Anda di halaman 1dari 68

The classification of macrophages and

dendritic cells (DCs) has mainly been


based on cell morphology, phenotype
and/or select functional attributes.
However, many of these markers
are not unique to a specific cell type,
which has resulted in much debate
as to whether a given mononuclear
phagocyte should be classified as a
DC or a macrophage. Reporting in
Cell, Reis e Sousa and colleagues show
that the precursors of conventional
DCs (cDCs) in mice are marked by
dendritic cell natural killer lectin
group receptor 1 (DNGR1; encoded
by CLEC9A), and they describe a
model in which these cells and their
progeny are genetically labelled, which
facilitates the identification of cDCs
in mice on the basis of ontogeny
rather than phenotype or function.
Previous studies have shown
that plasmacytoid DCs (pDCs)
and specific cDC subsets express
DNGR1. Now the authors show that
DNGR1 is also expressed by bone
marrow progenitor cells that pheno-
typically resemble
common DC precursors (CDPs).
To test the differentiation potential
of these cells, lineage-negative
CD115
+
DNGR1
+
bone marrow cells
were transferred to congenic mice.
In contrast with unfractionated bone
marrow, which gave rise to a variety
of lymphoid and myeloid lineages,
lineage-negative CD115
+
DNGR1
+

bone marrow cells almost exclusively
generated CD11c
+
MHC class II
+

cDCs, but not pDCs. This suggests
that DNGR1 expression marks pre-
cursor cells that have a cDC-restricted
differentiation potential. Of note, on
the basis of the data in the paper, the
authors suggested that the acronym
CDPs should therefore be defined as
conventional DC precursors.
Next, the authors generated
mice in which DNGR1-expressing
cells and their progeny are indel-
ibly marked with enhanced yellow
fluorescent protein (YFP), termed
Clec9a
+/cre
Rosa
+/EYFP
mice. As
expected, the expression of YFP in
these mice was restricted to CDPs,
to resident CD8
+
and CD11b
+

cDC subsets in the spleen and
skin-draining lymph nodes, and to
migratory CD103
+
cDCs in the skin.
No YFP expression was observed in
CD169
+
metallophilic macrophages,
in Langerhans cells, or in LY6C
hi
and
LY6C
low
monocytes. Of note, pDCs
(SIGLEC-H
+
B220
+
) expressed only
low levels of YFP, which, together
with the results of the transfer study
described above, suggests that these
cells arise from a distinct pDC-
specific precursor cell. However,
CD8
+
CD205

cells, which have


previously been reported to resemble
pDCs, expressed high levels of YFP,
which suggests that they arise from
CDPs.
Further investigation of these mice
showed that cDCs in non-lymphoid
tissues, including the lungs, intestines
and kidneys, can be specifically
identified by their DNGR1 expression
history. In the lungs, CD103
+
cells,
CD103

CD11b

cells and CD11b


+

cells were labelled with YFP, even
though the CD103

CD11b

and
CD11b
+
cell subsets lacked DNGR1
expression. By contrast, pDCs and
CD64
+
cells (which have been argued
to represent monocyte progeny)
were inefficiently labelled with YFP.
Similarly, in the small intestine,
CD11b

CD103
+
cells, CD11b
+
CD103
+

cells and CD11b
+
CD103

cells (a cell
subset which has previously been
suggested to arise from monocytes)
all expressed YFP, which indicates that
they descend from CDPs. By contrast,
monocyte-derived CD11c
low
CD64
+

cells were poorly labelled with YFP.
Interestingly, CD64
+
cells specifically
in the kidneys were also labelled
with YFP, which suggests that the
expression of CD64 does not dif-
ferentiate between CDP-derived and
monocyte-derived cells in this tissue
site. Further analysis showed that
CD64
+
CD11b
low
F4/80
hi
kidney cells
had phenotypic and functional
properties that are typical of cDCs.
Finally, Clec9a
+/cre
Rosa
+/EYFP
mice
were shown to faithfully trace
CDP-derived cells, but not monocyte-
derived cells that resemble cDCs,
during inflammation following
infection with Listeria monocytogenes
or following dextran sulphate sodium
treatment to induce colitis.
So, Clec9a
+/cre
Rosa
+/EYFP
mice
represent an in vivo model to identify
cDCs on the basis of their onto-
genetic descendence from a commit-
ted precursor cell and have been used
in this study to confirm that cDCs are
an independent leukocyte lineage.
Olive Leavy
DENDRI TI C CELLS
Tracing the origins of cDCs
ORIGINAL RESEARCH PAPER Schraml, B. U.
etal. Genetic tracing via DNGR-1 expression
history defines dendritic cells as a hematopoietic
lineage. Cell 154, 843858 (2013)
a model...
which
facilitates the
identification
of cDCs in mice
on the basis
of ontogeny
rather than
phenotype or
function
G
E
T
T
Y
/
P
H
O
T
O
D
I
S
C
RESEARCH HI GHLI GHTS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013
2013 Macmillan Publishers Limited. All rights reserved
The molecular mechanisms that
underlie allergic inflammatory
diseases remain unclear and this is
an area of active debate. Reporting
in Science, Millien et al. now suggest
that hyperactivation of an antifungal
pathway involving fibrinogen cleav-
age and Toll-like receptor 4 (TLR4)
signalling contributes to allergic
airway inflammation.
Wild-type mice that were intra-
nasally challenged daily for 2 weeks
with a fungal proteinase derived from
Aspergillus oryzae (PAO) developed
canonical features of asthma, includ-
ing airway hyperresponsiveness,
eosinophilia, enhanced mucin 5AC
(Muc5ac) expression, airway goblet
cell hyperplasia and production of
the T helper 2 (T
H
2) cell-associated
cytokines interleukin-4 (IL-4), IL-5
and IL-13.
By contrast, Tlr4
/
mice chal-
lenged with PAO showed reduced
or attenuated disease symptoms, but
equivalent levels of IL-4 production
compared with wild-type mice.
Similarly, the induction of allergic
lung inflammation, but not IL-4 pro-
duction, in response to allergen chal-
lenge with proteinase-free ovalbumin
or with the conidia of Aspergillus
niger was TLR4 dependent. These
data indicate that TLR4 is required
for the development of allergic
airway inflammation, irrespective of
allergen proteinase content, but not
for IL-4 production in the lungs.
Alveolar macrophages from
PAO-treated mice had increased
TLR4-dependent expression of
several genes that are associated with
antifungal immunity. In addition,
PAO-activated bone marrow-derived
macrophages (BMDMs) from wild-
type mice, but not from Tlr4
/
mice,
controlled fungal growth (known
as fungistasis) when the conidia of
A.niger were added to the cell cul-
tures. Of note, this fungistatic activity
only occurred in the presence of fetal
bovine serum. These data suggest
that PAO functions through both a
serum factor and TLR4 to induce
macrophage antifungal immunity.
Further investigations showed
that stimulation of BMDMs with the
endogenous proteinase thrombin,
which converts the serum factor
fibrinogen to fibrin, resulted in
similar fungistatic activity to that
observed after PAO stimulation.
Furthermore, fibrinogen cleavage
products (FCPs) generated by
incubation of fibrinogen with PAO
or thrombin induced fungistasis
when added to BMDM cultures with-
out fetal bovine serum. Moreover, the
proteinase inhibitor hirudin reduced
the fungistatic activity of PAO-
stimulated BMDMs in the presence
of fetal bovine serum. Interestingly,
FCPs also induced TLR4-dependent
MUC5AC and IL-13 receptor-1
(IL-13R1) expression, as well as
fungistatic activity, by ex vivo airway
epithelial cells. These data suggest
that exogenous and endogenous
proteinases can cleave fibrinogen to
generate TLR4 ligands that prime
epithelial cells to respond to IL-13.
But does this pathway have a role
in allergic airway disease? Mice chal-
lenged with high-dose FCPs showed
modest eosinophil recruitment and
Muc5ac expression in the lungs but
not airway hyperresponsiveness or
IL-4 production. Furthermore, the
inclusion of hirudin during allergen
challenge greatly reduced PAO-
induced airway hyperresponsiveness,
eosinophilia, Muc5ac expression
and IL-13 production. Thus, both
FCPTLR4 signalling and cytokine
signalling from T
H
2 cells are required
for the full induction of allergic
airway disease.
Taken together, these data sug-
gest that overwhelming exposure
to endogenous or exogenous pro-
teinases may lead to hyperactivation
of an antifungal pathway and may
drive allergic airway inflammation
through both fibrinogen and TLR4-
dependent and TLR4-independent
pathways.
Olive Leavy
ASTHMA AND ALLERGY
A fibrinogen root to airway inflammation
ORIGINAL RESEARCH PAPER Millien, V. O. et al.
Cleavage of fibrinogen by proteinases elicits
allergic responses through Toll-like receptor 4.
Science 341, 792796 (2013)
exogenous and
endogenous
proteinases
can cleave
fibrinogen to
generate TLR4
ligands that
prime innate
immune cells
to respond to
IL-13
G
E
T
T
Y
/
P
h
o
t
o
d
i
s
c
RESEARCH HI GHLI GHTS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013
Nature Reviews Immunology | AOP, published online 16 September 2013; doi:10.1038/nri3538
2013 Macmillan Publishers Limited. All rights reserved
Nucleic acids from pathogens are
important activators of innate
immune responses, but inappropri-
ate responses to self nucleic acids
can lead to autoimmune disease.
This study now shows that oxidative
damage to DNA potentiates its detec-
tion by the immune system, which
has implications for both self and
non-self recognition.
A series of experiments showed
that ultraviolet irradiation of DNA
has a dose-dependent effect on
type I interferon (IFN) production
by immune and non-immune cells
exposed to this DNA in mice and
humans. The authors observed the
generation of intracellular reac-
tive oxygen species (ROS) in cells
exposed to ultraviolet radiation,
which resulted in a dose-dependent
increase in oxidative damage to
DNA, as measured by 8-hydroxygua-
nine (8-OHG), which is the oxidation
product of guanine. The 8-OHG
content of DNA correlated with its
ability to stimulate IFN production.
The authors went on to investigate
how ROS production by myeloid cells
might affect the detection of ingested
pathogen DNA. The type I IFN
response of mouse dendritic cells that
had been transfected with bacterial
or viral genomic DNA was enhanced
in a dose-dependent manner when
the DNA was first incubated with
hydrogen peroxide. So, in addition to
directly damaging pathogens, ROS
production might also indirectly
facilitate their immune detection.
The oxidative burst of neutrophils
can be followed by the expulsion of
self genomic DNA to form neutro-
phil extracellular traps (NETs). NET
DNA had a higher 8-OHG content
than control neutrophil DNA and
induced higher levels of type I IFN
production when transfected into
monocytes. At high concentrations,
oxidized self DNA induced a typeI
IFN response in monocytes even
when no mechanism of intracellular
delivery was used, which indicates
that extracellular oxidized DNA that
has been released from neutrophils
might stimulate surrounding
immune cells.
The immunostimulatory capac-
ity of oxidized self DNA could
have implications for systemic
lupus erythematous (SLE), which
involves antibody responses
against self nucleic acids. MRLlpr
mice (which are a model of SLE)
produced higher levels of typeI
IFN in response to intravenous
injection of ultraviolet-damaged self
DNA compared with control mice,
presumably as a result of increased
antibody-mediated uptake of DNA.
This mechanism could account for
the ultraviolet photosensitivity of
patients with SLE. Indeed, injection
of ultraviolet-damaged DNA into
the ears of MRLlpr mice resulted
in skin lesions similar to those
found in patients with SLE.
Finally, the authors showed that
the recognition of oxidized DNA
involves signalling through the
cytosolic cyclic GAMP synthase
(cGAS)STING (stimulator of
interferon genes protein) pathway.
8-OHG-containing DNA was
degraded more slowly by the
cytosolic exonuclease TREX1 than
unmodified DNA, which resulted in
the accumulation of oxidized DNA
and therefore increased signalling
through the cGASSTING pathway.
Together these results show
that not only foreign DNA but also
self DNA induces a potentiated
immune response after oxidation,
which shows that it is a bona fide
damage-associated molecular pattern.
Kirsty Minton
PATTERN RECOGNI TI ON RECEPTORS
DNA damage drives detection
ORIGINAL RESEARCH PAPER Gehrke, N. et al.
Oxidative damage of DNA confers resistance to
cytosolic nuclease TREX1 degradation and
potentiates STING-dependent immune sensing.
Immunity http://dx.doi.org/10.1016/j.immuni.
2013.08.004 (2013)
FURTHER READING Broz, P. & Monack, D. M.
Newly described pattern recognition receptors
team up against intracellular pathogens.
Nature Rev. Immunol. 13, 551565 (2013)
The 8-OHG
content
of DNA
correlated with
its ability to
stimulate IFN
production
Watercolour image courtesy of Christiane Ahlemeyer, Institute of Clinical Chemistry
and Clinical Pharmacology, University of Bonn, Germany.
RESEARCH HI GHLI GHTS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013
Nature Reviews Immunology | AOP, published online 16 September 2013; doi:10.1038/nri3539
2013 Macmillan Publishers Limited. All rights reserved
Effector and memory T cells show an
increased propensity to traffic back
to the tissue sites in which they were
originally activated. Dendritic cells
(DCs) have been shown to drive the
selective trafficking of T cells to the
skin and small intestine by inducing
T cell expression of tissue-specific
homing molecules. Two studies now
describe the mechanisms by which
lung DCs can promote T cell migra-
tion to both the lungs and the small
intestine.
Luster and colleagues found that
CD4
+
T cells homed more efficiently
to the lungs if they were activated by
lung DCs than if they were activated
by DCs from various other tissue
sites. They showed that lung DCs
promote homing to the lungs
partly through the induction of
CC-chemokine receptor 4 (CCR4)
expression by T cells, which enables
T cell recruitment to the lungs in
response to CC-chemokine ligand17
(CCL17) and CCL22.
The physiological relevance of
this process was shown by transfer-
ring antigen-specific wild-type or
CCR4-deficient CD4
+
T cells that
had been activated by DCs from
different tissue sites into mice
infected with influenza virus.
Mice that received wild-type Tcells
activated by lung DCs showed
reduced weight loss and cleared
their infection more rapidly than
mice that received CCR4-deficient
Tcells activated by lung DCs or than
mice that received wild-type Tcells
activated by DCs not from the lungs.
Thus, DC-mediated imprinting
of T cell homing to the lungs is
important to drive more effective
immune responses to pathogens in
the airways.
Mehandru and colleagues
compared the ability of DCs from
different tissues to imprint gut-
homing molecules on CD4
+
Tcells
and found that DCs from the lungs,
but not from the spleen or skin-
draining lymph nodes, could induce
T cell expression of the gut-homing
molecules CCR9 and 47 integrin
as efficiently as DCs from the
mesenteric lymph nodes (MLNs).
Similarly to what has been shown
for DCs from the MLNs, the induc-
tion of 47 integrin expression on
Tcells by lung DCs was dependent
on retinoic acid and transforming
growth factor--mediated signal-
ling. However, although imprinting
with intestinal homing molecules
has been reported to be exclusively
driven by CD103
+
intestinal DCs,
both CD103
+
and CD103

