in CFP rather than a decreased adaptive response.

We cannot rule out the possibility that M.tb-specific T cell function in the elderly may be impaired within the microenvironment of M.tb infected tissues, especially the lung.

However, published data for adaptive immune responses in the lungs of aged human subjects are extremely limited. Our lab has previously shown that ex vivo responses to

M.tb antigens are comparable between blood and lung among different strains of mice with different TB disease susceptibility, suggesting that our observations of no difference in M.tb-specific T cell function in the blood of different age groups may hold true in the lung (273). While previous work in an intravenous route M.tb infection model has shown that aged mice are more likely than young mice to die from a primary M.tb infection and delayed recruitment of CD4 T cells to the spleen contributes to this susceptibility, no

65 study has demonstrated a clear defect in CD4 or CD8 T cell immunity to M.tb in aged mice using a pulmonary infection model (178–180).

In the case of general in vivo T cell immune function in elderly humans, several studies have previously shown that the impaired DTH response in the skin to recall antigens in the elderly is a general phenomenon to numerous antigens, not just PPD from M.tb (274,

275). Agius et al have shown that in the case of Candida albicans antigens, the reduced skin DTH response in the elderly is due to defective activation of dermal blood vessels resulting from decreased TNF- secretion by cutaneous macrophages (275). In this study, the intrinsic ability of peripheral blood T cells to migrate or of isolated cutaneous macrophages to secrete TNF- upon in vitro stimulation were not impaired in the elderly.

More recently, Vukmanovic-Stejic et al. showed that the impaired DTH to varicella zoster virus (VZV) in the elderly is associated with an antigen-nonspecific over exuberant proinflammatory response to injected saline, which was particularly associated with transient increased HLA-DR+, CD14+ and/or CD16+ mononuclear phagocytes likely recruited from the blood (177). Contrary to the DTR to Candida albicans, early dermal blood vessel activation in response to VZV antigens was not impaired in the elderly (177,

275). Treatment with a p38 MAP kinase inhibitor significantly reduced plasma C reactive protein levels, decreased peripheral blood monocyte secretion of IL-6 and TNF- and enhanced the VZV-specific DTR response in the skin of old subjects (177).

Interestingly, we observed an age-dependent increase in nonclassical monocyte proportion in elderly subjects with LTBI. We hypothesize that this monocyte subset may play a role in the susceptibility of the elderly to develop active TB, possibly by impairing

T cell responses at the site of infection, as in the cited studies in the skin (177). It has been shown that in adults with pulmonary TB, the proportion of CD16+ monocytes is increased and this correlates with disease severity and TNF- plasma levels (276, 277).

Moreover, this monocyte subpopulation is more permissive for M.tb growth, more migratory and is inhibited in the ability to stimulate T cell proliferation and cytokine production (277). Therefore, we hypothesize that nonclassical monocytes in the elderly may play an excessively inflammatory role at the site of infection, be more permissive to

M.tb growth and impair T cell responses at the site of infection. Our work does not preclude the possibility that the anti-mycobacterial functions of classical and intermediate monocytes are also impaired in the elderly, which has been suggested by one study (278).

Our study is limited by sample size and available outcomes data, and thus our ability to comment on immunological correlates of susceptibility of the elderly to die from active

TB is limited. We did observe a significantly increased monocyte/lymphocyte ratio in the elderly with active TB, relative to all other groups. In adults and children, an elevated monocyte/lymphocyte ratio has prospectively been associated with risk of developing active TB and dying from disease (264, 265). In adults, a very low monocyte/lymphocyte ratio has also been associated with risk of developing active TB (264). Thus, our additional observation of a decreased monocyte/lymphocyte ratio in elderly with LTBI

67 relative to controls and adults with LTBI suggests its potential role in the risk of the elderly to develop active TB. Further research is needed to identify biomarkers that associate with poor outcomes in the elderly with TB, together with partially known risk factors for their more difficult diagnosis and treatment (279, 280).

Our study is one of the first in decades to study immune function in the elderly with

LTBI and active TB. Importantly, our study also included a M.tb-uninfected cohort to account for biological effects due solely to age. With regard to M.tb-specific T cell frequency and production of cytokines important for M.tb control, our data show no evidence for M.tb-specific T cell dysfunction in the elderly with M.tb infection relative to adults, and our data is suggestive of an age-dependent alteration in monocyte proportion and phenotype that may correlate with susceptibility to active TB. While we do not provide a quantification due to our low sample size, our data are also consistent with the conclusion that the QFT and T-SPOT.TB tests provide concordant results in the elderly with LTBI or active TB. Although the sensitivities of these tests have been previously compared for diagnosing active TB in the elderly (281, 282), no published study has directly compared these tests in the elderly population with LTBI.

Our study has several limitations. The principle limitation is our low sample size in our elderly LTBI and both active TB groups. Therefore, our results are suggestive but not definitive for the parameters we measured. Together with this limitation, it was also difficult to control for the higher prevalence of diabetes, cancer and immunosuppressive

68 treatment in the elderly in study recruitment. However, our multivariate analyses showed that most differences we observed between age groups among controls were due to age, independent of these three confounders and others for which we have data. Importantly, diabetes, cancer and immunosuppressive treatment did not confound our conclusions for monocyte phenotypes in the elderly with LTBI, as only 1 subject from this group for whom we collected monocyte data had diabetes, and this subject had the lowest nonclassical monocyte proportion of the group (data not shown). Additionally, in our small cohort, we could not distinguish between active TB developed soon after exposure to M.tb or disease which occurred years after initial exposure, termed reactivation TB.

Finally, our elderly LTBI and active TB subjects were on average treated longer with TB medications before study recruitment than the adults, and the tissue distribution of TB disease was different between our adult and elderly subjects. Our low sample size precludes our ability to estimate the effect of these differences in M.tb infected subjects for our monocyte data.

While our principal line of interpreting our results assumes that the elderly are biologically more susceptible to develop active TB infection, it is important to consider the possibility that the elderly are not more susceptible than younger adults. In this scenario, the higher exposure of the elderly to M.tb when active TB was more prevalent in the past is the dominant cause of their observed higher case rate today. Our study’s results are not inconsistent with this possibility, as we observed no strong evidence of an interaction between age and TB status among the immune parameters we measured. It is

69 easy to make biological susceptibility the default assumption because of the many ways in which immune function is generally impaired in the elderly. However, there are several cohort studies that consider the case rate of active TB throughout the lifespan

(114–116). These studies allow one to compare the case rate of TB within distinct decade-long birth cohorts over time, wherein a similar degree of M.tb exposure existed within each birth cohort. If waning immunity in old age caused an excess of reactivation

TB, we should observe an increase in active TB rates at some point in advanced age. In almost every such cohort study, this was not observed (114–116). There was a decay of active TB rates throughout the lifespan, well into old age (75+ years) (115, 116). Most of these cohort studies were of populations in currently developed countries, and thus represent longitudinal TB trends in a background of a dramatic reduction of transmission to very low levels over the course of the study. These studies do not preclude the possibility that elderly individuals may be at higher risk of developing active TB than younger adults if first exposed to M.tb in old age. However, most technically feasible study designs to attempt to answer this question would be unethical as preventive treatment must be offered to those who are known to be recently infected by a contact with active TB. Even determining whether an elderly person has been infected by a recent contact is difficult as one recent study from South Korea has shown that there are no differences in TST or IGRA positivity between elderly contacts of active TB cases and elderly contacts of patients with non-TB pulmonary disease (122). Most likely only a study of coincident active TB among contacts, together with molecular confirmation of

70 transmission, could answer the question of whether the elderly are more susceptible than younger adults to developing active TB after a recent exposure.

This discussion also highlights the fact that there are virtually no laboratory tools to assess the risk that an M.tb infected elderly person will develop active TB (122). Adverse effects of treatment are a key determinant in the decision whether to treat someone with

LTBI to prevent active TB (14). The elderly have a higher risk of hepatotoxicity during

LTBI treatment, but isoniazid or shorter course rifapentine-based regiments can be safe when cautiously and carefully used (118–121). As discussed in Chapter 1, recent infection is the strongest clinical risk factor for developing active TB among HIV- persons. A laboratory test that could verify recent infection in M.tb infected persons could greatly aid the assessment of risk of developing active TB in all populations, with particular benefit in the elderly population where the risks of treatment are slightly increased. In Chapter 3 we describe our efforts to discover blood transcriptomic biomarkers of recent M.tb infection.

Table 2.1. Study Population Characteristics. Due to our complex study design and

available patient populations, not all covariates are balanced across all 6 study groups. p-

values are from chi-squared test for categorical data and ANOVA for continuous

variables. †Other recorded comorbidities included asthma, rheumatoid arthritis, chronic

kidney disease, COPD, bronchiectasis, frequent pneumonias, emphysema, psoriasis (1

subject), depressive disorder (1), narcotic painkiller use (1), thoracic aneurysm (1),

microcytic anemia-thalassemia (1), Bechet’s syndrome (1), CNS Toxoplasmosis (1 Adult

Active TB), SIADH (1), gastrointestinal bypass surgery (1) and liver transplant (1).

Adult Elderly Adult Elderly All TB/LTBI Adult Elderly Latent Latent Active Active groups groups p- Control Control TB TB TB TB p-value value n 63 53 31 9 6 7 24 24 16 4 3 4 Sex (%) 0.818 (38.1) (45.3) (51.6) (44.4) (50.0) (57.1) 34.21 73.23 31.48 75.56 33.33 76.29 Age (mean (range)) (25-44) (65-92) (25-42) (66-87) (26-38) (65-89) Ethnicity (%) <0.001 0.383 6 2 17 4 1 5 Asian (9.5) (3.8) (54.8) (44.4) (16.7) (71.4) 10 7 9 4 3 0 Black (15.9) (13.2) (29.0) (44.4) (50.0) (0.0) 0 2 3 1 2 1 Other (0.0) (3.8) (9.7) (11.1) (33.3) (14.3) 47 42 2 0 0 1 White (74.6) (79.2) (6.5) (0.0) (0.0) (14.3) Previous BCG Vaccination 0.424 0.078 (%) 30 15 12 5 1 1 No (47.6) (28.3) (38.7) (55.6) (16.7) (14.3) 13 14 13 3 2 4 Unknown (20.6) (26.4) (41.9) (33.3) (33.3) (57.1) 20 24 6 1 3 2 Yes (31.7) (45.3) (19.4) (11.1) (50.0) (28.6)

Continued

Table 2.1 Continued Adult Elderly Adult Elderly All TB/LTBI Adult Elderly Latent Latent Active Active groups groups p- Control Control TB TB TB TB p-value value Tobacco Use (%) 0.002 0.178 36 23 15 4 4 2 No (57.1) (43.4) (48.4) (44.4) (66.7) (28.6) 3 13 1 2 2 2 Previous (4.8) (24.5) (3.2) (22.2) (33.3) (28.6) 15 6 1 0 0 1 Unknown (23.8) (11.3) (3.2) (0.0) (0.0) (14.3) 9 11 14 3 0 2 Yes (14.3) (20.8) (45.2) (33.3) (0.0) (28.6) Alcohol Use (%) <0.001 0.054 18 28 21 5 6 3 No (28.6) (52.8) (67.7) (55.6) (100.0) (42.9) 0 0 0 2 0 0 Previous (0.0) (0.0) (0.0) (22.2) (0.0) (0.0) 17 6 4 0 0 2 Unknown (27.0) (11.3) (12.9) (0.0) (0.0) (28.6) 28 19 6 2 0 2 Yes (44.4) (35.8) (19.4) (22.2) (0.0) (28.6) 29 9 6 7 0.688 On TB/LTBI treatment (93.5) (100.0) (100.0) (100.0) Enrolled within 30 days of 0.002 29 4 6 4 TB/LTBI treatment start (93.5) (44.4) (100.0) (57.1) (%) 2 14 0 3 1 1 0.020 Diabetes Mellitus (%) <0.001 (3.2) (26.4) (0.0) (33.3) (16.7) (14.3) 7.25 7.71 7.35 7.60 6.88 7.47 0.960 WBC (mean (sd)) 0.964 (2.58) (2.90) (2.68) (2.92) (0.69) (2.34) 40.17 39.61 43.46 37.11 38.07 38.11 0.005 HCT (mean (sd)) 0.125 (5.18) (9.69) (4.15) (4.00) (7.60) (7.42) 4.10 4.03 4.47 3.76 4.10 3.65 <0.001 Albumin (mean (sd)) 0.002 (0.75) (0.35) (0.29) (0.71) (0.91) (0.61) On Immunosuppressive 2 16 1 0 1 1 0.359 <0.001 Therapy (%) (3.2) (30.2) (3.2) (0.0) (16.7) (14.3) 1 1 0 0 1 0 Positive HIV Status (%) 0.131 (1.6) (1.9) (0.0) (0.0) (16.7) (0.0) 0 0 3 3 0 0 0.108 Hepatitis B Infection (%) <0.001 (0.0) (0.0) (9.7) (33.3) (0.0) (0.0) 1 12 0 0 0 1 History of Cancer (%) <0.001 (1.6) (22.6) (0.0) (0.0) (0.0) (14.3) Has Other Comorbidity † 9 15 3 1 2 3 0.119 0.102 (%) (14.3) (28.3) (9.7) (11.1) (33.3) (42.9)

Table 2.2. Active TB Subject Characteristics. p-value is from chi-squared test.

Adult Active TB Elderly Active TB p-value n 6 7 Culture Confirmed M.tb (%) 4 (66.7) 6 (85.7) 0.879 Primary Body Sites Affected (%) Pulmonary 1 (16.7) 5 (71.4) Pleural 1 (16.7) 2 (28.6) Pulmonary and Musculoskeletal 1 (16.7) 0 (0.0) Musculoskeletal 1 (16.7) 0 (0.0) CNS TB and Pulmonary NTM 1 (16.7) 0 (0.0) Uveitis 1 (16.7) 0 (0.0)

Table 2.3. Multivariate linear models of nonclassical monocyte proportion in adult

(n=31) and elderly (n=35) controls based on clinical covariates.

Sum of Independent Variable F Value P Value Squares Model 1. Dependent variable:

Nonclassical monocyte proportion Age Group 360.3 4.09 0.048 Diabetes 557.3 6.32 0.015 History of Cancer 31.4 0.36 0.55 On Immunosuppressive Therapy 66.0 0.75 0.39 Has Other Comorbidity 29.8 0.34 0.56 Residuals 5287.6 Model 2. Dependent variable:

Nonclassical monocyte proportion Age Group 144.9 1.75 0.19 Diabetes 117.1 1.41 0.24 Ethnicity 926.6 2.79 0.037 Alcohol Use 267.5 1.61 0.21 History of Cancer 71.7 0.86 0.36 On Immunosuppressive Therapy 25.3 0.30 0.58 Has Other Comorbidity 25.0 0.30 0.59 HIV Status 79.1 0.95 0.33 Sex 44.4 0.53 0.47 Previous BCG Vaccination 19.7 0.12 0.89 Tobacco Use 316.4 1.27 0.30 Residuals 3900.0

Figure 2.1. Study Subject Exclusion Flow Chart.

Continued

Figure 2.2. T cell cytokine responses to M.tb antigens are not altered by age in those with LTBI or active TB. Number of IFN-+ T cells, measured in spot forming units

(SFUs), responding to ESAT-6 (A) and CFP-10 (B) from M.tb. IFN- (C), IL-10 (D) and

IL-4 (E) from culture supernatants of PBMCs stimulated with M.tb culture filtrate

Figure 2.2 continued

protein. Boxplots are medians with interquartile ranges. Total subjects for each panel are

157 (A), 157 (B), 100 (C) and 164 (D, E). ** p<0.01 by Mann-Whitney U test.

Figure 2.3. Global T cell CCR7 phenotype is not altered by age in those with LTBI or active TB but PD-1 phenotype may be increased in elderly with active TB. Percent of CD4+ (A) and CD8+ (B) lymphocytes expressing CCR7. Percent of CD4+ (C) and

CD8+ (D) lymphocytes expressing PD-1. For all comparisons between age groups not displayed, there were no significant differences (p>0.05). Boxplots are medians with interquartile ranges. Total subjects for each panel are 135 (A), 134 (B), 143 (C) and 142

(D). Other Kruskal-Wallis p values were p=0.0001 (B) and p=0.03 (C). * p<0.05, ** p<0.01, ***p<0.001 by Dunn’s test.

Figure 2.4. Monocyte/lymphocyte ratio is increased in elderly with active TB and decreased in elderly with LTBI, and old age and M.tb infection skew monocytes towards a nonclassical phenotype. Monocyte/lymphocyte ratio (A) and percent monocytes with nonclassical (CD14dimCD16+) phenotype (B). Boxplots are medians with interquartile ranges. Total subjects for each panel are 91 (A) and 91 (B). Kruskal-

Wallis p values were p=0.002 (A) and p=0.003 (B). * p<0.05, ** p<0.01, ***p<0.001 by

Dunn’s test, or p value otherwise displayed.

Chapter 3 Blood RNA Signatures Predict Recent Tuberculosis Exposure in Mice, Macaques and Humans

Abstract

Tuberculosis (TB) is the leading cause of death due to a single infectious disease.

Knowing when a person was infected with Mycobacterium tuberculosis (M.tb) is critical as recent infection is the strongest clinical risk factor for progression to TB disease in immunocompetent individuals. However, time since M.tb infection is challenging to determine in routine clinical practice. To define a biomarker for recent TB exposure, we determined whether gene expression patterns in blood RNA correlated with time since

M.tb infection or exposure. First, we found RNA signatures that accurately discriminated early and late time periods after experimental infection in mice and cynomolgus macaques. Next, we found a blood RNA signature that identified recently exposed individuals in two independent human cohorts. However, for M.tb infected adolescents and adult household contacts of TB cases, our RNA signature of recent infection was unable to provide prognostic information for TB disease progression, possibly because of its brief duration. Our work supports the need for future longitudinal studies of recent TB contacts to identify biomarkers of recent infection that have sufficient duration to provide

82 prognostic information of TB disease risk in individuals and to help map recent transmission in communities.

Introduction

Tuberculosis (TB) is the leading killer due to a single infectious disease, causing over 1 million deaths per year (283). Despite renewed efforts to combat the TB epidemic, the current decline in TB incidence of 1.5% per year has fallen far short of the needed 4-5% annual decline to meet the 2020 goals for the World Health Organization’s (WHO) End

TB Strategy (284). While approximately ¼ (1.7 billion) of the world’s population has been infected with its causative agent Mycobacterium tuberculosis (M.tb), only 5 to 10% of infected individuals will develop active TB disease during their lifespan, with the remainder controlling the infection in a state known as latent TB infection (LTBI) (13,

285). Recent global workshops have reemphasized targeting transmission of TB as critical to accelerating efforts to reduce the burden of TB disease throughout the world

(286, 287). Two critical areas for understanding and preventing TB transmission are knowing where and when transmission occurs, and preventing infected individuals from progressing to active TB disease and thereafter transmitting the bacteria via the airborne route (288, 289).

Historically, successful control of TB in nations has followed from a reduction in transmission to very low levels (116, 288). Studies of close contacts, and in particular

83 household contacts, of active TB cases are a critical tool for identifying new active TB cases from recent transmission and targeting therapy for preventing both subsequent disease and transmission. However, in high incidence countries where most of the burden of disease resides, more than 80% of TB transmission occurs outside of the home (94,

95). Genotyping M.tb isolates from active TB cases coupled with comparative genomic analysis has permitted population-level identification of hotspots of localized transmission, but these data are mostly available retrospectively and thus do not allow real-time monitoring of TB transmission in a community, particularly in areas of high incidence (106). It thus remains unknown whether with current methods TB transmission can be appreciably disrupted in high incidence settings. This is in contrast to low incidence settings where both contact studies and targeting specific higher incidence communities have been effective (290).

Recent infection is the single strongest clinical risk factor for developing active TB disease in immunocompetent persons, who comprise the vast majority of LTBI and active

TB cases (96–101). However, time since exposure or infection is very difficult to ascertain in the clinical setting, and its estimate is often unreliable (209). Moreover, there are no known validated biomarkers of recent exposure or infection beyond conversion on a tuberculin skin test (TST) or IFN- release assay (IGRA), which requires longitudinal sampling. At the same time, treating all LTBI+ individuals in areas of high TB incidence to prevent the development of active TB is not feasible and would entail unnecessary risk to the vast majority of LTBI+ individuals who will never develop disease. Prospective 84 gene expression-based (RNA) signatures of risk of developing active TB disease have been recently identified for LTBI+ adolescents and adult healthy household contacts

(HHCs) (85, 86). While the positive predictive value of these RNA signatures of risk of active TB is higher than TST/IGRAs, they are still significantly less than ideal: to prevent one case of active TB, ~37-64 LTBI+ people not at risk need to be treated (vs. ~85 for

TST/IGRA) (82, 85, 86). It is currently unknown whether these RNA signatures correlate with time since infection. Importantly, their positive predictive value for TB progression and the number needed to treat could be dramatically improved if combined with accurate knowledge of time since infection in the same individual.

Building on this prior work, we assess RNA expression as a potential biomarker of recent exposure or infection with M.tb. Using our murine data and recently published studies in cynomolgus macaques and humans (85, 86, 190, 291), we show for the first time that

RNA expression predicts recent infection/exposure in all three species. Moreover, in both macaques and humans, these RNA signatures of recent infection/exposure are independent of the recently identified signatures of individual prospective TB disease risk. However, in LTBI+ adolescents and adult HHCs, our RNA signature of recent infection was unable to provide prognostic information of TB disease risk, possibly because of its likely duration of only a few months. Our work supports the need for future longitudinal studies of recent TB contacts to identify biomarkers of recent infection that have sufficient duration to provide prognostic information of TB disease risk in individuals and to help map recent transmission in communities. 85

Materials and Methods

Study Design

The objective of this study was to identify blood RNA correlates of time since M.tb infection or exposure. We first infected mice with M.tb via the aerosol route and measured genome-wide RNA expression at pre-specified time points. Unsupervised analysis revealed potential discrimination between mice sacrificed at early time points (1-

2 months) vs. late time points (3-5 months). Cross-validation without hyperparameter tuning identified an unbiased RNA signature that accurately predicted early vs. late time period post-infection. We then retrospectively mined data from a prospective M.tb infected cynomolgus macaque cohort and a prospective healthy household contact human cohort to identify RNA signatures that predicted these same time periods post-infection.

The human RNA signature was validated in an independent cohort, adolescents who were recently infected with M.tb during longitudinal sampling.

Mice

Specific pathogen-free, 6-12 week old, female C57BL/6 wild-type mice (The Jackson

Laboratory, Bar Harbor, ME) were maintained in ventilated cages inside a biosafety level

3 (BSL3) facility and provided with sterile food and water ad libitum. All protocols were

86 approved by The Ohio State University’s Institutional Laboratory Animal Care and Use

Committee.

Mouse aerosol infection and blood collection

M.tb Erdman (ATCC no. 35801) was obtained from the American Type Culture

Collection. Stocks were grown according to published methods (207). Mice were infected with M.tb Erdman using an inhalation exposure system (Glas-Col) calibrated to deliver

50 to 100 CFUs to the lungs of each mouse, as previously described (207, 292). At specific time points post-M.tb infection, infected and age-matched uninfected mice were sacrificed and blood collected (400 L) from the heart into 1.2 mL Tempus reagent and stored at -80°C. No formal randomization was employed for choosing cages of mice to be sacrificed at each time point. For the M.tb infected mice, sample size per time point was determined by using the number we routinely use for well-powered molecular and immunological studies in inbred mice. No blinding was performed for the mouse study.

RNA processing and microarray hybridization

Whole blood was collected into Applied Biosystems Tempus Blood RNA Tubes and extracted with Applied Biosystems RNA Isolation kits. Globin message reduction was conducted using the Ambion GLOBINclear Mouse/Rat kit. RNA was quantified using a

NanoDrop 1000TM spectrophotometer (NanoDrop Technologies) and RNA integrity 87

(RIN) determined by a 2100 Bioanalyzer (Agilent). Samples with RIN ≥ 6.5 were submitted for hybridization onto Illumina mouse WG 6-V2 beadchips and scanned on an

Illumina Beadstation system. Microarray data will become available in the Gene

Expression Omnibus (GEO) database upon publication.

Microarray data pre-processing

For our murine data, Illumina BeadStudio/GenomeStudio software was used to subtract background and scale average signal intensity for each sample to the global median average intensity across all samples. Probes with a detection P value ≤ 0.01 in at least

10% of mice were filtered for analysis. Thereafter R scripts were used to quantile normalize the data, set all values <10 to 10 and log2 transform the data. Probes were filtered by two-fold change in expression from the median in at least 10% of samples. For the macaque data (GSE84152), microarray data pre-processing was performed as previously described (291). The data from the human adolescent cohort of IGRA converters (GSE116014) was pre-processed identically as the macaque data, except that data were quantile normalized and no batch correction was performed. When these adolescent data were used to validate the 6-gene signature, the data were downloaded at the gene-level using the R MetaIntegrator package, before additional pre-processing

(293).

RNA-seq data pre-processing

Human data from the Grand Challenges 6-74 (GC6-74) cohort were downloaded at the gene count level from GEO (GSE94438). Genes with read count ≤ 5 in 50% of samples were excluded. Data were quantile normalized and log2 transformed. To facilitate comparisons with a common RNA-seq alignment pipeline, gene counts were obtained from the ARCHS4 resource when comparing data from the Adolescent Cohort Study

(GSE79362) and GC6-74 cohorts using the 6-gene signature (294). These data were otherwise processed identically.

Machine learning predictions

For predicting time since infection in mice, we used the Random Forest algorithm in R with default parameter values (295). Out-of-bag predictions were used to estimate model accuracy, which corresponds approximately to 3-fold cross-validation.

To predict time since infection in macaques, we randomly partitioned the macaques into training (70%) and test (30%) sets. We compared several different machine learning algorithms using the R caret package (296). These included: Random Forest

(R ranger package (297)), Gradient Boosted Machines (R gbm package (298)), Support

Vector Machines using Polynomial (R kernlab package (299)) or RBF kernels (R kernlab package (299)) and Regularized Logistic Regression (R glmnet package (300)).

9-fold cross validation was used in the training set to optimize model hyperparameters and assess predictive performance, with all samples related to individual macaques being partitioned into the same held-out fold to ensure unbiased cross-validation. The caret package implementation did not permit 10-fold cross validation for this dataset, as in humans, but the results should be equivalent. Only Regularized Logistic

Regression was used for predictions in the test set and Regularized Linear Regression for predicting each time point post-infection after Regularized Logistic Regression was shown to be superior in predicting time period post-infection.

To predict time since TB exposure, time since IGRA conversion or prospective risk of

TB in humans, we used 10-fold cross validation on the training set (either GC6-74 or

Adolescent IGRA converter cohort), with each subject’s samples partitioned into the same held-out fold, to optimize Regularized Logistic Regression model hyperparameters before predicting on the test set. Prior to performing this procedure for time since TB exposure on the GC6-74 training set, we performed feature selection on genes by a

Wilcoxon test (P < 0.05). Where longitudinal data were available for individual macaques or persons, each time point was considered as an independent sample.

Forward search to discover parsimonious 6-gene signature

A forward search was performed in the GC6-74 Gambia and Ethiopia training set on genes selected by a Wilcoxon test (P < 0.05) using the R MetaIntegrator package as

90 previously described (293, 301). The stopping threshold for increase in AUC with the addition of each gene was varied until a signature comprising less than 10 genes and including both upregulated and downregulated genes at 6 months post-enrollment (vs. baseline) was obtained. The final signature’s score is calculated on normalized log2 expression values as a difference between upregulated and downregulated genes: (RP11-

552F3.12 + PYURF + TRIM7 + TUBGCP4) – (ZNF608 + BEAN1). When applying this score to microarray data, multiple detected probes that mapped to these genes, using the

R biomaRt package, were averaged (302). Genes without corresponding detected probes were omitted from the calculation.

Cell type deconvolution, pathway and transcriptional module analysis

Cell type proportions in blood were estimated from RNA-seq data as previously described using the R MetaIntegrator package (190, 293, 303). Gene-level expression for this deconvolution was obtained from the ARCHS4 resource (294). For pathway analysis, the 250 genes comprising the signature of time since exposure to an active TB case (6 months vs. baseline) were analyzed using canonical pathway analysis with QIAGEN’s

Ingenuity® Pathway Analysis platform (IPA®, QIAGEN Redwood City, www.qiagen.com/ingenuity). To compare transcriptional modules that were concordantly or discordantly regulated between mice, macaques and humans at early and late post- exposure time points, we used the R disco and tmod packages with transcriptional modules from Li et al. (304–306). Genes used in this analysis included all detected

91 probes (mice and macaques) and genes (humans). Differential expression and ortholog assignment were performed as previously described (307). The 6-month time point in the

GC6-74 cohort was taken as the early time point in humans based on the results in Figure

3.7 and Figure 3.8, as this time point had the highest 6-gene score (data not shown).

Statistical analysis

All statistical analyses were performed in R (version 3.4.3). Prediction performance was evaluated using receiver operator characteristic (ROC) curves. Statistical significance of the area under the curve (AUC) was assessed using the one-sided Wilcoxon test via the R

Verification package (308). ROC graphs and confidence intervals were obtained via the

R pROC package (309). Pearson test was used for correlation analysis. Fischer’s exact test (two-sided) was used to determine statistical significance of comparisons between proportions in evaluating the independence of the time since infection signatures from risk of TB disease in macaques. We used linear mixed models to assess the significance of cell type proportion changes with time since TB exposure via the R lme4 package

(310). Subject and site were included as random effects and time since exposure and site as fixed effects. These two-sided P values were obtained via the Satterthwaite approximation. The IPA canonical pathway P values were calculated by a one-sided

Fisher’s Exact Test, with P < 0.01 considered as significant. The transcriptional module

P values were calculated using the CERNO statistical test, with P < 0.05 considered as

92 significant after Benjamini-Hochberg correction (307). For all other statistical tests, P <

0.05 was considered as significant.

