REVIEWS

Heterogeneity in tuberculosis
Anthony M. Cadena1, Sarah M. Fortune2 and JoAnne L. Flynn1
Abstract | Infection with Mycobacterium tuberculosis, the causative agent of tuberculosis (TB),
results in a range of clinical presentations in humans. Most infections manifest as a clinically
asymptomatic, contained state that is termed latent TB infection (LTBI); a smaller subset of
infected individuals present with symptomatic, active TB. Within these two seemingly binary
states, there is a spectrum of host outcomes that have varying symptoms, microbiologies, immune
responses and pathologies. Recently, it has become apparent that there is diversity of infection
even within a single individual. A good understanding of the heterogeneity that is intrinsic to TB
— at both the population level and the individual level — is crucial to inform the development of
intervention strategies that account for and target the unique, complex and independent nature
of the local host–pathogen interactions that occur in this infection. In this Review, we draw on
model systems and human data to discuss multiple facets of TB biology and their relationship to
the overall heterogeneity observed in the human disease.

Tuberculin skin test

Classically, Mycobacterium tuberculosis infection in two states6. The biological and immunological under‑
A test that involves the humans is thought to result in one of two clinically pinnings of this variability are not well understood, and
induction of a delayed-type defined states: latent infection (termed latent tubercu‑ the clinical ramifications of this variability in human TB
hypersensitivity reaction by an losis (TB) infection (LTBI)) or active disease. LTBI is remain unknown.
intradermal injection of
characterized by the presence of immunological sen‑ This Review provides a framework for understanding
purified protein derivative,
which is a mixture of sitivity to mycobacterial antigen (as determined by a the heterogeneity observed in M. tuberculosis infection
Mycobacterium tuberculosis-­ tuberculin skin test or an interferon‑γ (IFNγ) release by separating host, granuloma and bacterial features that
derived proteins. The assay) in the absence of the clinical symptoms of dis‑ contribute to the overall spectrum of TB. In examining
tuberculin skin test is also ease, which can be extremely varied but most often each of these factors as they relate to the variability of
known as the Mantoux test
and is used as a diagnostic tool
include cough, fever or weight loss. LTBI accounts for TB, we highlight the intricate network of interactions
for M. tuberculosis infection, 90% of human infections and has an estimated burden that ultimately influences infection outcome and empha‑
but it does not distinguish of more than 2 billion individuals worldwide1,2. By con‑ size the need to consider all three domains for effective
latent infection from active trast, active TB is diagnosed in patients who have clinical intervention.
tuberculosis.
signs and symptoms of TB, and show microbiological
evidence of M. tuberculosis infection. Pulmonary TB is Host heterogeneity
typified by a chronic cough, fever, sustained weight loss, TB: a spectrum of clinical and pathological outcomes.
wasting and haemoptysis (that is, coughing up blood or It is widely accepted that the severity of active TB can
1
Department of Microbiology
and Molecular Genetics, blood-stained mucus)3. Microbiological confirmation — be highly varied and can show different patterns of lung
450 Technology Drive, either by culturing M. tuberculosis from sputum or other involvement 7–9. It is now appreciated that what is consid‑
University of Pittsburgh relevant samples, or by identifying the organism through ered to be LTBI can also encompass a range of infection
School of Medicine, nucleic acid testing or acid-fast staining — is required for outcomes6,10,11. The large reservoir of individuals with
Pittsburgh, Pennsylvania
15261, USA.
the unequivocal diagnosis of TB. asymptomatic LTBI represents a nuanced continuum
2
Department of Immunology In 2015, there were 10.4 million new cases of TB of bacterial persistence and host containment, ranging
and Infectious Diseases, infection and 1.8 million deaths due to TB4; the active from cleared infection to low-grade TB6,10. This concept
Harvard T. H. Chan School of disease state is estimated to occur in approximately of a spectrum of infection for LTBI extends the defi‑
Public Health,
5% of initial infections in the first 18 months with a nition of subclinical, latent TB beyond a single status
655 Huntington Avenue,
Boston, Massachusetts remaining risk of 5% (due to the reactivation of LTBI) and enables us to better differentiate the risk of LTBI
02115, USA. throughout the lifetime of the individual5. Although re­­activation in an individual, prioritize preventive treat‑
Correspondence to J.L.F. these definitions of active TB and LTBI have been the ment and emphasize the heterogeneity of host responses
[email protected] basis of decades of clinical practice, the TB field has to M. tuberculosis infection. Individuals who have a ster‑
doi:10.1038/nri.2017.69 more recently embraced a newer paradigm that recog­ ilized or extremely well-contained infection are the least
Published online 24 Jul 2017 nizes a spectrum of infection outcomes within these likely to suffer reactivation of infection and presumably

NATURE REVIEWS | IMMUNOLOGY VOLUME 17 | NOVEMBER 2017 | 691

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

are the last to require intervention. By contrast, individ‑ degree of pathology and bacterial burden. However,
uals with LTBI who are harbouring a low-grade, sub­ there were also differences in both the lungs and the
clinical infection are at a higher risk of reactivation and blood of monkeys in each of these two groups. Some
are more likely to require treatment. monkeys with clinically active disease had bi­­lateral lobe
Evidence supporting this concept has been observed involvement, pulmonary cavitation, tuberculous pneu‑
in a human study using [ 18F]fluorodeoxyglucose monia and extrapulmonary disease, whereas other
positron emission tomography and computed tomography monkeys with active disease had a pathology that was
(FDG PET–CT)12. The authors of this study found that confined to the thoracic lymph nodes and a single lung
among 35 antiretroviral therapy-naive, HIV‑1‑positive lobe14. There was similar heterogeneity observed in the
adults with LTBI, 10 patients with pulmonary irregu‑ macaques that maintained LTBI; four animals appeared
larities indicative of subclinical TB disease had a signif‑ to only have lymph node involvement, whereas three
Acid-fast staining icantly higher risk of developing active TB than did the animals had evidence of a Ghon complex. Of particular
A method for staining remaining 25 patients who had no subclinical pathol‑ interest were five animals that were classified as having
mycobacteria for microscopic
visualization, as the Gram stain
ogy as detected by FDG PET–CT. Specifically, 4 of the an intermediate disease state (between active disease
is not useful for mycobacteria. 10 patients with evidence of subclinical disease devel‑ and LTBI) and thus were analogous to the ten human
Acid-fast staining relies on oped active disease, whereas none of the 25 patients patients described above12. Four of these animals had
phenolic compounds that in the second cohort developed active disease. The 10 disease that was limited to the lungs and lymph nodes,
interact with the lipid-rich cell
patients who were at a higher risk of disease had radio‑ but intermittently showed M. tuberculosis growth in cul‑
walls of mycobacteria, and the
resistance of this interaction to logical evidence of active nodules, infiltrates or fibrotic tures of bronchoalveolar lavage fluid samples or gastric
acid alcohol is the basis of the scars, whereas the 25 participants who did not develop aspirate samples (a surrogate for human sputum sam‑
term ‘acid-fast’. disease had either normal lung parenchyma (n = 10) or pling), thereby precluding their inclusion in the LTBI
only discrete nodules (n = 15). Importantly, although the cohort. These monkeys were deemed to have subclin‑
Positron emission
tomography
group with pulmonary irregularities had an elevated risk ical disease and were termed ‘percolators’ on the basis
An imaging method that relative to the group without subclinical pathology, there of occasional M. tuberculosis-positive samples14. Overall,
depends on the 3D detection were six individuals in the higher-risk group who did these studies in humans and macaques support the idea
of radiation (positrons) from a not develop active disease during the period of study, of the biological heterogeneity of TB and suggest that
probe that is typically localized
further emphasizing the variability in disease progres‑ there may be clinical benefit in appreciating the vari‑
by uptake and retention in a
specific cell or by a specific sion and host outcome. This study challenges the binary ability of host outcomes in TB by segregating patients
process in vivo. This uptake classification of active disease and LTBI, and instead according to disease risk.
provides functional information emphasizes the importance of heterogeneity within LTBI
about the organ of interest. in which a subset of individuals harbour a low-grade, Peripheral transcriptional signatures reveal dynamic
Computed tomography
subclinical and asymptomatic disease state that places and variable disease states. Clinical heterogeneity has
An imaging method that uses them at increased risk of developing active TB. also been observed in the context of recent studies that
computer-processed These radiological findings in humans support our have examined whole-blood transcriptional signatures
combinations of many X‑ray earlier observations of a spectrum of disease in the related to host disease status15–17. In a landmark study
images taken from different
macaque model of TB13,14 (BOX 1). A detailed study of published in 2010, Berry et al.15 reported a 393‑transcript
angles to produce
cross-sectional (tomographic) experimentally infected cynomolgus macaques that signature that was unique to patients with active TB
images (virtual ‘slices’) of quantitatively assessed parameters of infection among (from both intermediate-burden and high-burden areas)
specific areas of a scanned animals with clinically defined active infection or LTBI relative to subjects with LTBI and healthy controls. This
object, resulting in a 3D found marked heterogeneity in disease pathology, signature reflected the upregulated transcription of
representation of an organ in a
living subject. This provides
extrapulmonary dissemination and tissue involvement14. IFN-inducible genes (that is, genes that are induced by
structural information about There were quantifiable differences between macaques both type I and type II IFNs) in blood neutrophils from
the organ of interest. with active TB and those with LTBI, including the patients with active TB that correlated with the extent
Pulmonary cavitation
The formation of a cavity in the
lung. A cavity is an abnormal, Box 1 | Modelling heterogeneity in animal models of Mycobacterium tuberculosis infection
gas-filled space with a lining
wall that has developed within Animal models of Mycobacterium tuberculosis infection are crucial to dissect the features of infection pathogenesis,
and replaced the normal lung tuberculosis (TB) pathology and immunology, and bacterial virulence. Several animal models have been used
architecture. In tuberculous throughout the past few decades to study TB, including zebrafish120, mice121 and non-human primates122–124. Each system
disease, these cavities are has both benefits and limitations as a model of TB, and although none perfectly recapitulates human M. tuberculosis
formed when necrosis invades infection, they all contribute to our understanding of this disease. The model recognized to perhaps best recapitulate
through the wall of an airway,
the multiple facets of variability in TB is the cynomolgus macaque. Although ethical considerations, limited reagents
dilating and distorting the
structure, and leading to the
and cost may detract from the general use of cynomolgus macaques, the range of outcomes both clinically and
discharge of necrotic debris pathologically is remarkably similar to those observed in humans13,14. There is a 50:50 ratio of latent infection and active
into the bronchial tree. disease in adult cynomolgus macaques infected with low-dose (<25 colony-forming units) virulent M. tuberculosis
Erdman when using criteria identical to those used for human diagnosis, including immunological sensitivity, culture
Ghon complex positivity of gastric aspirates and/or bronchoalveolar lavage fluid samples, signs of disease and increased erythrocyte
A term for pathological lesions sedimentation rates13,14. Importantly, within these binary definitions, infected macaques recapitulate the entire
in latent tuberculosis infection spectrum of human disease at both the local (lung) level and the overall host level, including the ability to reactivate
that consist of an latent infection following immune suppression with simian immunodeficiency virus125, or treatment with antibodies
often-calcified granuloma and
against tumour necrosis factor74 or CD4 (REF. 126).
an associated lymph node.