DCs from
the lungs induced CCR9 and 47
integrin expression on T cells.
In addition, the authors showed
that T cells activated in an antigen-
specific manner by lung DCs could
migrate to the intestinal lamina
propria. Furthermore, intranasal
immunization with antigens
expressed by the gastrointestinal
pathogen Salmonella enterica subsp.
enterica serovar Typhimurium was
more effective than subcutaneous
immunization in protecting mice
against subsequent enteric infection
with this pathogen.
Taken together, these studies
show that lung DCs can promote
immunity to mucosal pathogens
by driving T cell homing to the
lungs and intestines. They also add
further support to the much older
concept of a common mucosal
immune system.
Yvonne Bordon
MUCOSAL I MMUNOLOGY
Air miles for T cells
ORIGINAL RESEARCH PAPERS Mikhak, Z.,
Strassner, J. P. & Luster, A. D. Lung dendritic cells
imprint T cell lung homing and promote lung
immunity through the chemokine receptor CCR4.
J. Exp. Med. 210, 18551869 (2013) | Ruane, D. et al.
Lung dendritic cells induce migration of
protective T cells to the gastrointestinal tract.
J.Exp. Med. 210, 18711888 (2013)
N
P
G
lung DCs
can promote
immunity
to mucosal
pathogens by
driving Tcell
homing to
the lungs and
intestines
RESEARCH HI GHLI GHTS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013
Nature Reviews Immunology | AOP, published online 16 September 2013; doi:10.1038/nri3537
2013 Macmillan Publishers Limited. All rights reserved
Although it is not completely under-
stood, the pain that is associated with
acute bacterial infections is thought
to be secondary to activation of the
immune system. Clifford Woolf and
colleagues now show that bacteria
directly activate pain responses
by triggering nociceptor neurons.
Furthermore, activation of these
sensory neurons by bacteria leads to
the release of neuropeptides that can
suppress immune responses to the
infection.
The authors aimed to inves-
tigate the mechanisms by which
Staphylococcus aureus (which is a
common cause of wound infections)
induces pain. Pain thresholds were
assessed by infecting mice with
S.aureus and then measuring their
sensitivity to mechanical, heat- or
cold-associated stress. Mice showed
signs of hyperalgesia within 1 hour
of S. aureus infection, with the pain
response peaking 6 hours after infec-
tion and beginning to decrease after
24 hours. Surprisingly, the kinetics of
the pain response did not correlate
with the kinetics of tissue swelling
or with the kinetics of immune cell
recruitment. By contrast, bacterial
loads in the tissue closely correlated
with pain hypersensitivity, which
suggests that bacteria may directly
interact with sensory neurons. In
keeping with this idea, pain percep-
tion during S.aureus infection was
not decreased in mice deficient for
Toll-like receptor signalling compo-
nents, or in mice lacking neutrophils,
monocytes or lymphocytes. Indeed,
antibody-mediated depletion of neu-
trophils or monocytes led to higher
bacterial loads and to increased pain
hypersensitivity.
To test whether bacteria directly
induce pain, the authors applied
heat-killed bacteria to dorsal root
ganglia (DRG) sensory neurons.
Various strains of heat-killed bac-
teria, including S. aureus, induced
robust calcium fluxes in DRG
neurons. Additional experiments
suggested that bacterial N-formylated
peptides trigger mechanical (but
not heat) hyperalgesia by activating
nociceptors. Furthermore, they
showed that -haemolysin, which
is a pore-forming toxin produced
by S.aureus, binds to and activates
a subset of nociceptor neurons and
directly induces mechanical, heat and
cold hypersensitivity.
The authors proceeded to examine
how the activation of nociceptors
by bacteria can modulate immune
responses. Conditional ablation of
nociceptors abolished pain responses
during S. aureus infection and this
was associated with increased local
inflammation, despite there being
similar bacterial tissue loads in noci-
ceptor-deficient and control mice.
Microarray analyses showed that
receptors for the neuropeptides calci-
tonin gene-related peptide (CGRP),
galanin and somatostatin are highly
expressed by neutrophils, monocytes
and macrophages. Furthermore,
the treatment of macrophages with
these neuropeptides suppressed their
production of tumour necrosis factor
in response to S. aureus. Finally, the
exposure of DRG neurons to super-
natant from S. aureus cultures or to
-haemolysin promoted the release of
CGRP in a dose-dependent manner.
This study shows that bacte-
rial products can directly activate
nociceptors to induce pain and
the release of immunosuppressive
neuropeptides. The authors suggest
that pathogenic bacterial strains have
evolved to trigger nociceptors in
order to suppress the host immune
system and increase their own spread
in infected tissues.
Yvonne Bordon