Data availability

Mouse microarray data will be made available in the Gene Expression Omnibus database upon publication. Published data used in this study are available in the Gene Expression

Omnibus database under accession numbers GSE79362, GSE84152, GSE94438 and

GSE116014.

Code availability

Source code for all analyses is publicly available in a GitHub repository: https://github.com/remi10001/TB.

Results

Blood genome-wide RNA expression accurately discriminates early vs. late M.tb infection time periods in C57BL/6 mice

While several published studies have made genome-scale measurements of the in vivo host response to M.tb at several time points in mice (311–313), none have addressed the

93 question of whether these parameters can predict infection time point. To determine whether it is possible to predict time since M.tb infection in mice via a blood RNA signature, we measured genome-wide RNA expression in whole blood in C57BL/6 mice following low dose aerosol M.tb infection. Mouse cohorts were sacrificed every month post-infection for 5 months (n = 4 per time point) along with age-matched uninfected mice (n = 1-2 per time point). While M.tb colony forming units were not measured, it is well characterized that in this mouse strain lung bacterial burden increases exponentially from the day of M.tb infection until the peak of the adaptive immune response in the lungs at 1 month post-infection, thereafter remaining stable for approximately 300 days

(153, 314, 315). Thus, lung CFUs do not predict time since infection in this model after one month post-infection.

Principal component analysis (PCA) of our whole dataset revealed that the blood transcriptional state of M.tb infection during the first five months was distinct from that of uninfected mice, with uninfected and infected mice being entirely separable along the

1st principal component (21.8% of data variance; Figure 3.1A). When we performed

PCA on only M.tb infected mice, we found that early (30-60 days) and late (90-150 days) time periods were transcriptionally distinct, being separable along the 1st and 2nd principal components (18.8% and 16.5% of data variance, respectively; Figure 3.1B). Only 1 mouse from the 60 day time point clustered with the late time period along the 2nd principal component.

To find a predictive RNA signature of time since M.tb infection, we used the Random

Forest Classifier algorithm, without hyperparameter tuning due to low sample size, to predict early (30-60 days) vs. late (90-150 days) infection time period. Using out-of-bag predictions (approximately 3-fold cross-validation) to obtain an unbiased estimate of predictive performance, we found that we could predict early vs. late infection time period with 0.99 area under the curve (AUC) (95% CI: 0.96 – 1.00, P = 1.6 x 10-5; 87.5% sensitivity, 91.7% specificity for early infection; Figure 3.1C). To assess whether each month post-infection could be predicted accurately, we performed Random Forest

Regression with 3-fold cross-validation and confirmed that days 30 and 60 were predicted to be earlier time points than days 90-150 (Figure 3.1D). Days 90-150 were not resolved. Although low group size precludes confident quantification of the degree to which days 30 and 60 can be separated, 3 out of 4 mice within both early time points were predicted in the correct order. Taken together, these data indicate that we can broadly discriminate early and late M.tb infection in this cohort of C57BL/6 mice based on the whole blood transcriptomic response, and it may even be possible to discriminate between the first two months of infection.

Blood RNA signature discriminates early vs. late M.tb infection time periods in cynomolgus macaques

While inbred mice are a suitable model for studying molecular components of the immune response to M.tb, they do not replicate the variable clinical outcomes of M.tb

95 infection in humans. Cynomolgus macaques, an outbred non-human primate model for

TB, do exhibit heterogeneity in clinical outcomes, with approximately half of macaques progressing to symptomatic active TB disease that can be verified radiologically and bacteriologically within the first 6 months of infection, and the remainder controlling the infection in a latent state (137, 143). The lung pathology of M.tb infection in cynomolgus macaques also better replicates several features of human lung pathology than mice

(143).

To determine whether our findings in the murine model translated to the more human- like cynomolgus macaque model of M.tb infection, we mined publically available data from a longitudinal study of M.tb infection in macaques (291). In that study, cynomolgus macaques were infected with a low dose of M.tb in the lung, and their blood was sampled at 11 time points post-infection and 2 time points pre-infection for genome-wide RNA expression analysis. Importantly, while the study’s authors provided a broad, unsupervised analysis of their data according to time periods of infection, they did not assess our hypothesis that blood genome-wide RNA expression predicts time period or time point post-infection (291). To test our hypothesis and allow comparison with our mouse data and recently available human data, we restricted our analysis to 8 time points from 20 days through 180 days (6 months) post-infection. To permit comparison of different computational models and allow a final unbiased estimate of predictive performance, we randomly divided the 38 macaques from this study into a training set

96 and a test set, keeping the ratio of macaques with latent and active TB balanced in both groups (Figure 3.2).

Using 9-fold cross-validation on the training set, we found that Regularized Logistic

Regression, a linear method, was not inferior to several nonlinear classification methods in predicting early (20-56 days) vs. late (90-180 days) infection time period (Figure 3.3).

We thus chose Regularized Logistic Regression to find a predictive RNA signature of time since M.tb infection in cynomolgus macaques. We found that this model predicted early (20-56 days) vs. late (90-180 days) infection time period with an AUC of 0.78 in the training set (95% CI: 0.72-0.85, P = 5.6 x 10-13; 9-fold cross-validation; Figure 3.4A), and an AUC of 0.81 in the test set (95% CI: 0.71-0.91, P = 1.6 x 10-7; Figure 3.4A).

Importantly, our model was trained and tested on macaques irrespective of their present or future TB disease status. If our model partially predicted disease status rather than only time period post-infection, the proportion of samples from macaques with active disease would differ between predicted and actual early time period samples. However, we found that there was no change in the proportion of samples from macaques with active disease in the predicted early time periods relative to the actual early time periods, in both the training and test sets (P = 1.0, P = 1.0 respectively; Figure 3.4B). This was also true focusing on late time period predictions (P = 1.0 training, P = 1.0 test; data not shown).

Next, to assess whether each month post-infection could be predicted in cynomolgus macaques, we performed Regularized Linear Regression with 9-fold cross-validation on the training set and confirmed that days 20-56 were predicted as earlier time points than days 90-180, in both the training set and in the test set (Figure 3.4C-D). As in our murine model analysis days 90-180 were not resolved. Quantitatively, the median absolute error

(MAE) of the model was 38.5 days (Pearson’s r = 0.48, P = 1.6x10-13) on the training set and 35.7 days (r = 0.54, P = 1.1x10-7) on the test set. Probes selected and used by the final trained regression model to predict in the test set are shown in Table 3.1. To assess whether we could predict specific time point of infection within the first 3 months, as suggested by our murine data, we trained a model on only time points from 20-90 days

(Figure 3.4E-F). The MAE of this model was 15.8 days on the training set and 14.3 days on the test set (r = 0.52, P = 1.9x10-10 and r = 0.46, P = 4.7x10-4, respectively).

Taken together, these data indicate that we can broadly discriminate early and late M.tb infection in this cohort of cynomolgus macaques based on the whole blood transcriptomic response, and that we can moderately discriminate between the first two months of infection. These predictions do not depend on disease status, and the accuracy of the predictions is quantitatively lower in cynomolgus macaques than in C57BL/6 mice, as reflected by the AUC analyses.

Blood RNA expression of 250 genes predicts time since active TB exposure in humans.

To determine whether our findings in mice and cynomolgus macaques could translate to humans, several points are important to consider. While a recent study in the United

States and Canada showed that recent contacts of active TB cases are at highest risk of

TB disease in the first 1-3 months after the diagnosis of the TB index case, studies in other countries and other time periods show that the highest risk is in the first 1-2 years, with most cases accruing more than 2-3 months after a documented exposure (96, 99,

102–105). A recent vaccine study in rhesus macaques showed that BCG-induced immunity to M.tb delays IGRA conversion in a repeated limiting-dose M.tb challenge model (316). Therefore, the natural course of infection or reinfection and RNA correlates of recent infection in humans could be the same or delayed relative to our analysis in models of infection in M.tb-naïve animals. This could depend on local transmission burden and the likelihood of pre-existing immunity to M.tb, whether from BCG or a prior

M.tb exposure (317).

Whereas the day of infection is known in animal models, the precise time of exposure resulting in infection is difficult to determine in humans, even in careful clinical studies.

One surrogate for time of infection in humans is time of IGRA or TST conversion in people who were known to be IGRA/TST negative previously. This would synchronize a human study cohort to the time of an initial systemic T cell response to M.tb. To test this

99 hypothesis we accessed public data from South African adolescents who acquired latent

M.tb infection during longitudinal blood sampling every 6 months (190). We found that

Regularized Logistic Regression was unable to predict the first time point of known

IGRA conversion from 6 months post-first known IGRA conversion (0.54 AUC, 95% CI:

0.27-0.82, P = 0.64 on test set; Figure 3.5A). Notably, the biological event of actual

IGRA conversion in this cohort could have occurred anytime between the first time point of IGRA positivity and the preceding 6 months. Given our findings in mice and macaques that the RNA signature of time since M.tb infection occurs within a brief window of 2-3 months, we interpret these findings to mean that sampling blood every 6 months in humans is unlikely to constitute a cohort where actual time of IGRA conversion is synchronized sufficiently to discover an RNA signature of time since IGRA conversion.

Another study design that could identify RNA correlates of recent infection in humans is a household contact study wherein healthy contacts of active TB cases are enrolled within a certain time from the date of diagnosis of the active TB case and sampled longitudinally. Important limitations of this design that could reduce the power to detect

RNA correlates of recent infection are that the precise time of infection is not known and individuals who are IGRA+ at enrollment may have been infected either from the present exposure or in the more distant past. Cognizant of these limitations, we accessed publically available data from the Grand Challenges 6-74 (GC6-74) study of healthy household contacts (HHCs) of patients with active pulmonary TB (85). HHCs in this

100 cohort were enrolled within 2 months of the diagnosis of the active TB index case and had blood samples drawn at baseline, 6 months and/or 18 months post-enrollment (85).

Because our mouse and macaque analysis suggests that blood transcriptional changes are most prominent in an early 3 month window post-infection, we focused our first analysis on the baseline and 6 month time points. This included data from Gambian and Ethiopian cohorts but excluded data from the South African cohort because 6 month time points were not available for South Africa (85). We used the same training/test split as the authors in the Gambian cohort but randomly split the Ethiopian cohort 50/50 between our training and test sets. Importantly, with this training/test split and our data pre-processing, we could predict risk of TB with 0.72 AUC (95% CI: 0.60-0.83, P = 1.6 x 10-4; data not shown) in the training set by 10-fold cross-validation and 0.70 AUC (95% CI: 0.53-0.88,

P = 0.0071; data not shown) in the test set using Regularized Logistic Regression. From the GC6-74 and the Adolescent Cohort Study (ACS) we used the RISK4 genes (BLK,

CD1C, GAS6 and SEPT4), the post-hoc selected C1QC, TRAV27, ANKRD22, OSBPL10 genes and the 16 correlate of risk (COR) predictive genes together for this analysis (85,

86). When we used these same genes to train a model to predict time since TB exposure, we obtained no predictive performance, whether the model was trained with these gene sets separately or together (P > 0.05 for all test set predictions; Figure 3.5B). This suggests that genes selected for optimal prediction of prospective TB risk do not change across these two time points post-exposure.

101

To find a predictive RNA signature of time since TB exposure in these data, and as the study authors performed for TB risk prediction, we used the Wilcoxon test on the training set to select transcripts that differed in expression between baseline and 6 month time points (85). Using Regularized Logistic Regression we found that these genes predicted baseline vs. 6 month time points with 0.90 AUC (95% CI: 0.84-0.96, P = 1.9x10-13; 10- fold cross-validation; Figure 3.5C) in the training set and 0.69 AUC (95% CI: 0.56-0.81,

P = 0.0039; Figure 3.5C) in the test set. We further used the final genes selected by the model (250 genes, Table 3.2) on the training set to train a model to predict risk of TB. As expected, these genes exhibited no direct predictive performance for risk of TB on the training or test sets (P = 0.85, P = 0.07, respectively; Figure 3.5D). In summary, our findings with the household contact study design in humans parallel the results in macaques in that we can predict broad time period post-exposure via the whole blood transcriptomic response. Moreover, this transcriptomic signature of time period post- exposure to an active TB case is independent of the transcriptomic signature of risk of TB recently identified in the GC6-74 and ACS studies (85, 86).

Time since TB exposure in humans is associated with alteration in CD4+ T cell proportion and immune activation pathways.

Cell-type deconvolution algorithms have recently been used with genome-wide RNA expression data to help identify changes in immune cell proportions in the blood that are associated with TB disease, prospective TB disease risk and treatment success (190, 303).

102

To identify immune cell populations that are associated with time since TB exposure in the GC6-74 study, we used the leukocyte expression signature matrix ‘immunoStates’ and linear regression to infer leukocyte proportions for each subject’s sample (303). We found that the proportion of CD4+ α/β T cells was increased at 6 months vs. baseline time point in the Gambian and Ethiopian cohorts (P = 0.0079; linear mixed model; Figure

3.6A), but was not significantly changed at 18 months (P = 0.20 vs. baseline; linear mixed model, included South African cohort; Figure 3.6B). We saw no significant differences in NK cell proportion over time in the Gambian and Ethiopian cohorts

(Figure 3.6C-D). Likewise, no other cell types estimated by the ‘immunoStates’ signature matrix showed significant differences over time in these cohorts (P > 0.05, linear mixed model, data not shown). This result with CD4+ α/β T cells and NK cells is consistent with the conclusion that the RNA signature of time since TB exposure is independent from the RNA signature of prospective TB risk, since both T cells and NK cells are known to decrease in circulation in active TB disease (90, 190).

Our RNA signature of baseline vs. 6 month time points post-exposure included 250 genes selected by Regularized Logistic Regression (Table 3.2). We utilized Ingenuity Pathway

Analysis (IPA) to identify pathways associated with these genes. The majority of enriched canonical pathways (-log(p-value)>2) were associated with immune cell signaling, including B cells (B cell receptor and PI3K signaling), T cells (T cell receptor,

PKC, regulation of IL-2 expression, 4-1BB and CD28 signaling), cytokines (IL-6, IL-

15, IL-12, TNF, IL-8, IL-10 and IL-17A), innate immune cells (dendritic cell maturation

103 and LPS-stimulated MAPK signaling) and humoral immunity (Fc Epsilon RI Signaling)

(Figure 3.6E, Table 3.3). Other enriched canonical pathways were related to cellular injury and toxicity (apoptosis), metabolism, nervous system signaling, PPAR signaling, cell cycle regulation and intracellular & second messenger signaling (Table 3.3).

Considering the overall direction of change in the immune pathways between 6 month vs. baseline time points, the upregulation of several pro-inflammatory signaling pathways

(IL-6, IL-8, FLT3 signaling, PI3K signaling in B Lymphocytes and Dendritic Cell maturation) and decrease in anti-inflammatory signaling (PPAR signaling) suggests that an increase in peripheral blood immune activation occurs at the 6 month time point after exposure (Figure 3.6E).

To compare transcriptional modules altered in humans to those altered in mice and macaques, we used the recently defined disco score to identify concordantly and discordantly altered modules between these species (307). Several modules related to T cells, NK cells and monocytes were enriched (adjusted P < 0.05) in each pairwise comparison between two species (human vs. mouse, human vs. monkey, and monkey vs. mouse) (Figure 3.6F). Several B cell-related modules were uniquely concordantly regulated between macaques and mice (Figure 3.6F).

104

Application of reduced 6-gene expression signature of time since active TB exposure to adolescent M.tb infection acquisition cohort confirms its identification of recent infection in humans.

Implementation of our newly discovered RNA signature of time since active TB exposure using qRT-PCR would require a more parsimonious gene set than the 250 genes heretofore described. To find a reduced gene signature we ran a forward search using the

MetaIntegrator R package (293). This method identified 6 genes, RP11-552F3.12,

PYURF, TRIM7, TUBGCP4, ZNF608 and BEAN1, that recapitulated the performance of the 250 gene signature on baseline vs. 6 month time point discrimination with 0.86 AUC

(95% CI: 0.80-0.93, P = 1.7x10-11; Figure 3.7A) in our GC6-74 training set and 0.68

AUC (95% CI: 0.55-0.81, P = 0.0055; Figure 3.7A) in the test set. Independent validation of this signature requires a cohort wherein recent M.tb infection is documented and time points are available to test whether the signature allows discrimination between recent and more remote infection. While the cohort of South African adolescents who acquired latent M.tb infection did not permit discovery of an RNA signature of recent

M.tb infection, we reasoned that the whole cohort would be powered for validation of our signature discovered in the GC6-74 household contact study design (190). Three genes,

TRIM7, ZNF608, TUBGCP4, from our 6-gene signature were represented by detected probes in the microarray used in this study. These 3 genes discriminated the first time point of known IGRA conversion from all pre-conversion time points (6 months and 12 months prior to known conversion) with 0.72 AUC (95% CI: 0.58-0.87, P = 0.0030;

105

Figure 3.7B). These 3 genes likewise discriminated the first time point of known IGRA conversion from all sampled time points (6, 12 months prior to conversion and 6, 12 months after known conversion) with 0.68 AUC (95% CI: 0.56-0.81, P = 0.0039; Figure

3.7B). Figure 3.8 shows the trajectory of the 3 gene score over time, being highest at the first time point of known IGRA conversion.

Given that time since active TB exposure is the single strongest clinical risk factor for developing TB disease in immunocompetent persons, the finding that time since exposure and risk of TB, as predicted by the blood transcriptomic response, are independent in the

GC6-74 study of healthy household contacts suggests that these signatures could be combined to possibly better predict risk of TB when the time of exposure is unknown

(99, 101, 209). While the GC6-74 study was not powered for this particular secondary analysis, we assessed whether the highest 6-gene score during longitudinal sampling allowed discrimination of subjects who did or did not progress to active TB disease during study follow-up. The highest 6-gene score did not discriminate progressors from non-progressors in the test set, whether the subjects were from South Africa (AUC 0.51,

95% CI: 0.40-0.63, P = 0.60; Figure 3.7C) or from Gambia or Ethiopia (AUC 0.63, 95%

CI: 0.46-0.80, P = 0.095; Figure 3.7C). The same result was observed in the ACS cohort of IGRA+ adolescents with unknown exposure history (AUC 0.55, 95% CI: 0.43-0.66, P

= 0.78; Figure 3.7C).

106

We additionally tested whether the 6-gene signature discriminated early from late time periods post-infection as defined in our analysis of animal models of M.tb infection. The

5 genes represented by detected probes of homologous genes in the mouse microarray data discriminated the early (30-60 days) vs. late (90-150 days) infection time period with

0.80 AUC (95% CI: 0.60 – 1.00, P = 0.012; Figure 3.7D). Only 1 gene was represented by a detected probe in the macaque microarray data, and it alone did not discriminate the early vs. late infection time period (0.52 AUC, 95% CI: 0.45 – 0.58, P = 0.70; Figure

3.7D). Of note, this study utilized a human microarray platform for the macaque samples, which may have contributed to reduced measurability of the macaque homologues to these human genes (291).

Discussion

Early clinical studies in the pre-antibiotic era in the relatively isolated Faroe Islands shed light on the clinical features of primary infection with M.tb in humans, which often include fever, elevated erythrocyte sedimentation rate, X-ray abnormalities and, less often, erythema nodosum (102, 107). With time of exposure to an active TB case pinpointed within a two week period, and sometimes to a single day, Poulsen determined that these clinical features accompany TST conversion within 6 weeks of exposure (96,

102, 107). While these clinical features of initial M.tb infection are transient and not specific to M.tb infection, a method to determine that a person is currently in the first 1-2 years post initial infection would have great prognostic value for near-future TB disease

107 and could allow real-time geospatial mapping of recent TB transmission in communities

(96, 97, 99, 102–105). In our proof of concept analysis, we sought to determine whether it is possible to develop an RNA-based blood test to detect recent exposure or infection with M.tb. For TB disease risk prediction, we hypothesized that such a test could complement the recently developed RNA signatures of TB disease risk that are based on detecting incipient tuberculosis, which is the asymptomatic phase of early TB disease during which pathology progresses gradually before full-blown clinical TB (83–86).

Using our mouse data and published macaque data, we have demonstrated a highly accurate RNA signature of recent infection with M.tb (1-2 vs. 3-5/6 months post- infection). Using the GC6-74 cohort of HHCs of patients with active pulmonary TB, we discovered 250-gene and 6-gene human RNA signatures of recent exposure (0-2 vs. 6-8 months post-diagnosis of index case) that validated within a held-out test set (85). Using an independent cohort of adolescents who acquired M.tb infection during 6-month longitudinal sampling, we demonstrated that the 6-gene signature could discriminate the first known time point of IGRA conversion from pre-conversion time points and from 6-

12 months later with modest accuracy (0.68 AUC) (190). However, this 6-gene signature was unable to provide prognostic information of TB risk in the GC6-74 cohort or the

ACS cohort. The incomplete time point sampling of most individuals, and 6-month sampling likely reduced the power to find an association between our 6-gene signature score and TB risk in these two studies. Nevertheless, we believe the sampling constraints and target populations of these studies, adults who are HHCs and adolescents with LTBI

108 of unknown exposure history, both in highly endemic areas, are mostly in line with what may be feasible for applying transcriptional signatures of TB risk for targeted treatment to reduce TB incidence (84). Given that early blood transcriptional changes occurred within a short 3 month window in our mouse and macaque analyses, and the human data analyzed are not inconsistent with this brief timeline, we believe that blood RNA signatures for recent M.tb infection are too brief in duration to yield a useful biomarker to improve prediction of TB risk for targeted preventive therapy.

A recent genetic study of early progression to TB disease (within 18 months) demonstrated that the genetic architecture of early progression and later reactivation disease are different (123). Because the vast majority of TB disease burden can be accounted for epidemiologically by recent infection (past 1-2 years), we hypothesize that, on average, the genetic and environmental factors influencing progression of disease have resolved by 2 years post initial infection in humans (96, 285). Therefore, we hypothesize that a biological correlate of recent infection that has the longest duration during that time when the outcome of early disease progression has not been resolved would have the highest chance of being useful as a complement to tests for incipient TB in predicting TB risk. Indeed, our estimated cell type and pathway analyses suggest that both cellular and molecular signatures of immune activation associated with recent exposure and could be interrogated by other modalities such as epigenetics. Immune cell differences between recently acquired and remotely acquired infection have been reported by others in single cohorts without longitudinal sampling (209, 210). The high enrichment of B cell

109 signaling in our signature is interesting, and a recent case control study in a single cohort showed that several IgG and IgA antibodies to M.tb antigens strongly discriminated

(AUC > 0.90) active TB contacts who converted on TST from non-converters both at first known conversion and 3 months prior (186). Using dense sampling (> 1/month) for

3 years in an individual, DNA methylation was recently shown to have more prolonged dynamics in human blood in response to chronic disease states than RNA expression, and thus represents an epigenetic modality to be considered (318).

Our analyses and these considerations suggest that sampling IGRA-, untreated HHCs every month (or more frequently) for one to two years, starting as soon as possible after the diagnosis of their respective index case and determining IGRA conversion events, would allow for the discovery of biosignatures of recent M.tb infection that could be useful for helping predict TB disease risk. Follow-up in such a cohort for TB progression would allow better assessment of how signatures of recent infection and signatures of incipient TB could be combined to improve TB risk prediction. The addition of chest X- rays with deep machine learning analysis could be useful to discover heretofore unknown, specific radiogenomic features of recent infection or incipient TB (150, 319).

After IGRA conversion, staggered sampling at different times could reduce the study’s burden on individual subjects and allow more precise estimation of the duration of any biomarker. Most follow-up in such a study would have to be performed on those who refuse preventive treatment, as treatment would need to be offered because recent infection is precisely documented. Another potential benefit of such a study is that

110 validated biomarkers that associate strongly with TST/IGRA conversion but precede conversion, such as currently unvalidated IgG and IgA markers, could be used to identify

M.tb infection before TST/IGRA conversion and thus reduce the burden of follow-up of recent contacts in TB control programs and potentially help reduce LTBI treatment time

(186).

If deployed in population screening efforts, a test for recent M.tb infection could also allow real-time geospatial mapping of recent TB transmission in communities. This could greatly help the application of current control methods to reduce TB transmission and disease in high incidence settings. While it is possible that our current 6-gene signature of recent M.tb infection could be evaluated in the future for this purpose, we think it would be more prudent to first find biomarkers of recent M.tb infection that have a longer duration and are useful for individual TB risk prediction. However, biomarkers of varying duration could be jointly useful for the application of mapping recent transmission.

Our results in mice, macaques and humans, together with recent literature, suggest that future longitudinal studies of HHCs may be successful at identifying more accurate biomarkers of time since M.tb infection in humans. Our study represents one of only a handful of studies since Poulsen’s early work showing that there are biological events in the early human response to M.tb infection that can be reproducibly measured (102).

Future biomarker studies may enable the study of early events of infection in humans

111 both routinely and ethically and permit the identification of immunological or other biological events that determine whether an exposed person will develop TB disease or control the infection (286, 320–322). This could greatly aid vaccine development for TB as no correlates of protection for TB are yet known (29). We also expect that more accurate biomarkers of time since M.tb infection will be excellent tools to help better understand the human phenotypes of IGRA reversion and persistent resistance to IGRA conversion (59, 321).

Our current analysis has some limitations. Because most transmission occurs outside the household contact setting, many individuals in the GC6-74 study were TST+ at enrollment (~51.4% in Ethiopia, ~36.3% in The Gambia), and follow-up TST in this study were incomplete, it is highly likely that many, and possibly the majority, of contacts in this study were not infected or re-infected from their index TB case (89, 94,

95). However, the 6-gene RNA signature discovered in this cohort validated in adolescents where recent M.tb infection was documented via IGRA conversion in 100% of study participants (190). Additionally, our current analysis excluded HIV co-infection.

We also do not know whether RNA signatures of time since M.tb infection would change if M.tb were multi-drug resistant.

112

Table 3.1. Probes comprising 50-probe RNA signature of time since M.tb infection in cynomolgous macaques (regression of 1-6 months post-infection). Fold change and P value calculated from combined training and test set (n = 151 20-56 days, n = 143 90-180 days). P value from Wilcoxon test with Benjamini-Hochberg adjustment for multiple comparisons (using all 9050 probes in dataset).

ILLUMINA HUMAN HT 12-V4 FOLD ADJUSTED P PROBE HGNC SYMBOL CHANGE (log2) VALUE ILMN_1782487 GBP1P1 0.888076306 1.27E-06 ILMN_1774077 GBP2 0.81467902 1.72E-06 ILMN_2384181 DHRS9 0.476184449 0.000295942 ILMN_1721411 PARP10 0.311322723 0.009214237 ILMN_3255931 CD200R1L-AS1 0.270860562 0.000152621 ILMN_1808811 SBNO2 0.258252511 0.000505836 ILMN_1671039 GALNT3 0.253536365 0.002811251 ILMN_2106265 GDPD1 0.223602305 0.005372741 ILMN_2364529 EZH2 0.219489305 0.006148626 ILMN_3240236 SMCR5 0.216148623 0.003378466 ILMN_1656678 SUZ12P1 0.20141694 0.023476095 ILMN_1719857 GRIPAP1 0.187215534 0.015313146 ILMN_1708105 EZH2 0.154248029 0.056986928 ILMN_1691458 SYPL1 0.15125568 0.001453633 ILMN_1679238 LENG1 0.136744209 0.028402055 ILMN_1663390 CDC20 0.132254336 0.275531923 ILMN_1707002 TMLHE 0.128219384 0.012704096 ILMN_1737124 PRPF4B 0.120887637 0.116433982 ILMN_2360401 LFNG; MIR4648 0.102729113 0.035603246 ILMN_2365383 ENO3 0.049757597 0.705748166 ILMN_1704656 PPP2R1B 0.006064383 0.612923812 ILMN_2407464 FASTK 0.004531819 0.925840792 ILMN_1666891 LINC01625 -0.04208753 0.646931534 ILMN_1712352 DOCK3 -0.075682412 0.126840062 ILMN_1777437 LEXM -0.07648176 0.291117806 Continued 113

Table 3.1 continued

ILLUMINA HUMAN HT 12-V4 FOLD ADJUSTED P PROBE HGNC SYMBOL CHANGE (log2) VALUE ILMN_1861177 NA -0.088124465 0.022958147 ILMN_2248863 ZBTB38 -0.102669328 0.399829358 ILMN_2308689 AGBL5 -0.113710846 0.053318672 TMEM189; TMEM189- ILMN_1677446 UBE2V1 -0.114941214 0.172691494 ILMN_1712748 GSKIP -0.160413305 0.027068956 ILMN_1837428 RAB27B -0.190007556 0.057284597 ILMN_1806612 OTX1 -0.198606522 0.003547382 ILMN_1780887 USP21 -0.25011937 7.30E-05 ILMN_1805916 NIPSNAP1 -0.254274495 1.38E-05 ILMN_1709750 SUSD1 -0.260007763 7.95E-06 ILMN_1789839 GTF3C1 -0.26282267 0.000948383 ILMN_1679826 CST7 -0.271229247 0.004948241 ILMN_1763842 PTRH1 -0.272495716 4.29E-05 ILMN_1730731 ERLIN1 -0.275520584 0.000222365 ILMN_1782688 THNSL1 -0.288122209 0.000222522 ILMN_1762436 UBB -0.29211585 2.06E-06 ILMN_1811957 CAMSAP1 -0.323874933 0.000222522 ILMN_2194627 GMCL1 -0.324950872 3.57E-06 ILMN_1805826 BIVM -0.328123407 3.18E-05 ILMN_2077680 CLDND2 -0.34401714 0.001484094 ILMN_1655307 FAM136A -0.354830562 1.82E-05 ILMN_1779190 ALKBH8 -0.377910569 4.63E-07 ILMN_1660732 PPP2R2B -0.451648394 3.05E-07 ILMN_1801119 BCL2 -0.50015043 2.78E-08 ILMN_1799134 KLRD1 -0.503507863 1.91E-07

114

Table 3.2. Genes comprising 250-gene RNA signature of time since exposure to active TB index case (6 months vs. baseline). Fold change and P value calculated from combined Ethiopia and Gambia training and test set (n = 104 baseline, n = 79 6 month). P value from Wilcoxon test with Benjamini-Hochberg adjustment for multiple comparisons

(using all 14842 genes in dataset).