692 | NOVEMBER 2017 | VOLUME 17 www.nature.com/nri

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

of lung disease as assessed by radiography 15,18. Notably, late (13 weeks) time points in infection26. In the second
whereas the majority of individuals with LTBI clustered study 27, an analysis of peripheral blood from cyno­
independently of the patients with active TB, 10–25% of molgus macaques revealed that the greatest transcrip‑
the subjects with LTBI had similar transcriptional pro‑ tional change occurred 3–8 weeks after infection27.
files to patients with active TB, and it was considered There was a positive correlation between the transcrip‑
likely that these patients had subclinical, active disease15. tional signature and inflammation as assessed by FDG
These shared transcriptional profiles observed in a sub‑ PET–CT. One of the most important findings was that
set of patients with clinically defined LTBI reiterate the even before infection, stronger type I IFN signatures
heterogeneous radiological findings described above and were observed in the macaques that went on to develop
further emphasize the varied nature of this disease. active TB, which suggests that there is an innate response
Further studies have examined blood gene-­expression linked to host susceptibility to progressive disease. These
profiles in patients with pulmonary TB at diagnosis and pre-existing differences in immune status may be due
throughout TB treatment 19,20. In the first study, Bloom to host genetics; several genetic polymorphisms have
et al.19 found a 320‑transcript signature in patients after been linked to TB risk (BOX 2). Collectively, these stud‑
2 weeks of treatment that subsequently diminished ies suggest a model in which the dynamic regulation of
by 6 months. Cliff et al.20 reported similar changes in inflammation shapes the trajectory of infection to create
gene networks following antibiotic treatment. Within a continuum of outcomes.
1 week of TB treatment, the expression of more than
1,200 genes was markedly downregulated, including Granuloma heterogeneity
the expression of multiple inflammatory markers such TB granulomas have varied structures and cellular
as the complement proteins C1q and C2. This initial compositions that influence host outcome. The patho‑
period of rapid transcriptional change was followed by logical hallmark of human TB is the granuloma, which
a slower upregulation of genes involved in the humoral is an organized and localized aggregate of immune
response20. Importantly, as was observed in the study by cells that consists of macrophages, lymphocytes and
Berry et al.15, several of the treated patients from both other host immune cells, and forms in response to
studies had notable differences in their gene-expression persistent stimuli28. These structures arise at the sites
profiles relative to the majority of treated individuals. of M. tuberculosis infection, and are the locations in
This host-specific variability once more highlights the which there is active interchange between the host and
heterogeneity that is intrinsic to TB and suggests an the pathogen. The formation of granulomas is crucial
oppor­tunity to stratify patients to different treatment for controlling and containing infection29,30, but gran‑
strategies, particularly as the standard treatment reg‑ ulomas may also contribute to early M. tuberculosis
imen can be problematic owing to its length and its proliferation and dissemination31–33. Studies of human
potential for toxicity and drug interactions21,22. autopsy specimens from more than 50 years ago have
A whole-blood gene expression analysis of patients revealed that in active disease and LTBI, granulomas
with pulmonary and extrapulmonary TB found that
transcriptional profiles are influenced by symptom sta‑
tus and the site of disease23. Individuals with the high‑ Box 2 | Host genetics and tuberculosis
est mean molecular distance to health24 had the highest
likelihood of presenting with one or more of the follow‑ Host genotype is increasingly being associated with
ing symptoms: fever, night sweats, chest pain, cough or variation in host susceptibility and outcome in
Mycobacterium tuberculosis infection. Recent studies
weight loss. These findings suggest that clinical illness is
have pinpointed several genetic factors that correlate
linked to the site of infection, bacterial burden and host with protection, susceptibility to disease and vaccine
response, which are reflected in the diversity of tran‑ responses54,127–133. Toll-interacting protein (TOLLIP) is an
scriptional profiles and host statuses. A recent study of example of a protein for which variation in its
adolescents with LTBI demonstrated that a whole-blood transcriptional expression was linked with susceptibility
transcriptional signature could identify those who were to tuberculosis (TB)127 and the efficacy of bacillus
at risk of developing active TB up to 12 months before Calmette–Guérin (BCG) vaccination128. In the first
clinical diagnosis; the signature was validated in a sep‑ study127, TOLLIP deficiency was associated with an
arate adult population25. These data further support the increased risk for TB that was mediated by a loss of
concept of a differential risk of reactivation in a popu‑ negative regulation of Toll-like receptor 2 (TLR2) and
TLR4 signalling. The subsequent related study128 found
lation25; this differential risk is possibly due to a spec‑
that a deficiency of this protein was correlated with
trum of disease states within asymptomatic infection. decreased interleukin‑2 production by BCG-specific
This study provides an opportunity to target drug treat‑ CD4+ T cells. Similar studies have determined an
ment to those in whom LTBI reactivation is most likely important link between host genetic factors, immune
to occur. responses and TB outcome. Using an inbred recombinant
In non-human primates, similar studies have inves‑ mouse panel known as the ‘Collaborative Cross’
Molecular distance to health tigated gene-expression changes specific to lung granu­ (REFS 134,135), Smith et al.130 found that host
A numerical score that lomas26 and longitudinal changes in gene-expression susceptibility to TB and BCG vaccine efficacy were
measures the global remarkably variable and genetically uncoupled from one
transcriptional difference of
profiles in the blood27. In the first study 26, granulomas
from rhesus macaques were found to undergo tran‑ another, such that protection to vaccination correlated
each patient relative to the
with the intrinsic quality of the immune response.
median in healthy controls. scriptional reprogramming from early (4 weeks) to