NEUROI MMUNOLOGY
Pain blame it on the bug, eh?
ORIGINAL RESEARCH PAPER Chiu, I. M. et al.
Bacteria activate sensory neurons that modulate
pain and inflammation. Nature http://dx.doi.org/
10.1038/nature12479 (2013)
bacterial
products
can directly
activate
nociceptors
to induce
pain and
the release
of immuno-
suppressive
neuropeptides
PIXTAL
RESEARCH HI GHLI GHTS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 1
Nature Reviews Immunology | AOP, published online 2 September 2013; doi:10.1038/nri3533
2013 Macmillan Publishers Limited. All rights reserved
Following injury, the regeneration of
skeletal muscle is supported by
muscle-associated macrophages.
These macrophages initially show
pro-inflammatory characteristics
and stimulate myoblast prolif-
eration, but later they switch to an
anti-inflammatory phenotype that
supports the terminal differentia-
tion of myoblasts and the growth
of new muscle fibres. Mounier et
al. now report that the metabolic
regulator AMP-activated protein
kinase (AMPK) is crucial for such
macrophage skewing during muscle
regeneration.
AMPK senses cellular energy
levels by monitoring ADP:ATP and
AMP:ATP ratios, and it regulates
many metabolic processes that
are involved in cellular energy
homeostasis. Previous studies
have associated increased AMPK
activity with decreased inflamma-
tory responses in macrophages, so
Mounier et al. reasoned that AMPK
could be involved in the macrophage
phenotype-switching that occurs
during muscle repair. As AMPK1 is
the only catalytic subunit of AMPK
that is expressed in macrophages,
they examined muscle regeneration
in AMPK1-deficient mice.
Histological analysis showed that
skeletal muscle repair was delayed in
AMPK1-deficient mice and that this
was associated with higher numbers
of necrotic myofibres, a decrease in
the size of new myofibres and lower
overall muscle mass. Closer compari-
son of the reparatory process in wild-
type and AMPK1-deficient animals
showed that AMPK1 deficiency
does not alter muscle cell homeosta-
sis or the fusion of muscle cells into
myofibres. Furthermore, mice in
which AMPK1 was deleted under
the control of a myeloid-specific pro-
moter also showed impaired skeletal
muscle regeneration. Thus, AMPK1
expression by macrophages, but not
by other cells in the tissue, is needed
for skeletal muscle repair.
The authors next investigated
the differentiation of bone marrow-
derived macrophages from AMPK1-
deficient mice in vitro. Notably,
although AMPK1-deficient macro-
phages showed normal acquisition of
a pro-inflammatory M1phenotype,
they showed impaired acquisition of
an anti-inflammatory M2 phenotype,
as determined by lower levels of
transforming growth factor-, CD163
and CD206 and higher expression of
inducible nitric oxide synthase. High
oxygen consumption rates have been
associated with the M2 macrophage
phenotype but, whereas wild-type
macrophages showed a marked
increase in oxygen consumption
rates under M2-polarizing condi-
tions, this increase was not seen in
AMPK1-deficient macrophages.
Further experiments in vivo
confirmed that AMPK1-deficient
macrophages do not switch from
an M1 to an M2 phenotype during
muscle repair. Although wild-type
macrophages showed transition to an
anti-inflammatory and pro-
reparatory phenotype following
phagocytosis of apoptotic and necrotic
myoblasts, AMPK1-deficient mac-
rophages showed impaired phagocytic
responses and a failure to undergo
this phenotypic switching. Finally,
wild-type macrophages treated with
an inhibitor of calcium/calmodulin-
dependent protein kinase kinase 2
(CAMKK2), which is an upstream
activator of AMPK, also failed to
switch to an M2phenotype follow-
ing the phagocytosis of apoptotic
myoblasts.
These data suggest that the phago-
cytosis of cellular debris by inflam-
matory macrophages in an injured
tissue is associated with activation of
the metabolic regulator AMPK. This
enzyme then functions as a molecular
switch to promote the acquisition
of the pro-reparatory macrophage
phenotype that is needed for tissue
regeneration.
Yvonne Bordon
MACROPHAGES
Metabolic master prompts a change of tack
ORIGINAL RESEARCH PAPER Mounier, R. et. al.
AMPK1 regulates macrophage skewing at the
time of resolution of inflammation during skeletal
muscle regeneration. Cell Metab. 18, 251264
(2013)
AMPK1
expression by
macrophages
is needed for
skeletal muscle
repair
T
H
I
N
K
S
T
O
C
K
RESEARCH HI GHLI GHTS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013
2013 Macmillan Publishers Limited. All rights reserved
SI GNALLI NG
NF-B signalosomes on the ER
This study shows that stimulation of a range of innate and
adaptive immune receptors results in the accumulation of
ubiquitylated components of the nuclear factor-B (NF-B)
signalling cascade on the cytoplasmic leaflet of the endoplasmic
reticulum (ER) membrane. ER membrane fractions from
stimulated cells could activate inhibitor of NF-B kinase (IKK)
in a cell-free system, which indicates that the ER membrane
anchors a signalosome that is sufficient to propagate NF-B
signalling. The ER-resident protein metadherin was shown to
associate with ubiquitylated NF-B signalling components,
and knockdown of metadherin in both B and T cells inhibited
the accumulation of ubiquitylated signalling components on the
ER and selectively decreased NF-B activation downstream of
various immune receptors. The results support a role for the ER
in outside-in signalling.
ORIGINAL RESEARCH PAPER Alexia, C. et al. The endoplasmic reticulum acts as a platform
for ubiquitylated components of nuclear factor B signaling. Sci. Signal. 291, ra79 (2013)
REPRODUCTI VE I MMUNOLOGY
How NK cells affect pregnancy outcome
Interactions between killer cell immunoglobulin-like
receptors (KIRs) expressed by maternal decidual natural
killer (NK) cells and HLA-C molecules expressed by fetal
trophoblast cells affect the extent of trophoblast invasion
of the maternal blood supply by unknown mechanisms. This
study reports that decidual NK cells expressing the activating
receptor KIR2DS1 produce greater amounts of granulocyte
macrophage colony-stimulating factor (GM-CSF) in response
to HLA-C2 than NK cells expressing the inhibitory receptor
KIR2DL1 or those expressing both KIR2DS1 and KIR2DL1.
Trophoblast cells were shown to express GM-CSF receptor-,
and stimulation with GM-CSF increased their migration
through fibronectin-coated transwells. The authors suggest
that women expressing KIR2DL1, with or without KIR2DS2,
who carry a HLA-C2
+
fetus will have decreased GM-CSF
production in the decidua and hence decreased trophoblast
invasion, which correlates with pregnancy disorders such as
pre-eclampsia and fetal growth restriction.
ORIGINAL RESEARCH PAPER Xiong, S. et al. Maternal uterine NK cell-activating receptor
KIR2DS1 enhances placentation. J. Clin. Invest. http://dx.doi.org/10.1172/JCI68991 (2013)
I MMUNE REGULATI ON
IL-27 induces immunosuppressive DCs
This study shows that, instead of directly affecting T cells
as was previously thought, interleukin-27 (IL-27) modulates
dendritic cells (DCs) to suppress T cells. Pretreatment of
DCs with IL-27 decreased their ability to promote the
differentiation of Thelper1 (T
H
1) and T
H
17 cells and increased
their ability to generate regulatory T cells. Consistent with the
increased induction of pathogenic T
H
cell subsets, chimeric
mice containing IL-27 receptor -chain (IL-27RA)-deficient
DCs developed faster onset and more severe experimental
autoimmune encephalomyelitis (EAE) than control mice.
Microarray analysis of IL-27-treated DCs showed upregulation of
expression of CD39, which reduced extracellular concentrations
of ATP and suppressed nucleotide-dependent activation of
NLRP3 (NOD-, LRR- and pyrin domain-containing protein 3).
Finally, vaccination with IL-27-conditioned DCs suppressed
EAE and reduced epitope spreading.
ORIGINAL RESEARCH PAPER Mascanfroni, I. D. et al. IL-27 acts on DCs to suppress
the T cell response and autoimmunity by inducing expression of the immunoregulatory
molecule CD39. Nature Immunol. http://dx.doi.org/10.1038/ni.2695 (2013)
IN BRIEF
RESEARCH HI GHLI GHTS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013
2013 Macmillan Publishers Limited. All rights reserved
The textbook tenet is that B cells arise
in the bone marrow. But FredAlt,
Duane Wesemann and their col-
leagues now show that B cells can
also develop in the mouse gut for a
short time period after birth.
The authors used recombination
activating gene 2 (Rag2)-reporter
mice, in which the Rag2 gene is
fused to the gene that encodes
green fluorescent protein (GFP), to
mark immature B cells undergoing
RAG2-mediated generation of B cell
receptor repertoires. Analysis of these
mice showed that approximately 3%
of total CD19
+
B cells in the small
intestinal lamina propria expressed
RAG2GFP. The RAG2
+
B lineage
cells were mainly located near to the
bases of the villi, whereas mature
Bcells were distributed throughout
the lamina propria but were not
found in the mesenteric lymph nodes
or among intraepithelial lymphocytes
and only in very low frequencies in
the large intestinal lamina propria.
Interestingly, the numbers of lamina
propria RAG2
+
Blineage cells gradu-
ally increased after birth, peaking at
about 1823 days, before decreasing
to undetectable levels by postnatal
day 35.
The RAG
+
B lineage cell
populations in the bone marrow
comprise pro-Bcells (Ig