FOLD ENSEMBL GENE HGNC ADJUSTED CHANGE ID SYMBOL P VALUE (log2) ENSG00000007376 RPUSD1 0.151 0.053 ENSG00000007520 TSR3 0.137 0.096 ENSG00000007541 PIGQ 0.113 0.193 ENSG00000010318 PHF7 0.114 0.066 ENSG00000025770 NCAPH2 0.097 0.131 ENSG00000028528 SNX1 -0.072 0.04 ENSG00000035141 FAM136A 0.092 0.058 ENSG00000042317 SPATA7 -0.147 0.287 ENSG00000065559 MAP2K4 -0.09 0.231 ENSG00000070669 ASNS -0.043 0.508 ENSG00000072201 LNX1 -0.381 0.076 ENSG00000072506 HSD17B10 0.125 0.119 ENSG00000077380 DYNC1I2 -0.109 0.038 ENSG00000079134 THOC1 -0.1 0.086 ENSG00000079246 XRCC5 -0.081 0.182 ENSG00000079482 OPHN1 -0.267 0.074 ENSG00000080007 DDX43 -0.182 0.606 ENSG00000080371 RAB21 -0.153 0.105 ENSG00000085415 SEH1L 0.073 0.116 ENSG00000086717 PPEF1 -0.173 0.355 ENSG00000091536 MYO15A 0.165 0.114 ENSG00000095464 PDE6C -0.19 0.334 Continued 115

Table 3.2 continued FOLD ENSEMBL GENE HGNC ADJUSTED CHANGE ID SYMBOL P VALUE (log2) ENSG00000100258 LMF2 0.101 0.086 ENSG00000100603 SNW1 -0.097 0.039 ENSG00000103044 HAS3 -0.195 0.131 ENSG00000105135 ILVBL 0.159 0.064 ENSG00000105251 SHD 0.223 0.32 ENSG00000105254 TBCB 0.124 0.081 ENSG00000107175 CREB3 0.157 0.043 ENSG00000108561 C1QBP 0.148 0.025 ENSG00000108788 MLX 0.126 0.019 ENSG00000108830 RND2 0.314 0.088 ENSG00000109586 GALNT7 -0.213 0.059 ENSG00000109756 RAPGEF2 -0.167 0.086 ENSG00000110375 UPK2 0.338 0.063 ENSG00000111348 ARHGDIB -0.069 0.19 ENSG00000111961 SASH1 -0.346 0.117 ENSG00000112167 SAYSD1 0.136 0.019 ENSG00000112576 CCND3 0.125 0.038 ENSG00000112877 CEP72 -0.114 0.238 ENSG00000113141 IK -0.111 0.129 ENSG00000113966 ARL6 -0.238 0.156 ENSG00000114395 CYB561D2 0.11 0.082 ENSG00000117616 RSRP1 -0.165 0.129 ENSG00000118420 UBE3D 0.229 0.021 ENSG00000118690 ARMC2 -0.175 0.091 ENSG00000119772 DNMT3A -0.094 0.153 ENSG00000120509 PDZD11 0.161 0.082 ENSG00000121067 SPOP -0.068 0.104 ENSG00000121988 ZRANB3 -0.036 0.369 ENSG00000122203 KIAA1191 0.076 0.074 ENSG00000122642 FKBP9 -0.095 0.364 ENSG00000122679 RAMP3 0.303 0.347 Continued 116

Table 3.2 continued FOLD ENSEMBL GENE HGNC ADJUSTED CHANGE ID SYMBOL P VALUE (log2) ENSG00000123096 SSPN -0.312 0.074 ENSG00000124196 GTSF1L 0.304 0.16 ENSG00000124356 STAMBP -0.036 0.272 ENSG00000125124 BBS2 -0.141 0.038 ENSG00000125875 TBC1D20 0.026 0.322 ENSG00000125991 ERGIC3 0.154 0.043 ENSG00000126934 MAP2K2 0.114 0.124 ENSG00000127863 TNFRSF19 -0.406 0.031 ENSG00000128272 ATF4 0.105 0.083 ENSG00000128340 RAC2 -0.063 0.321 ENSG00000128567 PODXL -0.166 0.118 ENSG00000129472 RAB2B 0.164 0.045 ENSG00000129535 NRL 0.125 0.136 ENSG00000129925 TMEM8A 0.177 0.021 ENSG00000130021 PUDP 0.041 0.51 ENSG00000130203 APOE 0.351 0.096 ENSG00000130717 UCK1 0.128 0.059 ENSG00000131849 ZNF132 -0.085 0.348 ENSG00000132286 TIMM10B 0.088 0.041 ENSG00000133048 CHI3L1 -0.197 0.463 ENSG00000133678 TMEM254 0.194 0.019 ENSG00000134186 PRPF38B -0.271 0.039 ENSG00000134202 GSTM3 0.3 0.094 ENSG00000134250 NOTCH2 -0.161 0.04 ENSG00000134910 STT3A -0.069 0.24 ENSG00000135547 HEY2 0.233 0.198 ENSG00000136010 ALDH1L2 -0.171 0.232 ENSG00000137409 MTCH1 0.074 0.1 ENSG00000137807 KIF23 -0.23 0.21 ENSG00000137936 BCAR3 -0.173 0.201 ENSG00000138755 CXCL9 0.436 0.11 Continued 117

Table 3.2 continued FOLD ENSEMBL GENE HGNC ADJUSTED CHANGE ID SYMBOL P VALUE (log2) ENSG00000139899 CBLN3 0.146 0.246 ENSG00000140807 NKD1 0.165 0.235 ENSG00000141756 FKBP10 0.356 0.042 ENSG00000141873 SLC39A3 0.147 0.143 ENSG00000141968 VAV1 -0.075 0.236 ENSG00000143093 STRIP1 -0.103 0.053 ENSG00000143185 XCL2 0.319 0.253 ENSG00000143384 MCL1 -0.187 0.076 ENSG00000144648 ACKR2 0.186 0.344 ENSG00000145337 PYURF 0.146 0.159 ENSG00000146054 TRIM7 0.237 0.152 ENSG00000146232 NFKBIE 0.171 0.045 ENSG00000147364 FBXO25 0.129 0.061 ENSG00000148090 AUH -0.11 0.154 ENSG00000148429 USP6NL -0.151 0.051 ENSG00000148671 ADIRF 0.442 0.023 ENSG00000149531 FRG1BP -0.046 0.601 ENSG00000150433 TMEM218 0.101 0.049 ENSG00000151353 TMEM18 0.096 0.103 ENSG00000151640 DPYSL4 0.651 0.043 ENSG00000151694 ADAM17 -0.184 0.042 ENSG00000152763 WDR78 -0.305 0.047 ENSG00000152767 FARP1 0.409 0.031 ENSG00000157191 NECAP2 0.074 0.116 ENSG00000157315 TMED6 -0.123 0.353 ENSG00000158122 PRXL2C 0.136 0.064 ENSG00000158485 CD1B 0.326 0.061 ENSG00000158683 PKD1L1 -0.202 0.117 ENSG00000160688 FLAD1 0.136 0.038 ENSG00000160917 CPSF4 0.122 0.042 ENSG00000161249 DMKN 0.365 0.082 Continued 118

Table 3.2 continued FOLD ENSEMBL GENE HGNC ADJUSTED CHANGE ID SYMBOL P VALUE (log2) ENSG00000162734 PEA15 0.115 0.064 ENSG00000162878 PKDCC -0.037 0.804 ENSG00000162910 MRPL55 0.135 0.093 ENSG00000163026 WDCP 0.065 0.119 ENSG00000163393 SLC22A15 -0.19 0.109 ENSG00000163545 NUAK2 -0.095 0.25 ENSG00000163798 SLC4A1AP -0.15 0.074 ENSG00000163832 ELP6 0.139 0.046 ENSG00000164086 DUSP7 0.134 0.071 ENSG00000164850 GPER1 0.436 0.066 ENSG00000164976 MYORG 0.099 0.303 ENSG00000166025 AMOTL1 -0.299 0.109 ENSG00000166349 RAG1 -0.348 0.046 ENSG00000166452 AKIP1 0.175 0.061 ENSG00000166478 ZNF143 -0.072 0.121 ENSG00000166819 PLIN1 -0.1 0.576 ENSG00000166888 STAT6 -0.127 0.02 ENSG00000167094 TTC16 0.289 0.082 ENSG00000167208 SNX20 0.031 0.498 ENSG00000167306 MYO5B -0.363 0.058 ENSG00000167447 SMG8 -0.102 0.038 ENSG00000167578 RAB4B 0.111 0.133 ENSG00000167889 MGAT5B 0.117 0.43 ENSG00000167969 ECI1 0.167 0.083 ENSG00000168118 RAB4A 0.119 0.034 ENSG00000169047 IRS1 -0.275 0.037 ENSG00000169607 CKAP2L -0.269 0.291 ENSG00000170035 UBE2E3 0.102 0.103 ENSG00000170458 CD14 -0.147 0.227 ENSG00000170921 TANC2 -0.189 0.063 ENSG00000171222 SCAND1 0.122 0.207 Continued 119

Table 3.2 continued FOLD ENSEMBL GENE HGNC ADJUSTED CHANGE ID SYMBOL P VALUE (log2) ENSG00000171476 HOPX 0.309 0.044 ENSG00000171798 KNDC1 0.176 0.47 ENSG00000173040 EVC2 -0.133 0.43 ENSG00000173421 CCDC36 -0.169 0.456 ENSG00000173465 SSSCA1 0.156 0.118 ENSG00000174137 FAM53A 0.319 0.038 ENSG00000174307 PHLDA3 0.385 0.046 ENSG00000174775 HRAS 0.153 0.113 ENSG00000176124 DLEU1 0.21 0.019 ENSG00000177239 MAN1B1 0.102 0.167 ENSG00000177301 KCNA2 -0.175 0.319 ENSG00000177383 MAGEF1 -0.074 0.246 ENSG00000177595 PIDD1 0.148 0.083 ENSG00000178809 TRIM73 0.142 0.191 ENSG00000179523 EIF3J-DT -0.063 0.234 ENSG00000180353 HCLS1 -0.111 0.113 ENSG00000180914 OXTR -0.235 0.245 ENSG00000181789 COPG1 -0.107 0.118 ENSG00000182257 PRR34 0.109 0.252 ENSG00000182324 KCNJ14 0.068 0.429 ENSG00000183044 ABAT -0.185 0.094 ENSG00000183066 WBP2NL -0.243 0.171 ENSG00000184047 DIABLO 0.094 0.131 ENSG00000184281 TSSC4 0.103 0.104 ENSG00000184674 1.073 0.153 ENSG00000184811 TRARG1 0.498 0.023 ENSG00000185168 LINC00482 0.33 0.066 ENSG00000186197 EDARADD 0.221 0.141 ENSG00000186272 ZNF17 -0.14 0.051 ENSG00000186326 RGS9BP 0.222 0.15 ENSG00000187118 CMC1 0.199 0.041 Continued 120

Table 3.2 continued FOLD ENSEMBL GENE HGNC ADJUSTED CHANGE ID SYMBOL P VALUE (log2) ENSG00000187778 MCRS1 0.089 0.114 ENSG00000188026 RILPL1 -0.047 0.55 ENSG00000188056 TREML4 -0.417 0.359 ENSG00000189306 RRP7A 0.162 0.155 ENSG00000189369 GSPT2 -0.113 0.16 ENSG00000197299 BLM -0.076 0.329 ENSG00000197329 PELI1 -0.209 0.176 ENSG00000197557 TTC30A -0.19 0.218 ENSG00000197992 CLEC9A -0.367 0.103 ENSG00000198429 ZNF69 0.137 0.066 ENSG00000198547 C20orf203 0.297 0.176 ENSG00000198680 TUSC1 0.252 0.153 ENSG00000198939 ZFP2 -0.252 0.098 ENSG00000202198 0.668 0.019 ENSG00000203797 DDO 0.209 0.406 ENSG00000203950 RTL8A 0.132 0.087 ENSG00000204104 TRAF3IP1 -0.147 0.042 ENSG00000204237 OXLD1 0.149 0.065 ENSG00000204291 COL15A1 0.261 0.119 ENSG00000204568 MRPS18B 0.089 0.085 ENSG00000205643 CDPF1 0.147 0.044 ENSG00000206140 TMEM191C 0.327 0.038 ENSG00000214435 AS3MT -0.152 0.236 ENSG00000221823 PPP3R1 -0.099 0.13 ENSG00000223804 0.286 0.104 ENSG00000225163 LINC00618 0.374 0.038 ENSG00000225194 LINC00092 0.241 0.082 ENSG00000225975 LINC01534 0.219 0.045 ENSG00000226029 LINC01772 0.14 0.067 ENSG00000226124 FTCDNL1 0.245 0.055 ENSG00000227456 LINC00310 0.227 0.11 Continued 121

Table 3.2 continued FOLD ENSEMBL GENE HGNC ADJUSTED CHANGE ID SYMBOL P VALUE (log2) ENSG00000228549 0.487 0.112 ENSG00000229807 XIST 0.135 0.999 ENSG00000230438 SERPINB9P1 0.191 0.107 ENSG00000232810 TNF 0.186 0.025 ENSG00000233368 0.555 0.036 ENSG00000235863 B3GALT4 0.14 0.124 ENSG00000235999 0.116 0.26 ENSG00000239305 RNF103 -0.117 0.056 ENSG00000240849 TMEM189 0.093 0.145 ENSG00000241288 0.094 0.673 ENSG00000243959 RN7SL684P 0.412 0.022 ENSG00000249087 ZNF436-AS1 0.07 0.541 ENSG00000254184 -0.188 0.354 ENSG00000254802 0.179 0.204 ENSG00000255833 TIFAB 0.261 0.076 ENSG00000257949 TEN1 0.206 0.037 ENSG00000258102 MAP1LC3B2 -0.237 0.249 ENSG00000258366 RTEL1 0.109 0.097 ENSG00000259120 SMIM6 0.302 0.11 ENSG00000259371 0.284 0.084 ENSG00000260006 0.204 0.05 ENSG00000260101 -0.155 0.172 ENSG00000260396 -0.292 0.133 ENSG00000261126 RBFADN -0.189 0.526 ENSG00000262468 LINC01569 0.204 0.054 ENSG00000263990 -0.191 0.068 ENSG00000267260 0.154 0.296 ENSG00000267397 -0.116 0.497 ENSG00000268362 0.157 0.372 ENSG00000269089 -0.093 0.756 ENSG00000269107 0.267 0.026 Continued 122

Table 3.2 continued FOLD ENSEMBL GENE HGNC ADJUSTED CHANGE ID SYMBOL P VALUE (log2) ENSG00000269439 0.182 0.252 ENSG00000269893 SNHG8 0.212 0.045 ENSG00000270069 MIR222HG -0.252 0.177 ENSG00000271009 -0.259 0.104 ENSG00000271966 -0.193 0.142 ENSG00000272142 0.317 0.038 ENSG00000272279 0.218 0.309 ENSG00000272688 0.197 0.123 ENSG00000273009 0.26 0.35 ENSG00000273243 -0.309 0.101 ENSG00000273290 0.054 0.406

123

Table 3.3. Top significantly enriched canonical pathways in 250-gene RNA signature of time since exposure to active TB index case (6 months vs. baseline) by

IPA. Ratio is the proportion of genes within a given pathway that is in the 250-gene signature.

-Log Canonical Pathways (p- Ratio Molecules value) Immune Cells, Cytokine Signaling, Humoral Immunity, Pathogen Influence Signaling MAP2K4,RAC2,MAP2K2,PPP3R1,I B Cell Receptor Signaling 4.72 0.0529 RS1,NFKBIE,CREB3,ATF4,HRAS, VAV1 MAP2K4,MAP2K2,IRS1,NFKBIE,C IL-6 Signaling 4.25 0.0597 D14,HRAS,TNF,MCL1 Regulation of IL-2 Expression in Activated and MAP2K4,MAP2K2,PPP3R1,NFKBI 3.71 0.0706 Anergic T Lymphocytes E,HRAS,VAV1 MAP2K4,RAC2,MAP2K2,IRS1,HR Fc Epsilon RI Signaling 3.61 0.056 AS,VAV1,TNF STAT6,MAP2K2,IRS1,CREB3,ATF FLT3 Signaling in Hematopoietic Progenitor Cells 3.6 0.0674 4,HRAS MAP2K4,MAP2K2,IRS1,NFKBIE,C LPS-stimulated MAPK Signaling 3.49 0.0645 D14,HRAS MAP2K2,PPP3R1,IRS1,NFKBIE,AT PI3K Signaling in B Lymphocytes 3.47 0.053 F4,HRAS,VAV1 MAP2K4,MAP2K2,PPP3R1,IRS1,H T Cell Receptor Signaling 3.03 0.0526 RAS,VAV1 MAP2K4,RAC2,PPP3R1,IRS1,NFK PKCθ Signaling in T Lymphocytes 2.97 0.0438 BIE,HRAS,VAV1

IL-15 Signaling 2.88 0.061 STAT6,MAP2K2,IRS1,HRAS,TNF

MAP2K4,MAP2K2,PPP3R1,IRS1,N CD28 Signaling in T Helper Cells 2.78 0.0472 FKBIE,VAV1 STAT6,MAP2K2,PPP3R1,IRS1,HR IL-3 Signaling 2.72 0.0562 AS MAP2K4,RND2,MAP2K2,IRS1,HR HMGB1 Signaling 2.62 0.0438 AS,TNF MAP2K4,IRS1,NFKBIE,CREB3,CD Dendritic Cell Maturation 2.61 0.0378 1B,ATF4,TNF TNFR2 Signaling 2.54 0.103 MAP2K4,NFKBIE,TNF Continued

124

Table 3.3 continued

Immune Cells, Cytokine Signaling, Humoral Immunity, Pathogen Influence Signaling

MAP2K4,STAT6,APOE,MAP2K2,I IL-12 Signaling and Production in Macrophages 2.51 0.0417 RS1,TNF 4-1BB Signaling in T Lymphocytes 2.42 0.0938 MAP2K4,MAP2K2,NFKBIE MAP2K4,RND2,RAC2,MAP2K2,CC IL-8 Signaling 2.4 0.0347 ND3,IRS1,HRAS IL-10 Signaling 2.32 0.058 MAP2K4,NFKBIE,CD14,TNF

Role of IL-17A in Arthritis 2.32 0.058 MAP2K4,MAP2K2,IRS1,NFKBIE

Role of MAPK Signaling in the Pathogenesis of 2.17 0.0526 MAP2K4,MAP2K2,HRAS,TNF Influenza Oncostatin M Signaling 2.15 0.075 MAP2K2,HRAS,CHI3L1

CD40 Signaling 2.13 0.0513 MAP2K4,MAP2K2,IRS1,NFKBIE

IL-17A Signaling in Airway Cells 2.13 0.0513 MAP2K4,MAP2K2,IRS1,NFKBIE

GM-CSF Signaling 2.11 0.0506 MAP2K2,PPP3R1,IRS1,HRAS Role of IL-17F in Allergic Inflammatory Airway 2.09 0.0714 MAP2K2,CREB3,ATF4 Diseases

Natural Killer Cell Signaling 2.08 0.0397 RAC2,MAP2K2,IRS1,HRAS,VAV1

MIF Regulation of Innate Immunity 2.06 0.0698 MAP2K4,NFKBIE,CD14 MAP2K2,PPP3R1,IRS1,NFKBIE,HR fMLP Signaling in Neutrophils 2.05 0.0391 AS Cell Injury, and Apoptosis & Toxicity MAP2K4,DIABLO,MAP2K2,NFKB Apoptosis Signaling 4.36 0.0737 IE,HRAS,TNF,MCL1 MAP2K4,ALDH1L2,MAP2K2,GST Xenobiotic Metabolism Signaling 2.72 0.0321 M3,IRS1,HRAS,SNW1,TNF,SCAN D1 Apelin Liver Signaling Pathway 2.68 0.115 MAP2K4,IRS1,TNF MAP2K4,DIABLO,NFKBIE,TNF,A Death Receptor Signaling 2.66 0.0543 RHGDIB

Induction of Apoptosis by HIV1 2.54 0.0667 MAP2K4,DIABLO,NFKBIE,TNF

Ceramide Signaling 2.52 0.0505 MAP2K4,DIABLO,IRS1,HRAS,TNF

MAP2K4,MAP2K2,IRS1,GSTM3,A NRF2-mediated Oxidative Stress Response 2.46 0.0355 TF4,HRAS,UBE2E3 Hypoxia Signaling in the Cardiovascular System 2.23 0.0548 NFKBIE,CREB3,ATF4,UBE2E3 Continued

125

Table 3.3 continued Metabolism Fatty Acid β-oxidation I 2.5 0.1 HSD17B10,AUH,ECI1 Isoleucine Degradation I 2.09 0.143 HSD17B10,AUH Asparagine Biosynthesis I 2.01 1 ASNS Cell Cycle Regulation, Growth, Transcriptional Regulation & Intracellular and Second Messenger Signaling MAP2K4,RND2,RAC2,MAP2K2,IR Germ Cell-Sertoli Cell Junction Signaling 2.75 0.04 S1,HRAS,TNF MAP2K4,MAP2K2,IRS1,NFKBIE,H Gα12/13 Signaling 2.55 0.0426 RAS,VAV1 MAP2K4,MAP2K2,PPP3R1,IRS1,N RANK Signaling in Osteoclasts 2.5 0.05 FKBIE Melanocyte Development and Pigmentation 2.45 0.0485 MAP2K2,IRS1,CREB3,ATF4,HRAS Signaling MAP2K4,MAP2K2,IRS1,HRAS,TN PAK Signaling 2.39 0.0472 F GPER1,RAPGEF2,MAP2K2,CREB3 Gαs Signaling 2.34 0.0459 ,ATF4 MAP2K4,RND2,RAC2,MAP2K2,IR Integrin Signaling 2.29 0.033 S1,HRAS,BCAR3 Notch Signaling 2.24 0.0811 ADAM17,NOTCH2,HEY2 MAP2K4,RND2,STAT6,IRS1,VAV1 Tec Kinase Signaling 2.19 0.0357 ,TNF GPER1,MAP2K2,NFKBIE,IRS1,CR G-Protein Coupled Receptor Signaling 2.17 0.0286 EB3,PDE6C,ATF4,HRAS MAP2K4,MAP2K2,IRS1,HRAS,TN Renin-Angiotensin Signaling 2.07 0.0394 F Neurotransmitters and Other Nervous System Signaling MAP2K4,MAP2K2,IRS1,CREB3,AT Neurotrophin/TRK Signaling 3.76 0.0723 F4,HRAS MAP2K4,RAC2,MAP2K2,IRS1,CR Endocannabinoid Developing Neuron Pathway 3.57 0.0551 EB3,ATF4,HRAS MAP2K4,MAP2K2,IRS1,CREB3,AT NGF Signaling 2.85 0.0488 F4,HRAS MAP2K4,RND2,MAP2K2,HRAS,T Cholecystokinin/Gastrin-mediated Signaling 2.38 0.0467 NF ErbB4 Signaling 2.17 0.0526 ADAM17,MAP2K2,IRS1,HRAS MAP2K2,PPP3R1,CREB3,ATF4,HR Synaptic Long Term Potentiation 2.12 0.0407 AS GDNF Family Ligand-Receptor Interactions 2.05 0.0488 MAP2K4,MAP2K2,IRS1,HRAS MAP2K4,RAC2,MAP2K2,PPP3R1, Opioid Signaling Pathway 2 0.0292 CREB3,ATF4,HRAS Continued 126

Table 3.3 continued Nuclear Receptor Signaling MAP2K2,NFKBIE,HRAS,SNW1,TN PPAR Signaling 3.37 0.0612 F,SCAND1 Cancer MAP2K2,IRS1,NFKBIE,CREB3,AT Prostate Cancer Signaling 3.3 0.0594 F4,HRAS HSD17B10,IRS1,CREB3,ATF4,HR Estrogen-Dependent Breast Cancer Signaling 2.81 0.0588 AS MAP2K4,CCND3,MAP2K2,IRS1,C Endocannabinoid Cancer Inhibition Pathway 2.34 0.0385 REB3,ATF4 Disease Specific Pathways

Role of Macrophages, Fibroblasts and Endothelial MAP2K4,MAP2K2,PPP3R1,IRS1,N 2.42 0.029 Cells in Rheumatoid Arthritis FKBIE,CREB3,ATF4,HRAS,TNF

127

Figure 3.1. Blood genome-wide RNA expression discriminates early vs. late M.tb infection time periods in C57BL/6 mice. Principal component analysis of genome-wide

RNA expression measured via microarray in (A) all mice (n = 6 uninfected mice, n = 20

M.tb infected mice) stratified by infection status and (B) only M.tb infected mice stratified by time period post-infection. (C) ROC curve for out-of-bag performance of

Random Forest Classifier predicting time period post-infection (1-2 months vs. 3-5

128 months; P from Wilcoxon test, 95% confidence interval shown). (D) Random Forest

Regression out-of-bag predictions of monthly time point post-infection. Fit curve calculated via the Loess method with 95% CI shown.

129

Figure 3.2. Training and test set partition for cohort of cynomolgus macaques.

Active = Developed active TB during the 6 months of study follow-up. Latent = did not develop active TB in this study. n corresponds to individual macaques, each of which underwent longitudinal sampling.

130

Figure 3.3. Comparison of different machine algorithms to predict time period of

M.tb infection in cynomolgus macaques. Random hyperparameter search and 9-fold cross-validation of macaques were used on the training set to evaluate models to predict time period of infection from microarray data. Median (point), interquartile ranges

(boxes), and ranges (whiskers) are shown for predictions on each independent fold for the best performing model for each algorithm. glmnet = Regularized Logistic Regression, 131 gbm = Gradient Boosted Machines, svmPoly = Support Vector Machines with

Polynomial kernel, svmRadial = Support Vector Machines with RBF kernel, ranger =

Random Forest. Sens=Sensitivity, Spec=Specificity, ROC= Area under the curve.

132

Continued

Figure 3.4. Blood RNA signature discriminates early vs. late M.tb infection time periods in cynomolgus macaques. (A) ROC curves for Regularized Logistic Regression prediction of time period post-infection (20-56 days vs. 90-180 days) from RNA expression in cynomolgus macaques on 9-fold cross-validation in the training set (blue curve; n = 107 early time period samples, n = 103 late time period samples) and final model prediction on test set (red curve; n = 44 early time period, n = 40 late time period) 133

Figure 3.4 continued

(P from Wilcoxon test). (B) Comparison between early (20-56 days) (n = 107 train, n =

44 test) vs. predicted early (n = 104 train, n = 50 test) time period samples in proportion of samples from macaques that develop active TB (P from Fischer’s Exact test).

Regularized Linear Regression predictions of time point post-infection for (C) 9-fold cross-validation in the training set (n = 210) and for (D) final model prediction on the test set (n = 84). (E-F) Predictions from models trained and evaluated only on samples from the first 90 days post-infection (n = 134 train, n = 55 test). Boxplots represent medians with interquartile ranges for the predictions at each time point (best fit line shown, P from

Pearson test).

134

Figure 3.5. Blood RNA expression of 250 genes predicts time since active TB exposure in humans. ROC curves for prediction of time since first known IGRA+ (0 vs.

6 months) in South African adolescents who acquire M.tb infection for 10-fold cross- validation in the training set (blue curve; n = 17 0 month samples, n = 21 6 month

135 samples) and final model prediction on the test set (red curve; n = 10 0 month, n = 9 6 month) using Regularized Logistic Regression. (B) ROC curves for Regularized Logistic

Regression prediction of time since active TB exposure (baseline vs. 6 months post- enrollment) in GC6-74 Gambia and Ethiopia test set (n = 37 baseline samples, n = 31 6 months samples) using expression of genes from published signatures that predict prospective risk of active TB. (C) ROC curves for Regularized Logistic Regression prediction of time since active TB exposure for 10-fold cross-validation on the Gambia and Ethiopia training set (blue curve; n = 67 baseline, n = 48 6 months) and for final model prediction (contains 250 genes) on the Gambia and Ethiopia test set (red curve; n =

37 baseline, n = 31 6 months). (D) ROC curves for prediction of prospective risk of TB for 10-fold cross-validation on the Gambia and Ethiopia training set (blue curve; n = 67 baseline, n = 48 6 months) and for final model prediction on the test set (red curve; n =

37 baseline, n = 31 6 months) using the 250-gene set that predicted time since active TB exposure. P values for all ROC curves are from Wilcoxon test, and 95% confidence intervals are shown.