NATURE REVIEWS | IMMUNOLOGY VOLUME 17 | NOVEMBER 2017 | 693

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Caseum exhibit morphological heterogeneity 34,35. In addition to that poorly contain their bacteria are probably respon‑
A hallmark feature of human the classic caseous granuloma, granulomas can be non- sible for allowing bacterial dissemination, worsening
tuberculous granulomas that necro­tizing, neutrophil-­rich, mineralized, completely pathology and driving the onset of active disease37,40–42.
results from a distinctive type fibrotic or cavitary 30. In most of these types, the basic In a macaque model of TB, Lin et al.37 demonstrated
of central necrotic breakdown
known as caseous necrosis.
granuloma architecture exhibits the following structure: that animals with clinically active disease had both ster‑
The term caseum derives from a central acellular necrotic core, termed the caseum, sur‑ ile granulomas and regions of severe pathology (that is,
the ‘cheese-like’ appearance of rounded by a diverse population of macrophages that consolidations and TB‑associated pneumonia) that had
the necrotic area. is itself circumscribed by a lymphocytic cuff of CD4+ very different profiles of bacterial killing. These findings
and CD8+ T cells and B cells, and may have a peripheral shift the view of host control from a systemic response to
Consolidations
Pathological processes by
fibrotic edge29,32. Granulomas mainly contain macro­ a localized response within individual granulomas. The
which the pulmonary phages that are at various stages of activation, and mechanisms that establish diverging granuloma fates
infiltration of cells, fluid or T cells and B cells, but they can also contain neutrophils, are unclear, but involve complex interactions between
other material leads to the loss dendritic cells and fibroblasts30,32 (FIG. 1). tissue inflammation, host immune responses and bac‑
of aeration and of the normal
spongy consistency, causing
The important lesson learned from appreciating terial processes. We discuss these factors in greater
parenchymal tissue to have a the heterogeneity of human TB granulomas is that detail below.
more firm, solid texture. Such a each separate granuloma represents a localized micro­
change is most commonly environment that can be independently influenced Inflammation: a delicate and dynamic balance.
associated with
by the quality of the localized immune response; the Following successful infection, M. tuberculosis initiates
infection-induced inflammatory
infiltrates and leads to
pathogenicity, state and number of bacteria; the extent a clash of pro-inflammatory and anti-inflammatory sig‑
pneumonia. of immuno­p athology; and the overall host disease nals within the lungs that is vital in establishing the gran‑
status36,37. Recent studies using both animal models uloma and in influencing its eventual trajectory 30,41,43–47.
of TB14,37 and surgically resected tissue from humans The resulting ‘tug‑of‑war’ between these mediators of
with TB38,39 support these findings and reiterate the inflammation can both promote and limit bacterial
inherent complexity of the disease. Importantly, this dissemination 6,30,48. Skewing towards a robust pro-­
granuloma-specific heterogeneity is crucial in deter‑ inflammatory state can lead to remodelling within the
mining host outcome, as only one or a few granulomas granuloma, liquefaction (or softening) of caseum49 and

T cell B cell

Neutrophil NK cell

Epithelioid Apoptotic infected

macrophage epithelioid
macrophage

Macrophage Apoptotic
infected
macrophage

Necrotic
infected Dendritic
macrophage cell

Mycobacterium
tuberculosis

Figure 1 | A classical tuberculosis granuloma. The hallmark tuberculosis granuloma is a highly organized collection of
immune cells that aggregate around a central necrotic core. Reproduced from REF. 32 © Macmillan
NaturePublishers Limited.
Reviews | Immunology
NK, natural killer.

694 | NOVEMBER 2017 | VOLUME 17 www.nature.com/nri

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

the destruction of the surrounding lung parenchyma. we demonstrated that the vast majority of granulomas
Such processes are linked to the onset of active disease40, begin with a single bacterium37. The bacteria grow
and are necessary for cavitation into neighbouring air‑ within the nascent granuloma up to 4 weeks following
ways and successful transmission of M. tuberculosis 49,50. infection, at which time the adaptive immune response
By comparison, the resolution of inflammation within is engaged. After 4 weeks, bacterial killing is seen in
the granuloma and the lungs is associated with better the majority of granulomas, and approximately 10%
host outcome40, a reduced risk of reactivation51 and a of granulomas are sterilized by 11 weeks; sterilization
better long-term prognosis after treatment 52. increased even in the macaques that were developing
In a seminal study published in 2016, Malherbe et al.52 active TB. However, there was substantial variability in
observed that residual inflammation detected by FDG the frequency of granulomas that were sterilized in each
PET–CT in patients with seemingly cured pul­monary monkey, again supporting the variability of granuloma
TB was associated with the presence of M. tuber­culosis outcomes between individuals.
mRNA in both sputum and bronchoalveolar lavage fluid So, what is the variability of the host adaptive
samples. Patients from two separate cohorts in South immune response in granulomas? In a pilot study that
Africa and South Korea displayed a range of radio­graphic examined lung granulomas from patients with chronic
responses, with the majority exhibiting pul­monary pulmonary TB, Subbian et al.38 found significant var‑
inflammation that persisted a year after treatment despite iability in the density of T cells and in the extent of
receiving a curative regimen. This study highlights the fibrosis, which varied in accordance with immune acti‑
spectrum of host outcomes, even following treatment, as vation and bacterial load. The authors suggested that
well as the heterogeneous and potentially predictive role the observed heterogeneity was at least partly driven by
of inflammation in TB outcome. differential immune profiles in individual granuloma
A second study published in 2016 examined lipid microenvironments that reflected the maturation state
inflammatory pathways within human and rabbit of the granuloma. A similar study 62 that carefully meas‑
granulomas46. Marakalala et al.46 showed that granu‑ ured T cell cytokine responses in discrete granulomas
lomas have highly organized regions of inflammatory from infected macaques noted considerable hetero­
signalling that are closely linked to function: specifi‑ geneity across and within animals. As was observed by
cally, pro-­inflammatory, antibacterial lipid mediators Subbian et al.38 in the resected human tissue, Gideon
are concentrated in the centre of the granulomas, and et al.62 found extensive variability in the total numbers
anti-inflammatory, tissue-preserving mediators are in and phenotypes of T cells, as well as a wide range of
the periphery 46. These observations are consistent with a cytokine profiles and bacterial burdens within individ‑
previous study of human and macaque granulomas that ual granulomas, even within the same macaque. Fewer
found similar spatial compartmentalization of inflamma‑ than 10% of granuloma T cells were found to produce
tory programmes organized around differential popula‑ cytokines following M. tuberculosis antigen stimulation,
tions of macrophages53. Marakalala et al.46 proposed that and the majority of these cells produced only a single
the precise localization and balance of these inflammatory cytokine62; the dominant cytokines included IFNγ,
boundaries influences individual granuloma fate, which interleukin‑2 (IL‑2), tumour necrosis factor (TNF),
collectively affects host outcome. This model is broadly IL‑10 and IL‑17. However, when viewed as a whole,
consistent with earlier studies from several groups that granulomas were found to be multifunctional cytokine
have highlighted the importance of balance in the eicos‑ environments with different T cells contributing a
anoid inflammatory axis in TB54–60. Importantly, this broader cytokine repertoire.
hypothesis helps to explain the variation observed in TB Subsequent comparisons of differential T cell
granulomas, as slight differences in inflammatory path‑ responses with granuloma bacterial burden have
ways probably contribute to diverse granuloma archi‑ shown that a combination of pro-inflammatory and
tectures and functions, and have different consequences anti-inflammatory cytokines produced by different
for bacterial control. This large area of work ultimately T cells­— for example, IL‑10 with IL‑17 — best asso‑
suggests that beneficial inflammatory intervention at ciated with granuloma sterility, reiterating a necessity
the local granuloma level has the possibility to skew for balanced cytokine responses in achieving bac­terial
granuloma responses in favour of host resolution. clearance 30,61. However, we do not understand the
mechanisms that lead to the heterogeneity of adap‑
The heterogeneity of immune responses in TB granu- tive immune responses in granulomas or the extent
lomas. Closely linked to the inflammatory axis are the to which these differences are driven by the described
innate and adaptive immune responses that occur within variability in innate immune responses. The relatively
the TB granuloma. These responses are an integral com‑ small population of granuloma T cells that express
ponent of host defence and bacterial containment34,43,44,61, cytokines raises questions about the remainder of the
and they are intimately linked with granuloma out‑ T cells. For example, are they non‑TB‑specific T cells
come37,62–64. The particular immune cells involved in that are just recruited in response to granuloma‑
TB immunity 34,65–67, and their localization and kinet‑ tous inflammation? Does the initial interaction of
ics32,41,61,68, have been extensively reviewed in the past M. tuberculosis with a specific type of phagocyte skew
few years. Here, we discuss how the heterogeneity of the T cell responses in the granuloma that emerges? Is the
the immune response in TB influences the local granu‑ local adaptive immune response modulated by
loma microenvironment. In a recent study in macaques, the local initial innate immune response? Is there

NATURE REVIEWS | IMMUNOLOGY VOLUME 17 | NOVEMBER 2017 | 695

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

exhaustion of T cells in the granuloma, resulting in at 4–5 weeks post infection correlates with dissemina‑
few responding T cells? There are many potential tion69. Recent work in the zebrafish–Mycobacterium
mechanisms for the regulation of T cell responses marinum model has identified similar crucial events in
in granulomas, and further careful analysis of individ‑ granuloma formation and organization that depend on
ual granulomas may provide answers to these questions. the macrophage-specific reprogramming of adhesion
Nonetheless, the available data on granuloma T cells pathways that are dependent on epithelial cadherin
parallel the observations and data above, which indi‑ (E-cadherin; also known as cadherin 1)70. Disruption
cate that individual granulomas have unique and dis‑ of this axis resulted in disordered granuloma forma‑
tinct inflammatory signatures46 and killing potentials37, tion and increased host clearance following increased
and this suggests that there is an overall relationship neutrophil access. These observations once more
between local host immune responses, the subsequent link cellular responses with granuloma trajectory,
skewing of inflammation and differential bacterial and further suggest that variability may be encoded
containment that governs granuloma fate (FIG. 2). early in infection and influenced by multiple immune
There are recent studies40,51,69 that support this para­ components.
digm. In one study, Coleman et al.40 used serial FDG In a separate macaque study, we evaluated reac­
PET–CT imaging to evaluate granuloma dynamics in tivation risk in a large number (n = 26) of clinically latent
macaques that subsequently developed active disease cynomolgus macaques using FDG PET–CT imaging
or LTBI following a low-dose M. tuberculosis infec‑ before and during the administration of antibody spe‑
tion40. This revealed two features observed in the first cific for TNF51, which is a trigger for the reactivation of
3–6 weeks post infection that were associated with the infection in humans71–73 and macaques. We previously
eventual onset of active TB (months later): first, the showed that not all macaques undergo infection reacti­
development of new pulmonary granulomas (that is, vation during 8 weeks of TNF neutralization74, and this
dissemination); and second, increased inflammation allowed us to study which factors correlated with the
in pulmonary granulomas, as measured by FDG avid‑ risk of reactivation. FDG PET–CT scans performed
ity. These two observations not only emphasize the before TNF neutralization identified increased lung
importance of the early interactions between the host inflammation and the presence of an extrapulmonary
and pathogen in determining host outcome41, but also site of infection as features that predict reactivation
indicate that the early cellular response of a granuloma risk with 92% sensitivity and specificity 51. Using this
is linked to inflammation that then influences dissem‑ metric, we then classified a separate set of 25 clinically
ination risk. Very recent work in our laboratories has latent macaques as being at high or low risk of reactiva‑
confirmed these findings, demonstrating that only a tion, but necropsied them without TNF neutralization.
subset of granulomas disseminate to form new, pro‑ The macaques that were at high risk of infection had
ductive granulomas and showing that granuloma size at least one granuloma with a relatively high bacterial