Ig

),
pre-B cells (Ig
+
Ig

) and immature
B cells undergoing receptor editing
(Ig
+
Ig
+
). Similar relative levels of
these three subsets were found in the
gut and the bone marrow. Further
investigation of repertoire diversity
indicated that the immunoglobulin
heavy chain (IgH) repertoires of the
gut and the bone marrow cells were
indistinguishable, but the immuno-
globulin light chain (IgL) repertoires
were distinctive. The authors
proposed that the lamina propria IgL
repertoires were generated by receptor
editing in RAG2
+
immature B cells
in response to commensal micro-
organisms. In support of this idea,
colonization of germ-free mice with
commensal microorganisms led to
increases in RAG1 and RAG2 expres-
sion and increased the percentages
of pro-Bcells relative to total Bcells
in the gut and the bone marrow.
Moreover, there was a commensal
bacteria-dependent increase in Ig
usage a marker of receptor editing
specifically in the lamina propria.
So, B cell development and diver-
sification can occur in the intestinal
mucosa in response to colonization of
the intestinal microbiota at weaning.
Whether this process enhances overall
antibody diversity or whether it helps
to eliminate antibody reactivity to
commensal bacteria and self antigens
will require further study.
Lucy Bird
B CELL DEVELOPMENT
A window of opportunity
ORIGINAL RESEARCH PAPER Wesemann, D. R.
et al. Microbial colonization influences early
B-lineage development in the gut lamina propria.
Nature 501, 112115 (2013)
FURTHER READING Schlissel, M. B cell
development in the gut. Nature 501, 4243 (2013)
B cell
development
and
diversification
can occur in
the intestinal
mucosa in
response to
colonization of
the intestinal
microbiota
G
E
T
T
Y
/
P
h
o
t
o
d
i
s
c
RESEARCH HI GHLI GHTS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013
2013 Macmillan Publishers Limited. All rights reserved
Atherosclerosis results from a maladaptive inflamma-
tory response that is initiated by the intramural reten-
tion of cholesterol-rich, apolipoprotein B-containing
lipoproteins in susceptible areas of the arterial vascu-
lature
1
. Lipoproteins that are sequestered in the arte-
rial wall are susceptible to various modifications (such
as oxidation, enzymatic and non-enzymatic cleavage,
and aggregation), which render these particles pro-
inflammatory and which induce the activation of the
overlying endothelium. The ensuing immune response
is mediated by the recruitment of monocyte-derived
cells into the subendothelial space, where these cells dif-
ferentiate into mononuclear phagocytes that ingest the
accumulated normal and modified lipoproteins, which
transforms them into the cholesterol-laden foam cells.
Foam cells, typically classified as a type of macrophage,
persist in plaques, which promotes disease progression.
Although macrophage clearance of lipoproteins is likely
to be beneficial at the outset of this immune response,
there is little negative feedback following uptake and
thus these cells become grossly engorged with lipids.
The resulting dysregulation of lipid metabolism alters
the macrophage phenotype and compromises crucial
immune functions.
Notably, macrophages that accumulate in athero-
sclerotic plaques seem to have a diminished capacity
to migrate, which contributes to their failure to resolve
inflammation and to the progression of these lesions
to more advanced, complex plaques in which other
immune cell subsets and vascular smooth muscle
cells participate in the inflammatory process
2
. In these
advanced plaques, macrophages continue to be major
contributors to the inflammatory response through
their secretion of pro-inflammatory mediators (includ-
ing chemokines, cytokines and reactive oxygen and
nitrogen species) and matrix-degrading proteases, and
through their eventual death by necrosis or apoptosis.
Dying macrophages release their lipid contents and
tissue factors, which leads to the formation of a pro-
thrombotic necrotic core. This necrotic core is a key
component of unstable plaques and contributes to their
rupture and the ensuing intravascular blood clot that
underlies myocardial infarction andstroke.
Although many cell types, including endothelial cells,
monocytes, dendritic cells (DCs), lymphocytes, eosino-
phils, mast cells and smooth muscle cells, contribute
to the formation of atherosclerotic plaques, foam cells
are so central to the pathophysiology of atherosclerosis
that emphasis has long been placed on understanding
the mechanisms of monocyte recruitment into plaques
and on identifying strategies to reduce monocyte influx
to retard plaque progression. However, it has become
apparent that the recruitment of monocytes and other
leukocytes into the artery may also be crucial to promote
atherosclerosis regression and inflammation resolution
3
.
In addition, studies in some models of atherosclerosis
regression have shown that macrophage retention can
be reversed
46
, which led to the identification of path-
ways that promote macrophage accumulation in, or
egress from, the inflamed plaque. These advances have
shown that both the quantity and the phenotype of mac-
rophages influence the inflammatory state of the plaque,
and have potentially identified new targets for plaque
intervention. In this Review, we discuss the key roles of
Department of Medicine,
Leon H.Charney Division of
Cardiology, Marc and
RutiBell Vascular Biology
and Disease Program,
NewYork University School
of Medicine, New York,
New York 10016, USA.
Correspondence to
K.J.M.and E.A.F.
e-mails:
kathryn.moore@nyumc.org;
edward.fisher@nyumc.org
doi:10.1038/nri3520
Published online
2 September 2013
Foam cells
Macrophages in the arterial
wall that ingest oxidized
low-density lipoprotein and
assume a foamy appearance.
These cells secrete various
substances that are involved in
plaque growth.
Myocardial infarction
An episode of acute cardiac
ischaemia that leads to death
of heart muscle cells. It is
usually caused by a thrombotic
atherosclerotic plaque.
Macrophages in atherosclerosis:
a dynamic balance
Kathryn J.Moore, Frederick J.Sheedy and Edward A.Fisher
Abstract | Atherosclerosis is a chronic inflammatory disease that arises from an imbalance
in lipid metabolism and a maladaptive immune response driven by the accumulation of
cholesterol-laden macrophages in the artery wall. Through the analysis of the progression
and regression of atherosclerosis in animal models, there is a growing understanding that
the balance of macrophages in the plaque is dynamic and that both macrophage numbers
and the inflammatory phenotype influence plaque fate. In this Review, we summarize
recently identified pro- and anti-inflammatory pathways that link lipid and inflammation
biology with the retention of macrophages in plaques, as well as factors that have the
potential to promote their egress from these sites.
REVIEWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 709
2013 Macmillan Publishers Limited. All rights reserved
Atherosclerosis regression
A decrease in atherosclerotic
plaque size that is typically
accompanied by a reduction in
lipid levels, immune cells and
inflammatory gene expression.
macrophages in the initiation, progression and resolu-
tion of atherosclerotic inflammation, with a focus on
how the dynamics of macrophage recruitment, egress
and death alter the fate of theplaque.
Circulating monocytes and their recruitment
Hypercholesterolaemia and monocytosis. Hyper-
cholesterolaemia is associated with increased numbers
of circulating monocytes in mice, pigs and rabbits
7,8
. In
apolipoprotein E-deficient (Apoe
/
) mice, the numbers of
circulating monocytes are ~50% higher than in wild-type
mice
9,10
. How does hypercholesterolaemia cause mono-
cytosis? Studies using mice have shown that cholesterol
enrichment of haematopoietic stem and progenitor cells
(HSPCs; precursors of monocytes and neutrophils)
increases their expression of the common -subunit of
the interleukin-3 (IL-3) and the granulocyte/macrophage
colony-stimulating factor (GM-CSF) receptor and
thus, HSPC proliferation
11
. Notably, the expression of
factors that promote cholesterol efflux (high-density
lipoprotein (HDL) and APOE) in hypercholesterolemic
mouse models corrected HSPC proliferation.
Circulating monocytes in mice consist of two
major subsets, LY6C
hi
and LY6C
low
monocytes (BOX1).
Interestingly, the monocytosis in hypercholesterolemic
mice primarily derives from an increase in the more
inflammatory LY6C
hi
subset, which constitutes the
majority of cells recruited to progressing atherosclerotic
plaques and which is thought to be the source of the
M1 macrophages (also known as classically activated
macrophages) that are found in the plaques
911
. The
basis for this cell bias has been postulated to be due to
a hypercholesterolaemia-induced impairment of a pro-
cess in which LY6C
hi
cells are converted to LY6C
low
cells
9
;
however, this remains an area of active investigation.
Recruitment of monocytes into athero-prone arterial
sites. The early steps of atherogenesis have been the
subject of numerous reviews (for example, REFS1214)
and will only be briefly covered here. Atherosclerotic
plaques are not randomly distributed, but tend to form
at the inner curvatures and branch points of arteries,
where laminar flow is either disturbed or insufficient to
maintain the normal, quiescent state of the endothelium.
Box 1 | Characteristics of monocyte and macrophage subsets
LY6C
hi
monocytes
Express high levels of CC-chemokine receptor 2
Thought to be pro-inflammatory because of their recruitment to sites of inflammation, including to atherosclerotic plaques
Normally represent 50% of monocytes in mice, but their frequency is increased in hyperlipidaemia
Thought to correspond to the CD14
+
CD16

monocyte subset in humans, which represent 95% of monocytes in humans


Proposed to be precursors of M1 macrophages
LY6C
low
monocytes
Express high levels of CX
3
C-chemokine receptor 1
Thought to patrol the vasculature in a homeostatic function
Thought to correspond to the CD14
low
CD16
+
monocyte subset in humans
Proposed to be precursors of M2 macrophages
M1 macrophages
Classical activation by lipopolysaccharide (or by other Toll-like receptor ligands) and interferon-
Enriched in progressing plaques
Secrete pro-inflammatory cytokines, such as interleukin-1 (IL-1), IL-12 and tumour necrosis factor
Produce high levels of inducible nitric oxide synthase and nitric oxide
Express the pro-inflammatory transcription factors nuclear factor-B, activator protein 1 and hypoxia-inducible factor 1
Express MHC classII molecules and the co-stimulatory molecules CD80 and CD86
M2 macrophages
Alternative activation by IL-4 and IL-13
Enriched in regressing plaques
High endocytic activity
Take up and oxidize fatty acids
Secrete anti-inflammatory cytokines, such as IL-1 receptor antagonist and IL-10
Express high levels of arginase 1 and have increased secretion of collagen, which promotes tissue repair
Express the transcription factors Krppel-like factor 4, peroxisome proliferator activated receptor- and signal
transducer and activator of transcription 6 (STAT6)
Express CD163, mannose receptor 1 (also known as CD206) and FIZZ1
Mox macrophages
Induced by oxidized phospholipids and nitrosylated fatty acids
Enriched in progressing plaques
Express high levels of reactive oxygen species and haem oxygenase 1, and the transcription factor nuclear erythroid 2-
related factor 2 (NRF2)
REVI EWS
710 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Retention
molecules
Netrin 1
Semaphorin 3E
Cytokines
IL-1
IL-6
TNF
Chemokines
CCL2
CCL5
CXCL1
Nature Reviews | Immunology
T cell
DC
Oxidized
LDL
LDL
SR-A1
Foam cell Macrophage
CCR1
CX
3
CR1
CX
3
CL1
CXCL1
and CCL5
P-selectin
PSGL1
Crawling
(patrolling)
Capture Rolling Arrest
LDL
Endothelial
cell
Extravasation
ICAM1
LFA1
LFA1
VCAM1 ICAM1
VLA4
20%
PSGL1
low
LFA1
hi
GR1

LY6C
low
PSGL1
hi
LFA1
low
80%
Media
Lumen
Intima
CCR5
CD36
GR1
+
LY6C
hi
monocyte GR1

LY6C
low
monocyte
Leukocyte adhesion cascade
The key steps that are involved
in leukocyte adhesion to the
endothelium. These include
rolling (which is mediated by
selectins), activation (which is
mediated by chemokines) and
arrest (which is mediated by
integrins). Recent additional
steps have been defined that
include capture (also known
as tethering), slow rolling,
adhesion strengthening and
spreading, intravascular
crawling, and paracellular and
transcellular transmigration.
Firm adhesion
The interactions of rolling
leukocytes with chemokines
or lipid mediators, such
as leukotriene B4, at the
endothelial surface leads
to the activation of leukocyte
integrins another family of
adhesion molecules. After
they are activated, integrins
mediate the high-affinity
adhesive interactions between
leukocytes and endothelial
cells, which results in the
arrest and firm adhesion
of rolling leukocytes.
This activation of the endothelium leads to increased
permeability to lipoproteins and an accumulation of
extracellular matrix proteins that cause a poorly under-
stood diffuse intimal thickening and the retention of the
atherogenic APOB lipoproteins. The arterial intima at
these athero-prone sites contains an increased number
of myeloid cells that have features ofDCs
15
.
The activation of the endothelium also promotes the
recruitment of circulating monocytes (FIG.1). In addi-
tion to the bone marrow origin of these monocytes, it
has recently been recognized that splenic HSPCs can
be an extramedullary myelopoietic source of monocytes,
which are mobilized to inflammatory sites, including to
atherosclerotic plaques
16
. The steps that regulate mono-
cyte entry into the arterial intima are apparently inde-
pendent of the source of the cells and depend on the
upregulation on activated endothelial cells of molecules
that mediate the arrest of circulating monocytes by the
leukocyte adhesion cascade
17
. The capture and rolling
phases of this cascade depend on the immobilization
of chemokines, particularly CC-chemokine ligand 5
(CCL5) and CXC-chemokine ligand 1 (CXCL1), on
endothelial cell glycosaminoglycans, and on P-selectin,
which is expressed on the luminal side of endothelial
cells. Very recent results have shown that the arrest of
LY6C
hi
monocytes through CCL5 depends not only on
its interaction with CC-chemokine receptor 5 (CCR5)
but also on its interaction with CCR1 (REF.18). Vascular
cell adhesion molecule 1 (VCAM1) and intercellu-
lar adhesion molecule 1 (ICAM1), which bind to the
integrins very late antigen 4 (VLA4; also known as
41 integrin) and lymphocyte function-associated anti-
gen1 (LFA1; also known as L2 integrin), respectively,
are important for the firm adhesion of monocytes to the
luminal surface of the endothelium. Comparatively more
LFA1 is expressed by LY6C
low
cells than by LY6C
hi
cells,
which may underlie the greater tendency of LY6C
low

cells to adhere to, but not to enter, the vasculature
19
.
Figure 1 | Mechanisms regulating monocyte recruitment and accumulation in plaques. Hyperlipidaemia
increases the number of GR1
+
LY6C
hi
monocytes, which constitute 80% of the monocytes recruited to mouse
atherosclerotic plaques, with the remainder being the GR1