136

Continued

Figure 3.6. Time since TB exposure in humans is associated with alteration in CD4+

T cell proportion and immune activation pathways. Changes in CD4+ T cell percentages (A,B) and NK cell percentages (C, D) in GC6-74 healthy household contacts cohort at baseline (n = 104 in A,C; n = 272 in B, D), 6 month (A, C; n = 79) and 18 month (B, D; n = 64) time points after active TB exposure were determined by cell-type deconvolution (P from linear mixed model). Boxplots represent medians with 137

Figure 3.6 continued

interquartile ranges. (E) Top immunity related enriched canonical pathways in the 250- gene RNA signature of time since exposure to active TB index case (6 months vs. baseline) by IPA (P from Fisher’s Exact test). (F) Enriched transcriptional modules that are concordantly or discordantly regulated during recent M.tb exposure or infection between mice, macaques or humans by disco analysis (P from CERNO statistical test).

Continued

138

Figure 3.6 continued

139

Figure 3.7. Application of reduced 6-gene signature of time since active TB exposure to adolescent M.tb infection acquisition cohort confirms its identification of recent infection in humans. ROC curves for 6-gene score prediction of time since active TB exposure in the Gambia and Ethiopia training set (blue curve; n = 67 baseline samples, n

= 48 6 months samples) and for the Gambia and Ethiopia test set (red curve; n = 37 baseline, n = 31 6 months). (B) ROC curves for discrimination between time of first 140 known IGRA+ and all pre-conversion time points (blue curve; n = 27 0 month, n = 24 pre-conversion) and between time of first known IGRA+ and all other time points (green curve; n = 27 0 month, n = 24 pre-conversion and n = 31 6 or 12 months after known conversion) in South African adolescents who acquire M.tb infection using 3-gene score from genes detected in microarray data. (C) ROC curves for prediction of prospective risk of TB using highest 6-gene score observed per individual in the ACS cohort (n = 74 nonprogressors, n = 31 progressors), GC6-74 Gambia and Ethiopia test set (n = 49 nonprogressors, n = 11 progressors) and GC6-74 South Africa cohort (n = 141 nonprogressors, n = 39 progressors). (D) ROC curves for discrimination of early and late time periods post-infection in mice (blue curve; n = 8 early mice, n = 12 late mice) and macaques (green curve; n = 151 early samples, n = 143 late samples) using genes from the 6-gene signature that were detected in the respective microarrays. P values for all

ROC curves are from Wilcoxon test, and 95% confidence intervals are shown.

141

Figure 3.8. Trajectory of 3-gene signature for recent M.tb infection before and after

IGRA conversion in adolescents who acquire M.tb infection. One sample (score =

0.47) from 360 days after known conversion is omitted but was included in analyses of

Figure 5B. Boxplots represent medians with interquartile ranges, and the blue line connects medians. n = 7 -360 days, n = 17 -180 days, n = 27 0 days, n = 30 180 days.

142

Chapter 4 Discussion

Implications for addressing TB in the elderly

In Chapter 2 we examined the hypothesis that there are immunological differences between the elderly and younger adults in their response to M.tb that may explain the higher case rate of tuberculosis in the elderly in developed nations. At the outset of this study we expected that M.tb-specific immunity would be impaired in the elderly given general knowledge of immune function impairment in the elderly, as well as the observation that responsiveness to TST is reduced with advancing age past 65 years old

(231, 232). However, we observed no strong evidence that the peripheral blood immune response specific to M.tb in elderly individuals is altered relative to younger adults. This is consistent with the available strong epidemiologic data based on longitudinal birth cohorts that TB disease incidence continues to decline throughout the lifespan well into old age (114–116). A recent review of epidemiological experiments assessing the absolute risk of TB in those with LTBI in the setting of various immunosuppressive conditions including HIV, TNF inhibition and organ transplant further provided evidence that the vast majority of people who are TST/IGRA+ do not harbor bacteria capable of causing disease (323). It is then expected that as people age, fewer and fewer people harbor bacteria capable of causing TB disease. Therefore, as expected from the very low

143 positive predictive value of TST/IGRA, these tests for immune sensitization to M.tb antigens do not reflect current viable infection but rather an infection sometime in the past, and only possibly in the present.

Consistent with our work, others have demonstrated that although TST+ results are dramatically reduced in patients over the age of 80 who have culture-positive TB, their reactivity to the IGRA is not diminished relative to younger adults (271, 272). Therefore, waning of TST reactivity in old age probably reflects general changes in skin-specific immunity and does not reflect changes in systemic immunity to M.tb. This is supported by similar observations and associated mechanistic studies from others testing for skin responses to antigens from multiple different human pathogens (274, 275, 177).

While there are not specific published studies that formally assess the predictive value of

TST/IGRA for incident TB in the elderly who are not given preventive treatment, in a study of 26,970 nursing home residents over 65 years of age in Arkansas, nearly all of whom underwent TST testing, those who never tested positive had a 60-fold lower TB incidence than those who tested positive on their first test and a 150-fold lower TB incidence than those who converted to positive (324). Therefore, the very high negative predictive value of TST, and probably IGRA, are almost certainly maintained in the elderly. Together, these observations challenge the assumption that the elderly are a high- risk group for developing TB disease. Moreover, we have no laboratory tests that provide useful information on which elderly people are at risk of TB disease from a remote

144 infection, with evidence that unless a conversion is documented, IGRA/TST may not provide any information in elderly contacts of TB cases (110, 324).

No unique guidance is given in WHO guidelines for LTBI management for elderly individuals (14). The elderly have a higher risk of hepatotoxicity during LTBI treatment, but isoniazid or shorter course rifapentine-based regiments can be safe when cautiously and carefully used (118–121, 324). Given that elderly individuals are at lower risk than at any point in their lives of developing TB disease in the absence of new exposures, we do not believe that the elderly should be screened for LTBI unless there is a high suspicion that they have been recently exposed and infected. In the setting of no recent exposures, the harms of treatment outweigh the benefits even more for the elderly than for younger individuals. However, the elderly are at much more risk of developing severe complications and dying if they do develop TB disease. This suggests that if we could know an elderly person was going to develop TB disease, then the benefits of preventive treatment would be higher for them than for younger people. Unfortunately, there are no laboratory tests that provide helpful information in this regard. In nursing homes where the risk of TB transmission can be heightened, treatment for presumed LTBI and monitoring for TST/IGRA conversion events can be helpful (324). But for elderly persons in the community without clinical or epidemiologic evidence of recent infection, we believe there is no role for these measures. A test for recent infection could be very helpful to inform the role for TB preventive services in the elderly, particularly in contact

145 investigations where the amount of exposure of an elderly case to an infectious individual is low.

While in theory TB in the elderly poses a challenge for complete elimination of all TB in countries of low TB incidence, we do not believe that further TB in the elderly can be prevented safely with available drug regimens when recent transmission is successfully controlled. Residual TB in the elderly from infection when they were younger does not pose a major public health threat despite its severe detrimental impact on elderly individuals and their families when it does rarely occur. If much safer (and shorter) drug regimens for preventing TB disease were developed, then the question of treating elderly people from the general population who are TST/IGRA+ could be reconsidered.

Progress towards developing a test for recent M.tb infection

The need for a test to detect recent M.tb infection to help direct preventive treatment for

TB disease in the elderly reflects a general need for such a test for all individuals at risk for M.tb infection. Such a test has the potential to make an important contribution to other attempts at quantifying TB disease risk, such as the tests for incipient TB that are furthest in development. This is due to the fact that tests for recent infection and tests for incipient

TB provide independent information reflecting different, but both key, aspects of the underlying biology of TB disease. Recent infection is a prerequisite for most TB disease, likely because the majority of biological events that determine whether a person will progress to TB disease occur during this time (1-2 years after infection). Incipient TB 146 appears to be a measurable phase of TB disease progression in most affected people, notwithstanding how brief it usually is (< 3 months) (84). Incipient TB occurs in a subset of those who are recently infected but also probably in the smaller portion of people whose disease atypically develops long after initial infection. This is supported by the evidence that TB pathology at time of diagnosis does not depend on when someone was first infected (151, 152).

In pursuit of the goal for a useful noninvasive test for recent M.tb infection, we have shown that there are blood RNA expression correlates with recent M.tb infection in mice, macaques and humans. The timeline of when these occurred (within the first few months of infection) and their duration (only 1-3 months) appeared similar between all three species in the cohorts we analyzed. Importantly, we are the first to assess whether a validated biomarker of recent M.tb infection associates with TB disease risk in humans.

Our work suggests that the duration of such a test is likely very important for its prognostic value. No prior study for biomarkers of recent exposure included any longitudinal analysis and thus were unable to investigate this important possibility.

Longitudinal analysis is critical because in many settings, especially low incidence settings, there are likely to be many clinical, epidemiological and demographic differences between groups at risk of recent vs. remote infection which confounds an analysis evaluating such a biomarker. While the timing of a susceptible person’s incubation period for developing active TB disease could be influenced by many factors already determined at the time of their exposure and initial infection, including host and

147 bacterial genetics, infectious dose and present immune status, our present knowledge of a person’s risk of TB disease is dynamic over time for that individual. Hence, there is a need to demonstrate that biomarkers of recent infection are indeed valid at the individual level, in that they revert to their previous state, or one indicating immunological protection, once the biological events of TB pathogenesis have run their course and the person is no longer at substantial risk of TB disease. Therefore, as proposed in Chapter 3, we believe that future studies aimed at discovering biomarkers of recent M.tb infection should be longitudinal in nature whenever possible, despite the cost involved. Our analysis builds on and adds value to recent large, well conducted prospective studies for assessing TB disease risk. We believe it is now better to conduct well-powered, rigorous, longitudinal studies with fully phenotyped individuals that will enable broad discovery of potential biomarkers for recent M.tb infection and as full an evaluation of their validity as possible, rather than more smaller cross-sectional studies that do not have the ability to rigorously test the validity of these biomarkers. The importance of longitudinal analysis is also reflected in the fact that a cross-sectional view of TB case rates in the elderly can mistakenly lead one to believe that the older a person is, the higher their current risk of

TB relative to when they were younger, which we have discussed is erroneous.

Our work contributes to a model we would like to propose for TB susceptibility in humans. We term this model the sequential model of infection. It may be widely applicable to other infectious diseases. This model is bookended by two phenomena that are widely applicable to most people, 1) the development of a T cell response measurable

148 by TST or IGRA and 2) the inflammatory type I and II interferon signaling signature that can foretell the clinical onset of TB diseases and is widely observed among those with current clinical disease. We have reviewed literature on biological events that can but do not necessarily occur for most people in between these two common events of TB disease progression. We believe there can be separate branches in the host and bacterial response to infection between these two signposts, but most disease starts at one signpost

(IGRA/TST conversion) and converges at the end (incipient TB and then TB disease) regardless of the specific path in between. Along the way in a branch of infection there is the possibility for the bacteria to be rendered incapable of causing disease due to genetic, environmental and/or stochastic factors or an interaction thereof, beyond which there can be no more progression in that sequence. Or the bacteria can become uncontrollable and cause disease due to some combination of these factors, or progress to the next step in a particular branch. Possible events in these branches include disseminating to other organs, differing pathological events in those organs, reseeding or dissemination to different parts of the lung and different potential pathologies thereby. These different potential events could be measured, albeit there could be substantial heterogeneity between people and setting as to which occur and their precise timing. Bacterial genetics and host-bacteria genetic interactions likely influence the timing of these events and possibly the pathologies that develop. The general truth would still hold that most measurable events will occur within the 1-2 year period post-infection during which people are most at risk of disease progression.

149

In chapter 3 we proposed a longitudinal study design for discovering biomarkers of recent M.tb infection. This study would involve sampling IGRA-, untreated HHCs every month (or more frequently) for one to two years, starting as soon as possible after the diagnosis of their respective index case and determining IGRA conversion events along the way. Further, less frequent follow-up at more remote times than 2 years in some participants would allow verification of whether biomarkers of a long duration during the first 2 years eventually subside. This proposed longitudinal study design with frequent sampling has the potential to measure many different, heretofore unknown immunological events in the blood at various timings post-infection in people. It may be that no one immunological event is common to all people and has sufficient duration to be sensitive enough for improving targeted preventive treatment. Our sequential model for infection predicts this to be the case. However, if we can find when different events occur in subsets of people, then we could potentially discover multiple unique events of different timing, duration and generalizability that could be combined into a composite outcome signifying recent infection in the past 1-2 years. This is a departure from a “one size fits all” approach to a stratified approach that in combination could apply to most people. This may seem like a tall order for implementation if many laboratory tests must be performed simultaneously to provide information on recent infection in an individual.

However, we believe the major difficulty is the resource-intensive study design to discover these biomarkers. If our hypothesis is correct, we would likely be able to narrow the field down to a small number of biomolecules to measure, and it is possible, though unlikely, there could be one biomarker of sufficient duration and commonality to be

150 useful. This theoretical discussion argues for large, rigorous studies and measuring as much as possible from each participant, rather than many smaller studies of inferior study design and fewer measurements.

The substantial heterogeneity in the human immune response to recent M.tb infection may appear overwhelming and not amenable to the approach we have proposed.

However, our work described in Chapter 3 is the clearest demonstration so far that this heterogeneity, though likely substantial, is tractable. Our analysis of recently infected, outbred macaques shows that host genetic heterogeneity is not insurmountable in finding biomarkers of recent infection. Our analysis in humans show that the variability in timing and magnitude of M.tb exposure, as well as bacterial genetics, is likewise tractable.

Nevertheless, we believe there are inherent properties of the blood RNA response to recent infection, namely its brief duration, that make it unsuitable as a sole biomarker for recent infection.

Our study and the few previous studies on identifying biomarkers of recent M.tb infection reveal heterogeneous patterns that support our sequential model of infection and proposed study design for identifying useful biomarkers. In two studies, remote infection was unique among recent infection, disease, treated disease and uninfected controls, possibly representing an immune response only developing in those who have controlled their infection (211, 214). These biomarker were found by screening IGRA responses to multiple different M.tb antigens or IgA and IgM to M.tb HBHA. In the case of the two

151 studies investigating an effector T cell subset or MDSCs, recent infection and active disease were similar, possibly representing an effector response for combating high bacterial load, though not necessarily reflecting the ultimate outcome of infection for those recently infected but without clinical disease (209, 210). For these markers remote infection, uninfected controls and cured disease would be expected to be similar. Finally, there is the pattern that we have observed with blood RNA and others have observed for

IgM to M.tb HspX, wherein recent infection is distinct from all entities (212, 213). We believe that these data reflect aspects of the initial immune response that are common to recently infected individuals who develop TB or do not develop TB and that are not maintained in TB disease. Biomarkers reflecting this pattern are likely to be of the shortest duration as they are only present in one state that is not persistent, recent infection, while the other two patterns can be in states that are longer-lived, either controlled infection or sustained disease. All three of these distinct patterns can contribute independent information for a useful future test(s) for recent M.tb infection. It is important that future studies of biomarkers of recent M.tb infection are designed so as to be able to discover biomarkers reflecting each of these distinct patterns.

In summary, while several studies have identified potential biomarkers of recent M.tb infection, no published study has proceeded beyond the discovery phase in a single cross- sectional cohort (209–214). Our multi-species analysis has evaluated blood RNA with longitudinal cohorts and an independent human cohort, and importantly has assessed the association between our blood RNA signature of recent infection and TB disease risk.

152

Our current understanding of the biology of recent M.tb infection in mammals, including humans, and these studies on identifying biomarkers lead us to argue for future large, longitudinal cohort studies where many different aspects of peripheral immunity are measured so as to identify useful biomarkers of recent infection that together are widely applicable.

Biology to probe in future studies of recent M.tb infection

We observed that the timeline of blood RNA correlates to recent M.tb infection were similar between all three species in the cohorts we analyzed. Importantly, in both mice and macaques, blood RNA expression did not change dramatically between 3-6 months post-infection, which was the period after the signature of recent infection was most prominent. Therefore, we believe that these animal models could be valuable to find other biomarkers of recent M.tb infection that may have a more useful duration for potential application in humans. While any potential biomarker would need to be studied and validated directly in humans, further study in animal models could help to prioritize which biomolecules to measure in difficult to obtain human samples. After all, human samples must be preserved properly and differently for accurate quantitation of different biomolecules. Our proposed human study design and model for the heterogeneity in host responses reflective of recent infection could be piloted in animal models. For example, outbred mice could be used together with a combination of diverse M.tb strains during the first 6 months to a year of infection, or even over the lifespan of the mice, as has

153 begun to be investigated for genetic and molecular signatures of TB disease progression

(325–327). At a smaller scale nonhuman primates could also be used, with the advantage that longitudinal measurements, especially in the blood, are more feasible for these large animals.

While the RNA correlates of recent M.tb infection we identified in humans are unlikely to be useful for predicting TB disease risk, they may represent a future avenue of research into the biology of M.tb infection. We have shown through cell-type deconvolution analysis that much of the RNA response is likely explained by alterations in cell-type proportion in the blood during the initial immune response to infection. Alterations in the activation state and other transcriptional regulation at the single cell level may also play a role in the changes we observed in whole blood RNA. Our 6-gene sample of robust correlates of recent M.tb infection included RP11-552F3.12, PYURF, TRIM7,

TUBGCP4, ZNF608 and BEAN1. TRIM7 is an ubiquitin protein ligase, BEAN1 physically interacts with an ubiquitin protein ligase, TUBGCP4 binds -tubulin and is involved in microtubule nucleation at the centrosome and ZNF608 is a transcription factor whose mouse homologue represses RAG1 and RAG2 expression in developing thymocytes (328–335). RP11-552F3.12 and PYURF are much less well characterized. It was recently shown that TRIM7 mediates c-Jun/AP-1 activation by Ras signaling and thus influences cellular proliferation and apoptosis (328). Both TRIM7 and TUBGCP4 are upregulated during recent M.tb infection; we believe this reflects increased proliferation, whether recently or currently, among blood infiltrating immune cells,

154 although it could simply be a result of altered cell type composition in the blood. The downregulation of ZNF608 during recent M.tb infection likely reflects the increased CD4

T cell infiltration into the blood at this time, as we would expect low expression in mature T cells, but it could also reflect cell-intrinsic downregulation of this transcription factor (335). Overall, we believe the correlation of these genes with recent infection reflects increased cellular proliferation and infiltration of immune cells required for M.tb control during recent infection in humans. While we do not believe these genes play a direct and critical role in M.tb control in humans, and their discovery was not designed for that purpose, we cannot rule out this possibility.

In Chapter 3 we briefly discussed the potential role for DNA methylation and antibodies in discovering novel biomarkers of increased duration for recent M.tb infection. We will now describe these two potential biological avenues in more detail. As cited in Chapter 3, there is increasing evidence that at the individual level DNA methylation has more prolonged dynamics in human blood than RNA expression in response to chronic disease states (318). This observation likely results from the fundamental mechanisms of DNA methylation maintenance and function in the cell.

DNA methylation is an epigenetic modification of the genome that can occur at cytosine and adenosine bases, with cytosine methylation being much more extensively studied in mammals (336, 337). In 1975, DNA methylation was first proposed as a potential mechanism whereby cells could retain differentiation state information during

155 development (338, 339). This hypothesis has since been confirmed through numerous studies, and one general mechanism through which DNA methylation exerts its effect is through repression of gene transcription (336, 340). In mammals, cytosine is almost exclusively methylated in a CpG context, which allows methylation state to be mitotically inherited (336, 340). Therefore, DNA methylation provides a mechanism for cellular memory and is different between cells of different developmental lineages. Thus, genome-wide DNA methylation encodes cell type composition information in multicellular tissues like blood (341, 342). It has been shown that DNA methylation in long-term memory CD8 T cells can retain information from the cell’s proliferation and effector function during initial vaccination for more than a decade in humans (343).

These principles suggest that measuring DNA methylation in the blood at different times post-infection or exposure to M.tb will sample the changes in differentiation history of white blood cells at each time point as well as track the alteration in proportion of different cell types, including effector and memory lymphocytes, over time. Thus, we hypothesize that there exist blood DNA methylation correlates of recent M.tb infection that have a longer duration than blood RNA correlates.

There is increasing evidence that both M.tb and BCG alter host DNA methylation and other epigenetic marks in vitro and in vivo. There are differences in monocyte and granulocyte DNA methylation between individuals with LTBI vs. active TB (344). In a subset of individuals whose monocyte-derived macrophages exhibit enhanced restriction of M.tb replication, BCG vaccination alters human PBMC methylation (345). It has also

156 been shown directly in vitro that M.tb infection rapidly alters DNA methylation of human macrophages and dendritic cells (346–349). Finally, BCG vaccination in humans and mice epigenetically reprograms monocytes and macrophages and confers heterologous protection against experimental live virus vaccine infection in humans and protection against M.tb in mice (350, 351).

In light of the fundamental properties of DNA methylation and increasing evidence for the effect of M.tb infection on host DNA methylation, we have recently conducted a mouse study similar to but larger than that described in Chapter 3 wherein two strains of mice were infected with M.tb and cohorts sacrificed at several time points during the first

5 months of infection for DNA methylation profiling at the genome-wide level. This study also included a much larger cohort of age-matched uninfected mice to control for the known confounder that DNA methylation is the most accurate biological correlate of aging in mammals, including mice as recently demonstrated (352, 353). We will soon obtain the next generation sequencing results quantifying DNA methylation in this cohort, to evaluate our hypothesis that there exist blood DNA methylation correlates of recent M.tb infection that have a longer duration than the blood RNA correlates we have discovered. If our hypothesis is true, the generalizability of our result to nonhuman primates and humans could be verified using the approach we have shown in Chapter 3 and described herein for future validation in larger human studies.

157

In addition to DNA methylation, B cells and antibodies likely represent another avenue for discovering correlates of recent M.tb infection of sufficient duration to be useful. The most significantly enriched immune pathway in our broad analysis of gene expression and recent M.tb exposure in humans was B cell signaling. While antibodies to some infections and/or vaccines essentially have an infinite half-life relative to the lifespan, such as for measles, for others antibody quantitative levels decay over time since initial infection and/or vaccination (354, 355). Very few studies of antibodies to M.tb in humans have followed individuals over time (356). The cross-sectional studies we previously cited highlight that antibodies, just as the T cell response and our RNA response, to different M.tb antigens are likely to have varied kinetics and correlations to TB disease states (212–214). In fact, a study discussed in Chapter 3 showed that there are IgG and

IgA antibodies to several M.tb antigens that discriminate TST/IGRA converters from nonconverters 3 months prior to first known conversion (186). This is further evidence that the kinetics of antibodies can potentially be different from that of IGRA T cell responses.

A potential limitation of antibody-based tests is cross-reactivity to environmental mycobacteria, which could alter associations depending on the specificity of the antigen as well as geographical setting. Even for antigens not strictly specific to M.tb, our proposed study design will be able to verify their usefulness in the settings where it is conducted. This potential for cross-reactivity especially limits the role of antibodies for diagnosis of active TB, since non-TB pulmonary diseases can be associated with

158 hyperglobulinemia, and thus an increase in nonspecific anti-mycobacterial antibodies

(357). This is less of an issue for a test for recent M.tb infection because the target population is clinically healthy persons.

The study design here proposed for improved discovery and validation of biomarkers of recent M.tb infection would be especially suitable to discovering and validating known and unknown antibody-based biomarkers. As performed by Weiner and colleagues, proteome microarrays to evaluate antibody responses to thousands of M.tb antigens would be powerful if applied to such a cohort (186). As has been recently demonstrated for immunological correlates of TB disease, the application of additional high parameter measurement technologies, such as CyTOF, proteomics, metabolomics, ChIP-Seq and single cell genomics would increase the likelihood of finding biomarkers of varying timing, duration and generalizability so as to be useful for determining recent infection

(190, 358). A single cell approach is likely not currently feasible for human field studies, but it could be used for discovering unknown correlates in animal pilot experiments.

Concluding Remarks

Our work and others suggest that there is no “one size fits all” approach to preventing TB disease in highly endemic areas. It is likely that a more personalized approach, using all available information, including any useful new laboratory tests developed such as those for incipient TB or recent infection, will help drive preventive treatment to be more

159 effective in the future. However, we must remember that the incidence of TB declined in several historically high burden countries, now industrialized, to levels significantly below what is currently seen in the highest burden areas before modern antituberculous chemotherapy was discovered and implemented (7, 12). Therefore, while new medical intervention strategies have the potential to help individuals, it’s unlikely that they are the sole means needed to reduce the toll of TB disease on the world’s people. The need for improvement in poverty, economic and living conditions, education, physical infrastructure, general health systems and other areas tied to the wellbeing of individuals and families cannot be ignored. Medicine, biomedical science and these general measures clearly interact, such as the example of being able to apply some of the newly developed technologies for incipient TB. Those who most need new useful interventions and technology need to be able to receive them in the locations and settings in which they live. Reducing human death and suffering from TB is a multi-disciplinary and worldwide effort.

160

References

1. R. Ault, V. Dwivedi, E. Koivisto, J. Nagy, K. Miller, K. Nagendran, I. Chalana, X. Pan, S.-H. Wang, J. Turner, Altered monocyte phenotypes but not impaired peripheral T cell immunity may explain susceptibility of the elderly to develop tuberculosis, Exp. Gerontol. 111, 35–44 (2018).

2. R. C. Ault, C. A. Headley, A. E. Hare, B. J. Carruthers, A. Mejias, J. Turner, Blood RNA Signatures Predict Recent Tuberculosis Exposure in Mice, Macaques and Humans, bioRxiv , 830794 (2019).

3. M. J. Allison, D. Mendoza, A. Pezzia, Documentation of a Case of Tuberculosis in Pre-Columbian America, Am. Rev. Respir. Dis. 107, 985–991 (1973).

4. D. Morse, D. R. Brothwell, P. J. Ucko, Tuberculosis in Ancient Egypt, Am. Rev. Respir. Dis. 90, 524–541 (1964).

5. T. Paulson, Epidemiology: A mortal foe, Nature 502, S2–S3 (2013).

6. World Health Organization, Global Tuberculosis Report 2012 (World Health Organization, Geneva, Switzerland; http://apps.who.int/medicinedocs/en/m/abstract/Js19908en/).

7. J. F. Murray, The White Plague: Down and Out, or Up and Coming?—J. Burns Amberson Lecture, Am. Rev. Respir. Dis. 140, 1788–1795 (1989).

8. C. Hinshaw, W. H. Feldman, K. H. Pfuetze, TREATMENT OF TUBERCULOSIS WITH STREPTOMYCIN: A Summary of Observations on One Hundred Cases, J. Am. Med. Assoc. 132, 778–782 (1946).

9. R. Koch, Die Aetiologie der Tuberculose, Berl. Klin. Wochenschr. 19, 221–230 (1882).

10. R. Koch, Die Aetiologie der Tuberculose, Mitth. Aus Dem Kais. Gesundheitsamt 2, 1–88 (1884).

11. Medical Research Council, Long-term chemotherapy in the treatment of chronic pulmonary tuberculosis with cavitation: A report to the Medical Research Council by their Tuberculosis Chemotherapy Trials Committee, Tubercle 43, 201–267 (1962). 161

12. World Health Organization, Global Tuberculosis Report 2019 (World Health Organization, Geneva, Switzerland; http://www.who.int/tb/publications/global_report/en/).

13. E. Vynnycky, P. E. M. Fine, Lifetime Risks, Incubation Period, and Serial Interval of Tuberculosis, Am. J. Epidemiol. 152, 247–263 (2000).

14. World Health Organization, Latent tuberculosis infection: updated and consolidated guidelines for programmatic management. (World Health Organization, Geneva, Switzerland, 2018; http://www.ncbi.nlm.nih.gov/books/NBK531235/).

15. Y. Hamada, J. Lujan, K. Schenkel, N. Ford, H. Getahun, Sensitivity and specificity of WHO’s recommended four-symptom screening rule for tuberculosis in people living with HIV: a systematic review and meta-analysis, Lancet HIV 5, e515–e523 (2018).

16. C. S. Restrepo, R. Katre, A. Mumbower, Imaging Manifestations of Thoracic Tuberculosis, Radiol. Clin. North Am. 54, 453–473 (2016).

17. K. Dheda, T. Gumbo, N. R. Gandhi, M. Murray, G. Theron, Z. Udwadia, G. B. Migliori, R. Warren, Global control of tuberculosis: from extensively drug-resistant to untreatable tuberculosis, Lancet Respir. Med. 2, 321–338 (2014).

18. K. Dheda, T. Gumbo, G. Maartens, K. E. Dooley, R. McNerney, M. Murray, J. Furin, E. A. Nardell, L. London, E. Lessem, G. Theron, P. van Helden, S. Niemann, M. Merker, D. Dowdy, A. Van Rie, G. K. H. Siu, J. G. Pasipanodya, C. Rodrigues, T. G. Clark, F. A. Sirgel, A. Esmail, H.-H. Lin, S. R. Atre, H. S. Schaaf, K. C. Chang, C. Lange, P. Nahid, Z. F. Udwadia, C. R. Horsburgh, G. J. Churchyard, D. Menzies, A. C. Hesseling, E. Nuermberger, H. McIlleron, K. P. Fennelly, E. Goemaere, E. Jaramillo, M. Low, C. M. Jara, N. Padayatchi, R. M. Warren, The epidemiology, pathogenesis, transmission, diagnosis, and management of multidrug-resistant, extensively drug-resistant, and incurable tuberculosis, Lancet Respir. Med. 5, 291–360 (2017).

19. P. Nahid, S. E. Dorman, N. Alipanah, P. M. Barry, J. L. Brozek, A. Cattamanchi, L. H. Chaisson, R. E. Chaisson, C. L. Daley, M. Grzemska, J. M. Higashi, C. S. Ho, P. C. Hopewell, S. A. Keshavjee, C. Lienhardt, R. Menzies, C. Merrifield, M. Narita, R. O’Brien, C. A. Peloquin, A. Raftery, J. Saukkonen, H. S. Schaaf, G. Sotgiu, J. R. Starke, G. B. Migliori, A. Vernon, Official American Thoracic Society/Centers for Disease Control and Prevention/Infectious Diseases Society of America Clinical Practice Guidelines: Treatment of Drug-Susceptible Tuberculosis, Clin. Infect. Dis. 63, e147–e195 (2016).