Mycobacterium tuberculosis Inflammation Granuloma fate

and host response

Neutrophil
Resolved
granuloma Epithelioid
macrophage
Pro-inflammatory and Macrophage
anti-inflammatory
mediators
Apoptotic
infected
macrophage
Necrotic
infected
macrophage
Dendritic cell
Local infection site(s) T cell
Unresolved
granuloma
B cell
Mycobacterium
tuberculosis

Figure 2 | Granuloma fate is influenced by a complex and dynamic exchange of host and bacterial
Nature features.
Reviews | Immunology
In the lungs, a crucial interplay between the bacteria and host immune cells influences inflammatory programmes that
contribute to granuloma outcome. This process is highly dynamic and iterative, with multiple components having
pleiotropic, knock‑on and feedback effects on inflammation and the host–pathogen interaction.

696 | NOVEMBER 2017 | VOLUME 17 www.nature.com/nri

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

Theoretical granuloma equations variability might influence host–pathogen interactions.

We then briefly review the extensive literature on the
a+b+c=x a = pro-inflammatory cytokines ways in which differences in granuloma course can
Sterile granulomas
a+ c –e=x b = anti-inflammatory cytokines also affect bacterial cell state — here termed ‘reactive’
bacterial variability — and consequently the efficacy of
a+b+e=y c = activated host cells
antibiotic treatment.
a + 2b + c = y Stable granulomas d = dysregulated host cells
a+b+c+d=y e = excessive immunopathology Is there deterministic bacterial variation? In consider‑
a+b+c+d+e=z f = bacterial virulence or ing how differences in the course of infection or granu‑
immunomodulation loma fate emerge, one possibility is that variation in the
2a + b + e = z Disseminating
granulomas bacterial population helps to shape granuloma trajectory.
a+b+d+f=z It is clear from studies of fixed genetic variants that dif‑
2a + b + 2c = z ferences between bacteria can influence host responses.
For example, compared with the M. tuberculosis H37Rv
Figure 3 | Multiple ‘equations’ can determine granuloma fate. There are multiple
Nature Reviews | Immunology
strain, M. tuberculosis strains from the East Asian lin‑
pathways that lead to both resolved and unresolved granuloma outcomes in human eage induce a distinct immunopathological response
tuberculosis. These are fluid and dynamic processes that change as the structures encounter in mouse models, with larger areas of pneumonia and
different contexts of immune responses, inflammation and bacterial persistence. relatively less granulomatous lung tissue79,80. These dif‑
ferences in immunopathology have been linked, at least
partially, to differences in the production of distinct cell
burden, suggesting that, first, the robust control of infec‑ wall glycolipids81, which result in the differential activa‑
tion at all sites is essential for preventing reactivation, tion of Toll-like receptors and resulting inflammatory
and second, the reactivation risk is granuloma specific, responses82–85.
once more returning our focus to the individual granu‑ Although the importance of differences between
loma and its unique immune response. Further support M. tuberculosis strains has become clear, it is less clear
for this idea came from data indicating that only a subset whether bacterial variation contributes to differences
of granulomas in each macaque had dynamic changes in granuloma trajectory in a given individual. Bacterial
in FDG avidity or size during TNF neutralization; this barcoding analyses have demonstrated that the majority
suggests that some granulomas are more TNF dependent of TB granulomas are founded by a single bacterium37.
than others in the same animal. It is important to note From studies in macaques, there seems to be an early
that the stable granulomas during TNF neutralization wave of dissemination during the first 6–8 weeks of
were not necessarily sterile, yet they did not seem to be infection in which most granulomas are established,
affected by the loss of TNF. although later dissemination is not uncommon69. Thus,
Thus, our current understanding of granuloma-­ there is a very tight bottleneck on the bacterial popula‑
specific host responses can be considered as a set of tion early in infection at the time of secondary granu‑
mathematical equations that can be solved in many loma formation. However, this is only likely to result in a
different ways by each granuloma to reach the result of significant bacterial founder effect if there is biologically
infection control (FIG. 3). In this scenario, modulating meaningful variation in the bacterial population.
one aspect of the immune response could cause one or a In many infections, the infecting pathogen popula‑
few granulomas to lose control of the infection, leading tion is large and already diverse, or there is sufficient
to dissemination and disease, while other granulomas genomic instability that within a short period of time
are unaffected. Indeed, computational modelling of there is tremendous opportunity for genetic diversifica‑
TB granulomas has demonstrated that there are many tion of the pathogen population86,87. However, M. tuber-
pathways to containment or loss of control, and the culosis is comparatively genetically monomorphic,
ultimate outcome depends on the particular combina‑ and the infectious dose is very low 88. Whole-genome
tion of different factors75–78. Again, it seems from these sequencing of macaque granulomas and human samples
data that only one or a few granulomas need to fail — indicates that even after extended periods of infection,
either during chronic infection or LTBI, or even in early little genetic diversity accumulates in the bacterial popu‑
infection — to result in active TB. Collectively, these lation36,89,90. The epidemiological guide is that fewer than
findings provide a basis to explain the heterogeneity five single-nucleotide polymorphisms separate bacteria
among infected hosts, as we believe variable granuloma within a given individual36,89,90, making it unlikely that
contexts — particularly early in infection41 — initiate genetic diversification of the M. tuberculosis population
heterogeneous disease states (FIG. 4). followed by sharp bottlenecking creates sufficiently
distinct pathogen populations to drive differences in
Bacterial heterogeneity granuloma fate.
The heterogeneity in granuloma course raises ques‑ By contrast, there is a surprising amount of pheno‑
tions both about the contribution of bacterial processes typic heterogeneity within M. tuberculosis populations.
to granuloma fate and about the impact of the hetero­ More than 50 years ago, Bigger 91 described the existence
geneous granuloma state on the bacterial population. of a small, phenotypically distinct subpopulation of bac‑
For the purpose of this Review, we define ‘deterministic’ teria (in his studies of Staphylococcus pyogenes) that were
bacterial heterogeneity as the ways in which bacterial recognizable as non-growing, highly antibiotic-tolerant

NATURE REVIEWS | IMMUNOLOGY VOLUME 17 | NOVEMBER 2017 | 697

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Spectrum of TB

All resolved Neutrophil

granulomas Clearance
Epithelioid
macrophage

Macrophage

Latent
Apoptotic
infected
macrophage
b
Necrotic
infected
macrophage
≥1 unresolving Subclinical
granuloma with or low- Dendritic cell
viable M. tuberculosis grade TB
T cell

B cell
Mycobacterium

Active
tuberculosis

≥1 poorly controlled
and unresolving Severe TB
granuloma with many
viable M. tuberculosis

Figure 4 | Individual granulomas establish variable host outcomes and contribute to the overall spectrum of
tuberculosis. a | An individual cynomolgus macaque in which all granulomas have resolved, thus leading
Nature to an | Immunology
Reviews
asymptomatic, contained outcome, has the lowest risk of reactivation of infection, and has effectively cleared most or all of
the Mycobacterium tuberculosis infection. b | An individual macaque that has one or more granulomas that contain viable
M. tuberculosis but can be asymptomatic for clinical tuberculosis (TB). These granulomas are not actively disseminating or
progressing to worse pathologies, but they may have a persistent low level of inflammation. c | An individual macaque that
has one or more very poorly controlled granulomas that are actively disseminating bacteria to other sites (indicated by
outward-pointing arrows) can develop progressive disease and potentially worse forms of active TB.