LY6C
low
patrolling monocytes. These monocyte subsets
use different chemokinechemokine receptor pairs to infiltrate the intima, which is facilitated by endothelial adhesion
molecules, including selectins, intercellular adhesion molecule 1 (ICAM1) and vascular adhesion molecule 1 (VCAM1).
The recruited monocytes differentiate into macrophages or dendritic cells (DCs) in the intima, where they interact
with atherogenic lipoproteins. Macrophages avidly take up native and modified (for example, oxidized) low-density
lipoprotein (LDL) via macropinocytosis or scavenger receptor-mediated pathways (including via scavenger receptor A1
(SR-A1) and CD36), which results in the formation of the foam cells that are a hallmark of the atherosclerotic plaque.
These foam cells secrete pro-inflammatory cytokines (including interleukin-1 (IL-1), IL-6, and tumour necrosis factor
(TNF)) and chemokines (such as CC-chemokine ligand 2 (CCL2), CCL5 and CXC-chemokine ligand 1 (CXCL1)), as well
as macrophage retention factors (such as netrin 1 and semaphorin 3E) that amplify the inflammatory response. CX
3
CL1,
CX
3
C-chemokine ligand 1; CX
3
CR1, CX
3
C-chemokine receptor 1; LFA1, lymphocyte function-associated antigen 1;
PSGL1, P-selectin glycoprotein ligand 1; VLA4, very late antigen 4.
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 711
2013 Macmillan Publishers Limited. All rights reserved
Pattern recognition
receptors
(PRRs). Host receptors (such
as Toll-like receptors) that can
sense pathogen-associated or
damage-associated molecular
patterns and that can initiate
signalling cascades (which
involve activation of nuclear
factor-B) that lead to an
innate immune response.
The transmigration of monocytes across the endothe-
lium into plaques is mediated by chemokines that are
secreted by endothelial cells, intimal macrophages and
smooth muscle cells. Although several chemokines
have been implicated in atherosclerosis
20
, the three
major chemokine receptorchemokine pairs that
are thought to be involved in monocyte transmigra-
tion are CCR2CCL2, CX
3
C-chemokine receptor 1
(CX
3
CR1)CX
3
C-chemokine ligand 1 (CX
3
CL1)
and CCR5CCL5 (REF. 10). Indeed, the elimination
of these three chemokine axes led to a ~90% reduc-
tion in atherosclerosis in Apoe
/
mice
21
. In addition
to these chemokines, CD31 (also known as PECAM1;
an endothelial cell surface immunoglobulin-like adhe-
sion molecule) and VCAM1 may also have a role in
monocyte transmigration into atherosclerotic plaques.
It should also be noted that CCR2 and CX
3
CR1, in addi-
tion to their effects on transmigration, indirectly influ-
ence the number of monocytes that enter the plaques:
in particular, CCR2 is required for the extravasation
of LY6C
hi
cells from the bone marrow and CX
3
CR1
promotes their survival by inhibiting apoptosis
22,23
.
In addition to the factors described above, emerg-
ing evidence suggests that neuronal guidance cues are
involved in monocyte recruitment in atherosclerosis.
We recently reported that members of the netrin,
semaphorin and ephrin families are expressed by arte-
rial endothelial cells and that they are differentially
regulated under conditions that promote or protect
from atherosclerosis
24
; for example, the expression of
ephrinB2 is upregulated under pro-atherosclerotic
conditions and is a chemoattractant, which increases
leukocyte recruitment to athero-prone arterial sites
in the absence of additional chemokines
24
. By con-
trast, the expression of netrin 1 and semaphorin 3A,
which inhibit the chemokine-directed migration of
human and murine monocytes invitro, are decreased
in athero-prone regions, and the inhibition of these
molecules by blocking peptides in wild-type mice
increased leukocyte adhesion to the endothelium
24
.
Although further studies in hyperlipidemic mouse
models are needed, the data suggest that the coor-
dinated regulation of positive and negative guidance
cues facilitates leukocyte infiltration of the endothe-
lium. Notably, these neuronal guidance cues have
additional roles in atherosclerosis as they regulate the
chemostasis of plaque macrophages
25,26
(see below).
Therefore, overall, the recruitment of circulating
monocytes into plaques requires the integration of at
least three discrete processes, namely, their capture, roll-
ing and transmigration, and each step is regulated by
multiple, and sometimes overlapping, molecular factors.
The fates of these recruited monocytes in the plaques are
addressed in the sectionsbelow.
Foam cell formation in atherosclerosis
Lipoprotein uptake. Lipoprotein uptake by monocyte-
derived macrophages is thought to be one of the earliest
pathogenic events in the nascent plaque and results in
the development of foam cells (FIG.2). The mechanisms
of foam cell formation have been intensely studied
(reviewed in REF.27). Although macrophages can clear
APOB-containing lipoproteins through the low-density
lipoprotein (LDL) receptor, the expression of this recep-
tor is downregulated early during foam cell formation by
the increased cellular cholesterol levels. These observa-
tions led to the early hypothesis that lipoproteins must
become modified in the artery wall and that they must
be taken up by other mechanisms. Multiple means of
LDL modification have now been identified that facili-
tate cholesterol loading of macrophages invitro (FIG.2);
however, the physiologically relevant pathways of foam
cell formation invivo remain an area ofdebate.
A prevailing paradigm has been that increased
oxidative stress in the artery wall promotes modifica-
tions of LDL, which generates damage signals that
are recognized by pattern recognition receptors (PRRs)
on cells of the innate immune system. This hypothesis
is supported by the presence of oxidized LDL in both
human and mouse atheromas, and of natural antibodies
(predominantly IgM) that recognize oxidation-specific
epitopes of LDL
28
. A variety of mechanisms mediated by
enzymes (such as 12/15-lipoxygenase and myeloperoxi-
dase) and by free radicals (such as superoxide, hydro-
gen peroxide and nitric oxide) have been identified that
could promote LDL oxidation in the arterywall
28
, and
invitro preparations of such modified LDLs are avidly
endocytosed by macrophages
29,30
.
Scavenger receptors, which are a type of PRR
expressed by macrophages, have an important role in
atherosclerosis and were originally characterized by
their ability to recognize and process modified LDL
27
.
Numerous scavenger receptor family members
including scavenger receptor A1 (SR-A1; encoded by
MSR1), macrophage receptor with collagenous struc-
ture (MARCO; also known as SR-A2), CD36 (also
known as platelet glycoprotein 4), scavenger receptorB1
(SR-B1), lectin-like oxidized LDL receptor 1 (LOX1),
scavenger receptor expressed by endothelial cells 1
(SREC1) and scavenger receptor for phosphatidylserine
and oxidized LDL (SR-PSOX; also known as CXCL16)
can bind to oxidized LDL and can promote foam cell
formation
31
. SR-A1 and CD36 mediate 7590% of the
degradation of LDL that has been modified by acety-
lation or oxidation by macrophages invitro
29
. These
receptors internalize the lipoproteins and, in the late
endolysosomal compartment, the cholesteryl esters
of the lipoproteins are hydrolysed to free cholesterol
and fatty acids. Free cholesterol in the endolysosomal
compartment is then trafficked to the endoplasmic
reticulum (ER), where it undergoes re-esterification
by acetyl-coenzymeA:cholesterol acetyltransferase 1
(ACAT1) to cholesteryl fatty acid esters that provide
the foam of the foam cells
32
.
Combined deficiency of SR-A1 and CD36 reduced
foam cell formation in Apoe
/
mice; however, this effect
was incomplete, which suggests that there are addi-
tional mechanisms of macrophage cholesterol uptake
in vivo
33,34
. Despite this redundancy in cholesterol
uptake mechanisms, plaques in mice that are deficient
in both CD36 and APOE (Cd36
/
Apoe
/
mice) and in
mice that are deficient in SR-A1, CD36 and APOE
REVI EWS
712 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
ER stress
Apoptosis
Cytokines and
chemokines
NLRP3
inammasome
activation
Nature Reviews | Immunology
Lipid eux
Pro-inammatory
signalling
Lipoprotein uptake
Macro-
pinocytosis
Phagocytosis
aggregated LDL
SR-A1 LOX1
LDL
SR-B1 CD36
Oxidized LDL
Cholesterol-
rich lipid raf
VLDL
ABCA1
ABCG1
Nascent HDL
Lipid-poor
APOA1
Mature HDL
Acid
lipolysis
Lipophagy
Phagophore
Autophagosome
Lipolysis
Lipid
droplets
NCEH1
ACAT1
LXRRXR
ER
Nucleus
Lysosome
Endosome
Free
cholesterol
Cholesterol
crystals
Lysosomal
dysfunction
CD36
TLR4TLR6
TLR4
NF-B
(Msr
/
Cd36
/
Apoe
/
mice) have reduced signs of
inflammation, macrophage apoptosis and secondary
necrosis, which suggests that these scavenger receptors
have roles beyond lipid uptake
33,34
. Nevertheless, the
invivo relevance of oxidative processes in atheroscle-
rosis remains speculative. Several well-powered human
clinical trials of antioxidant vitamins, such as vitaminE
and vitaminC, have failed to show a reduction of cardio-
vascular events
35
, which encourages the field to consider
alternative mechanisms for foam cell formation.
Modification by various proteases and lipases that
are present in the intima can also mediate LDL modi-
fications, particularly the aggregation of LDL. The
extracellular matrix glycoproteins contribute to this
process by retaining the lipoproteins and by modulat-
ing the activity of various enzymes, including group IIA
secretory phospholipase A2 (PLA2G2A), PLA2G5 and
PLA2G10, as well as secretory sphingomyelin
27
. These
lipolytic enzymes produce modified forms of LDL
that are taken up via pathways that are independent of