20. M. Z. Imperial, P. Nahid, P. P. J. Phillips, G. R. Davies, K. Fielding, D. Hanna, D. Hermann, R. S. Wallis, J. L. Johnson, C. Lienhardt, R. M. Savic, A patient-level pooled

162 analysis of treatment-shortening regimens for drug-susceptible pulmonary tuberculosis, Nat. Med. 24, 1708–1715 (2018).

21. A. Calmette, L’infection bacillaire et la tuberculose chez l’homme et chez les animaux: processus d’infection, et de defense, étude biologique et expérimentale (Masson, et cie, Paris, 1920).

22. S. A. Plotkin, S. L. Plotkin, The development of vaccines: how the past led to the future, Nat. Rev. Microbiol. 9, 889–893 (2011).

23. G. A. Colditz, T. F. Brewer, C. S. Berkey, M. E. Wilson, E. Burdick, H. V. Fineberg, F. Mosteller, Efficacy of BCG Vaccine in the Prevention of Tuberculosis: Meta-analysis of the Published Literature, JAMA 271, 698–702 (1994).

24. J. A. Sterne, L. C. Rodrigues, I. N. Guedes, Does the efficacy of BCG decline with time since vaccination?, Int. J. Tuberc. Lung Dis. Off. J. Int. Union Tuberc. Lung Dis. 2, 200–207 (1998).

25. N. E. Aronson, M. Santosham, G. W. Comstock, R. S. Howard, L. H. Moulton, E. R. Rhoades, L. H. Harrison, Long-term Efficacy of BCG Vaccine in American Indians and Alaska Natives: A 60-Year Follow-up Study, JAMA 291, 2086–2091 (2004).

26. A. L. Katelaris, C. Jackson, J. Southern, R. K. Gupta, F. Drobniewski, A. Lalvani, M. Lipman, P. Mangtani, I. Abubakar, Effectiveness of BCG Vaccination Against Mycobacterium tuberculosis Infection in Adults: A Cross-sectional Analysis of a UK- Based Cohort, J. Infect. Dis. , doi:10.1093/infdis/jiz430.

27. B. B. Trunz, P. Fine, C. Dye, Effect of BCG vaccination on childhood tuberculous meningitis and miliary tuberculosis worldwide: a meta-analysis and assessment of cost- effectiveness, The Lancet 367, 1173–1180 (2006).

28. E. Nemes, H. Geldenhuys, V. Rozot, K. T. Rutkowski, F. Ratangee, N. Bilek, S. Mabwe, L. Makhethe, M. Erasmus, A. Toefy, H. Mulenga, W. A. Hanekom, S. G. Self, L.-G. Bekker, R. Ryall, S. Gurunathan, C. A. DiazGranados, P. Andersen, I. Kromann, T. Evans, R. D. Ellis, B. Landry, D. A. Hokey, R. Hopkins, A. M. Ginsberg, T. J. Scriba, M. Hatherill, Prevention of M. tuberculosis Infection with H4:IC31 Vaccine or BCG Revaccination, N. Engl. J. Med. 379, 138–149 (2018).

29. O. Van Der Meeren, M. Hatherill, V. Nduba, R. J. Wilkinson, M. Muyoyeta, E. Van Brakel, H. M. Ayles, G. Henostroza, F. Thienemann, T. J. Scriba, A. Diacon, G. L. Blatner, M.-A. Demoitié, M. Tameris, M. Malahleha, J. C. Innes, E. Hellström, N. Martinson, T. Singh, E. J. Akite, A. Khatoon Azam, A. Bollaerts, A. M. Ginsberg, T. G. Evans, P. Gillard, D. R. Tait, Phase 2b Controlled Trial of M72/AS01E Vaccine to Prevent Tuberculosis, N. Engl. J. Med. 379, 1621–1634 (2018).

163

30. D. S. Burke, Of postulates and peccadilloes: Robert Koch and vaccine (tuberculin) therapy for tuberculosis, Vaccine 11, 795–804 (1993).

31. E. J. C. Goldstein, E. Lee, R. S. Holzman, Evolution and Current Use of the Tuberculin Test, Clin. Infect. Dis. 34, 365–370 (2002).

32. C. V. Pirquet, FREQUENCY OF TUBERCULOSIS IN CHILDHOOD, J. Am. Med. Assoc. LII, 675–678 (1909).

33. S. Nayak, B. Acharjya, Mantoux test and its interpretation, Indian Dermatol. Online J. 3, 2–6 (2012).

34. F. B. Seibert, The Purification and Properties of the Purified Protein Derivative of Tuberculin, Am. Rev. Tuberc. 30, 713–720 (1934).

35. F. B. Seibert, J. T. Glen, Tuberculin purified protein derivative: preparation and analyses of a large quantity for standard, Am Rev Tuberc 44, 9–24 (1941).

36. Centers for Disease Control and Prevention, Targeted Tuberculosis Testing and Interpreting Tuberculin Skin Test Results (2018) (available at https://www.cdc.gov/tb/publications/factsheets/testing/skintestresults.htm).

37. I. Abubakar, F. Drobniewski, J. Southern, A. J. Sitch, C. Jackson, M. Lipman, J. J. Deeks, C. Griffiths, G. Bothamley, W. Lynn, H. Burgess, B. Mann, A. Imran, S. Sridhar, C.-Y. Tsou, V. Nikolayevskyy, M. Rees-Roberts, H. Whitworth, O. M. Kon, P. Haldar, H. Kunst, S. Anderson, A. Hayward, J. M. Watson, H. Milburn, A. Lalvani, D. Adeboyeku, N. Bari, J. Barker, H. Booth, F. Chua, D. Creer, M. Darmalingam, R. N. Davidson, M. Dedicoat, A. Dunleavy, J. Figueroa, M. Haseldean, N. Johnson, S. Losewicz, J. Lord, J. Moore-Gillon, G. Packe, M. Pareek, S. Tiberi, A. Pozniak, F. Sanderson, Prognostic value of interferon-γ release assays and tuberculin skin test in predicting the development of active tuberculosis (UK PREDICT TB): a prospective cohort study, Lancet Infect. Dis. 18, 1077–1087 (2018).

38. S. Machingaidze, C. Wiysonge, Y. Gonzalez-Angulo, M. Hatherill, S. Moyo, W. Hanekom, H. Mahomed, The Utility of an Interferon Gamma Release Assay for Diagnosis of Latent Tuberculosis Infection and Disease in Children: A Systematic Review and Meta-analysis, Pediatr. Infect. Dis. J. 30, 694–700 (2011).

39. J. A. Seddon, J. Paton, Z. Nademi, D. Keane, B. Williams, A. Williams, S. B. Welch, S. Liebeschutz, A. Riddell, J. Bernatoniene, S. Patel, N. Martinez-Alier, P. McMaster, B. Kampmann, The impact of BCG vaccination on tuberculin skin test responses in children is age dependent: evidence to be considered when screening children for tuberculosis infection, Thorax 71, 932–939 (2016).

164

40. A. Zwerling, M. A. Behr, A. Verma, T. F. Brewer, D. Menzies, M. Pai, The BCG World Atlas: A Database of Global BCG Vaccination Policies and Practices, PLOS Med. 8, e1001012 (2011).

41. C. S. L. Arlehamn, D. M. McKinney, C. Carpenter, S. Paul, V. Rozot, E. Makgotlho, Y. Gregg, M. van Rooyen, J. D. Ernst, M. Hatherill, W. A. Hanekom, B. Peters, T. J. Scriba, A. Sette, A Quantitative Analysis of Complexity of Human Pathogen-Specific CD4 T Cell Responses in Healthy M. tuberculosis Infected South Africans, PLOS Pathog. 12, e1005760 (2016).

42. J. Bustamante, S. Boisson-Dupuis, L. Abel, J.-L. Casanova, Mendelian susceptibility to mycobacterial disease: Genetic, immunological, and clinical features of inborn errors of IFN-γ immunity, Semin. Immunol. 26, 454–470 (2014).

43. S. Sakai, K. D. Kauffman, M. A. Sallin, A. H. Sharpe, H. A. Young, V. V. Ganusov, D. L. Barber, CD4 T Cell-Derived IFN-γ Plays a Minimal Role in Control of Pulmonary Mycobacterium tuberculosis Infection and Must Be Actively Repressed by PD-1 to Prevent Lethal Disease, PLOS Pathog. 12, e1005667 (2016).

44. P. Andersen, M. Munk, J. Pollock, T. Doherty, Specific immune-based diagnosis of tuberculosis, The Lancet 356, 1099–1104 (2000).

45. J. Rothel, S. Jones, L. Corner, J. Cox, P. Wood, A sandwich enzyme immunoassay for bovine interferon-γ and its use for the detection of tuberculosis in cattle, Aust. Vet. J. 67, 134–137 (1990).

46. P. R. Wood, L. A. Corner, P. Plackett, Development of a simple, rapid in vitro cellular assay for bovine tuberculosis based on the production of gamma interferon, Res. Vet. Sci. 49, 46–49 (1990).

47. P. J. Converse, S. L. Jones, J. Astemborski, D. Vlahov, N. M. H. Graham, Comparison of a Tuberculin Interferon-γ Assay with the Tuberculin Skin Test in High- Risk Adults: Effect of Human Immunodeficiency Virus Infection, J. Infect. Dis. 176, 144–150 (1997).

48. M. Harboe, S. Nagai, M. E. Patarroyo, M. L. Torres, C. Ramirez, N. Cruz, Properties of proteins MPB64, MPB70, and MPB80 of Mycobacterium bovis BCG., Infect. Immun. 52, 293–302 (1986).

49. H. Li, J. C. Ulstrup, T. O. Jonassen, K. Melby, S. Nagai, M. Harboe, Evidence for absence of the MPB64 gene in some substrains of Mycobacterium bovis BCG., Infect. Immun. 61, 1730–1734 (1993).

165

50. P. Andersen, A. B. Andersen, A. L. Sørensen, S. Nagai, Recall of long-lived immunity to Mycobacterium tuberculosis infection in mice., J. Immunol. 154, 3359–3372 (1995).

51. F.-X. Berthet, P. B. Rasmussen, I. Rosenkrands, P. Andersen, B. Gicquel, A Mycobacterium tuberculosis operon encoding ESAT=6 and a novel low-molecular-mass culture filtrate protein (CFP-10), Microbiology, 144, 3195–3203 (1998).

52. J. M. Pollock, P. Andersen, The Potential of the ESAT-6 Antigen Secreted by Virulent Mycobacteria for Specific Diagnosis of Tuberculosis, J. Infect. Dis. 175, 1251– 1254 (1997).

53. C. Aagaard, I. Brock, A. Olsen, T. H. M. Ottenhoff, K. Weldingh, P. Andersen, Mapping Immune Reactivity toward Rv2653 and Rv2654: Two Novel Low-Molecular- Mass Antigens Found Specifically in the Mycobacterium tuberculosis Complex, J. Infect. Dis. 189, 812–819 (2004).

54. G. H. Mazurek, P. A. LoBue, C. L. Daley, J. Bernardo, A. A. Lardizabal, W. R. Bishai, M. F. Iademarco, J. S. Rothel, Comparison of a Whole-Blood Interferon γ Assay With Tuberculin Skin Testing for Detecting Latent Mycobacterium tuberculosis Infection, JAMA 286, 1740–1747 (2001).

55. M. X. Rangaka, K. A. Wilkinson, J. R. Glynn, D. Ling, D. Menzies, J. Mwansa- Kambafwile, K. Fielding, R. J. Wilkinson, M. Pai, Predictive value of interferon-γ release assays for incident active tuberculosis: a systematic review and meta-analysis, Lancet Infect. Dis. 12, 45–55 (2012).

56. R. Diel, R. Loddenkemper, A. Nienhaus, Predictive Value of Interferon-γ Release Assays and Tuberculin Skin Testing for Progression From Latent TB Infection to Disease State, Chest 142, 63–75 (2012).

57. C. M. Stein, M. Nsereko, L. L. Malone, B. Okware, H. Kisingo, S. Nalukwago, K. Chervenak, H. Mayanja-Kizza, T. R. Hawn, W. H. Boom, Long-term Stability of Resistance to Latent Mycobacterium tuberculosis Infection in Highly Exposed Tuberculosis Household Contacts in Kampala, Uganda, Clin. Infect. Dis. 68, 1705–1712 (2019).

58. C. M. Stein, S. Zalwango, L. L. Malone, B. Thiel, E. Mupere, M. Nsereko, B. Okware, H. Kisingo, C. L. Lancioni, C. M. Bark, C. C. Whalen, M. L. Joloba, W. H. Boom, H. Mayanja-Kizza, Resistance and Susceptibility to Mycobacterium tuberculosis Infection and Disease in Tuberculosis Households in Kampala, Uganda, Am. J. Epidemiol. 187, 1477–1489 (2018).

59. L. L. Lu, M. T. Smith, K. K. Q. Yu, C. Luedemann, T. J. Suscovich, P. S. Grace, A. Cain, W. H. Yu, T. R. McKitrick, D. Lauffenburger, R. D. Cummings, H. Mayanja- 166

Kizza, T. R. Hawn, W. H. Boom, C. M. Stein, S. M. Fortune, C. Seshadri, G. Alter, IFN- γ-independent immune markers of Mycobacterium tuberculosis exposure, Nat. Med. , 1 (2019).

60. R. F. Corpe, S. Grzybowski, F. M. MacDonald, M. M. Newman, A. H. Niden, A. B. Organick, W. Lester, Preventive Treatment in Tuberculosis, Am. Rev. Respir. Dis. 91, 297–298 (1965).

61. D. Jenkins, F. F. Davidson, Isoniazid Chemoprophylaxis of Tuberculosis, Calif. Med. 116, 1–5 (1972).

62. H. P. Lambert, The Chemoprophylaxis of Tuberculosis, Am. Rev. Respir. Dis. 80, 648–658 (1959).

63. E. M. Lincoln, The Effect of Antimicrobial Therapy on the Prognosis of Primary Tuberculosis in Children,, Am. Rev. Tuberc. 69, 682–689 (1954).

64. G. W. Comstock, How much isoniazid is needed for prevention of tuberculosis among immunocompetent adults?, Int. J. Tuberc. Lung Dis. Off. J. Int. Union Tuberc. Lung Dis. 3, 847–850 (1999).

65. Efficacy of various durations of isoniazid preventive therapy for tuberculosis: five years of follow-up in the IUAT trial, Bull. World Health Organ. 60, 555–564 (1982).

66. T. Diallo, M. Adjobimey, R. Ruslami, A. Trajman, O. Sow, J. Obeng Baah, G. B. Marks, R. Long, K. Elwood, D. Zielinski, M. Gninafon, D. A. Wulandari, L. Apriani, C. Valiquette, F. Fregonese, K. Hornby, P.-Z. Li, P. C. Hill, K. Schwartzman, A. Benedetti, D. Menzies, Safety and Side Effects of Rifampin versus Isoniazid in Children, N. Engl. J. Med. 379, 454–463 (2018).

67. D. Menzies, M. Adjobimey, R. Ruslami, A. Trajman, O. Sow, H. Kim, J. Obeng Baah, G. B. Marks, R. Long, V. Hoeppner, K. Elwood, H. Al-Jahdali, M. Gninafon, L. Apriani, R. C. Koesoemadinata, A. Kritski, V. Rolla, B. Bah, A. Camara, I. Boakye, V. J. Cook, H. Goldberg, C. Valiquette, K. Hornby, M.-J. Dion, P.-Z. Li, P. C. Hill, K. Schwartzman, A. Benedetti, Four Months of Rifampin or Nine Months of Isoniazid for Latent Tuberculosis in Adults, N. Engl. J. Med. 379, 440–453 (2018).

68. T. R. Sterling, M. E. Villarino, A. S. Borisov, N. Shang, F. Gordin, E. Bliven- Sizemore, J. Hackman, C. D. Hamilton, D. Menzies, A. Kerrigan, S. E. Weis, M. Weiner, D. Wing, M. B. Conde, L. Bozeman, C. R. Horsburgh, R. E. Chaisson, Three Months of Rifapentine and Isoniazid for Latent Tuberculosis Infection, N. Engl. J. Med. 365, 2155– 2166 (2011).

69. S. Swindells, R. Ramchandani, A. Gupta, C. A. Benson, J. Leon-Cruz, N. Mwelase, M. A. Jean Juste, J. R. Lama, J. Valencia, A. Omoz-Oarhe, K. Supparatpinyo, G. 167

Masheto, L. Mohapi, R. O. da Silva Escada, S. Mawlana, P. Banda, P. Severe, J. Hakim, C. Kanyama, D. Langat, L. Moran, J. Andersen, C. V. Fletcher, E. Nuermberger, R. E. Chaisson, One Month of Rifapentine plus Isoniazid to Prevent HIV-Related Tuberculosis, N. Engl. J. Med. 380, 1001–1011 (2019).

70. M. J. Saunders, T. Wingfield, M. A. Tovar, M. R. Baldwin, S. Datta, K. Zevallos, R. Montoya, T. R. Valencia, J. S. Friedland, L. H. Moulton, R. H. Gilman, C. A. Evans, A score to predict and stratify risk of tuberculosis in adult contacts of tuberculosis index cases: a prospective derivation and external validation cohort study, Lancet Infect. Dis. 17, 1190–1199 (2017).

71. E. W. Tiemersma, M. J. van der Werf, M. W. Borgdorff, B. G. Williams, N. J. D. Nagelkerke, Natural History of Tuberculosis: Duration and Fatality of Untreated Pulmonary Tuberculosis in HIV Negative Patients: A Systematic Review, PLoS ONE 6 (2011), doi:10.1371/journal.pone.0017601.

72. K. Romanowski, B. Baumann, C. A. Basham, F. A. Khan, G. J. Fox, J. C. Johnston, Long-term all-cause mortality in people treated for tuberculosis: a systematic review and meta-analysis, Lancet Infect. Dis. 19, 1129–1137 (2019).

73. B. Allwood, M. van der Zalm, G. Makanda, K. Mortimer, A. A. F.s, E. Uzochukwu, E. Denise, G. Diane, H. Graeme, I. Olena, J. Rupert, M. F. M, M. Jamilah, M. Stellah, van K. Sanne, R. Andrea, S. Ingrid, S. Cari, von D. Dalene, W. Naomi, W. Robert, The long shadow post-tuberculosis, Lancet Infect. Dis. 19, 1170–1171 (2019).

74. S. Datta, C. A. Evans, Healthy survival after tuberculosis, Lancet Infect. Dis. 19, 1045–1047 (2019).

75. World Health Organization, High-priority target product profiles for new tuberculosis diagnostics (World Health Organization, Geneva, Switzerland, 2014; http://www.who.int/tb/publications/tpp_report/en/).

76. E. MacLean, T. Broger, S. Yerlikaya, B. L. Fernandez-Carballo, M. Pai, C. M. Denkinger, A systematic review of biomarkers to detect active tuberculosis, Nat. Microbiol. 4, 748–758 (2019).

77. J. Minion, E. Leung, E. Talbot, K. Dheda, M. Pai, D. Menzies, Diagnosing tuberculosis with urine lipoarabinomannan: systematic review and meta-analysis, Eur. Respir. J. 38, 1398–1405 (2011).

78. C. Liu, Z. Zhao, J. Fan, C. J. Lyon, H.-J. Wu, D. Nedelkov, A. M. Zelazny, K. N. Olivier, L. H. Cazares, S. M. Holland, E. A. Graviss, Y. Hu, Quantification of circulating Mycobacterium tuberculosis antigen peptides allows rapid diagnosis of active disease and treatment monitoring, Proc. Natl. Acad. Sci. 114, 3969–3974 (2017).

168

79. L. Paris, R. Magni, F. Zaidi, R. Araujo, N. Saini, M. Harpole, J. Coronel, D. E. Kirwan, H. Steinberg, R. H. Gilman, E. F. Petricoin, R. Nisini, A. Luchini, L. Liotta, Urine lipoarabinomannan glycan in HIV-negative patients with pulmonary tuberculosis correlates with disease severity, Sci. Transl. Med. 9 (2017), doi:10.1126/scitranslmed.aal2807.

80. M. P. R. Berry, C. M. Graham, F. W. McNab, Z. Xu, S. A. A. Bloch, T. Oni, K. A. Wilkinson, R. Banchereau, J. Skinner, R. J. Wilkinson, C. Quinn, D. Blankenship, R. Dhawan, J. J. Cush, A. Mejias, O. Ramilo, O. M. Kon, V. Pascual, J. Banchereau, D. Chaussabel, A. O’Garra, An interferon-inducible neutrophil-driven blood transcriptional signature in human tuberculosis, Nature 466, 973–977 (2010).

81. H. Warsinske, R. Vashisht, P. Khatri, Host-response-based gene signatures for tuberculosis diagnosis: A systematic comparison of 16 signatures, PLOS Med. 16, e1002786 (2019).

82. World Health Organization, Development of a Target Product Profile (TPP) and a framework for evaluation for a test for predicting progression from tuberculosis infection to active disease (World Health Organization, Geneva, Switzerland, 2017; http://apps.who.int/iris/bitstream/handle/10665/259176/WHO-HTM-TB-2017.18- eng.pdf;jsessionid=EBD2B5F9B500750ECB57D8E796BFD533?sequence=1).

83. F. Cobelens, S. Kik, H. Esmail, D. M. Cirillo, C. Lienhardt, A. Matteelli, From latent to patent: rethinking prediction of tuberculosis, Lancet Respir. Med. 5, 243–244 (2017).

84. R. K. Gupta, C. T. Turner, C. Venturini, H. Esmail, M. X. Rangaka, A. Copas, M. Lipman, I. Abubakar, M. Noursadeghi, Concise whole blood transcriptional signatures for incipient tuberculosis: A systematic review and patient-level pooled meta-analysis, Lancet Respir. Med. 0 (2020), doi:10.1016/S2213-2600(19)30282-6.

85. S. Suliman, E. G. Thompson, J. Sutherland, J. Weiner, M. O. C. Ota, S. Shankar, A. Penn-Nicholson, B. Thiel, M. Erasmus, J. Maertzdorf, F. J. Duffy, P. C. Hill, E. J. Hughes, K. Stanley, K. Downing, M. L. Fisher, J. Valvo, S. K. Parida, G. van der Spuy, G. Tromp, I. M. O. Adetifa, S. Donkor, R. Howe, H. Mayanja-Kizza, W. H. Boom, H. M. Dockrell, T. H. M. Ottenhoff, M. Hatherill, A. Aderem, W. A. Hanekom, T. J. Scriba, S. H. E. Kaufmann, D. E. Zak, G. Walzl, G. Walzl, G. F. Black, G. van der Spuy, K. Stanley, M. Kriel, N. Du Plessis, N. Nene, T. Roberts, L. Kleynhans, A. Gutschmidt, B. Smith, N. Nene, A. G. Loxton, N. N. Chegou, G. Tromp, D. Tabb, T. H. M. Ottenhoff, M. R. Klein, M. C. Haks, K. L. M. C. Franken, A. Geluk, K. E. van Meijgaarden, S. A. Joosten, W. H. Boom, B. Thiel, H. Mayanja-Kizza, M. Joloba, S. Zalwango, M. Nsereko, B. Okwera, H. Kisingo, S. H. E. Kaufmann, S. K. Parida, R. Golinski, J. Maertzdorf, J. Weiner, M. Jacobson, H. Dockrell, S. Smith, P. Gorak-Stolinska, Y.-G. Hur, M. Lalor, J.- S. Lee, A. C. Crampin, N. French, B. Ngwira, A. Ben-Smith, K. Watkins, L. Ambrose, F. Simukonda, H. Mvula, F. Chilongo, J. Saul, K. Branson, S. Suliman, T. J. Scriba, H. 169

Mahomed, E. J. Hughes, N. Bilek, M. Erasmus, O. Xasa, A. Veldsman, K. Downing, M. Fisher, A. Penn-Nicholson, H. Mulenga, B. Abel, M. Bowmaker, B. Kagina, W. K. Chung, W. A. Hanekom, J. Sadoff, D. Sizemore, S. Ramachandran, L. Barker, M. Brennan, F. Weichold, S. Muller, L. Geiter, D. Kassa, A. Abebe, T. Mesele, B. Tegbaru, D. van Baarle, F. Miedema, R. Howe, A. Mihret, A. Aseffa, Y. Bekele, R. Iwnetu, M. Tafesse, L. Yamuah, M. Ota, J. Sutherland, P. Hill, R. Adegbola, T. Corrah, M. Antonio, T. Togun, I. Adetifa, S. Donkor, P. Andersen, I. Rosenkrands, M. Doherty, K. Weldingh, G. Schoolnik, G. Dolganov, T. Van, F. Kafaar, L. Workman, H. Mulenga, T. J. Scriba, E. J. Hughes, N. Bilek, M. Erasmus, O. Xasa, A. Veldsman, Y. Cloete, D. Abrahams, S. Moyo, S. Gelderbloem, M. Tameris, H. Geldenhuys, W. Hanekom, G. Hussey, R. Ehrlich, S. Verver, L. Geiter, Four-Gene Pan-African Blood Signature Predicts Progression to Tuberculosis, Am. J. Respir. Crit. Care Med. 197, 1198–1208 (2018).

86. D. E. Zak, A. Penn-Nicholson, T. J. Scriba, E. Thompson, S. Suliman, L. M. Amon, H. Mahomed, M. Erasmus, W. Whatney, G. D. Hussey, D. Abrahams, F. Kafaar, T. Hawkridge, S. Verver, E. J. Hughes, M. Ota, J. Sutherland, R. Howe, H. M. Dockrell, W. H. Boom, B. Thiel, T. H. M. Ottenhoff, H. Mayanja-Kizza, A. C. Crampin, K. Downing, M. Hatherill, J. Valvo, S. Shankar, S. K. Parida, S. H. E. Kaufmann, G. Walzl, A. Aderem, W. A. Hanekom, A blood RNA signature for tuberculosis disease risk: a prospective cohort study, The Lancet 387, 2312–2322 (2016).

87. F. J. Duffy, E. Thompson, K. Downing, S. Suliman, H. Mayanja-Kizza, W. H. Boom, B. Thiel, J. Weiner III, S. H. E. Kaufmann, D. Dover, D. L. Tabb, H. M. Dockrell, T. H. M. Ottenhoff, G. Tromp, T. J. Scriba, D. E. Zak, G. Walzl, the G.-74 Consortium, S. H. E. Kaufmann, S. K. Parida, R. Golinski, J. Maertzdorf, J. I. Weiner, M. Jacobson, G. McEwen, G. Walzl, G. Black, G. van der Spuy, K. Stanley, M. Kriel, N. Du Plessis, N. Nene, A. Loxton, N. N. Chegou, S. Suliman, T. Scriba, H. Mahomed, M. Erasmus, O. Xasa, A. Veldsman, J. Hughes, K. Downing, A. Penn-Nicholson, H. Mulenga, B. Abel, M. Bowmaker, B. Kagina, W. K. C, W. Hanekom, T. H. M. Ottenhoff, M. R. Klein, M. C. Haks, K. L. Franken, A. Geluk, V. Meijgaarden, K. E, S. A. Joosten, D. van Baarle, F. Miedema, W. H. Boom, B. Thiel, J. Sadoff, D. Sizemore, S. Ramachandran, L. Barker, M. Brennan, F. Weichold, S. Muller, L. Geiter, G. Schoolnik, G. Dolganov, T. Van, H. Mayanja-Kizza, M. Joloba, S. Zalwango, M. Nsereko, B. Okwera, H. Kisingo, H. Dockrell, S. Smith, P. Gorak-Stolinska, Y.-G. Hur, M. Lalor, J.-S. Lee, A. C. Crampin, N. French, B. Ngwira, A. B. Smith, K. Watkins, L. Ambrose, F. Simukonda, H. Mvula, F. Chilongo, J. Saul, K. Branson, D. Kassa, A. Abebe, T. Mesele, B. Tegbaru, R. Howe, A. Mihret, A. Aseffa, Y. Bekele, R. Iwnetu, M. Tafesse, L. Yamuah, M. Ota, J. Sutherland, P. Hill, R. Adegbola, T. Corrah, M. Antonio, T. Togun, I. Adetifa, S. Donkor, P. Andersen, I. Rosenkrands, M. Doherty, K. Weldingh, A Serum Circulating miRNA Signature for Short-Term Risk of Progression to Active Tuberculosis Among Household Contacts, Front. Immunol. 9 (2018), doi:10.3389/fimmu.2018.00661.

88. A. Penn-Nicholson, T. Hraha, E. G. Thompson, D. Sterling, S. K. Mbandi, K. M. Wall, M. Fisher, S. Suliman, S. Shankar, W. A. Hanekom, N. Janjic, M. Hatherill, S. H. 170

E. Kaufmann, J. Sutherland, G. Walzl, M. A. D. Groote, U. Ochsner, D. E. Zak, T. J. Scriba, A. and G.-74 cohort study Groups, Discovery and validation of a prognostic proteomic signature for tuberculosis progression: A prospective cohort study, PLOS Med. 16, e1002781 (2019).

89. J. Weiner, J. Maertzdorf, J. S. Sutherland, F. J. Duffy, E. Thompson, S. Suliman, G. McEwen, B. Thiel, S. K. Parida, J. Zyla, W. A. Hanekom, R. P. Mohney, W. H. Boom, H. Mayanja-Kizza, R. Howe, H. M. Dockrell, T. H. M. Ottenhoff, T. J. Scriba, D. E. Zak, G. Walzl, S. H. E. Kaufmann, Metabolite changes in blood predict the onset of tuberculosis, Nat. Commun. 9, 5208 (2018).