cells; he termed these cells ‘persisters’. More recently, of these populations has been hard to address experi‑
considerable efforts to identify phenotypically distinct mentally because genetic mechanisms to perturb these
subpopulations of M. tuberculosis cells have led to the distinct cell states have not been identified.
recognition that there are likely to be many phenotypic
variants in any given M. tuberculosis population and that Reactive bacterial variation. The concept of function‑
these variants arise through a variety of mechanisms92,93. ally important bacterial heterogeneity that arises in
This variability has typically been described in terms of distinct granuloma environments is much better estab‑
cell-to-cell differences in antibiotic susceptibility; it is lished in the TB field than is the concept of determinis‑
unclear to what extent this variation results in differences tic bacterial variation. Mitchison and colleagues96 were
between bacterial cells in terms of their pathogenicity. among the first proponents of this model, which they
It is also unclear whether there are mechanisms to gen‑ invoked to explain the results of the early trials of com‑
erate variation in bacterial pathogenicity at a sufficiently bination drug treatment for TB that were conducted in
high frequency but also of sufficiently long duration to the 1950s and 1960s. In the earliest studies of various
plausibly drive granuloma fate. Asymmetric growth and antimicrobial regimens, investigators recognized that
division generates high-frequency variants, but their treatment failure due to the emergence of drug resistance
phenotypes change rapidly over successive cycles of cell occurred frequently even when patients were treated
division94. Studies describing subpopulations of drug-­ with two fully effective antibiotics. Given the limited
resistant cells that arise at a relatively high frequency and capacity of the bacterium for genetic diversification due
show semi-stable drug resistance across several genera‑ to its low mutation rate and relatively small population
tions suggest that epigenetic mechanisms may exist and size, these data were difficult to explain unless there were
fill this gap95. However, the relative importance of any subpopulations of organisms that were not exposed to or

698 | NOVEMBER 2017 | VOLUME 17 www.nature.com/nri

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

functionally susceptible to both drugs96. From these data which it is difficult to obtain transcripts. Several studies
and experimental studies of drug efficacy, Mitchison93 have successfully probed bacterial gene expression and
proposed that in a given individual there are different pheno­typic state of M. tuberculosis isolated from sputum.
subpopulations of organisms that have distinct drug Although these studies capture bacteria only from the
susceptibilities based on growth rate differences or granulomas that progress, even in M. tuberculosis iso‑
granuloma pH. lated from sputum there is evidence of metabolic repro‑
The Mitchison model did not claim that these func‑ gramming 105–109. These results are broadly consistent
tionally distinct bacterial subpopulations were spatially with studies in mice in which bacterial reporter strains
segregated in distinct granulomas. However, the model have provided single-cell measures of pathogen state that
was consistent with pioneering work by Medlar et al.97 suggest that the adaptive immune response drives the
who undertook histopathological and microbiological emergence of a subpopulation of metabolically active but
examination of thousands of TB granulomas obtained non-growing bacterial cells110,111. It remains to be deter‑
at autopsy, and noted that some ‘closed’ granulomas mined whether this is more nuanced: that is, whether
(that were not accessible to an airway) contained specific immune environments drive distinct bacte‑
micro­s copically visible but unculturable bacteria, rial responses or whether there is a feedback loop such
whereas other ‘open’ granulomas were teeming with that the emergent bacterial populations then promote
readily culturable organisms. Subsequent work con‑ different immune reactions.
firmed these findings98,99, and led investigators to fur‑
ther suggest that at least a proportion of the bacterial Variable responses to drug treatment. Although the
population in closed granulomas was not dead but contribution of specific bacterial populations — in dis‑
instead viable and unculturable100. These findings are tinct metabolic states — to the rate of progression to
early evidence of the spectrum of granulomas in a given disease remains unclear, there is strong evidence that
individual as reflected in a variety of bacterial states; non-replicating or differentially culturable bacteria are
closed granulomas were first described in samples from more tolerant of antibiotics, which makes granuloma
patients who died of active TB and who also had open state an important determinant of treatment response.
granulomas characterized by high numbers of readily Current treatment for drug-sensitive, active TB remains
culturable organisms. This and other work led investi‑ the standard 6‑month regimen of isoniazid, rifampin,
gators to postulate that under certain types of environ‑ pyrazinamide and ethambutol. This regimen provides
mental stress — most notably hypoxia — the bacterium cure rates of 90–95% in trial TB control programmes21,
enters a state of so‑called ‘non-­replicating persistence’, but several attempts to shorten the treatment period
which is characterized by metabolic remodelling and have failed112. It is thought that the prolonged dura‑
multidrug tolerance101. Importantly, these changes in tion of TB treatment is dictated by the fact that at any
bacterial cell state correlate with changes in cell wall given time some bacteria are functionally tolerant to the
composition that alter the acid fastness of the organism administered drugs, and this is fundamentally a result of
and thus its ability to be detected via Ziehl–Neelsen granuloma heterogeneity. The search for antibiotics that
staining 102. Acid-fast staining-negative TB bacilli have are active against bacteria across a range of granuloma
been associated with stages of infection in which the environments has led to the identification of new classes
bacterial population is not actively replicating 103, of antibiotics such as the nitroimidazoles113–115 and may
although this association is imperfect as large repli­ explain the clinical potency of drugs that target energy
cating populations of weakly acid-fast bacteria have metabolism such as bedaquiline116,117.
been described in some animal models104. Granuloma state also influences the activity and con‑
As these discrepancies suggest, there is likely to be centrations of different antibiotics independent of bac‑
a range of bacterial states in vivo that are only crudely terial state. This was first recognized for the first-line
probed by metrics such as aggregate growth rate and antibiotic pyrazinamide, which is a crucial agent for
acid fastness. Studies suggest that even in sputum — so bringing the duration of TB treatment down to 6 months
presumably coming from open granulomas — there are (a reduction compared with the very lengthy treatments
subpopulations of ‘differentially culturable’ M. tubercu- of earlier times118). Pyrazinamide is only active under
losis bacilli105,106. However, it has been surprisingly chal‑ acidic conditions and thus is thought to be particu‑
lenging to develop molecular definitions for the state of larly useful against bacteria in inflammatory environ‑
M. tuberculosis in TB granulomas in humans or non-­ ments. More recently, it has become clear that different
human primates, and even more difficult to establish drugs do not access all TB granulomas equally. Using
the molecular mechanisms by which bacterial subpop‑ matrix-­assisted laser desorption ionization (MALDI)
ulations arise and are maintained. In part, this question mass spectrometry, Prideaux et al.119 demonstrated that
has become hard to address because of the success of some drugs — for example, the second-line agent moxi­
antibiotic therapy. floxacin — diffuse poorly into caseum, whereas other
Unlike Medlar et al.97, researchers today do not have drugs, such as rifampicin, efficiently penetrate all granu‑
access to thousands of human TB granulomas. Even in lomas. Thus, differences in granuloma fate and anatomy
macaque granulomas, it has proved technically difficult can result in uneven drug delivery, and at a minimum
to probe M. tuberculosis state, for example, by tran‑ provide a framework for constructing drug regimens
scriptional profiling, in individual granulomas, because that are both more effective and better protected against
many granulomas have relatively few organisms, from the emergence of drug resistance.

NATURE REVIEWS | IMMUNOLOGY VOLUME 17 | NOVEMBER 2017 | 699

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Conclusions to define the features of successful and unsuccessful

In this Review, we have developed a model for the bio‑ immune responses within a given individual. We antic‑
logical heterogeneity of TB infection, which manifests at ipate that these types of study will be most useful as a
the level of individual granulomas through variability in foundation for rational vaccine design, and potentially
inflammation, the local adaptive immune response and help investigators to determine whether it is necessary to
bacterial state. Granuloma heterogeneity has implications target one or multiple paths to achieve bacterial contain‑
for how we study TB, and highlights the importance of ment or sterility. Although there remains much work to
tissue-level analyses rather than analyses of circulating be done, we are optimistic that embracing this paradigm
immune cells or even tissue samples that contain mul‑ of granuloma heterogeneity is central to skewing the
tiple granulomas. It also indicates a great opportunity host–pathogen ‘arms race’ in favour of humans.