Figure 2 | Mechanisms controlling macrophage lipoprotein uptake and efflux. Macrophages internalize native
low-density lipoprotein (LDL) and very low-density lipoprotein (VLDL) as well as oxidized lipoproteins in the plaque via
macropinocytosis, phagocytosis of aggregated LDL and scavenger receptor-mediated uptake (including by scavenger
receptor A1 (SR-A1), lectin-like oxidized LDL receptor 1 (LOX1), SR-B1 and CD36). The internalized lipoproteins and
their associated lipids are digested in the lysosome, which results in the release of free cholesterol that can travel to the
plasma membrane and be effluxed from the cell or to the endoplasmic reticulum (ER) membrane. In the ER, it can then be
esterified by acetyl-coenzyme A:cholesterol acetyltransferase 1 (ACAT1) and is ultimately stored in this form in cytosolic
lipid droplets. These stored lipids can be mobilized for efflux either via lipolysis by neutral cholesterol ester hydrolase 1
(NCEH1) or via lipophagy, which is a form of autophagy, resulting in the delivery of lipid droplets to lysosomes. The
accumulation of cellular cholesterol activates the liver X receptor (LXR)retinoid X receptor (RXR) heterodimeric
transcription factor that upregulates expression of the ATP-binding cassette subfamily A member 1 (ABCA1) and ABCG1.
This mediates the transfer of free cholesterol to lipid-poor apolipoprotein A1 (APOA1) to form nascent high-density
lipoprotein (HDL) or more lipidated HDL particles in which free cholesterol has been esterified and stored in the core of
the particle (known as mature HDL). Excessive free cholesterol accumulation can induce cholesterol crystal formation in
the lysosome to activate the NLRP3 (NOD-, LRR- and pyrin domain-containing 3) inflammasome, and may also interfere
with the function of the ER (inducing ER stress), which, if prolonged, results in cell death by apoptosis. In addition, lipid
rafts are enriched in sphingomyelin, which forms a complex with the free cholesterol. As the cholesterol content of lipid
rafts increases, pro-inflammatory Toll-like receptor 4 (TLR4) signalling is promoted, which can also be induced by oxidized
LDL, through a heterotrimeric complex composed of CD36TLR4TLR6. This signalling results in the activation of nuclear
factor-B (NF-B) and in the production of pro-inflammatory cytokines and chemokines.
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 713
2013 Macmillan Publishers Limited. All rights reserved
Pinocytosis
Also known as fluid-phase
endocytosis. A process of
engulfment of extracellular
fluid and its solutes.
It can be mediated by
an actin-dependent
mechanism that results in the
engulfment of large volumes
(macropinocytosis) or by other
mechanisms that result in the
engulfment of smaller volumes
(micropinocytosis).
Efferocytosis
The process of macrophage
clearance of apoptotic cells.
ATP-binding cassette
subfamily A member 1
(ABCA1). A member of a
superfamily of proteins that
transport various molecules
across extracellular and
intracellular membranes using
the energy of ATP hydrolysis.
Eukaryotic ABC genes are
classified in seven families,
from ABCA to ABCG, on the
basis of gene organization and
primary sequence homology.
Functional characterization can
be partly made by differential
sensitivity to inhibitory drugs.
Autophagy
An evolutionarily conserved
process in which acidic
double-membrane vacuoles
sequester intracellular contents
(such as damaged organelles
and macromolecules) and
target them for degradation,
through fusion to secondary
lysosomes.
NLRP3 inflammasome
A molecular complex
containing NLRP3 (NOD-, LRR-
and pyrin domain-containing 3)
and the adaptor molecule
ASC that controls the activity
of caspase 1. Formation of this
complex results in the cleavage
of the highly pro-inflammatory
cytokines pro-interleukin-1
(IL-1) and pro-IL-18, thereby
producing active IL-1 and
IL-18.
scavenger receptors
36
. Evidence from mouse models sup-
ports a role for PLA2 family members in atherosclerosis
progression
37
, and circulating PLA2 levels in humans
correlate with coronary artery disease risk
38,39
, which
identifies it as a promising therapeutic target, although
further validation is required.
Finally, although a role for native LDL in foam
cell formation was initially discounted, recent studies
have shown that, in the arterial intima, LDL under-
goes pinocytosis by macrophages when it is at concen-
trations similar to those that occur in hyperlipidemic
conditions, which results in foam cell formation
40
. This
receptor-independent endocytic pathway also delivers
cholesterol to the endolysosomal compartment and
stimulates cholesterol esterification. Thus, rather than
the originally envisioned single modification model
in which LDL and other APOB-containing lipoproteins
would be rendered atherogenic it is probable that
multiple, simultaneous pathways contribute to foam cell
formation invivo.
Defective cholesterol trafficking. Macrophage choles-
terol metabolism can become overwhelmed during
excessive cholesterol uptake, which results in patho-
logical processes. When stored in the cell as cholesteryl
ester, cholesterol is fairly inert; however, free cholesterol
can be toxic to cells. Enrichment of ER membranes
with free cholesterol can result in defective cholesterol
esterification by ACAT1 in macrophages, which pro-
motes the further accumulation of free cholesterol. In
addition, free cholesterol enrichment of cell membranes
can enhance inflammatory signalling from lipid rafts,
particularly Toll-like receptor (TLR) signalling and acti-
vation of nuclear factor-B (NF-B)
4143
. Furthermore,
trafficking of free cholesterol out of lysosomes may also
become defective in these macrophages, which consti-
tutes a barrier to cholesterol efflux and further amplifies
inflammation
44
. Such dysregulation in lipid metabolism
contributes to ER stress in macrophages, which, if pro-
longed and combined with other insults, can ultimately
result in apoptotic cell death
45
. Efficient clearance
of apoptotic cells by surrounding macrophages (the
process of efferocytosis) requires intact lipid metabo-
lism pathways (such as cholesterol esterification and
efflux) in the engulfing cell to deal with the ingested
lipids from the apoptotic bodies. Thus, as macrophage
lipid metabolism becomes dysregulated, the increase in
macro phage apoptosis combined with defective effero-
cytosis results in secondary necrosis and in the release
of cellular components and lipids that form the necrotic
core
46
. This feature of advanced atherosclerotic plaques,
along with the thinning of the fibrous cap, may increase
the vulnerability of plaques to rupture.
Lipid efflux. Cells respond to excessive lipid accumula-
tion by increasing the expression of pathways that pro-
mote the removal of cholesterol and other lipids from
the cell. In foam cells several macrophage transporters
facilitate the efflux of lipids including ATP-binding
cassette subfamily A member1 (ABCA1), ABCG1 and
SR-B1 (FIG. 2) although passive diffusion from the
plasma membrane also occurs
47
. ABCA1 promotes cho-
lesterol efflux to lipid-poor APOA1, which is the building
block of HDL, whereas ABCG1 promotes efflux to mature
HDL particles. The genes encoding ABCA1 and ABCG1
are transcriptionally upregulated in response to elevated
cellular cholesterol levels by liver X receptors (LXRs),
which are ligand-activated nuclear receptors that function
as sterol sensors; for example, LXR activation by choles-
terol derivatives (such as oxysterols) or by desmosterol
(which is a molecule similar to cholesterol)
48
, promotes
macrophage cholesterol efflux via ABCA1 and ABCG1
and also has anti-inflammatory effects
49
. Thus, synthetic
LXR agonists have been actively investigated for the
treatment of atherosclerosis.
In addition, autophagy has a crucial role in mac-
rophage cholesterol efflux: lipid droplets in macrophages
and other cell types are targeted to and hydrolysed
by the autophagy machinery in a process known as
lipophagy
50
. Fusion of the autophagosome with the
lysosome degrades cholesteryl esters and makes free
and modified cholesterol available for efflux through an
ABCA1-dependent pathway
51
(FIG.2). The protective role
of autophagy has been shown in studies in Apoe
/
mice
in which the deletion of components of the autophagy
machinery enhanced atherosclerosis
52,53
. Furthermore,
autophagy regulates innate and adaptive immune
responses (discussed below), including inflammasome
activation, antigen presentation and Tcell activation
5355
.
Thus, pathways that stimulate the efflux of cholesterol
from the macrophage have two atheroprotective func-
tions: they promote lipid homeostasis and they protect
against inflammation.
Innate immune activation
Evidence supports the idea that innate immune activa-
tion is a central process in the pathogenesis of atheroscle-
rosis. As reviewed above, dysregulated lipid metabolism
contributes to the development of foam cells. Such aber-
rations and the resulting endogenous danger ligands that
accumulate in atherosclerotic plaques trigger PRRs that
are expressed by macrophages, including NOD-like
receptors (NLRs), scavenger receptors and TLRs, thereby
activating the inflammatory response.
NLRs and inflammasome activation. Cholesterol crys-
tals are present in atherosclerotic plaques and are found
in both extracellular spaces and within plaque macro-
phages. Although previously thought to be a feature
of advanced plaques, a recent study using combined
confocal-reflection microscopy showed the presence of
cholesterol crystals in early lesions in Apoe
/
mice
56,57