90. T. J. Scriba, A. Penn-Nicholson, S. Shankar, T. Hraha, E. G. Thompson, D. Sterling, E. Nemes, F. Darboe, S. Suliman, L. M. Amon, H. Mahomed, M. Erasmus, W. Whatney, J. L. Johnson, W. H. Boom, M. Hatherill, J. Valvo, M. A. D. Groote, U. A. Ochsner, A. Aderem, W. A. Hanekom, D. E. Zak, other members of the A. cohort study Team, Sequential inflammatory processes define human progression from M. tuberculosis infection to tuberculosis disease, PLOS Pathog. 13, e1006687 (2017).

91. ClinicalTrials.gov, The Correlate of Risk Targeted Intervention Study (CORTIS), Identifier NCT02735590 (available at https://clinicaltrials.gov/ct2/show/NCT02735590).

92. F. J. Duffy, J. I. Weiner, S. Hansen, D. L. Tabb, S. Suliman, E. Thompson, J. Maertzdorf, S. Shankar, G. Tromp, S. Parida, D. Dover, M. K. Axthelm, J. S. Sutherland, H. M. Dockrell, T. H. M. Ottenhoff, T. J. Scriba, L. J. Picker, G. Walzl, S. H. E. Kaufmann, D. E. Zak, T. G.-74 Consortium, S. H. E. Kaufmann, S. K. Parida, R. Golinski, J. Maertzdorf, J. 3rd Weiner, M. Jacobson, G. McEwen, G. Walzl, G. F. Black, G. van der Spuy, K. Stanley, M. Kriel, N. Du Plessis, N. Nene, A. G. Loxton, N. N. Chegou, S. Suliman, T. Scriba, M. Fisher, H. Mahomed, J. Hughes, K. Downing, A. Penn-Nicholson, H. Mulenga, B. Abel, M. Bowmaker, B. Kagina, W. Kwong, C. W. Hanekom, T. H. M. Ottenhoff, M. R. Klein, M. C. Haks, K. L. Franken, A. Geluk, K. E. van Meijgaarden, S. A. Joosten, D. van Baarle, F. Miedema, W. H. Boom, B. Thiel, J. Sadoff, D. Sizemore, S. Ramachandran, L. Barker, M. Brennan, F. Weichold, S. Muller, L. Geiter, G. Schoolnik, G. Dolganov, T. Van, H. Mayanja-Kizza, M. Joloba, S. Zalwango, M. Nsereko, B. Okwera, H. Kisingo, H. M. Dockrell, S. Smith, P. Gorak- Stolinska, Y.-G. Hur, M. Lalor, J.-S. Lee, A. C. Crampin, N. French, B. Ngwira, A. B. Smith, K. Watkins, L. Ambrose, F. Simukonda, H. Mvula, F. Chilongo, J. Saul, K. Branson, D. Kassa, A. Abebe, T. Mesele, B. Tegbaru, R. Howe, A. Mihret, A. Aseffa, Y. Bekele, R. Iwnetu, M. Tafesse, L. Yamuah, M. Ota, J. Sutherland, P. Hill, R. Adegbola, T. Corrah, M. Antonio, T. Togun, I. Adetifa, S. Donkor, P. Andersen, I. Rosenkrands, M. Doherty, K. Weldingh, Immunometabolic Signatures Predict Risk of Progression to Active Tuberculosis and Disease Outcome, Front. Immunol. 10 (2019), doi:10.3389/fimmu.2019.00527.

171

93. M. J. Saunders, T. Wingfield, S. Datta, R. Montoya, E. Ramos, M. R. Baldwin, M. A. Tovar, B. E. W. Evans, R. H. Gilman, C. A. Evans, A household-level score to predict the risk of tuberculosis among contacts of patients with tuberculosis: a derivation and external validation prospective cohort study, Lancet Infect. Dis. (2019), doi:10.1016/S1473-3099(19)30423-2.

94. S. Verver, R. M. Warren, Z. Munch, M. Richardson, G. D. van der Spuy, M. W. Borgdorff, M. A. Behr, N. Beyers, P. D. van Helden, Proportion of tuberculosis transmission that takes place in households in a high-incidence area, The Lancet 363, 212–214 (2004).

95. T. A. Yates, P. Y. Khan, G. M. Knight, J. G. Taylor, T. D. McHugh, M. Lipman, R. G. White, T. Cohen, F. G. Cobelens, R. Wood, D. A. J. Moore, I. Abubakar, The transmission of Mycobacterium tuberculosis in high burden settings, Lancet Infect. Dis. 16, 227–238 (2016).

96. M. A. Behr, P. H. Edelstein, L. Ramakrishnan, Revisiting the timetable of tuberculosis, BMJ 362, k2738 (2018).

97. G. J. Fox, S. E. Barry, W. J. Britton, G. B. Marks, Contact investigation for tuberculosis: a systematic review and meta-analysis, Eur. Respir. J. 41, 140–156 (2013).

98. P. Kasaie, J. R. Andrews, W. D. Kelton, D. W. Dowdy, Timing of Tuberculosis Transmission and the Impact of Household Contact Tracing. An Agent-based Simulation Model, Am. J. Respir. Crit. Care Med. 189, 845–852 (2014).

99. M. R. Reichler, A. Khan, T. R. Sterling, H. Zhao, J. Moran, J. McAuley, P. Bessler, B. Mangura, I. Bakhtawar, C. LeDoux, J. McAuley, J. Beison, M. Fitzgerald, M. Naus, M. Nakajima, N. Schluger, Y. Hirsch-Moverman, J. Moran, H. Blumberg, J. Tapia, L. Singha, E. Hershfeld, B. Roche, B. Mangura, A. Sevilla, T. Sterling, T. Chavez-Lindell, F. Maruri, S. Dorman, W. Cronin, E. Munk, A. Khan, Y. Yuan, B. Chen, F. Yan, Y. Shen, H. Zhao, H. Zhang, P. Bessler, M. Fagley, M. Reichler, M. Reichler, T. Sterling, J. Tapia, C. Hirsch, C. Luo, Risk and Timing of Tuberculosis Among Close Contacts of Persons with Infectious Tuberculosis, J. Infect. Dis. 218, 1000–1008 (2018).

100. R. Sloot, M. F. Schim van der Loeff, P. M. Kouw, M. W. Borgdorff, Risk of Tuberculosis after Recent Exposure. A 10-Year Follow-up Study of Contacts in Amsterdam, Am. J. Respir. Crit. Care Med. 190, 1044–1052 (2014).

101. I. Sutherland, Recent studies in the epidemiology of tuberculosis, based on the risk of being infected with tubercle bacilli, Adv. Tuberc. Res. Fortschritte Tuberkuloseforschung Progres Explor. Tuberc. 19, 1–63 (1976).

102. A. Poulsen, Some clinical features of tuberculosis, Acta Tuberc. Scand. 33, 37–92; concl (1957). 172

103. T. Gedde-Dahl, Tuberculous infection in the light of tuberculin matriculation, Am. J. Hyg. 56, 139–214 (1952).

104. O. R. McCarthy, Asian immigrant tuberculosis--the effect of visiting Asia, Br. J. Dis. Chest 78, 248–253 (1984).

105. H.-A. Hatherell, X. Didelot, S. L. Pollock, P. Tang, A. Crisan, J. C. Johnston, C. Colijn, J. L. Gardy, Declaring a tuberculosis outbreak over with genomic epidemiology, Microb. Genomics 2 (2016), doi:10.1099/mgen.0.000060.

106. J. L. Zelner, M. B. Murray, M. C. Becerra, J. Galea, L. Lecca, R. Calderon, R. Yataco, C. Contreras, Z. Zhang, J. Manjourides, B. T. Grenfell, T. Cohen, Identifying Hotspots of Multidrug-Resistant Tuberculosis Transmission Using Spatial and Molecular Genetic Data, J. Infect. Dis. 213, 287–294 (2016).

107. A. Poulsen, Some clinical features of tuberculosis. 1. Incubation period, Acta Tuberc. Scand. 24, 311–346 (1950).

108. A. Wallgren, The time-table of tuberculosis, Tubercle 29, 245–251 (1948).

109. G. J. Churchyard, K. L. Fielding, J. J. Lewis, L. Coetzee, E. L. Corbett, P. Godfrey- Faussett, R. J. Hayes, R. E. Chaisson, A. D. Grant, A Trial of Mass Isoniazid Preventive Therapy for Tuberculosis Control, N. Engl. J. Med. 370, 301–310 (2014).

110. H. Esmail, C. E. Barry, D. B. Young, R. J. Wilkinson, The ongoing challenge of latent tuberculosis, Phil Trans R Soc B 369, 20130437 (2014).

111. S. H. Ferebee, Controlled chemoprophylaxis trials in tuberculosis. A general review, Bibl. Tuberc. 26, 28–106 (1970).

112. M. W. Borgdorff, M. Sebek, R. B. Geskus, K. Kremer, N. Kalisvaart, D. van Soolingen, The incubation period distribution of tuberculosis estimated with a molecular epidemiological approach, Int. J. Epidemiol. 40, 964–970 (2011).

113. P. D. Hart, I. Sutherland, BCG and vole bacillus vaccines in the prevention of tuberculosis in adolescence and early adult life., Br. Med. J. 2, 293–295 (1977).

114. W. H. Frost, THE AGE SELECTION OF MORTALITY FROM TUBERCULOSIS IN SUCCESSIVE DECADES, Am. J. Epidemiol. 141, 4–9 (1995).

115. K. Tocque, M. A. Bellis, C. M. Tam, S. L. Chan, Q. Syed, T. Remmington, P. D. O. Davies, Long-term Trends in Tuberculosis, Am. J. Respir. Crit. Care Med. 158, 484–488 (1998).

173

116. H. G. Wiker, T. Mustafa, G. A. Bjune, M. Harboe, Evidence for waning of latency in a cohort study of tuberculosis, BMC Infect. Dis. 10, 37–46 (2010).

117. Centers for Disease Control and Prevention, Reported Tuberculosis in the United States, 2016 (2017) (available at https://www.cdc.gov/tb/statistics/reports/2016/pdfs/2016_Surveillance_FullReport.pdf).

118. P.-C. Chan, P.-H. Lee, M.-J. Lu, Y.-C. Huang, T.-Y. Feng, W.-W. Chen, C.-C. Lee, Y.-F. Huang, C.-Y. Chiang, Tolerability of rifapentine-based regimens in latent tuberculosis infection treatment in the elderly, Eur. Respir. J. 53, 1802396 (2019).

119. L. Gao, H. Zhang, H. Xin, J. Liu, S. Pan, X. Li, L. Guan, F. Shen, Z. Liu, D. Wang, X. Guan, J. Yan, H. Li, B. Feng, X. Cao, Y. Chen, W. Cui, Z. Zhang, Y. Ma, X. Chen, X. Zhou, Q. Jin, Short-course regimens of rifapentine plus isoniazid to treat latent tuberculosis infection in older Chinese patients: a randomised controlled study, Eur. Respir. J. 52, 1801470 (2018).

120. C. S. Noh, H. I. Kim, H. Choi, Y. Kim, C.-H. Kim, J.-H. Choi, I. G. Hyun, M. S. Baek, Completion rate of latent tuberculosis infection treatment in patients aged 65 years and older, Respir. Med. 157, 52–58 (2019).

121. J. J. Saukkonen, D. L. Cohn, R. M. Jasmer, S. Schenker, J. A. Jereb, C. M. Nolan, C. A. Peloquin, F. M. Gordin, D. Nunes, D. B. Strader, J. Bernardo, R. Venkataramanan, T. R. Sterling, An Official ATS Statement: Hepatotoxicity of Antituberculosis Therapy, Am. J. Respir. Crit. Care Med. 174, 935–952 (2006).

122. J. S. Eom, I. Kim, W.-Y. Kim, E.-J. Jo, J. Mok, M.-H. Kim, K. Lee, K. U. Kim, H.- K. Park, M. K. Lee, Household tuberculosis contact investigation in a tuberculosis- prevalent country: Are the tuberculin skin test and interferon-gamma release assay enough in elderly contacts?, Medicine (Baltimore) 97, e9681 (2018).

123. Y. Luo, S. Suliman, S. Asgari, T. Amariuta, Y. Baglaenko, M. Martínez-Bonet, K. Ishigaki, M. Gutierrez-Arcelus, R. Calderon, L. Lecca, S. R. León, J. Jimenez, R. Yataco, C. Contreras, J. T. Galea, M. Becerra, S. Nejentsev, P. A. Nigrovic, D. B. Moody, M. B. Murray, S. Raychaudhuri, Early progression to active tuberculosis is a highly heritable trait driven by 3q23 in Peruvians, Nat. Commun. 10, 1–10 (2019).

124. J. B. Torrelles, L. S. Schlesinger, Integrating Lung Physiology, Immunology and Tuberculosis, Trends Microbiol. 25, 688–697 (2017).

125. S. B. Cohen, B. H. Gern, J. L. Delahaye, K. N. Adams, C. R. Plumlee, J. K. Winkler, D. R. Sherman, M. Y. Gerner, K. B. Urdahl, Alveolar Macrophages Provide an Early Mycobacterium tuberculosis Niche and Initiate Dissemination, Cell Host Microbe 24, 439-446.e4 (2018).

174

126. C. J. Martin, A. M. Cadena, V. W. Leung, P. L. Lin, P. Maiello, N. Hicks, M. R. Chase, J. L. Flynn, S. M. Fortune, Digitally Barcoding Mycobacterium tuberculosis Reveals In Vivo Infection Dynamics in the Macaque Model of Tuberculosis, mBio 8 (2017), doi:10.1128/mBio.00312-17.

127. B. A. Norris, J. D. Ernst, Mononuclear cell dynamics in M. tuberculosis infection provide opportunities for therapeutic intervention, PLOS Pathog. 14, e1007154 (2018).

128. M. Samstein, H. A. Schreiber, I. M. Leiner, B. Sušac, M. S. Glickman, E. G. Pamer, Essential yet limited role for CCR2+ inflammatory monocytes during Mycobacterium tuberculosis-specific T cell priming, eLife 2, e01086 (2013).

129. A. J. Wolf, B. Linas, G. J. Trevejo-Nuñez, E. Kincaid, T. Tamura, K. Takatsu, J. D. Ernst, Mycobacterium tuberculosis Infects Dendritic Cells with High Frequency and Impairs Their Function In Vivo, J. Immunol. 179, 2509–2519 (2007).

130. A. J. Wolf, L. Desvignes, B. Linas, N. Banaiee, T. Tamura, K. Takatsu, J. D. Ernst, Initiation of the adaptive immune response to Mycobacterium tuberculosis depends on antigen production in the local lymph node, not the lungs, J. Exp. Med. 205, 105–115 (2008).

131. S. K. C. Ganchua, A. M. Cadena, P. Maiello, H. P. Gideon, A. J. Myers, B. F. Junecko, E. C. Klein, P. L. Lin, J. T. Mattila, J. L. Flynn, Lymph nodes are sites of prolonged bacterial persistence during Mycobacterium tuberculosis infection in macaques, PLOS Pathog. 14, e1007337 (2018).

132. J. A. Myers, The Natural History of Tuberculosis in the Human Body: Forty-Five Years of Observation, JAMA 194, 1086–1092 (1965).

133. M. A. Behr, W. R. Waters, Is tuberculosis a lymphatic disease with a pulmonary portal?, Lancet Infect. Dis. 14, 250–255 (2014).

134. J. Barrios-Payán, M. Saqui-Salces, M. Jeyanathan, A. Alcántara-Vazquez, M. Castañon-Arreola, G. Rook, R. Hernandez-Pando, Extrapulmonary Locations of Mycobacterium tuberculosis DNA During Latent Infection, J. Infect. Dis. 206, 1194– 1205 (2012).

135. M. Beigier-Bompadre, G. N. Montagna, A. A. Kühl, L. Lozza, J. W. Iii, A. Kupz, A. Vogelzang, H.-J. Mollenkopf, D. Löwe, S. Bandermann, A. Dorhoi, V. Brinkmann, K. Matuschewski, S. H. E. Kaufmann, Mycobacterium tuberculosis infection modulates adipose tissue biology, PLOS Pathog. 13, e1006676 (2017).

136. W. R. Bishai, Rekindling old controversy on elusive lair of latent tuberculosis, The Lancet 356, 2113–2114 (2000).

175

137. S. V. Capuano, D. A. Croix, S. Pawar, A. Zinovik, A. Myers, P. L. Lin, S. Bissel, C. Fuhrman, E. Klein, J. L. Flynn, Experimental Mycobacterium tuberculosis Infection of Cynomolgus Macaques Closely Resembles the Various Manifestations of Human M. tuberculosis Infection, Infect. Immun. 71, 5831–5844 (2003).

138. A. A. Chackerian, J. M. Alt, T. V. Perera, C. C. Dascher, S. M. Behar, Dissemination of Mycobacterium tuberculosis is influenced by host factors and precedes the initiation of T-cell immunity, Infect. Immun. 70, 4501–4509 (2002).

139. B. Das, S. S. Kashino, I. Pulu, D. Kalita, V. Swami, H. Yeger, D. W. Felsher, A. Campos-Neto, CD271+ Bone Marrow Mesenchymal Stem Cells May Provide a Niche for Dormant Mycobacterium tuberculosis, Sci. Transl. Med. 5, 170ra13-170ra13 (2013).

140. R. Hernández-Pando, M. Jeyanathan, G. Mengistu, D. Aguilar, H. Orozco, M. Harboe, G. Rook, G. Bjune, Persistence of DNA from Mycobacterium tuberculosis in superficially normal lung tissue during latent infection, The Lancet 356, 2133–2138 (2000).

141. J. Mayito, I. Andia, M. Belay, D. A. Jolliffe, D. P. Kateete, S. T. Reece, A. R. Martineau, Anatomic and Cellular Niches for Mycobacterium tuberculosis in Latent Tuberculosis Infection, J. Infect. Dis. 219, 685–694 (2019).

142. O. Neyrolles, R. Hernández-Pando, F. Pietri-Rouxel, P. Fornès, L. Tailleux, J. A. B. Payán, E. Pivert, Y. Bordat, D. Aguilar, M.-C. Prévost, C. Petit, B. Gicquel, Is Adipose Tissue a Place for Mycobacterium tuberculosis Persistence?, PLOS ONE 1, e43 (2006).

143. P. L. Lin, M. Rodgers, L. Smith, M. Bigbee, A. Myers, C. Bigbee, I. Chiosea, S. V. Capuano, C. Fuhrman, E. Klein, J. L. Flynn, Quantitative Comparison of Active and Latent Tuberculosis in the Cynomolgus Macaque Model, Infect. Immun. 77, 4631–4642 (2009).

144. J. Chan, S. Mehta, S. Bharrhan, Y. Chen, J. M. Achkar, A. Casadevall, J. Flynn, The role of B cells and humoral immunity in Mycobacterium tuberculosis infection, Semin. Immunol. 26, 588–600 (2014).

145. E. Guirado, L. Schlesinger, Modeling the Mycobacterium tuberculosis Granuloma – the Critical Battlefield in Host Immunity and Disease, Front. Immunol. 4 (2013), doi:10.3389/fimmu.2013.00098.

146. P. L. Lin, C. B. Ford, M. T. Coleman, A. J. Myers, R. Gawande, T. Ioerger, J. Sacchettini, S. M. Fortune, J. L. Flynn, Sterilization of granulomas is common in active and latent tuberculosis despite within-host variability in bacterial killing, Nat. Med. 20, 75–79 (2014).

176

147. R. L. Hunter, Pathology of post primary tuberculosis of the lung: An illustrated critical review, Tuberculosis 91, 497–509 (2011).

148. R. L. Hunter, The Pathogenesis of Tuberculosis: The Early Infiltrate of Post-primary (Adult Pulmonary) Tuberculosis: A Distinct Disease Entity, Front. Immunol. 9 (2018), doi:10.3389/fimmu.2018.02108.

149. R. L. Hunter, C. Jagannath, J. K. Actor, Pathology of postprimary tuberculosis in humans and mice: Contradiction of long-held beliefs, Tuberculosis 87, 267–278 (2007).

150. H. Esmail, R. Lai, M. Lesosky, K. Wilkinson, C. Graham, A. Coussens, T. Oni, J. Warwick, Q. Said-Hartley, C. Koegelenberg, G. Walzl, J. Flynn, D. Young, C. Barry, A. O’Garra, R. Wilkinson, Characterization of progressive HIV-associated tuberculosis using 2-deoxy-2-[18F]fluoro-D-glucose positron emission and computed tomography, Nat. Med. 22, 1090–1093 (2016).

151. E. Geng, B. Kreiswirth, J. Burzynski, N. W. Schluger, Clinical and Radiographic Correlates of Primary and Reactivation Tuberculosis: A Molecular Epidemiology Study, JAMA 293, 2740–2745 (2005).

152. B. E. Jones, R. Ryu, Z. Yang, M. D. Cave, J. M. Pogoda, M. Otaya, P. F. Barnes, Chest Radiographic Findings in Patients with Tuberculosis with Recent or Remote Infection, Am. J. Respir. Crit. Care Med. 156, 1270–1273 (1997).

153. W. P. Gill, N. S. Harik, M. R. Whiddon, R. P. Liao, J. E. Mittler, D. R. Sherman, A replication clock for Mycobacterium tuberculosis, Nat. Med. 15, 211–214 (2009).

154. U. S. Gautam, T. W. Foreman, A. N. Bucsan, A. V. Veatch, X. Alvarez, T. Adekambi, N. A. Golden, K. M. Gentry, L. A. Doyle-Meyers, K. E. Russell-Lodrigue, P. J. Didier, J. L. Blanchard, K. G. Kousoulas, A. A. Lackner, D. Kalman, J. Rengarajan, S. A. Khader, D. Kaushal, S. Mehra, In vivo inhibition of tryptophan catabolism reorganizes the tuberculoma and augments immune-mediated control of Mycobacterium tuberculosis, Proc. Natl. Acad. Sci. 115, E62–E71 (2018).

155. S. Srivastava, J. D. Ernst, Direct recognition of infected cells by CD4 T cells is required for control of intracellular Mycobacterium tuberculosis in vivo, J. Immunol. Baltim. Md 1950 191, 1016–1020 (2013).

156. W. H. Boom, R. S. Wallis, K. A. Chervenak, Human Mycobacterium tuberculosis- reactive CD4+ T-cell clones: heterogeneity in antigen recognition, cytokine production, and cytotoxicity for mononuclear phagocytes., Infect. Immun. 59, 2737–2743 (1991).

157. N. V. Serbina, C.-C. Liu, C. A. Scanga, J. L. Flynn, CD8+ CTL from Lungs of Mycobacterium tuberculosis-Infected Mice Express Perforin In Vivo and Lyse Infected Macrophages, J. Immunol. 165, 353–363 (2000). 177

158. J. D. Yang, D. Mott, R. Sutiwisesak, Y.-J. Lu, F. Raso, B. Stowell, G. H. Babunovic, J. Lee, S. M. Carpenter, S. S. Way, S. M. Fortune, S. M. Behar, Mycobacterium tuberculosis-specific CD4+ and CD8+ T cells differ in their capacity to recognize infected macrophages, PLOS Pathog. 14, e1007060 (2018).

159. L. Abel, J. Fellay, D. W. Haas, E. Schurr, G. Srikrishna, M. Urbanowski, N. Chaturvedi, S. Srinivasan, D. H. Johnson, W. R. Bishai, Genetics of human susceptibility to active and latent tuberculosis: present knowledge and future perspectives, Lancet Infect. Dis. 18, e64–e75 (2018).

160. R. Martínez-Barricarte, J. G. Markle, C. S. Ma, E. K. Deenick, N. Ramírez-Alejo, F. Mele, D. Latorre, S. A. Mahdaviani, C. Aytekin, D. Mansouri, V. L. Bryant, F. Jabot- Hanin, C. Deswarte, A. Nieto-Patlán, L. Surace, G. Kerner, Y. Itan, S. Jovic, D. T. Avery, N. Wong, G. Rao, E. Patin, S. Okada, B. Bigio, B. Boisson, F. Rapaport, Y. Seeleuthner, M. Schmidt, A. Ikinciogullari, F. Dogu, G. Tanir, P. Tabarsi, M. R. Bloursaz, J. K. Joseph, A. Heer, X.-F. Kong, M. Migaud, T. Lazarov, F. Geissmann, B. Fleckenstein, C. L. Arlehamn, A. Sette, A. Puel, J.-F. Emile, E. van de Vosse, L. Quintana-Murci, J. P. D. Santo, L. Abel, S. Boisson-Dupuis, J. Bustamante, S. G. Tangye, F. Sallusto, J.-L. Casanova, Human IFN-γ immunity to mycobacteria is governed by both IL-12 and IL-23, Sci. Immunol. 3 (2018), doi:10.1126/sciimmunol.aau6759.

161. S. Boisson-Dupuis, N. Ramirez-Alejo, Z. Li, E. Patin, G. Rao, G. Kerner, C. K. Lim, D. N. Krementsov, N. Hernandez, C. S. Ma, Q. Zhang, J. Markle, R. Martinez-Barricarte, K. Payne, R. Fisch, C. Deswarte, J. Halpern, M. Bouaziz, J. Mulwa, D. Sivanesan, T. Lazarov, R. Naves, P. Garcia, Y. Itan, B. Boisson, A. Checchi, F. Jabot-Hanin, A. Cobat, A. Guennoun, C. C. Jackson, S. Pekcan, Z. Caliskaner, J. Inostroza, B. T. Costa- Carvalho, J. A. T. de Albuquerque, H. Garcia-Ortiz, L. Orozco, T. Ozcelik, A. Abid, I. A. Rhorfi, H. Souhi, H. N. Amrani, A. Zegmout, F. Geissmann, S. W. Michnick, I. Muller- Fleckenstein, B. Fleckenstein, A. Puel, M. J. Ciancanelli, N. Marr, H. Abolhassani, M. E. Balcells, A. Condino-Neto, A. Strickler, K. Abarca, C. Teuscher, H. D. Ochs, I. Reisli, E. H. Sayar, J. El-Baghdadi, J. Bustamante, L. Hammarström, S. G. Tangye, S. Pellegrini, L. Quintana-Murci, L. Abel, J.-L. Casanova, Tuberculosis and impaired IL-23–dependent IFN-γ immunity in humans homozygous for a common TYK2 missense variant, Sci. Immunol. 3 (2018), doi:10.1126/sciimmunol.aau8714.

162. H. P. Gideon, J. Phuah, A. J. Myers, B. D. Bryson, M. A. Rodgers, M. T. Coleman, P. Maiello, T. Rutledge, S. Marino, S. M. Fortune, D. E. Kirschner, P. L. Lin, J. L. Flynn, Variability in Tuberculosis Granuloma T Cell Responses Exists, but a Balance of Pro- and Anti-inflammatory Cytokines Is Associated with Sterilization, PLOS Pathog. 11, e1004603 (2015).

163. M. A. Sallin, K. D. Kauffman, C. Riou, E. D. Bruyn, T. W. Foreman, S. Sakai, S. G. Hoft, T. G. Myers, P. J. Gardina, A. Sher, R. Moore, T. Wilder-Kofie, I. N. Moore, A. 178

Sette, C. S. L. Arlehamn, R. J. Wilkinson, D. L. Barber, Host resistance to pulmonary Mycobacterium tuberculosis infection requires CD153 expression, Nat. Microbiol. 3, 1198 (2018).

164. I. Comas, J. Chakravartti, P. M. Small, J. Galagan, S. Niemann, K. Kremer, J. D. Ernst, S. Gagneux, Human T cell epitopes of Mycobacterium tuberculosis are evolutionarily hyperconserved, Nat. Genet. 42, 498–503 (2010).

165. M. Coscolla, R. Copin, J. Sutherland, F. Gehre, B. de Jong, O. Owolabi, G. Mbayo, F. Giardina, J. D. Ernst, S. Gagneux, M. tuberculosis T Cell Epitope Analysis Reveals Paucity of Antigenic Variation and Identifies Rare Variable TB Antigens, Cell Host Microbe 18, 538–548 (2015).

166. J. L. Flynn, J. Chan, Immunology of Tuberculosis, Annu. Rev. Immunol. 19, 93–129 (2001).

167. T. D. Bold, N. Banaei, A. J. Wolf, J. D. Ernst, Suboptimal Activation of Antigen- Specific CD4+ Effector Cells Enables Persistence of M. tuberculosis In Vivo, PLOS Pathog. 7, e1002063 (2011).

168. J. G. Egen, A. G. Rothfuchs, C. G. Feng, M. A. Horwitz, A. Sher, R. N. Germain, Intravital Imaging Reveals Limited Antigen Presentation and T Cell Effector Function in Mycobacterial Granulomas, Immunity 34, 807–819 (2011).

169. S. Srivastava, P. S. Grace, J. D. Ernst, Antigen Export Reduces Antigen Presentation and Limits T Cell Control of M. tuberculosis, Cell Host Microbe 19, 44–54 (2016).

170. A. Alpert, Y. Pickman, M. Leipold, Y. Rosenberg-Hasson, X. Ji, R. Gaujoux, H. Rabani, E. Starosvetsky, K. Kveler, S. Schaffert, D. Furman, O. Caspi, U. Rosenschein, P. Khatri, C. L. Dekker, H. T. Maecker, M. M. Davis, S. S. Shen-Orr, A clinically meaningful metric of immune age derived from high-dimensional longitudinal monitoring, Nat. Med. 25, 487 (2019).

171. J. J. Goronzy, C. M. Weyand, Successful and Maladaptive T Cell Aging, Immunity 46, 364–378 (2017).

172. J. J. C. Thome, N. Yudanin, Y. Ohmura, M. Kubota, B. Grinshpun, T. Sathaliyawala, T. Kato, H. Lerner, Y. Shen, D. L. Farber, Spatial Map of Human T Cell Compartmentalization and Maintenance over Decades of Life, Cell 159, 814–828 (2014).