1. Getahun, H., Matteelli, A., Chaisson, R. E. & 20. Cliff, J. M. et al. Distinct phases of blood gene 38. Subbian, S. et al. Lesion-specific immune response in
Raviglione, M. Latent Mycobacterium tuberculosis expression pattern through tuberculosis treatment granulomas of patients with pulmonary tuberculosis:
infection. N. Engl. J. Med. 372, 2127–2135 (2015). reflect modulation of the humoral immune response. a pilot study. PLoS ONE 10, e0132249 (2015).
2. Corbett, E. L. et al. The growing burden of J. Infect. Dis. 207, 18–29 (2013). 39. Lenaerts, A., Barry, C. E. III & Dartois, V.
tuberculosis: global trends and interactions with the 21. Zumla, A., Nahid, P. & Cole, S. T. Advances in the Heterogeneity in tuberculosis pathology,
HIV epidemic. Arch. Intern. Med. 163, 1009–1021 development of new tuberculosis drugs and treatment microenvironments and therapeutic responses.
(2003). regimens. Nat. Rev. Drug Discov. 12, 388–404 Immunol. Rev. 264, 288–307 (2015).
3. Lawn, S. D. & Zumla, A. I. Tuberculosis. Lancet 378, (2013). 40. Coleman, M. T. et al. Early changes by
57–72 (2011). 22. Koul, A., Arnoult, E., Lounis, N., Guillemont, J. & 18
fluorodeoxyglucose positron emission tomography
4. World Health Organization. Global Tuberculosis Andries, K. The challenge of new drug discovery for coregistered with computed tomography predict
Report 2015 (World Health Organization, 2015). tuberculosis. Nature 469, 483–490 (2011). outcome after Mycobacterium tuberculosis infection
5. Andrews, J. R. et al. Risk of progression to active 23. Blankley, S. et al. The transcriptional signature of in cynomolgus macaques. Infect. Immun. 82,
tuberculosis following reinfection with Mycobacterium active tuberculosis reflects symptom status in extra- 2400–2404 (2014).
tuberculosis. Clin. Infect. Dis. 54, 784–791 (2012). pulmonary and pulmonary tuberculosis. PLoS ONE 11, This study reveals that early granuloma
6. Lin, P. L. & Flynn, J. L. Understanding latent e0162220 (2016). dissemination and inflammation influence the
tuberculosis: a moving target. J. Immunol. 185, 24. Banchereau, R. et al. Host immune transcriptional clinical outcome of infection in infected cynomolgus
15–22 (2010). profiles reflect the variability in clinical disease macaques.
7. Chen, R. Y. et al. PET/CT imaging correlates with manifestations in patients with Staphylococcus aureus 41. Cadena, A. M., Flynn, J. L. & Fortune, S. M.
treatment outcome in patients with multidrug- infections. PLoS ONE 7, e34390 (2012). The importance of first impressions: early events in
resistant tuberculosis. Sci. Transl Med. 6, 265ra166 25. Zak, D. E. et al. A blood RNA signature for Mycobacterium tuberculosis infection influence
(2014). tuberculosis disease risk: a prospective cohort study. outcome. mBio 7, e00342‑16 (2016).
8. Lenzini, L., Rottoli, P. & Rottoli, L. The spectrum of Lancet 387, 2312–2322 (2016). 42. Lieberman, T. D. et al. Genomic diversity in autopsy
human tuberculosis. Clin. Exp. Immunol. 27, 26. Mehra, S. et al. Transcriptional reprogramming in samples reveals within-host dissemination of HIV-
230–237 (1977). nonhuman primate (rhesus macaque) tuberculosis associated Mycobacterium tuberculosis. Nat. Med.
9. Poulsen, A. Some clinical features of tuberculosis. granulomas. PLoS ONE 5, e12266 (2010). 22, 1470–1474 (2016).
Acta Tuberc. Scand. 33, 37–92 (1957). 27. Gideon, H. P., Skinner, J. A., Baldwin, N., Flynn, J. L. & 43. Cooper, A. M., Mayer-Barber, K. D. & Sher, A.
10. Barry, C. E. 3rd et al. The spectrum of latent Lin, P. L. Early whole blood transcriptional signatures Role of innate cytokines in mycobacterial infection.
tuberculosis: rethinking the biology and intervention are associated with severity of lung inflammation in Mucosal Immunol. 4, 252–260 (2011).
strategies. Nat. Rev. Microbiol. 7, 845–855 (2009). cynomolgus macaques with Mycobacterium 44. Lerner, T. R., Borel, S. & Gutierrez, M. G. The innate
11. Young, D. B., Gideon, H. P. & Wilkinson, R. J. tuberculosis infection. J. Immunol. 197, 4817–4828 immune response in human tuberculosis.
Eliminating latent tuberculosis. Trends Microbiol. 17, (2016). Cell. Microbiol. 17, 1277–1285 (2015).
183–188 (2009). 28. Williams, G. T. & Williams, W. J. Granulomatous 45. Lin, P. L. et al. Early events in Mycobacterium
12. Esmail, H. et al. Characterization of progressive HIV- inflammation — a review. J. Clin. Pathol. 36, tuberculosis infection in cynomolgus macaques.
associated tuberculosis using 2‑deoxy‑2-[18F]fluoro- 723–733 (1983). Infect. Immun. 74, 3790–3803 (2006).
D‑glucose positron emission and computed 29. Russell, D. G., Cardona, P. J., Kim, M. J., Allain, S. & 46. Marakalala, M. J. et al. Inflammatory signaling in
tomography. Nat. Med. 22, 1090–1093 (2016). Altare, F. Foamy macrophages and the progression of human tuberculosis granulomas is spatially organized.
This study provides evidence of subclinical, active the human tuberculosis granuloma. Nat. Immunol. 10, Nat. Med. 22, 531–538 (2016).
disease in a subset of adults infected with HIV‑1 943–948 (2009). By analysing the proteomes of human and rabbit
and latent TB, reiterating the spectrum of disease 30. Flynn, J. L., Chan, J. & Lin, P. L. Macrophages and granulomas, this study reveals that
seen in human TB. control of granulomatous inflammation in tuberculosis. pro-inflammatory and anti-inflammatory
13. Capuano, S. V. et al. Experimental Mycobacterium Mucosal Immunol. 4, 271–278 (2011). programmes occur simultaneously, but in physically
tuberculosis infection of cynomolgus macaques closely 31. Davis, J. M. & Ramakrishnan, L. The role of the distinct compartments.
resembles the various manifestations of human granuloma in expansion and dissemination of early 47. Fallahi-Sichani, M., El‑Kebir, M., Marino, S.,
M. tuberculosis infection. Infect. Immun. 71, tuberculous infection. Cell 136, 37–49 (2009). Kirschner, D. E. & Linderman, J. J. Multiscale
5831–5844 (2003). 32. Ramakrishnan, L. Revisiting the role of the granuloma computational modeling reveals a critical role for
14. Lin, P. L. et al. Quantitative comparison of active and in tuberculosis. Nat. Rev. Immunol. 12, 352–366 TNF-α receptor 1 dynamics in tuberculosis granuloma
latent tuberculosis in the cynomolgus macaque model. (2012). formation. J. Immunol. 186, 3472–3483 (2011).
Infect. Immun. 77, 4631–4642 (2009). 33. Flynn, J. L. Mutual attraction: does it benefit the host 48. Guirado, E. & Schlesinger, L. S. Modeling the
15. Berry, M. P. et al. An interferon-inducible neutrophil- or the bug? Nat. Immunol. 5, 778–779 (2004). Mycobacterium tuberculosis granuloma — the critical
driven blood transcriptional signature in human 34. O’Garra, A. et al. The immune response in battlefield in host immunity and disease.
tuberculosis. Nature 466, 973–977 (2010). tuberculosis. Annu. Rev. Immunol. 31, 475–527 Front. Immunol. 4, 98 (2013).
Using transcriptional profiling of whole blood, (2013). 49. Kaplan, G. et al. Mycobacterium tuberculosis growth
this study identifies a discriminatory, IFN-inducible, 35. Canetti, G. The Tubercle Bacillus in the Pulmonary at the cavity surface: a microenvironment with failed
neutrophil-mediated signature of active TB. This Lesion of Man; Histobacteriology and its Bearing on immunity. Infect. Immun. 71, 7099–7108 (2003).
work also provides new insights into the spectrum the Therapy of Pulmonary Tuberculosis (Springer, 50. Ernst, J. D. The immunological life cycle of tuberculosis.
of TB by revealing a subset of patients with latent 1955). Nat. Rev. Immunol. 12, 581–591 (2012).
TB who have signatures that overlap closely with 36. Ford, C. B. et al. Use of whole genome sequencing to 51. Lin, P. L. et al. PET CT identifies reactivation risk in
the signature of active disease. estimate the mutation rate of Mycobacterium cynomolgus macaques with latent M. tuberculosis.
16. Maertzdorf, J. et al. Functional correlations of tuberculosis during latent infection. Nat. Genet. 43, PLoS Pathog. 12, e1005739 (2016).
pathogenesis-driven gene expression signatures in 482–486 (2011). 52. Malherbe, S. T. et al. Persisting positron emission
tuberculosis. PLoS ONE 6, e26938 (2011). 37. Lin, P. L. et al. Sterilization of granulomas is common tomography lesion activity and Mycobacterium
17. Ottenhoff, T. H. et al. Genome-wide expression in active and latent tuberculosis despite within-host tuberculosis mRNA after tuberculosis cure. Nat. Med.
profiling identifies type 1 interferon response variability in bacterial killing. Nat. Med. 20, 75–79 22, 1094–1100 (2016).
pathways in active tuberculosis. PLoS ONE 7, e45839 (2014). The authors of this paper show that in spite of a
(2012). This study is the first to demonstrate that the standard curative 6‑month regimen, there were
18. Cliff, J. M., Kaufmann, S. H., McShane, H., van majority of granulomas are seeded by a single multiple patients with persistent pulmonary
Helden, P. & O’Garra, A. The human immune response bacillus and that there is differential bacterial inflammation that was coincident with the
to tuberculosis and its treatment: a view from the killing of granulomas within the host that is detection of M. tuberculosis mRNA. These
blood. Immunol. Rev. 264, 88–102 (2015). independent of host status. These findings shift observations highlight the variability of treatment
19. Bloom, C. I. et al. Detectable changes in the blood the focus to the local granuloma level and outcome in individual granulomas, even after a year
transcriptome are present after two weeks of suggest that the individual trajectories of of successful treatment, and implicate a
antituberculosis therapy. PLoS ONE 7, e46191 granulomas influence the clinical outcome of complementary, ongoing requirement for immunity
(2012). infection. in maintaining sterility.