and showed that macrophage engulfment of choles-
terol crystals induces the NLRP3 inflammasome (FIG.2).
Uptake of pre-formed crystals by human and mouse
macrophages induces lysosomal destabilization as well
as the release of proteases and/or reactive oxygen species
into the cytosol that activate NLRP3 (NOD-, LRR- and
pyrin domain-containing 3), which leads to the pro-
cessing and secretion of the cytokine IL-1
56,5861
. The
potential importance of this pathway in atherogenesis
was shown using LDL receptor (Ldlr)
/
mice, in which
REVI EWS
714 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
M1 macrophages
Macrophages that are
activated by Toll-like
receptor ligands (such as
lipopolysaccharide) and
interferon- and that express,
among others, inducible nitric
oxide synthase and nitric
oxide.
M2 macrophages
Macrophages that are
stimulated by interleukin-4
(IL-4) or IL-13 and that express
arginase 1, the mannose
receptor 1 (also known as
CD206) and the IL-4 receptor
-chain.
transplantation with bone marrow cells deficient in
IL-1 or in components of the NLRP3 inflammasome
complex led to reduced plaque formation
56
. A subse-
quent study using Apoe
/
mice with somatic deficiency
of Nlrp3 failed to show protection from atherosclerosis
62
.
Although the reasons for this discrepancy will need to
be investigated, a potential confounding factor may have
been the different cholesterol contents of the Western
diet used in the two studies (0.3% versus 1.25%).
In addition to pre-formed cholesterol crystals, recent
work indicates that loading of macrophages with cho-
lesterol can lead to the denovo formation of intracel-
lular cholesterol crystals that trigger NLRP3 (REF.63).
CD36 has a crucial role in the accrual and the nuclea-
tion of cholesterol crystals within macrophages that
have been treated with oxidized LDL, as well as in the
ensuing lysosomal disruption and NLRP3 inflamma-
some activation
63
. Consequently, macrophages lacking
CD36 failed to induce IL-1 production in response to
oxidized LDL, and targeting CD36 in atherosclerotic
mice decreased serum IL-1 levels and plaque choles-
terol crystal accumulation. Notably, the CD36-mediated
uptake of amyloid-forming peptides that are implicated
in Alzheimers disease and type2 diabetes also activates
NLRP3. This suggests that there is a common pathway
of lysosomal-mediated NLRP3 activation that occurs in
the cell after the new aggregation and transformation
of these soluble ligands into their pathogenic forms
63
.
Although not yet investigated, other crystalline or
amyloid substances in atherosclerotic plaques, such as
calcium phosphate crystals or serum amyloid A
64
, may
also represent damage-associated molecular patterns
(DAMPs) that could trigger the inflammasome and
IL-1 secretion.
TLR signalling. The participation of TLR signalling
pathways in the promotion of atherosclerosis is sup-
ported by mouse studies in which the whole body
deletion of Tlr2 orTlr4 (REFS6568) or of the adaptor
proteins used by these TLRs, including IL-1 recep-
tor-associated kinase 4 (IRAK4)
69,70
, TNF receptor-
associated factor 6 (TRAF6)
71
, TIR-domain-containing
adaptor protein inducing IFN (TRIF; also known
as TICAM1)
72
and myeloid differentiation primary-
response protein 88 (MYD88)
65,73
, confers protection
from atherosclerosis. This finding has initiated inves-
tigations of the endogenous ligands that accumulate
during hypercholesterolaemia and in plaques that
may trigger these microbial-sensing pathways in mac-
rophages. Among the candidates proposed, oxidized
LDL species have been extensively studied as ligands
for both the scavenger receptors and the TLRs, and
the extent of oxidation influences their recognition by
these receptors (FIG.2); for example, minimally oxidized
LDL is recognized by CD14TLR4MD2 and initiates
cytoskeletal rearrangements, as well as tumour necrosis
factor (TNF), IL-6 and IL-10 production
74
. Moderately
oxidized LDL that is recognized by CD36 signals via
a heterodimer of TLR4 and TLR6, which results in
NF-B activation and in the expression of chemokines
that promote monocyte recruitment to atherosclerotic
lesions
67
. Finally, oxidized phospholipids and saturated
fatty acids induce cooperative signalling of CD36 and
TLR2 that promotes apoptosis in macrophages under-
going prolonged ER stress
75
. However, in addition to
these ligand-initiated signalling pathways, the enrich-
ment of macrophage plasma membranes with free
cholesterol can also lead to the sustained activation of
various TLRs, including TLR3 and TLR4 (REFS43,76).
Thus, numerous pathways may contribute to the initia-
tion and the maintenance of TLR-induced macrophage
inflammation in atherosclerotic plaques.
Macrophage polarization and plasticity
One consequence of the TLR-dependent activation of
monocyte-derived cells entering the plaque might be
their polarization to M1 macrophages. These inflamma-
tory macrophages secrete pro-atherosclerotic cytokines
(such as IL-6 and IL-12), as well as reactive oxygen and
nitrogen species that would exacerbate oxidative stress
in the plaque
77
(BOX1). Histological analysis of human
plaques showed M1 macrophages to be enriched in
lipids and localized to areas that are distinct from
those in which the less inflammatory M2 macrophages
(also known as alternatively activated macrophages)
are localized
78
. Studies of M1 and M2 macrophages
that have been polarized invitro and in mouse mod-
els of atherosclerosis have led to a simplified view that
M1macrophages promote plaque inflammation and
M2 macrophages resolve plaque inflammation.
However, the phenotypic range of macrophages invivo
is likely to be complex, as macrophages encounter a
microenvironment of diverse, and even opposing, sig-
nals; for example, in addition to inducing TLR signal-
ling that can lead to M1 polarization, oxidized LDL has
also been reported to induce the expression of the M2
macrophage phenotypic marker arginase 1 via the acti-
vation of peroxisome proliferator activated receptor-
(PPAR)
79
. In addition, oxidized phospholipids present
in oxidized LDL induce a macrophage phenotype that
is distinct from M1 or M2 macrophage phenotypes
and that has been termed Mox; these macrophages are
characterized by the increased expression of nuclear
factor erythroid 2-related factor 2 (NRF2; also known
as NFE2L2)-dependent genes and reactive oxygen spe-
cies
80
. It is probable that T helper 1 (T
H
1) and T
H
2 cells
in plaques secrete macrophage-polarizing factors
81
that
also contribute to the balance of M1 and M2 macro-
phages. Nonetheless, the factors in the plaque micro-
environment that promote the polarization of these cells
invivo remain incompletely defined.
The recent identification of transcriptional pro-
grammes that regulate macrophage polarization has
provided some insights into the effects of M1 and M2
macrophages on atherogenesis. Whole body or bone
marrow-specific deletion of the transcription factor
NR4A1 (also known as NUR77), which has been sug-
gested to control the LY6C
low
patrolling monocyte phe-
notype and to favour M2 macrophage differentiation
82
,
resulted in increased polarization of macrophages to
an M1 macrophage phenotype and an acceleration of
atherosclerosis in Apoe
/
and Ldlr
/
mice
83,84
, although
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 715
2013 Macmillan Publishers Limited. All rights reserved
this result has been inconsistent
85
. Similarly, the targeted
deletion of the transcription factor Krppel-like factor4
(KLF4), which promotes M2 macrophage polarization
and inhibits M1 macrophage polarization
86
, enhanced
both pro-inflammatory M1 macrophage activation and
foam cell formation, and accelerated atherosclerosis
in Apoe
/
mice
87
. Notably, the expression of KLF4 in
macro phages is reduced by pro-inflammatory cytokines
and oxidized phospholipids found in plaques
87
, which
suggests that the KLF4-driven M2 macrophage phe-
notype may be repressed during atherogenesis and
that this contributes to disease progression when such
signals predominate. Indeed, the administration of
the M2-polarizing cytokine IL-13 to Ldlr
/
mice was
shown to drive plaque macrophages to M2-like cells
and to inhibit atherosclerosis progression
88
. Moreover,
an enrichment of M2 macrophages has been shown to
occur in plaques in which the regression of atheroscle-
rosis in mice (TABLE1) is induced by aggressive lowering
of lipids or raising of HDL levels
4,5,89
(discussed further
below). Collectively, these studies suggest that path-
ways that promote the M2 polarization of macrophages
protect against atherosclerosis.
The origin of M1 and M2 macrophages in plaques
remains an area of debate. Although it has been sug-
gested that LY6C
hi
monocytes are the precursors of
M1 macrophages, studies using Apoe
/
mice have
shown that M2 macrophages populate early lesions
(also known as fatty streaks), which are present at
the stage in which LY6C
hi
monocytes are thought
to be the predominant monocyte subset recruited
into plaques. However, as plaques progress to more
complex inflammatory lesions, the M1 macrophage
phenotype becomes more frequent
90
. Further studies
are needed to address the origins of M1 and M2 mac-
rophages in atherosclerosis, particularly whether the
recruitment of LY6C
low
monocytes thought to pref-
erentially become M2 macrophages predominate in
the earliest lesions, whether there is interconversion
between M1 and M2 macrophage phenotypes invivo,
or whether M2 macrophages are derived from the
proliferation of a small population of tissue-resident
M2 macrophages, as has recently described in other
disease models
91,92
. A better understanding of the
regulation of macrophage polarization is likely to
offer insights into pathways that could be used for
the potential manipulation of macrophage behaviour
towards an atheroprotectivestate.
Plaque macrophage retention and emigration
The number of macrophages in a plaque is kinetically
determined by monocyte recruitment and local prolifera-
tion, and is counterbalanced by the emigration and death
of macrophages. The factors that determine macrophage
recruitment to plaques were discussed above. With
regard to the local proliferation of monocyte-derived
macrophages, this probably occurs in the plaque, as has
been suggested by the assessment of proliferation mark-
ers in lesional macrophages and DCs
93
. Nevertheless,
the quantitative importance of macrophage prolifera-
tion in atherosclerosis progression remains to be deter-
mined. Of note, on the basis of a recent report showing
a lower percentage of proliferating macrophages in early
plaques compared with advanced plaques, it is likely to
be variable in different stages of the disease
117
.
Macrophage emigration has been shown to occur in
early atherosclerotic plaques, but the rate of macrophage
egress has been reported to decrease with atherosclero-
sis progression
94
(FIG.3). It is probable that plaque mac-
rophages are subject to both retention and emigration
signals, and that the balance of these forces contributes to
the net accumulation of plaque macrophages. These sig-
nals are only beginning to be defined. Cholesterol loading
of macrophages has been shown to increase the expres-
sion of the neuro-immune guidance cues netrin 1 and
semaphorin 3E, which both function to induce macro-
phage chemostasis invitro
25,26
. Macrophage expression
of these migration-inhibitory molecules is also induced
Table 1 | Selected mouse models of atherosclerosis progression and regression
Mouse model Important Features Lipoprotein profile Refs
Progression
Apoe
/
mice Spontaneous development of complex plaques when
mice are fed on a chow diet; and acceleration of plaque
formation when mice are fed on a Western diet
Intestinally derived remnant lipoprotein
particles
114,115
Ldlr
/
mice Development of plaques following feeding mice a
cholesterol and fat-enriched diet; and lipoprotein profile
similar to that of humans
VLDL and LDL 116
Regression
Aortic transplant mice Rapid regression of atherosclerosis; but requires surgical
procedure, for example, the transplantation of aortas from
Apoe
/
mice to wild-type mice
Lipid levels revert to the levels in wild-type
mice
103
Reversa mice An Ldlr
/
mouse-based platform that shows inducible
reversal of hyperlipidaemia after conditional inactivation
of Mttp
Lipid levels revert to nearly the levels that are
observed in wild-type mice
4
Reconstitution of Apoe
/
mice with APOE
The inducible regression of atherosclerosis by
adenoviral delivery of Apoe to the liver or by correcting a
hypomorphic allele of the Apoe gene
Lipid levels revert to nearly the levels that are
observed in wild-type mice
100,102
Apoe, apolipoprotein E; Ldlr, low-density lipoprotein receptor; Mttp, microsomal triglyceride transfer protein large subunit; VLDL, very low-density lipoprotein.