173. J. J. C. Thome, B. Grinshpun, B. V. Kumar, M. Kubota, Y. Ohmura, H. Lerner, G. D. Sempowski, Y. Shen, D. L. Farber, Longterm maintenance of human naive T cells through in situ homeostasis in lymphoid tissue sites, Sci. Immunol. 1 (2016), doi:10.1126/sciimmunol.aah6506.

179

174. Q. Qi, M. M. Cavanagh, S. L. Saux, H. NamKoong, C. Kim, E. Turgano, Y. Liu, C. Wang, S. Mackey, G. E. Swan, C. L. Dekker, R. A. Olshen, S. D. Boyd, C. M. Weyand, L. Tian, J. J. Goronzy, Diversification of the antigen-specific T cell receptor repertoire after varicella zoster vaccination, Sci. Transl. Med. 8, 332ra46-332ra46 (2016).

175. R. S. Herati, M. A. Reuter, D. V. Dolfi, K. D. Mansfield, H. Aung, O. Z. Badwan, R. K. Kurupati, S. Kannan, H. Ertl, K. E. Schmader, M. R. Betts, D. H. Canaday, E. J. Wherry, Circulating CXCR5+PD-1+ Response Predicts Influenza Vaccine Antibody Responses in Young Adults but not Elderly Adults, J. Immunol. 193, 3528–3537 (2014).

176. S. Mohanty, S. R. Joshi, I. Ueda, J. Wilson, T. P. Blevins, B. Siconolfi, H. Meng, L. Devine, K. Raddassi, S. Tsang, R. B. Belshe, D. A. Hafler, S. M. Kaech, S. H. Kleinstein, M. Trentalange, H. G. Allore, A. C. Shaw, Prolonged Proinflammatory Cytokine Production in Monocytes Modulated by Interleukin 10 After Influenza Vaccination in Older Adults, J. Infect. Dis. 211, 1174–1184 (2015).

177. M. Vukmanovic-Stejic, E. S. Chambers, M. Suárez-Fariñas, D. Sandhu, J. Fuentes- Duculan, N. Patel, E. Agius, K. E. Lacy, C. T. Turner, A. Larbi, V. Birault, M. Noursadeghi, N. A. Mabbott, M. H. A. Rustin, J. G. Krueger, A. N. Akbar, Enhancement of cutaneous immunity during aging by blocking p38 mitogen-activated protein (MAP) kinase–induced inflammation, J. Allergy Clin. Immunol. (2017), doi:10.1016/j.jaci.2017.10.032.

178. I. M. Orme, Aging and immunity to tuberculosis: increased susceptibility of old mice reflects a decreased capacity to generate mediator T lymphocytes., J. Immunol. 138, 4414–4418 (1987).

179. I. M. Orme, J. P. Griffin, A. D. Roberts, D. N. Ernst, Evidence for a Defective Accumulation of Protective T Cells in Old Mice Infected with Mycobacterium tuberculosis, Cell. Immunol. 147, 222–229 (1993).

180. S.-H. Wang, B. Carruthers, J. Turner, The Influence of Increasing Age on Susceptibility of the Elderly to Tuberculosis, Open Longev. Sci. 6 (2012) (available at https://benthamopen.com/ABSTRACT/TOLSJ-6-73).

181. D. E. Harrison, C. M. Astle, M. K. K. Niazi, S. Major, G. L. Beamer, Genetically diverse mice are novel and valuable models of age-associated susceptibility to Mycobacterium tuberculosis, Immun. Ageing 11, 24 (2014).

182. J. Y. Kawahara, E. B. Irvine, G. Alter, A Case for Antibodies as Mechanistic Correlates of Immunity in Tuberculosis, Front. Immunol. 10 (2019), doi:10.3389/fimmu.2019.00996.

183. L. L. Lu, A. W. Chung, T. R. Rosebrock, M. Ghebremichael, W. H. Yu, P. S. Grace, M. K. Schoen, F. Tafesse, C. Martin, V. Leung, A. E. Mahan, M. Sips, M. P. Kumar, J. 180

Tedesco, H. Robinson, E. Tkachenko, M. Draghi, K. J. Freedberg, H. Streeck, T. J. Suscovich, D. A. Lauffenburger, B. I. Restrepo, C. Day, S. M. Fortune, G. Alter, A Functional Role for Antibodies in Tuberculosis, Cell 167, 433-443.e14 (2016).

184. J. Phuah, E. A. Wong, H. P. Gideon, P. Maiello, M. T. Coleman, M. R. Hendricks, R. Ruden, L. R. Cirrincione, J. Chan, P. L. Lin, J. L. Flynn, Effects of B Cell Depletion on Early Mycobacterium tuberculosis Infection in Cynomolgus Macaques, Infect. Immun. 84, 1301–1311 (2016).

185. H. Li, X. Wang, B. Wang, L. Fu, G. Liu, Y. Lu, M. Cao, H. Huang, B. Javid, Latently and uninfected healthcare workers exposed to TB make protective antibodies against Mycobacterium tuberculosis, Proc. Natl. Acad. Sci. 114, 5023–5028 (2017).

186. J. Weiner, T. Domaszewska, S. Donkor, S. H. E. Kaufmann, P. C. Hill, J. S. Sutherland, Changes in transcript, metabolite and antibody reactivity during the early protective immune response in humans to Mycobacterium tuberculosis infection, Clin. Infect. Dis. , doi:10.1093/cid/ciz785.

187. D. Frasca, B. B. Blomberg, Effects of aging on B cell function, Curr. Opin. Immunol. 21, 425–430 (2009).

188. T. Pritz, J. Lair, M. Ban, M. Keller, B. Weinberger, M. Krismer, B. Grubeck‐ Loebenstein, Plasma cell numbers decrease in bone marrow of old patients, Eur. J. Immunol. 45, 738–746 (2015).

189. C. Henry, N.-Y. Zheng, M. Huang, A. Cabanov, K. T. Rojas, K. Kaur, S. F. Andrews, A.-K. E. Palm, Y.-Q. Chen, Y. Li, K. Hoskova, H. A. Utset, M. C. Vieira, J. Wrammert, R. Ahmed, J. Holden-Wiltse, D. J. Topham, J. J. Treanor, H. C. Ertl, K. E. Schmader, S. Cobey, F. Krammer, S. E. Hensley, H. Greenberg, X.-S. He, P. C. Wilson, Influenza Virus Vaccination Elicits Poorly Adapted B Cell Responses in Elderly Individuals, Cell Host Microbe 25, 357-366.e6 (2019).

190. R. R. Chowdhury, F. Vallania, Q. Yang, C. J. L. Angel, F. Darboe, A. Penn- Nicholson, V. Rozot, E. Nemes, S. T. Malherbe, K. Ronacher, G. Walzl, W. Hanekom, M. M. Davis, J. Winter, X. Chen, T. J. Scriba, P. Khatri, Y. Chien, A multi-cohort study of the immune factors associated with M. tuberculosis infection outcomes, Nature 560, 644–648 (2018).

191. S. Esin, G. Batoni, Natural Killer Cells: A Coherent Model for Their Functional Role in Mycobacterium tuberculosis Infection, J. Innate Immun. 7, 11–24 (2015).

192. C. G. Feng, M. Kaviratne, A. G. Rothfuchs, A. Cheever, S. Hieny, H. A. Young, T. A. Wynn, A. Sher, NK Cell-Derived IFN-γ Differentially Regulates Innate Resistance and Neutrophil Response in T Cell-Deficient Hosts Infected with Mycobacterium tuberculosis, J. Immunol. 177, 7086–7093 (2006). 181

193. A. P. Junqueira-Kipnis, A. Kipnis, A. Jamieson, M. G. Juarrero, A. Diefenbach, D. H. Raulet, J. Turner, I. M. Orme, NK Cells Respond to Pulmonary Infection with Mycobacterium tuberculosis, but Play a Minimal Role in Protection, J. Immunol. 171, 6039–6045 (2003).

194. E. E. Kroon, A. K. Coussens, C. Kinnear, M. Orlova, M. Möller, A. Seeger, R. J. Wilkinson, E. G. Hoal, E. Schurr, Neutrophils: Innate Effectors of TB Resistance?, Front. Immunol. 9 (2018), doi:10.3389/fimmu.2018.02637.

195. L. Huang, E. V. Nazarova, S. Tan, Y. Liu, D. G. Russell, Growth of Mycobacterium tuberculosis in vivo segregates with host macrophage metabolism and ontogenyOntogeny determines macrophage responses in tuberculosis, J. Exp. Med. 215, 1135–1152 (2018).

196. A. Ardain, R. Domingo-Gonzalez, S. Das, S. W. Kazer, N. C. Howard, A. Singh, M. Ahmed, S. Nhamoyebonde, J. Rangel-Moreno, P. Ogongo, L. Lu, D. Ramsuran, M. de la Luz Garcia-Hernandez, T. K Ulland, M. Darby, E. Park, F. Karim, L. Melocchi, R. Madansein, K. J. Dullabh, M. Dunlap, N. Marin-Agudelo, T. Ebihara, T. Ndung’u, D. Kaushal, A. S. Pym, J. K. Kolls, A. Steyn, J. Zúñiga, W. Horsnell, W. M. Yokoyama, A. K. Shalek, H. N. Kløverpris, M. Colonna, A. Leslie, S. A. Khader, Group 3 innate lymphoid cells mediate early protective immunity against tuberculosis, Nature 570, 528– 532 (2019).

197. A. Chackerian, J. Alt, V. Perera, S. M. Behar, Activation of NKT cells protects mice from tuberculosis, Infect. Immun. 70, 6302–6309 (2002).

198. S.-J. Kee, Y.-S. Kwon, Y.-W. Park, Y.-N. Cho, S.-J. Lee, T.-J. Kim, S.-S. Lee, H.- C. Jang, M.-G. Shin, J.-H. Shin, S.-P. Suh, D.-W. Ryang, Dysfunction of Natural Killer T Cells in Patients with Active Mycobacterium tuberculosis Infection, Infect. Immun. 80, 2100–2108 (2012).

199. A. Qaqish, D. Huang, C. Y. Chen, Z. Zhang, R. Wang, S. Li, E. Yang, Y. Lu, M. H. Larsen, W. R. Jacobs, L. Qian, J. Frencher, L. Shen, Z. W. Chen, Adoptive Transfer of Phosphoantigen-Specific γδ T Cell Subset Attenuates Mycobacterium tuberculosis Infection in Nonhuman Primates, J. Immunol. Baltim. Md 1950 198, 4753–4763 (2017).

200. L. Shen, J. Frencher, D. Huang, W. Wang, E. Yang, C. Y. Chen, Z. Zhang, R. Wang, A. Qaqish, M. H. Larsen, H. Shen, S. A. Porcelli, W. R. Jacobs, Z. W. Chen, Immunization of Vγ2Vδ2 T cells programs sustained effector memory responses that control tuberculosis in nonhuman primates, Proc. Natl. Acad. Sci. U. S. A. 116, 6371– 6378 (2019).

201. K. D. Kauffman, M. A. Sallin, S. G. Hoft, S. Sakai, R. Moore, T. Wilder-Kofie, I. N. Moore, A. Sette, C. S. L. Arlehamn, D. L. Barber, Limited Pulmonary Mucosal- Associated Invariant T Cell Accumulation and Activation during Mycobacterium

182 tuberculosis Infection in Rhesus Macaques, Infect. Immun. 86 (2018), doi:10.1128/IAI.00431-18.

202. C. H. Canan, N. S. Gokhale, B. Carruthers, W. P. Lafuse, L. S. Schlesinger, J. B. Torrelles, J. Turner, Characterization of lung inflammation and its impact on macrophage function in aging, J. Leukoc. Biol. 96, 473–480 (2014).

203. T. J. Piergallini, J. Turner, Tuberculosis in the elderly: Why inflammation matters, Exp. Gerontol. 105, 32–39 (2018).

204. J. Turner, A. A. Frank, I. M. Orme, Old mice express a transient early resistance to pulmonary tuberculosis that is mediated by CD8 T cells, Infect. Immun. 70, 4628–4637 (2002).

205. B. Vesosky, D. K. Flaherty, E. K. Rottinghaus, G. L. Beamer, J. Turner, Age dependent increase in early resistance of mice to Mycobacterium tuberculosis is associated with an increase in CD8 T cells that are capable of antigen independent IFN- gamma production, Exp. Gerontol. 41, 1185–1194 (2006).

206. B. Vesosky, D. K. Flaherty, J. Turner, Th1 Cytokines Facilitate CD8-T-Cell- Mediated Early Resistance to Infection with Mycobacterium tuberculosis in Old Mice, Infect. Immun. 74, 3314–3324 (2006).

207. B. Vesosky, E. K. Rottinghaus, C. Davis, J. Turner, CD8 T Cells in Old Mice Contribute to the Innate Immune Response to Mycobacterium tuberculosis via Interleukin-12p70-Dependent and Antigen-Independent Production of Gamma Interferon, Infect. Immun. 77, 3355–3363 (2009).

208. W. P. Lafuse, M. V. S. Rajaram, Q. Wu, J. I. Moliva, J. B. Torrelles, J. Turner, L. S. Schlesinger, Identification of an Increased Alveolar Macrophage Subpopulation in Old Mice That Displays Unique Inflammatory Characteristics and Is Permissive to Mycobacterium tuberculosis Infection, J. Immunol. 203, 2252–2264 (2019).

209. A. Halliday, H. Whitworth, S. H. Kottoor, U. Niazi, S. Menzies, H. Kunst, S. Bremang, A. Badhan, P. Beverley, O. M. Kon, A. Lalvani, Stratification of Latent Mycobacterium tuberculosis Infection by Cellular Immune Profiling, J. Infect. Dis. 215, 1480–1487 (2017).

210. N. du Plessis, L. Loebenberg, M. Kriel, F. von Groote-Bidlingmaier, E. Ribechini, A. G. Loxton, P. D. van Helden, M. B. Lutz, G. Walzl, Increased Frequency of Myeloid- derived Suppressor Cells during Active Tuberculosis and after Recent Mycobacterium tuberculosis Infection Suppresses T-Cell Function, Am. J. Respir. Crit. Care Med. 188, 724–732 (2013).

183

211. D. Goletti, O. Butera, V. Vanini, F. N. Lauria, C. Lange, K. L. M. C. Franken, C. Angeletti, T. H. M. Ottenhoff, E. Girardi, Response to Rv2628 latency antigen associates with cured tuberculosis and remote infection, Eur. Respir. J. 36, 135–142 (2010).

212. J. Castro-Garza, P. García-Jacobo, L. G. Rivera-Morales, F. D. Quinn, J. Barber, R. Karls, D. Haas, S. Helms, T. Gupta, H. Blumberg, J. Tapia, I. Luna-Cruz, A. Rendon, J. Vargas-Villarreal, L. Vera-Cabrera, C. Rodríguez-Padilla, Detection of anti-HspX antibodies and HspX protein in patient sera for the identification of recent latent infection by Mycobacterium tuberculosis, PLOS ONE 12, e0181714 (2017).

213. M. F. Rabahi, A. P. Junqueira-Kipnis, M. C. G. dos Reis, W. Oelemann, M. B. Conde, Humoral response to HspX and GlcB to previous and recent infection by Mycobacterium tuberculosis, BMC Infect. Dis. 7, 148 (2007).

214. M. Belay, M. Legesse, A. Mihret, T. H. M. Ottenhoff, K. L. Franken, G. Bjune, F. Abebe, IFN-γ and IgA against non-methylated heparin-binding hemagglutinin as markers of protective immunity and latent tuberculosis: Results of a longitudinal study from an endemic setting, J. Infect. 72, 189–200 (2016).

215. US Census Bureau, An Aging World: 2015 (2016) (available at https://www.census.gov/library/publications/2016/demo/P95-16-1.html).

216. C. López-Otín, M. A. Blasco, L. Partridge, M. Serrano, G. Kroemer, The Hallmarks of Aging, Cell 153, 1194–1217 (2013).

217. T. Niccoli, L. Partridge, Ageing as a Risk Factor for Disease, Curr. Biol. 22, R741– R752 (2012).

218. G. Gavazzi, K.-H. Krause, Ageing and infection, Lancet Infect. Dis. 2, 659–666 (2002).

219. T. T. Yoshikawa, Perspective: Aging and Infectious Diseases: Past, Present, and Future, J. Infect. Dis. 176, 1053–1057 (1997).

220. World Health Organization, Global Tuberculosis Report 2017 (World Health Organization, Geneva, Switzerland; http://apps.who.int/iris/bitstream/10665/259366/1/9789241565516-eng.pdf?ua=1).

221. Institute for Health Metrics and Evaluation, Global Burden of Disease (2016) (available at http://www.healthdata.org/gbd).

222. J. Negin, S. Abimbola, B. J. Marais, Tuberculosis among older adults – time to take notice, Int. J. Infect. Dis. 32, 135–137 (2015).

184

223. P. Glaziou, C. Sismanidis, K. Floyd, M. Raviglione, Global Epidemiology of Tuberculosis, Cold Spring Harb. Perspect. Med. 5, a017798 (2015).

224. N. S. Hochberg, C. R. Horsburgh, Prevention of Tuberculosis in Older Adults in the United States: Obstacles and Opportunities, Clin. Infect. Dis. 56, 1240–1247 (2013).

225. C. R. Horsburgh, M. O’Donnell, S. Chamblee, J. L. Moreland, J. Johnson, B. J. Marsh, M. Narita, L. S. Johnson, C. F. von Reyn, Revisiting Rates of Reactivation Tuberculosis, Am. J. Respir. Crit. Care Med. 182, 420–425 (2010).

226. R. Byng-Maddick, M. Noursadeghi, Does tuberculosis threaten our ageing populations?, BMC Infect. Dis. 16, 119 (2016).

227. G. Guzzetta, D. Kirschner, The Roles of Immune Memory and Aging in Protective Immunity and Endogenous Reactivation of Tuberculosis, PLOS ONE 8, e60425 (2013).

228. S. Menon, R. Rossi, L. Nshimyumukiza, A. Wusiman, N. Zdraveska, M. S. Eldin, Convergence of a diabetes mellitus, protein energy malnutrition, and TB epidemic: the neglected elderly population, BMC Infect. Dis. 16, 361 (2016).

229. S. Rajagopalan, Tuberculosis in Older Adults, Clin. Geriatr. Med. 32, 479–491 (2016).

230. L. Thrupp, S. Bradley, P. Smith, A. Simor, N. Gantz, K. Crossley, M. Loeb, L. Strausbaugh, L. Nicolle, S. L. Committee, Tuberculosis Prevention and Control in Long‐ Term–Care Facilities for Older Adults, Infect. Control Hosp. Epidemiol. 25, 1097–1108 (2004).

231. E. Dorken, S. Grzybowski, E. A. Allen, Significance of the Tuberculin Test in the Elderly, Chest 92, 237–240 (1987).

232. P. Van den Brande, M. Demedts, Four-Stage Tuberculin Testing in Elderly Subjects Induces Age-Dependent Progressive Boosting, Chest 101, 447–450 (1992).

233. D. A. Adambekov, V. L. Morozov, [T-suppressors and their functional activity in tuberculosis in middle-aged and elderly subjects], Probl. Tuberk. , 32–34 (1989).

234. S. Harada, [Lymphocyte function of pulmonary tuberculosis in the elderly], Kekkaku 64, 669–670 (1989).

235. S. Commandeur, K. E. van Meijgaarden, M. Y. Lin, K. L. M. C. Franken, A. H. Friggen, J. W. Drijfhout, F. Oftung, G. E. Korsvold, A. Geluk, T. H. M. Ottenhoff, Identification of Human T-Cell Responses to Mycobacterium tuberculosis Resuscitation- Promoting Factors in Long-Term Latently Infected Individuals, Clin. Vaccine Immunol. 18, 676–683 (2011). 185

236. J. Fox, S. Weisberg, An {R} Companion to Applied Regression (Sage, Thousand Oaks, Calif., Third., 2019; https://socialsciences.mcmaster.ca/jfox/Books/Companion/).

237. M. I. de Jonge, G. Pehau-Arnaudet, M. M. Fretz, F. Romain, D. Bottai, P. Brodin, N. Honoré, G. Marchal, W. Jiskoot, P. England, S. T. Cole, R. Brosch, ESAT-6 from Mycobacterium tuberculosis Dissociates from Its Putative Chaperone CFP-10 under Acidic Conditions and Exhibits Membrane-Lysing Activity, J. Bacteriol. 189, 6028–6034 (2007).

238. P. S. Renshaw, P. Panagiotidou, A. Whelan, S. V. Gordon, R. G. Hewinson, R. A. Williamson, M. D. Carr, Conclusive Evidence That the Major T-cell Antigens of theMycobacterium tuberculosis Complex ESAT-6 and CFP-10 Form a Tight, 1:1 Complex and Characterization of the Structural Properties of ESAT-6, CFP-10, and the ESAT-6·CFP-10 Complex IMPLICATIONS FOR PATHOGENESIS AND VIRULENCE, J. Biol. Chem. 277, 21598–21603 (2002).

239. Y. Li, L. Zhang, X. Liu, L. Wang, X. Wang, J. Wang, J. Qian, The role of in vitro interferonγ-release assay in differentiating intestinal tuberculosis from Crohn’s disease in China, J. Crohns Colitis 6, 317–323 (2012).

240. L. Richeldi, Tuberculosis Diagnostic Tests: Sensitivity, Specificity, and Comparing Apples with Apples, Am. J. Respir. Crit. Care Med. 174, 953a–9954 (2006).

241. B. Jamil, F. Shahid, Z. Hasan, N. Nasir, T. Razzaki, G. Dawood, R. Hussain, Interferonγ/IL10 ratio defines the disease severity in pulmonary and extra pulmonary tuberculosis, Tuberculosis 87, 279–287 (2007).

242. G. L. Beamer, D. K. Flaherty, B. D. Assogba, P. Stromberg, M. Gonzalez-Juarrero, R. de W. Malefyt, B. Vesosky, J. Turner, Interleukin-10 Promotes Mycobacterium tuberculosis Disease Progression in CBA/J Mice, J. Immunol. 181, 5545–5550 (2008).

243. M. G. Bonecini-Almeida, J. L. Ho, N. Boéchat, R. C. Huard, S. Chitale, H. Doo, J. Geng, L. Rego, L. C. O. Lazzarini, A. L. Kritski, W. D. Johnson, T. A. McCaffrey, J. R. L. e Silva, Down-Modulation of Lung Immune Responses by Interleukin-10 and Transforming Growth Factor β (TGF-β) and Analysis of TGF-β Receptors I and II in Active Tuberculosis, Infect. Immun. 72, 2628–2634 (2004).

244. V. P. Dwivedi, D. Bhattacharya, S. Chatterjee, D. V. R. Prasad, D. Chattopadhyay, L. V. Kaer, W. R. Bishai, G. Das, Mycobacterium tuberculosis Directs T Helper 2 Cell Differentiation by Inducing Interleukin-1β Production in Dendritic Cells, J. Biol. Chem. 287, 33656–33663 (2012).

245. J. P. Huynh, C.-C. Lin, J. M. Kimmey, N. N. Jarjour, E. A. Schwarzkopf, T. R. Bradstreet, I. Shchukina, O. Shpynov, C. T. Weaver, R. Taneja, M. N. Artyomov, B. T.

186

Edelson, C. L. Stallings, Bhlhe40 is an essential repressor of IL-10 during Mycobacterium tuberculosis infection, J. Exp. Med. 215, 1823–1838 (2018).

246. A. Pooran, M. Davids, A. Nel, A. Shoko, J. Blackburn, K. Dheda, IL-4 subverts mycobacterial containment in M. tuberculosis-infected human macrophages, Eur. Respir. J. , 1802242 (2019).

247. P. S. Redford, A. Boonstra, S. Read, J. Pitt, C. Graham, E. Stavropoulos, G. J. Bancroft, A. O’Garra, Enhanced protection to Mycobacterium tuberculosis infection in IL-10-deficient mice is accompanied by early and enhanced Th1 responses in the lung, Eur. J. Immunol. 40, 2200–2210 (2010).

248. S. C. Jameson, D. Masopust, Understanding Subset Diversity in T Cell Memory, Immunity 48, 214–226 (2018).

249. F. Sallusto, J. Geginat, A. Lanzavecchia, Central Memory and Effector Memory T Cell Subsets: Function, Generation, and Maintenance, Annu. Rev. Immunol. 22, 745–763 (2004).

250. L. M. McLane, M. S. Abdel-Hakeem, E. J. Wherry, CD8 T Cell Exhaustion During Chronic Viral Infection and Cancer, Annu. Rev. Immunol. 37, 457–495 (2019).

251. D. L. Barber, K. D. Mayer-Barber, C. G. Feng, A. H. Sharpe, A. Sher, CD4 T Cells Promote Rather than Control Tuberculosis in the Absence of PD-1–Mediated Inhibition, J. Immunol. 186, 1598–1607 (2011).

252. D. L. Barber, S. Sakai, R. R. Kudchadkar, S. P. Fling, T. A. Day, J. A. Vergara, D. Ashkin, J. H. Cheng, L. M. Lundgren, V. N. Raabe, C. S. Kraft, J. J. Nieva, M. A. Cheever, P. T. Nghiem, E. Sharon, Tuberculosis following PD-1 blockade for cancer immunotherapy, Sci. Transl. Med. 11, eaat2702 (2019).

253. E. Lázár-Molnár, B. Chen, K. A. Sweeney, E. J. Wang, W. Liu, J. Lin, S. A. Porcelli, S. C. Almo, S. G. Nathenson, W. R. Jacobs, Programmed death-1 (PD-1)– deficient mice are extraordinarily sensitive to tuberculosis, Proc. Natl. Acad. Sci. 107, 13402–13407 (2010).

254. C.-C. Tsai, J.-H. Chen, Y.-C. Wang, F.-Y. Chang, Re-activation of pulmonary tuberculosis during anti-programmed death-1 (PD-1) treatment, QJM Int. J. Med. 112, 41–42 (2019).

255. S. Koch, A. Larbi, E. Derhovanessian, D. Özcelik, E. Naumova, G. Pawelec, Multiparameter flow cytometric analysis of CD4 and CD8 T cell subsets in young and old people, Immun. Ageing 5, 6 (2008).

187

256. J. Yan, J. M. Greer, R. Hull, J. D. O’Sullivan, R. D. Henderson, S. J. Read, P. A. McCombe, The effect of ageing on human lymphocyte subsets: comparison of males and females, Immun. Ageing 7, 4 (2010).

257. K. L. Wong, W. H. Yeap, J. J. Y. Tai, S. M. Ong, T. M. Dang, S. C. Wong, The three human monocyte subsets: implications for health and disease, Immunol. Res. 53, 41–57 (2012).

258. A. A. Patel, Y. Zhang, J. N. Fullerton, L. Boelen, A. Rongvaux, A. A. Maini, V. Bigley, R. A. Flavell, D. W. Gilroy, B. Asquith, D. Macallan, S. Yona, The fate and lifespan of human monocyte subsets in steady state and systemic inflammation, J. Exp. Med. 214, 1913–1923 (2017).

259. J. Cros, N. Cagnard, K. Woollard, N. Patey, S.-Y. Zhang, B. Senechal, A. Puel, S. K. Biswas, D. Moshous, C. Picard, J.-P. Jais, D. D’Cruz, J.-L. Casanova, C. Trouillet, F. Geissmann, Human CD14dim Monocytes Patrol and Sense Nucleic Acids and Viruses via TLR7 and TLR8 Receptors, Immunity 33, 375–386 (2010).

260. M. Norelli, B. Camisa, G. Barbiera, L. Falcone, A. Purevdorj, M. Genua, F. Sanvito, M. Ponzoni, C. Doglioni, P. Cristofori, C. Traversari, C. Bordignon, F. Ciceri, R. Ostuni, C. Bonini, M. Casucci, A. Bondanza, Monocyte-derived IL-1 and IL-6 are differentially required for cytokine-release syndrome and neurotoxicity due to CAR T cells, Nat. Med. 24, 739–748 (2018).

261. A. Puchta, A. Naidoo, C. P. Verschoor, D. Loukov, N. Thevaranjan, T. S. Mandur, P. Nguyen, M. Jordana, M. Loeb, Z. Xing, L. Kobzik, M. J. Larché, D. M. E. Bowdish, TNF Drives Monocyte Dysfunction with Age and Results in Impaired Anti- pneumococcal Immunity, PLOS Pathog. 12, e1005368 (2016).

262. K. S. Rogacev, B. Cremers, A. M. Zawada, S. Seiler, N. Binder, P. Ege, G. Große- Dunker, I. Heisel, F. Hornof, J. Jeken, N. M. Rebling, C. Ulrich, B. Scheller, M. Böhm, D. Fliser, G. H. Heine, CD14++CD16+ Monocytes Independently Predict Cardiovascular Events: A Cohort Study of 951 Patients Referred for Elective Coronary Angiography, J. Am. Coll. Cardiol. 60, 1512–1520 (2012).

263. A. C. Shaw, S. Joshi, H. Greenwood, A. Panda, J. M. Lord, Aging of the innate immune system, Curr. Opin. Immunol. 22, 507–513 (2010).

264. V. Naranbhai, A. V. S. Hill, A. Karim, S. S, K. Naidoo, Q. Abdool Karim, G. M. Warimwe, H. McShane, H. Fletcher, Ratio of Monocytes to Lymphocytes in Peripheral Blood Identifies Adults at Risk of Incident Tuberculosis Among HIV-Infected Adults Initiating Antiretroviral Therapy, J. Infect. Dis. 209, 500–509 (2014).