700 | NOVEMBER 2017 | VOLUME 17 www.nature.com/nri

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

53. Mattila, J. T. et al. Microenvironments in tuberculous 74. Lin, P. L. et al. Tumor necrosis factor neutralization 98. Beck, F. & Yegian, D. A study of the tubercle bacillus in
granulomas are delineated by distinct populations of results in disseminated disease in acute and latent resected pulmonary lesions. Am. Rev. Tuberc. 66,
macrophage subsets and expression of nitric oxide Mycobacterium tuberculosis infection with normal 44–51 (1952).
synthase and arginase isoforms. J. Immunol. 191, granuloma structure in a cynomolgus macaque model. 99. Canetti, G. Anatomical and bacteriological changes in
773–784 (2013). Arthritis Rheum. 62, 340–350 (2010). tuberculous lesions under the influence of antibiotics
54. Tobin, D. M. et al. Host genotype-specific therapies 75. Marino, S. et al. Computational and empirical studies and chemotherapy. Bull. Int. Union Tuberc. 24,
can optimize the inflammatory response to predict Mycobacterium tuberculosis-specific T cells as 144–240 (1954).
mycobacterial infections. Cell 148, 434–446 (2012). a biomarker for infection outcome. PLoS Comput. Biol. 100. Salkin, D. & Wayne, L. G. The bacteriology of resected
55. Chen, M. et al. Lipid mediators in innate immunity 12, e1004804 (2016). tuberculous pulmonary lesions. I. The effect of interval
against tuberculosis: opposing roles of PGE2 and LXA4 76. Marino, S. et al. Macrophage polarization drives between reversal of infectiousness and subsequent
in the induction of macrophage death. J. Exp. Med. granuloma outcome during Mycobacterium tuberculosis surgery. Am. Rev. Tuberc. 74, 376–387 (1956).
205, 2791–2801 (2008). infection. Infect. Immun. 83, 324–338 (2015). 101. Wayne, L. G. & Sohaskey, C. D. Nonreplicating
56. Divangahi, M. et al. Mycobacterium tuberculosis 77. Kirschner, D. E. & Linderman, J. J. Mathematical and persistence of Mycobacterium tuberculosis.
evades macrophage defenses by inhibiting plasma computational approaches can complement Annu. Rev. Microbiol. 55, 139–163 (2001).
membrane repair. Nat. Immunol. 10, 899–906 experimental studies of host–pathogen interactions. 102. Vilcheze, C. & Kremer, L. Acid-fast positive and acid-
(2009). Cell. Microbiol. 11, 531–539 (2009). fast negative Mycobacterium tuberculosis: the Koch
57. Divangahi, M., Desjardins, D., Nunes-Alves, C., 78. Kirschner, D. E., Hunt, C. A., Marino, S., Fallahi- paradox. Microbiol. Spectr. http://dx.doi.org/10.1128/
Remold, H. G. & Behar, S. M. Eicosanoid pathways Sichani, M. & Linderman, J. J. Tuneable resolution as microbiolspec.TBTB2-0003-2015 (2017).
regulate adaptive immunity to Mycobacterium a systems biology approach for multi-scale, multi- 103. Seiler, P. et al. Cell-wall alterations as an attribute of
tuberculosis. Nat. Immunol. 11, 751–758 (2010). compartment computational models. Wiley Mycobacterium tuberculosis in latent infection.
58. Divangahi, M., Behar, S. M. & Remold, H. Dying to Interdiscip. Rev. Syst. Biol. Med. 6, 289–309 (2014). J. Infect. Dis. 188, 1326–1331 (2003).
live: how the death modality of the infected 79. Lopez, B. et al. A marked difference in pathogenesis 104. Obregon-Henao, A. et al. Cortisone-forced reactivation
macrophage affects immunity to tuberculosis. and immune response induced by different of weakly acid fast positive Mycobacterium
Adv. Exp. Med. Biol. 783, 103–120 (2013). Mycobacterium tuberculosis genotypes. Clin. Exp. tuberculosis in guinea pigs previously treated with
59. Mayer-Barber, K. D. et al. Host-directed therapy of Immunol. 133, 30–37 (2003). chemotherapy. Mycobact. Dis. 2, 1000116 (2012).
tuberculosis based on interleukin‑1 and type I 80. Ribeiro, S. C. et al. Mycobacterium tuberculosis 105. Mukamolova, G. V., Turapov, O., Malkin, J.,
interferon crosstalk. Nature 511, 99–103 (2014). strains of the modern sublineage of the Beijing family Woltmann, G. & Barer, M. R. Resuscitation-promoting
This study establishes a link between IL‑1 and are more likely to display increased virulence than factors reveal an occult population of tubercle bacilli in
type I IFNs that is mediated by eicosanoids. It also strains of the ancient sublineage. J. Clin. Microbiol. sputum. Am. J. Respir. Crit. Care Med. 181, 174–180
confirms the role of host-directed manipulation of 52, 2615–2624 (2014). (2010).
the eicosanoid balance in favour of prostaglandin 81. Reed, M. B. et al. A glycolipid of hypervirulent 106. Chengalroyen, M. D. et al. Detection and
E2 in resolving disease exacerbations in vivo. tuberculosis strains that inhibits the innate immune quantification of differentially culturable tubercle
60. Wallis, R. S. & Hafner, R. Advancing host-directed response. Nature 431, 84–87 (2004). bacteria in sputum from patients with tuberculosis.
therapy for tuberculosis. Nat. Rev. Immunol. 15, 82. Dormans, J. et al. Correlation of virulence, lung Am. J. Respir. Crit. Care Med. 194, 1532–1540
255–263 (2015). pathology, bacterial load and delayed type (2016).
61. Orme, I. M., Robinson, R. T. & Cooper, A. M. The hypersensitivity responses after infection with different 107. Garton, N. J. et al. Cytological and transcript analyses
balance between protective and pathogenic immune Mycobacterium tuberculosis genotypes in a BALB/c reveal fat and lazy persister-like bacilli in tuberculous
responses in the TB‑infected lung. Nat. Immunol. 16, mouse model. Clin. Exp. Immunol. 137, 460–468 sputum. PLoS Med. 5, e75 (2008).
57–63 (2015). (2004). 108. Walter, N. D. et al. Transcriptional adaptation of drug-
62. Gideon, H. P. et al. Variability in tuberculosis 83. Manca, C. et al. Differential monocyte activation tolerant Mycobacterium tuberculosis during
granuloma T cell responses exists, but a balance of underlies strain-specific Mycobacterium tuberculosis treatment of human tuberculosis. J. Infect. Dis. 212,
pro- and anti-inflammatory cytokines is associated pathogenesis. Infect. Immun. 72, 5511–5514 (2004). 990–998 (2015).
with sterilization. PLoS Pathog. 11, e1004603 84. Portevin, D., Gagneux, S., Comas, I. & Young, D. 109. Honeyborne, I. et al. Profiling persistent tubercule
(2015). Human macrophage responses to clinical isolates from bacilli from patient sputa during therapy predicts early
63. Cilfone, N. A., Perry, C. R., Kirschner, D. E. & the Mycobacterium tuberculosis complex discriminate drug efficacy. BMC Med. 14, 68 (2016).
Linderman, J. J. Multi-scale modeling predicts a between ancient and modern lineages. PLoS Pathog. 110. Sukumar, N., Tan, S., Aldridge, B. B. & Russell, D. G.
balance of tumor necrosis factor-α and interleukin‑10 7, e1001307 (2011). Exploitation of Mycobacterium tuberculosis reporter
controls the granuloma environment during 85. Carmona, J. et al. Mycobacterium tuberculosis strains strains to probe the impact of vaccination at sites of
Mycobacterium tuberculosis infection. PLoS ONE 8, are differentially recognized by TLRs with an impact on infection. PLoS Pathog. 10, e1004394 (2014).
e68680 (2013). the immune response. PLoS ONE 8, e67277 (2013). 111. Manina, G., Dhar, N. & McKinney, J. D. Stress and
64. Cilfone, N. A. et al. Computational modeling predicts 86. Grant, A. J. et al. Modelling within-host host immunity amplify Mycobacterium tuberculosis
IL‑10 control of lesion sterilization by balancing early spatiotemporal dynamics of invasive bacterial disease. phenotypic heterogeneity and induce nongrowing
host immunity-mediated antimicrobial responses with PLoS Biol. 6, e74 (2008). metabolically active forms. Cell Host Microbe 17,
caseation during Mycobacterium tuberculosis 87. Joseph, S. B., Swanstrom, R., Kashuba, A. D. & 32–46 (2015).
infection. J. Immunol. 194, 664–677 (2015). Cohen, M. S. Bottlenecks in HIV‑1 transmission: 112. Gillespie, S. H. et al. Four-month moxifloxacin-based
65. Srivastava, S., Ernst, J. D. & Desvignes, L. Beyond insights from the study of founder viruses. Nat. Rev. regimens for drug-sensitive tuberculosis. N. Engl.
macrophages: the diversity of mononuclear cells in Microbiol. 13, 414–425 (2015). J. Med. 371, 1577–1587 (2014).
tuberculosis. Immunol. Rev. 262, 179–192 (2014). 88. Jacobs, A. L. Infective dose in pulmonary tuberculosis. 113. Mukherjee, T. & Boshoff, H. Nitroimidazoles for the
66. Behar, S. M., Carpenter, S. M., Booty, M. G., Tubercle 22, 266–271 (1941). treatment of TB: past, present and future. Future
Barber, D. L. & Jayaraman, P. Orchestration of 89. Walker, T. M. et al. Whole-genome sequencing to Med. Chem. 3, 1427–1454 (2011).
pulmonary T cell immunity during Mycobacterium delineate Mycobacterium tuberculosis outbreaks: a 114. Matsumoto, M. et al. OPC‑67683, a nitro-dihydro-
tuberculosis infection: immunity interruptus. retrospective observational study. Lancet Infect. Dis. imidazooxazole derivative with promising action
Semin. Immunol. 26, 559–577 (2014). 13, 137–146 (2013). against tuberculosis in vitro and in mice. PLoS Med. 3,
67. Guirado, E., Schlesinger, L. S. & Kaplan, G. 90. Bryant, J. M. et al. Inferring patient to patient e466 (2006).
Macrophages in tuberculosis: friend or foe. transmission of Mycobacterium tuberculosis from 115. Stover, C. K. et al. A small-molecule nitroimidazopyran
Semin. Immunopathol. 35, 563–583 (2013). whole genome sequencing data. BMC Infect. Dis. 13, drug candidate for the treatment of tuberculosis.
68. Sia, J. K., Georgieva, M. & Rengarajan, J. Innate 110 (2013). Nature 405, 962–966 (2000).
immune defenses in human tuberculosis: an overview 91. Bigger, J. W. Treatment of staphylococcal infections 116. Koul, A. et al. Diarylquinolines target subunit c of
of the interactions between Mycobacterium with penicillin by intermittent sterilisation. Lancet mycobacterial ATP synthase. Nat. Chem. Biol. 3,
tuberculosis and innate immune cells. J. Immunol. 244, 497–500 (1944). 323–324 (2007).
Res. 2015, 747543 (2015). 92. Kester, J. C. & Fortune, S. M. Persisters and beyond: 117. Cole, S. T. Inhibiting Mycobacterium tuberculosis
69. Martin, C. J. et al. Digitally barcoding Mycobacterium mechanisms of phenotypic drug resistance and drug within and without. Phil. Trans. R. Soc. B Biol. Sci.
tuberculosis reveals in vivo infection dynamics in the tolerance in bacteria. Crit. Rev. Biochem. Mol. Biol. http://dx.doi.org/10.1098/rstb.2015.0506 (2016).
macaque model of tuberculosis. mBio 8, e00312‑17 49, 91–101 (2014). 118. Zhang, Y. & Mitchison, D. The curious characteristics
(2017). 93. Mitchison, D. A. How drug resistance emerges as a of pyrazinamide: a review. Int. J. Tuberc. Lung Dis. 7,
70. Cronan, M. R. et al. Macrophage epithelial result of poor compliance during short course 6–21 (2003).
reprogramming underlies mycobacterial granuloma chemotherapy for tuberculosis. Int. J. Tuberc. Lung 119. Prideaux, B. et al. The association between sterilizing
formation and promotes infection. Immunity 45, Dis. 2, 10–15 (1998). activity and drug distribution into tuberculosis lesions.
861–876 (2016). 94. Kieser, K. J. & Rubin, E. J. How sisters grow apart: Nat. Med. 21, 1223–1227 (2015).
71. Keane, J. et al. Tuberculosis associated with infliximab, mycobacterial growth and division. Nat. Rev. This study reveals that different drugs demonstrate
a tumor necrosis factor α-neutralizing agent. N. Engl. Microbiol. 12, 550–562 (2014). variable penetration, accumulation and spatial
J. Med. 345, 1098–1104 (2001). 95. Wakamoto, Y. et al. Dynamic persistence of antibiotic- distribution within tuberculous granulomas, thus
72. Bruns, H. et al. Anti-TNF immunotherapy reduces stressed mycobacteria. Science 339, 91–95 (2013). setting up separate niches for the development of
CD8+ T cell-mediated antimicrobial activity against 96. Fox, W., Ellard, G. A. & Mitchison, D. A. Studies on the drug resistance and bacterial persistence.
Mycobacterium tuberculosis in humans. J. Clin. Invest. treatment of tuberculosis undertaken by the British 120. Myllymaki, H., Bauerlein, C. A. & Ramet, M.
119, 1167–1177 (2009). Medical Research Council tuberculosis units, The zebrafish breathes new life into the study of
73. Maini, R. et al. Infliximab (chimeric anti-tumour 1946–1986, with relevant subsequent publications. tuberculosis. Front. Immunol. 7, 196 (2016).
necrosis factor α monoclonal antibody) versus placebo Int. J. Tuberc. Lung Dis. 3, S231–S279 (1999). 121. Kramnik, I. & Beamer, G. Mouse models of human TB
in rheumatoid arthritis patients receiving concomitant 97. Medlar, E. M., Bernstein, S. & Steward, D. M. pathology: roles in the analysis of necrosis and the
methotrexate: a randomised phase III trial. ATTRACT A bacteriologic study of resected tuberculous lesions. development of host-directed therapies. Semin.
Study Group. Lancet 354, 1932–1939 (1999). Am. Rev. Tuberc. 66, 36–43 (1952). Immunopathol. 38, 221–237 (2016).