265. V. Naranbhai, S. Kim, H. Fletcher, M. F. Cotton, A. Violari, C. Mitchell, S. Nachman, G. McSherry, H. McShane, A. V. Hill, S. A. Madhi, The association between 188 the ratio of monocytes:lymphocytes at age 3 months and risk of tuberculosis (TB) in the first two years of life, BMC Med. 12, 120 (2014).

266. A. C. Hearps, G. E. Martin, T. A. Angelovich, W.-J. Cheng, A. Maisa, A. L. Landay, A. Jaworowski, S. M. Crowe, Aging is associated with chronic innate immune activation and dysregulation of monocyte phenotype and function, Aging Cell 11, 867– 875 (2012).

267. J. Nyugen, S. Agrawal, S. Gollapudi, S. Gupta, Impaired Functions of Peripheral Blood Monocyte Subpopulations in Aged Humans, J. Clin. Immunol. 30, 806–813 (2010).

268. B. D. Pence, J. R. Yarbro, Aging impairs mitochondrial respiratory capacity in classical monocytes, Exp. Gerontol. 108, 112–117 (2018).

269. K. E. Powell, L. S. Farer, The Rising Age of the Tuberculosis Patient: A Sign of Success and Failure, J. Infect. Dis. 142, 946–948 (1980).

270. Yew Wing W., Yoshiyama Takashi, Leung Chi C., Chan Denise P., Epidemiological, clinical and mechanistic perspectives of tuberculosis in older people, Respirology (2018), doi:10.1111/resp.13303.

271. Y. Kobashi, K. Mouri, S. Yagi, Y. Obase, N. Miyashita, N. Okimoto, T. Matsushima, T. Kageoka, M. Oka, Clinical Utility of the QuantiFERON TB-2G Test for Elderly Patients With Active Tuberculosis, Chest 133, 1196–1202 (2008).

272. T. Mori, M. Sakatani, F. Yamagishi, T. Takashima, Y. Kawabe, K. Nagao, E. Shigeto, N. Harada, S. Mitarai, M. Okada, K. Suzuki, Y. Inoue, K. Tsuyuguchi, Y. Sasaki, G. H. Mazurek, I. Tsuyuguchi, Specific Detection of Tuberculosis Infection, Am. J. Respir. Crit. Care Med. 170, 59–64 (2004).

273. G. L. Beamer, D. K. Flaherty, B. Vesosky, J. Turner, Peripheral Blood Gamma Interferon Release Assays Predict Lung Responses and Mycobacterium tuberculosis Disease Outcome in Mice, Clin. Vaccine Immunol. 15, 474–483 (2008).

274. S. C. Castle, D. C. Norman, T. T. Perls, M. P. Chang, T. T. Yoshikawa, T. Makinodan, Analysis of cutaneous delayed-type hypersensitivity reaction and T cell proliferative response in elderly nursing home patients: an approach to identifying immunodeficient patients, Gerontology 36, 217–229 (1990).

275. E. Agius, K. E. Lacy, M. Vukmanovic-Stejic, A. L. Jagger, A.-P. Papageorgiou, S. Hall, J. R. Reed, S. J. Curnow, J. Fuentes-Duculan, C. D. Buckley, M. Salmon, L. S. Taams, J. Krueger, J. Greenwood, N. Klein, M. H. A. Rustin, A. N. Akbar, Decreased TNF-α synthesis by macrophages restricts cutaneous immunosurveillance by memory CD4+ T cells during aging, J. Exp. Med. 206, 1929–1940 (2009). 189

276. L. Balboa, M. M. Romero, J. I. Basile, C. A. S. y García, P. Schierloh, N. Yokobori, L. Geffner, R. M. Musella, J. Castagnino, E. Abbate, S. de la Barrera, M. C. Sasiain, M. Alemán, Paradoxical role of CD16+CCR2+CCR5+ monocytes in tuberculosis: efficient APC in pleural effusion but also mark disease severity in blood, J. Leukoc. Biol. 90, 69– 75 (2011).

277. C. Lastrucci, A. Bénard, L. Balboa, K. Pingris, S. Souriant, R. Poincloux, T. Al Saati, V. Rasolofo, P. González-Montaner, S. Inwentarz, E. J. Moraña, I. Kondova, F. A. Verreck, M. del C. Sasiain, O. Neyrolles, I. Maridonneau-Parini, G. Lugo-Villarino, C. Cougoule, Tuberculosis is associated with expansion of a motile, permissive and immunomodulatory CD16+ monocyte population via the IL-10/STAT3 axis, Cell Res. 25, 1333–1351 (2015).

278. J. M. Guerra-Laso, S. González-García, C. González-Cortés, C. Diez-Tascón, R. López-Medrano, O. M. Rivero-Lezcano, Macrophages from elders are more permissive to intracellular multiplication of Mycobacterium tuberculosis, AGE 35, 1235–1250 (2013).

279. D. C. Oshi, S. N. Oshi, I. Alobu, K. N. Ukwaja, Profile and treatment outcomes of tuberculosis in the elderly in southeastern Nigeria, 2011-2012, PloS One 9, e111910 (2014).

280. B. R. V. Velayutham, D. Nair, V. Chandrasekaran, B. Raman, G. Sekar, B. Watson, N. Charles, M. Malaisamy, A. Thomas, S. Swaminathan, Profile and response to antituberculosis treatment among elderly tuberculosis patients treated under the TB Control programme in South India, PloS One 9, e88045 (2014).

281. W. Bae, K. U. Park, E. Y. Song, S. J. Kim, Y. J. Lee, J. S. Park, Y.-J. Cho, H. I. Yoon, J.-J. Yim, C.-T. Lee, J. H. Lee, Comparison of the Sensitivity of QuantiFERON- TB Gold In-Tube and T-SPOT.TB According to Patient Age, PLOS ONE 11, e0156917 (2016).

282. F. Du, L. Xie, Y. Zhang, F. Gao, H. Zhang, W. Chen, B. Sun, W. Sha, Y. Fang, H. Jia, A. Xing, B. Du, L. Zheng, M. Gao, Z. Zhang, Prospective Comparison of QFT-GIT and T-SPOT.TB Assays for Diagnosis of Active Tuberculosis, Sci. Rep. 8, 5882 (2018).

283. World Health Organization, Global Tuberculosis Report 2018 (World Health Organization, Geneva, Switzerland; http://apps.who.int/medicinedocs/en/m/abstract/Js23553en/).

284. World Health Organization, Global strategy and targets for tuberculosis prevention, care and control after 2015 (World Health Organization, Geneva, Switzerland, 2014; http://www.who.int/tb/post2015_strategy/en/).

190

285. R. M. G. J. Houben, P. J. Dodd, The Global Burden of Latent Tuberculosis Infection: A Re-estimation Using Mathematical Modelling, PLOS Med. 13, e1002152 (2016).

286. S. Keshavjee, F. Amanullah, A. Cattamanchi, R. Chaisson, K. M. Dobos, G. J. Fox, H. E. Gendelman, R. Gordon, A. Hesseling, H. Le Van, B. Kampmann, B. Kana, G. Khuller, D. M. Lewinsohn, D. A. Lewinsohn, P. L. Lin, L. L. Lu, G. Maartens, A. Owen, M. Protopopova, J. Rengarajan, E. Rubin, P. Salgame, E. Schurr, J. A. Seddon, S. Swindells, D. M. Tobin, Z. Udwadia, G. Walzl, S. Srinivasan, R. Rustomjee, P. Nahid, Moving toward Tuberculosis Elimination. Critical Issues for Research in Diagnostics and Therapeutics for Tuberculosis Infection, Am. J. Respir. Crit. Care Med. 199, 564–571 (2018).

287. N. S. Shah, P. Kim, B. D. Kana, R. Rustomjee, Getting to Zero New Tuberculosis Infections: Insights From the National Institutes of Health/US Centers for Disease Control and Prevention/Bill & Melinda Gates Foundation Workshop on Research Needs for Halting Tuberculosis Transmission, J. Infect. Dis. 216, S627–S628 (2017).

288. G. Churchyard, P. Kim, N. S. Shah, R. Rustomjee, N. Gandhi, B. Mathema, D. Dowdy, A. Kasmar, V. Cardenas, What We Know About Tuberculosis Transmission: An Overview, J. Infect. Dis. 216, S629–S635 (2017).

289. D. W. Dowdy, A. D. Grant, K. Dheda, E. Nardell, K. Fielding, D. A. J. Moore, Designing and Evaluating Interventions to Halt the Transmission of Tuberculosis, J. Infect. Dis. 216, S654–S661 (2017).

290. J. P. Cegielski, D. E. Griffith, P. K. McGaha, M. Wolfgang, C. B. Robinson, P. A. Clark, W. L. Hassell, V. A. Robison, K. P. Walker, C. Wallace, Eliminating Tuberculosis One Neighborhood at a Time, Am. J. Public Health 103, 1292–1300 (2013).

291. H. P. Gideon, J. A. Skinner, N. Baldwin, J. L. Flynn, P. L. Lin, Early Whole Blood Transcriptional Signatures Are Associated with Severity of Lung Inflammation in Cynomolgus Macaques with Mycobacterium tuberculosis Infection, J. Immunol. 197, 4817–4828 (2016).

292. J. C. Cyktor, B. Carruthers, P. Stromberg, E. Flaño, H. Pircher, J. Turner, Killer Cell Lectin-Like Receptor G1 Deficiency Significantly Enhances Survival after Mycobacterium tuberculosis Infection, Infect. Immun. 81, 1090–1099 (2013).

293. W. A. Haynes, F. Vallania, C. Liu, E. Bongen, A. Tomczak, M. Andres-Terrè, S. Lofgren, A. Tam, C. A. Deisseroth, M. D. Li, T. E. Sweeney, P. Khatri, Empowering Multi-Cohort Gene Expression Analysis to Increase Reproducibility, Pac. Symp. Biocomput. Pac. Symp. Biocomput. 22, 144–153 (2016).

191

294. A. Lachmann, D. Torre, A. B. Keenan, K. M. Jagodnik, H. J. Lee, L. Wang, M. C. Silverstein, A. Ma’ayan, Massive mining of publicly available RNA-seq data from human and mouse, Nat. Commun. 9, 1366 (2018).

295. A. Liaw, M. Wiener, Classification and regression by random forest, R News 2, 18– 22 (2002).

296. M. Kuhn, J. Wing, S. Weston, A. Williams, C. Keefer, A. Engelhardt, T. Cooper, Z. Mayer, B. Kenkel, the R. C. Team, M. Benesty, R. Lescarbeau, A. Ziem, L. Scrucca, Y. Tang, C. Candan, and T. Hunt, caret: Classification and Regression Training (https://CRAN.R-project.org/package=caret, 2018; https://CRAN.R- project.org/package=caret).

297. M. N. Wright, A. Ziegler, ranger: A Fast Implementation of Random Forests for High Dimensional Data in C++ and R, J. Stat. Softw. 77 (2017), doi:10.18637/jss.v077.i01.

298. B. Greenwell, B. Boehmke, J. Cunningham, G. B. M. Developers (https://github.com/gbm-developers), gbm: Generalized Boosted Regression Models (https://CRAN.R-project.org/package=gbm, 2018; https://CRAN.R- project.org/package=gbm).

299. A. Karatzoglou, A. Smola, K. Hornik, A. Zeileis, kernlab - An S4 Package for Kernel Methods in R, J. Stat. Softw. 11, 1–20 (2004).

300. J. Friedman, T. Hastie, R. Tibshirani, Regularization Paths for Generalized Linear Models via Coordinate Descent, J. Stat. Softw. 33, 1–22 (2010).

301. T. E. Sweeney, L. Braviak, C. M. Tato, P. Khatri, Genome-wide expression for diagnosis of pulmonary tuberculosis: a multicohort analysis, Lancet Respir. Med. 4, 213– 224 (2016).

302. S. Durinck, P. T. Spellman, E. Birney, W. Huber, Mapping identifiers for the integration of genomic datasets with the R/Bioconductor package biomaRt, Nat. Protoc. 4, 1184–1191 (2009).

303. F. Vallania, A. Tam, S. Lofgren, S. Schaffert, T. D. Azad, E. Bongen, W. Haynes, M. Alsup, M. Alonso, M. Davis, E. Engleman, P. Khatri, Leveraging heterogeneity across multiple datasets increases cell-mixture deconvolution accuracy and reduces biological and technical biases, Nat. Commun. 9, 4735 (2018).

304. S. Li, N. Rouphael, S. Duraisingham, S. Romero-Steiner, S. Presnell, C. Davis, D. S. Schmidt, S. E. Johnson, A. Milton, G. Rajam, S. Kasturi, G. M. Carlone, C. Quinn, D. Chaussabel, A. K. Palucka, M. J. Mulligan, R. Ahmed, D. S. Stephens, H. I. Nakaya, B.

192

Pulendran, Molecular signatures of antibody responses derived from a systems biology study of five human vaccines, Nat. Immunol. 15, 195–204 (2014).

305. T. Domaszewska, J. Weiner, disco: Discordance and Concordance of Transcriptomic Responses (2018; https://CRAN.R-project.org/package=disco).

306. J. Weiner 3rd, T. Domaszewska, tmod: an R package for general and multivariate enrichment analysis (PeerJ Preprints, 2016; https://peerj.com/preprints/2420).

307. T. Domaszewska, L. Scheuermann, K. Hahnke, H. Mollenkopf, A. Dorhoi, S. H. E. Kaufmann, J. Weiner, Concordant and discordant gene expression patterns in mouse strains identify best-fit animal model for human tuberculosis, Sci. Rep. 7, 1–13 (2017).

308. NCAR, verification: Weather Forecast Verification Utilities (https://CRAN.R- project.org/package=verification, 2015; https://CRAN.R- project.org/package=verification).

309. X. Robin, N. Turck, A. Hainard, N. Tiberti, F. Lisacek, J.-C. Sanchez, M. Müller, pROC: an open-source package for R and S+ to analyze and compare ROC curves, BMC Bioinformatics 12, 77–84 (2011).

310. D. Bates, M. Mächler, B. Bolker, S. Walker, Fitting Linear Mixed-Effects Models Using lme4, J. Stat. Softw. 67, 1–48 (2015).

311. M. Gonzalez-Juarrero, L. C. Kingry, D. J. Ordway, M. Henao-Tamayo, M. Harton, R. J. Basaraba, W. H. Hanneman, I. M. Orme, R. A. Slayden, Immune Response to Mycobacterium tuberculosis and Identification of Molecular Markers of Disease, Am. J. Respir. Cell Mol. Biol. 40, 398–409 (2009).

312. H.-J. Mollenkopf, K. Hahnke, S. H. E. Kaufmann, Transcriptional responses in mouse lungs induced by vaccination with Mycobacterium bovis BCG and infection with Mycobacterium tuberculosis, Microbes Infect. 8, 136–144 (2006).

313. L. Shi, H. Salamon, E. A. Eugenin, R. Pine, A. Cooper, M. L. Gennaro, Infection with Mycobacterium tuberculosis induces the Warburg effect in mouse lungs, Sci. Rep. 5, srep18176 (2015).

314. G. L. Beamer, J. Turner, Murine models of susceptibility to tuberculosis, Arch. Immunol. Ther. Exp. (Warsz.) 53, 469–483 (2005).

315. Medina, North, Resistance ranking of some common inbred mouse strains to Mycobacterium tuberculosis and relationship to major histocompatibility complex haplotype and Nramp1 genotype, Immunology 93, 270–274 (1998).

193

316. K. Dijkman, C. C. Sombroek, R. A. W. Vervenne, S. O. Hofman, C. Boot, E. J. Remarque, C. H. M. Kocken, T. H. M. Ottenhoff, I. Kondova, M. A. Khayum, K. G. Haanstra, M. P. M. Vierboom, F. A. W. Verreck, Prevention of tuberculosis infection and disease by local BCG in repeatedly exposed rhesus macaques, Nat. Med. 25, 255 (2019).

317. J. R. Andrews, F. Noubary, R. P. Walensky, R. Cerda, E. Losina, C. R. Horsburgh, Risk of Progression to Active Tuberculosis Following Reinfection With Mycobacterium tuberculosis, Clin. Infect. Dis. 54, 784–791 (2012).

318. R. Chen, L. Xia, K. Tu, M. Duan, K. Kukurba, J. Li-Pook-Than, D. Xie, M. Snyder, Longitudinal personal DNA methylome dynamics in a human with a chronic condition, Nat. Med. 24, 1930–1939 (2018).

319. E. J. Hwang, S. Park, K.-N. Jin, J. I. Kim, S. Y. Choi, J. H. Lee, J. M. Goo, J. Aum, J.-J. Yim, C. M. Park, D. H. Kim, W. Woo, C. Choi, I. P. Hwang, Y. S. Song, L. Lim, K. Kim, J. Y. Wi, S. S. Oh, M.-J. Kang, Development and Validation of a Deep Learning– based Automatic Detection Algorithm for Active Pulmonary Tuberculosis on Chest Radiographs, Clin. Infect. Dis. 69, 739–747 (2019).

320. M. T. Coleman, P. Maiello, J. Tomko, L. J. Frye, D. Fillmore, C. Janssen, E. Klein, P. L. Lin, Early Changes by 18Fluorodeoxyglucose Positron Emission Tomography Coregistered with Computed Tomography Predict Outcome after Mycobacterium tuberculosis Infection in Cynomolgus Macaques, Infect. Immun. 82, 2400–2404 (2014).

321. P. L. Lin, J. L. Flynn, The End of the Binary Era: Revisiting the Spectrum of Tuberculosis, J. Immunol. 201, 2541–2548 (2018).

322. A. Singhania, R. J. Wilkinson, M. Rodrigue, P. Haldar, A. O’Garra, The value of transcriptomics in advancing knowledge of the immune response and diagnosis in tuberculosis, Nat. Immunol. 19, 1159–1168 (2018).

323. M. A. Behr, P. H. Edelstein, L. Ramakrishnan, Is Mycobacterium tuberculosis infection life long?, BMJ 367 (2019), doi:10.1136/bmj.l5770.

324. W. W. Stead, J. P. Lofgren, E. Warren, C. Thomas, Tuberculosis as an Endemic and Nosocomial Infection among the Elderly in Nursing Homes, N. Engl. J. Med. 312, 1483– 1487 (1985).

325. M. K. K. Niazi, N. Dhulekar, D. Schmidt, S. Major, R. Cooper, C. Abeijon, D. M. Gatti, I. Kramnik, B. Yener, M. Gurcan, G. Beamer, Lung necrosis and neutrophils reflect common pathways of susceptibility to Mycobacterium tuberculosis in genetically diverse, immune-competent mice, Dis. Model. Mech. 8, 1141–1153 (2015).

326. C. M. Smith, M. K. Proulx, A. J. Olive, D. Laddy, B. B. Mishra, C. Moss, N. M. Gutierrez, M. M. Bellerose, P. Barreira-Silva, J. Y. Phuah, R. E. Baker, S. M. Behar, H. 194

Kornfeld, T. G. Evans, G. Beamer, C. M. Sassetti, Tuberculosis Susceptibility and Vaccine Protection Are Independently Controlled by Host Genotype, mBio 7 (2016), doi:10.1128/mBio.01516-16.

327. C. M. Smith, M. K. Proulx, R. Lai, M. C. Kiritsy, T. A. Bell, P. Hock, F. Pardo- Manuel de Villena, M. T. Ferris, R. E. Baker, S. M. Behar, C. M. Sassetti, Functionally Overlapping Variants Control Tuberculosis Susceptibility in Collaborative Cross Mice, mBio 10 (2019), doi:10.1128/mBio.02791-19.

328. A. Chakraborty, M. E. Diefenbacher, A. Mylona, O. Kassel, A. Behrens, The E3 ubiquitin ligase Trim7 mediates c-Jun/AP-1 activation by Ras signalling, Nat. Commun. 6, 1–12 (2015).

329. V. Guillet, M. Knibiehler, L. Gregory-Pauron, M.-H. Remy, C. Chemin, B. Raynaud-Messina, C. Bon, J. M. Kollman, D. A. Agard, A. Merdes, L. Mourey, Crystal structure of γ-tubulin complex protein GCP4 provides insight into microtubule nucleation, Nat. Struct. Mol. Biol. 18, 915–919 (2011).

330. Y. Niimi, M. Takahashi, E. Sugawara, S. Umeda, M. Obayashi, N. Sato, T. Ishiguro, M. Higashi, Y. Eishi, H. Mizusawa, K. Ishikawa, Abnormal RNA structures (RNA foci) containing a penta-nucleotide repeat (UGGAA)n in the Purkinje cell nucleus is associated with spinocerebellar ataxia type 31 pathogenesis, Neuropathol. Off. J. Jpn. Soc. Neuropathol. 33, 600–611 (2013).

331. N. P. Reed, M. A. Henderson, E. M. Oltz, T. M. Aune, Reciprocal regulation of Rag expression in thymocytes by the zinc-finger proteins, Zfp608 and Zfp609, Genes Immun. 14, 7–12 (2013).

332. S. Scheidecker, C. Etard, L. Haren, C. Stoetzel, S. Hull, G. Arno, V. Plagnol, S. Drunat, S. Passemard, A. Toutain, C. Obringer, M. Koob, V. Geoffroy, V. Marion, U. Strähle, P. Ostergaard, A. Verloes, A. Merdes, A. T. Moore, H. Dollfus, Mutations in TUBGCP4 Alter Microtubule Organization via the γ-Tubulin Ring Complex in Autosomal-Recessive Microcephaly with Chorioretinopathy, Am. J. Hum. Genet. 96, 666–674 (2015).

333. T. Z. Tan, Q. H. Miow, R. Y.-J. Huang, M. K. Wong, J. Ye, J. A. Lau, M. C. Wu, L. H. Bin Abdul Hadi, R. Soong, M. Choolani, B. Davidson, J. M. Nesland, L.-Z. Wang, N. Matsumura, M. Mandai, I. Konishi, B.-C. Goh, J. T. Chang, J. P. Thiery, S. Mori, Functional genomics identifies five distinct molecular subtypes with clinical relevance and pathways for growth control in epithelial ovarian cancer, EMBO Mol. Med. 5, 1051– 1066 (2013).

195

334. X. Xu, D. Shang, H. Cheng, D. J. Klionsky, R. Zhou, Gene essentiality of Tubgcp4: dosage effect and autophagy regulation in retinal photoreceptors, Autophagy 15, 1834– 1837 (2019).

335. F. Zhang, L. R. Thomas, E. M. Oltz, T. M. Aune, Control of thymocyte development and recombination-activating gene expression by the zinc finger protein Zfp608, Nat. Immunol. 7, 1309–1316 (2006).

336. D. Schübeler, Function and information content of DNA methylation, Nature 517, 321–326 (2015).

337. T. P. Wu, T. Wang, M. G. Seetin, Y. Lai, S. Zhu, K. Lin, Y. Liu, S. D. Byrum, S. G. Mackintosh, M. Zhong, A. Tackett, G. Wang, L. S. Hon, G. Fang, J. A. Swenberg, A. Z. Xiao, DNA methylation on N6-adenine in mammalian embryonic stem cells, Nature 532, 329–333 (2016).

338. R. Holliday, J. E. Pugh, DNA modification mechanisms and gene activity during development, Science 187, 226–232 (1975).

339. A. D. Riggs, X inactivation, differentiation, and DNA methylation, Cytogenet. Cell Genet. 14, 9–25 (1975).

340. Z. D. Smith, A. Meissner, DNA methylation: roles in mammalian development, Nat. Rev. Genet. 14, 204–220 (2013).

341. E. A. Houseman, W. P. Accomando, D. C. Koestler, B. C. Christensen, C. J. Marsit, H. H. Nelson, J. K. Wiencke, K. T. Kelsey, DNA methylation arrays as surrogate measures of cell mixture distribution, BMC Bioinformatics 13, 86 (2012).

342. E. A. Houseman, M. L. Kile, D. C. Christiani, T. A. Ince, K. T. Kelsey, C. J. Marsit, Reference-free deconvolution of DNA methylation data and mediation by cell composition effects, BMC Bioinformatics 17, 259 (2016).

343. R. S. Akondy, M. Fitch, S. Edupuganti, S. Yang, H. T. Kissick, K. W. Li, B. A. Youngblood, H. A. Abdelsamed, D. J. McGuire, K. W. Cohen, G. Alexe, S. Nagar, M. M. McCausland, S. Gupta, P. Tata, W. N. Haining, M. J. McElrath, D. Zhang, B. Hu, W. J. Greenleaf, J. J. Goronzy, M. J. Mulligan, M. Hellerstein, R. Ahmed, Origin and differentiation of human memory CD8 T cells after vaccination, Nature 552, 362–367 (2017).

344. M. M. Esterhuyse, J. Weiner, E. Caron, A. G. Loxton, M. Iannaccone, C. Wagman, P. Saikali, K. Stanley, W. E. Wolski, H.-J. Mollenkopf, M. Schick, R. Aebersold, H. Linhart, G. Walzl, S. H. E. Kaufmann, Epigenetics and Proteomics Join Transcriptomics in the Quest for Tuberculosis Biomarkers, mBio 6, e01187-15 (2015).

196

345. D. Verma, V. R. Parasa, J. Raffetseder, M. Martis, R. B. Mehta, M. Netea, M. Lerm, Anti-mycobacterial activity correlates with altered DNA methylation pattern in immune cells from BCG-vaccinated subjects, Sci. Rep. 7, 12305 (2017).

346. A. Pacis, L. Tailleux, A. M. Morin, J. Lambourne, J. L. MacIsaac, V. Yotova, A. Dumaine, A. Danckaert, F. Luca, J.-C. Grenier, K. D. Hansen, B. Gicquel, M. Yu, A. Pai, C. He, J. Tung, T. Pastinen, M. S. Kobor, R. Pique-Regi, Y. Gilad, L. B. Barreiro, Bacterial infection remodels the DNA methylation landscape of human dendritic cells, Genome Res. 25, 1801–1811 (2015).

347. A. Pacis, F. Mailhot-Léonard, L. Tailleux, H. E. Randolph, V. Yotova, A. Dumaine, J.-C. Grenier, L. B. Barreiro, Gene activation precedes DNA demethylation in response to infection in human dendritic cells, Proc. Natl. Acad. Sci. 116, 6938–6943 (2019).

348. G. Sharma, S. Upadhyay, M. Srilalitha, V. K. Nandicoori, S. Khosla, The interaction of mycobacterial protein Rv2966c with host chromatin is mediated through non-CpG methylation and histone H3/H4 binding, Nucleic Acids Res. 43, 3922–3937 (2015).

349. G. Sharma, D. T. Sowpati, P. Singh, M. Z. Khan, R. Ganji, S. Upadhyay, S. Banerjee, V. K. Nandicoori, S. Khosla, Genome-wide non-CpG methylation of the host genome during M. tuberculosis infection, Sci. Rep. 6, srep25006 (2016).

350. R. J. W. Arts, S. J. C. F. M. Moorlag, B. Novakovic, Y. Li, S.-Y. Wang, M. Oosting, V. Kumar, R. J. Xavier, C. Wijmenga, L. A. B. Joosten, C. B. E. M. Reusken, C. S. Benn, P. Aaby, M. P. Koopmans, H. G. Stunnenberg, R. van Crevel, M. G. Netea, BCG Vaccination Protects against Experimental Viral Infection in Humans through the Induction of Cytokines Associated with Trained Immunity, Cell Host Microbe 23, 89- 100.e5 (2018).

351. E. Kaufmann, J. Sanz, J. L. Dunn, N. Khan, L. E. Mendonça, A. Pacis, F. Tzelepis, E. Pernet, A. Dumaine, J.-C. Grenier, F. Mailhot-Léonard, E. Ahmed, J. Belle, R. Besla, B. Mazer, I. L. King, A. Nijnik, C. S. Robbins, L. B. Barreiro, M. Divangahi, BCG Educates Hematopoietic Stem Cells to Generate Protective Innate Immunity against Tuberculosis, Cell 172, 176-190.e19 (2018).

352. S. Horvath, DNA methylation age of human tissues and cell types, Genome Biol. 14, 3156 (2013).

353. D. A. Petkovich, D. I. Podolskiy, A. V. Lobanov, S.-G. Lee, R. A. Miller, V. N. Gladyshev, Using DNA Methylation Profiling to Evaluate Biological Age and Longevity Interventions, Cell Metab. 25, 954-960.e6 (2017).

354. I. J. Amanna, N. E. Carlson, M. K. Slifka, Duration of humoral immunity to common viral and vaccine antigens, N. Engl. J. Med. 357, 1903–1915 (2007).

197

355. J. G. Petrie, S. E. Ohmit, E. Johnson, R. Truscon, A. S. Monto, Persistence of Antibodies to Influenza Hemagglutinin and Neuraminidase Following One or Two Years of Influenza Vaccination, J. Infect. Dis. 212, 1914–1922 (2015).

356. F. Abebe, M. Belay, M. Legesse, F. K. L. M. C., T. H. M. Ottenhoff, IgA and IgG against Mycobacterium tuberculosis Rv2031 discriminate between pulmonary tuberculosis patients, Mycobacterium tuberculosis-infected and non-infected individuals, PLoS ONE 13 (2018), doi:10.1371/journal.pone.0190989.

357. E. Julián, L. Matas, A. Pérez, J. Alcaide, M.-A. Lanéelle, M. Luquin, Serodiagnosis of Tuberculosis: Comparison of Immunoglobulin A (IgA) Response to Sulfolipid I with IgG and IgM Responses to 2,3-Diacyltrehalose, 2,3,6-Triacyltrehalose, and Cord Factor Antigens, J. Clin. Microbiol. 40, 3782–3788 (2002).

358. R. A. M. Dirks, H. G. Stunnenberg, H. Marks, Genome-wide epigenomic profiling for biomarker discovery, Clin. Epigenetics 8, 122 (2016).

198