NATURE REVIEWS | IMMUNOLOGY VOLUME 17 | NOVEMBER 2017 | 701

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

122. Scanga, C. A. & Flynn, J. L. Modeling tuberculosis in tuberculosis. Am. J. Respir. Crit. Care Med. http:// 135. Collaborative Cross Consortium. The genome
nonhuman primates. Cold Spring Harb. Perspect. dx.doi.org/10.1164/rccm.201611-2346OC (2017). architecture of the Collaborative Cross mouse genetic
Med. 4, a018564 (2014). 129. Graustein, A. D. et al. The SIGLEC14 null allele is reference population. Genetics 190, 389–401
123. Flynn, J. L., Gideon, H. P., Mattila, J. T. & Lin, P. L. associated with Mycobacterium tuberculosis- and (2012).
Immunology studies in non-human primate models of BCG-induced clinical and immunologic outcomes.
tuberculosis. Immunol. Rev. 264, 60–73 (2015). Tuberculosis (Edinb.) 104, 38–45 (2017). Acknowledgements
124. Pena, J. C. & Ho, W. Z. Monkey models of tuberculosis: 130. Smith, C. M. et al. Tuberculosis susceptibility and The authors gratefully acknowledge the intellectual contribu-
lessons learned. Infect. Immun. 83, 852–862 (2015). vaccine protection are independently controlled by tions of the members of the Flynn and Fortune laboratories,
125. Diedrich, C. R. et al. Reactivation of latent tuberculosis host genotype. mBio 7, e01516‑16 (2016). as well as P. Ling Lin, E. Klein, J. Mattila and C. Scanga for
in cynomolgus macaques infected with SIV is 131. Berrington, W. R. & Hawn, T. R. Mycobacterium helpful discussions. This work was supported, in part, by the
associated with early peripheral T cell depletion and tuberculosis, macrophages, and the innate immune US National Institutes of Health (T32 AI089443 to A.M.C.;
not virus load. PLoS ONE 5, e9611 (2010). response: does common variation matter? AI094745, HL110811, AI105422 and AI123093 to J.L.F.;
126. Lin, P. L. et al. CD4 T cell depletion exacerbates acute Immunol. Rev. 219, 167–186 (2007). and AI114674 to J.L.F. and S.M.F.), the Bill and Melinda
Mycobacterium tuberculosis while reactivation of 132. Casanova, J. L. & Abel, L. Genetic dissection of Gates Foundation (to J.L.F. and S.M.F.) and the Aeras Global
latent infection is dependent on severity of tissue immunity to mycobacteria: the human model. Fund (to J.L.F. and S.M.F.). Support was also provided by the
depletion in cynomolgus macaques. AIDS Res. Hum. Annu. Rev. Immunol. 20, 581–620 (2002). Burroughs Wellcome Foundation (to S.M.F.).
Retroviruses 28, 1693–1702 (2012). 133. Misch, E. A. & Hawn, T. R. Toll-like receptor
127. Shah, J. A. et al. Human TOLLIP regulates TLR2 and polymorphisms and susceptibility to human disease. Competing interests statement
TLR4 signaling and its polymorphisms are associated Clin. Sci. (Lond.) 114, 347–360 (2008). The authors declare no competing interests.
with susceptibility to tuberculosis. J. Immunol. 189, 134. Churchill, G. A. et al. The Collaborative Cross, a
1737–1746 (2012). community resource for the genetic analysis of Publisher’s note
128. Shah, J. A. et al. A functional TOLLIP variant is complex traits. Nat. Genet. 36, 1133–1137 Springer Nature remains neutral with regard to jurisdictional
associated with BCG-specific immune responses and (2004). claims in published maps and institutional affiliations.

702 | NOVEMBER 2017 | VOLUME 17 www.nature.com/nri

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.