REVI EWS
716 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Nature Reviews | Immunology
Circulating
monocyte
Collagen
Lumen
Intima
Media
Necrotic core
Progressing plaque
Chemokine
gradient
Foam cell
Macrophage
Adhesion
Reverse
transmigration
Chemostasis
UNC5B
To adventitial lymphatics
CCR7
ABCA1
Regressing plaque
SR-A1
Oxidized
LDL
CD36
LOX1
Migration
Proliferation
Lipid
unloading
LDL
Netrin 1
Semaphorin 3E
Cadherins
ER stress
Apoptosis
Defective
eerocytosis
Macrophage
emigration
Eective
eerocytosis
during hypoxia, which is intimately linked to athero-
sclerosis
26,95
; this occurs in mice
96
and has become rec-
ognized as a primary trigger of plaque inflammation.
Studies using Ldlr
/
mice with a bone marrow deficiency
of netrin 1 showed that they had reduced atherosclero-
sis progression and increased macrophage emigration
from lesions, which suggests that netrin 1 may function
to retain macrophages in plaques
25
. Similar experiments
using mice that lack semaphorin 3E in macrophages will
be needed to extend these findings and are in progress.
Other factors that inhibit cell movement (such as adhe-
sion molecules
97
) or the resolution of inflammation are
also likely to contribute to the retention of macrophages
in the plaque, and studies comparing mouse models of
atherosclerosis progression and regression are beginning
to uncover these signals (see below).
The signals that guide macrophages to exit plaques,
either by reverse transmigration through the endothelium
to the lumen or by migrating through the media to the
adventitial lymphatics, remain poorly defined. In stud-
ies in which macrophage emigration from plaques was
induced by normalizing the hyperlipidemic plasma profile
of mice in an aortic transplant model, the cells that emi-
grated expressed several markers that are characteristic of
both macrophages and DCs
98
; for example, the expression
of CCR7, which is the receptor for the chemokines CCL19
and CCL21 that regulate DC homing to the lymph nodes,
was upregulated in the emigrating CD68 (also known as
macrosialin)-expressing cells. Furthermore, blocking this
pathway led to substantial retention of these cells in the
plaque
98
. Further studies are needed to define other fac-
tors in this and other models of atherosclerosis regression.
Figure 3 | Pathways regulating macrophage retention and emigration in plaques. Imbalances in macrophage
lipid metabolism in the progressing plaque lead to the retention of macrophages and to chronic inflammation.
The accumulating lipid-laden macrophages express retention molecules (such as netrin 1 and its receptor UNC5B,
semaphorin3E and cadherins) that promote macrophage chemostasis. In this inflammatory milieu, these accumulating
macrophages experience endoplasmic reticulum (ER) stress, which, if prolonged, results in apoptosis. This cell death,
coupled with defective efferocytosis, results in the formation of the necrotic core that is characteristic of advanced plaques.
The mechanisms that promote lipid unloading of the foam cell, including the factors that upregulate ATP-binding cassette
subfamily A member 1 (ABCA1) expression on plaque macrophages and cholesterol efflux, reverse the accumulation of
these foam cells. This plaque regression is characterized by an upregulation of CC-chemokine receptor 7 (CCR7) on
myeloid-derived cells and a decrease in the expression of retention factors. The accumulating evidence summarized
in this Review supports the idea that the regulation of these macrophage migration factors contributes to macrophage
emigration from the plaque through reverse transmigration to the lumen or through trafficking to the adventitial
lymphatics. LDL, low-density lipoprotein; LOX1, lectin-like oxidized LDL receptor 1; SR-A1, scavenger receptor A1.
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 717
2013 Macmillan Publishers Limited. All rights reserved
In addition, the continued presence of macrophage
foam cells in the inflammatory lipid-rich environ-
ment of the plaque can eventually lead to cytotoxic-
ity from ER and oxidative stress
1
. Activation of ER
stress responses occurs as a result of free cholesterol
accumulation in macrophages and by saturated fatty
acid signalling via SR-A1, TLR2 and TLR4 (REF.75).
Prolonged ER stress leads to macrophage apoptosis,
which is observed in 2 to 4% of cells in mouse plaques,
with the highest levels in advanced plaques. In these
late-stage plaques, the ability of macrophages to clear
their dying counterparts through such receptors as
tyrosine protein kinase MER (MERTK) and LDLR-
related protein 1 (LRP1) becomes compromised, and
this has been partly attributed to cholesterol accumula-
tion in the engulfing cells
99
. This defective efferocytosis
contributes to secondary necrosis and to the forma-
tion and expansion of the lipid cores, which, in turn,
contribute to plaque vulnerability and to rupture
46
.
Therefore, it is possible that apoptosis, especially in
the context of efficient efferocytosis, also contributes
to net changes in macrophage or foam cell content, as
has been suggested in a recent study of atherosclero-
sis regression
100
; however, a mathematical analysis of
those data suggest that a rate tenfold higher than usual
would be required for the changes observed (S. Russell
and E.A.F., unpublished observations). In summary,
monocyte recruitment and cell retention, emigration
and death are all potential kinetic contributors to the
net plaque contents of macrophages and foam cells.
The quantitative effect of each of these processes will
probably vary during the different stages of the disease
and in different models of progression and regression,
as well as in co-morbid states, such as insulin resistance
or diabetes, and chronic kidney disease.
Lessons from models of plaque regression
The historical focus on atherosclerosis in both human
and animal studies has been on its progression, with the
prevailing view that, except for early lesions which are
dominated by foam cells, atherosclerosis was essentially
irreversible, although the mechanisms by which even
an immature plaque regressed remained undefined.
More recent discoveries, including finding that mac-
rophages can emigrate from plaques in some animal
models and that tissue-remodelling M2 macrophages
are present in human and animal plaques, suggest that
there is cause for optimism that clinical atherosclero-
sis regression could be achieved. Nevertheless, under-
standing the biology of atherosclerosis regression,
and the discovery of therapeutic targets to achieve it,
requires robust preclinical models. Therefore, several
mouse models of atherosclerosis, such as Apoe
/
mice
and Ldlr
/
mice, have been adapted for studies of ath-
erosclerosis regression (TABLE 1). Common to all mod-
els has been the finding that in the regressing plaque
there is a decline in the number of macrophages and,
in some, a change in their phenotypic characteristics,
with an enrichment of M2 macrophage characteris-
tics
46,89,101104
, which suggests that this is a common
signature of regressing plaques.
Transcriptomic profiling of macrophages that have
been isolated by laser capture microdissection
98
of pro-
gressing and regressing plaques in an aortic transplanta-
tion mouse model showed there to be >700 differentially
regulated genes
97
, including the recently described mac-
rophage retention factors semaphorin 3E and netrin1.
Other genes that are downregulated in macrophages in
regressing plaques include adhesion molecules, such as
members of the cadherin family
97
. By contrast, cellular
motility factors were upregulated. In addition, CCR7
was expressed at low levels in plaque macrophages
and was probably suppressed by hypercholesterolaemia
as a result of a serum-response element in its promoter
105
.
Notably, the transcription of Ccr7 was upregulated in
macrophages when plaques were placed in a regression
environment, thereby increasing the migratory capacity
of the cells. Taken together, the transcriptomic data from
the aortic transplantation model indicate that the emigra-
tion of macrophages from plaques is a highly regulated
process, and reflect coordinated changes in macrophage
retention and movement. Transcriptome analyses from
other models of atherosclerosis regression will be needed
to determine how conserved these changesare.
Therapeutic targeting of plaque macrophages
Therapies that alter macrophage content by reducing
macrophage recruitment to atherosclerotic plaques or by
promoting macrophage apoptosis, efferocytosis or emi-
gration have been proposed to have beneficial effects on
disease. However, the quantitative effect of each of these
processes on disease progression probably depends on
the stage of disease; for example, macrophage recruitment
dominates compared with emigration in disease progres-
sion, whereas macrophage emigration is increased in sev-
eral models of atherosclerosis regression. Furthermore,
the low level of macrophage apoptosis that is seen in early
atherosclerosis (typically ~24% of cells) increases as
plaques become more complex, with secondary necrosis
also becoming prominent as the efferocytosis of apoptotic
cells falters
1
. In addition, as seen in models of progression
and regression
5,90,97
, the inflammatory phenotype of the
macrophages (using the simplified scheme of M1 versus
M2 macrophages) is not constant, which probably reflects
the well-known plasticity of monocyte-derived cells in
response to microenvironmental changes. Therapies that
alter macrophage inflammation by increasing polarization
to an M2 macrophage phenotype, by increasing efferocyto-
sis or by increasing macrophage emigration would be pre-
dicted to be beneficial on the basis of preclinicalmodels.
The fact that new clinical targets are needed is obvious
from the failure of conventional risk factor management
to effectively eliminate the risk of cardiovascular disease,
with more than half of patients in controlled clinical trials
having heart attacks or strokes despite aggressive treat-
ments. A recent example of the discovery of a potential
target from mouse studies is our finding that neuronal
guidance molecules function as macrophage retention
factors in plaque progression and that their expression in
macrophages decreases in regressing plaques
25,26,97
. Thus,
it may be desirable to selectively deliver small interfering
RNAs (siRNAs), or other therapeutics directed against
REVI EWS
718 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
MicroRNA
A single-stranded RNA
molecule of approximately
2123 nucleotides in length
that regulates the expression
of other genes.
these and other factors that facilitate the emigration of
macrophages. There is considerable optimism that it is
possible to specifically target agents to modify the fac-
tors in plaques that are discussed above, on the basis of
recent studies using nanoparticles, including reconstituted
lipoproteins, to deliver siRNAs, imaging agents and small
molecules to plaques
106108
.
One obvious impediment to targeting macrophages
in plaques is that targeting a specific process, such as
monocyte recruitment, may be advantageous locally, but
not desirable systemically. In addition, even if monocyte
recruitment to plaques could be specifically blocked, the
timing of the inhibition may be crucial. An example of
this is that blocking monocyte recruitment via CCR2 may
be an effective strategy in atherosclerosis progression, but
recent investigations suggest that the shift in the pheno-
typic balance to the M2 macrophage phenotype during
atherosclerosis regression in the aortic transplant model
requires the recruitment of LY6C
hi
monocytes via CCR2
(Y. Vengrenyuk and E.A.F., unpublished observations),
which is similar to what has been observed in autoim-
mune encephalomyelitis and allergic skin reactions
109,110
.
Thus, the inhibition of CCR2 may impair atherosclerosis
regression. Similarly, timing may also be an issue for tar-
geting other chemokine receptors, such as CX
3
CR1 and
CCR5, which, together with CCR2, control over 90% of
monocyte entry into progressing plaques
21
.
The timing of strategies that therapeutically target
macrophage death is also an important issue it is
expected that in early plaque development, when effe-
rocytosis is efficient, increasing apoptosis would be
beneficial. By contrast, efferocytosis is impaired in more
complex plaques, and it is this that is relevant to clinical
events. In complex plaques increasing apoptosis would
lead to augmented release of macrophage lipid content
and tissue factors that would expand the necrotic core
and enhance its thrombogenicity. Current efforts are
focused on maintaining levels of efferocytosis through-
out plaque progression through the use of agents such as
IL-10 or LXR agonists, which also have additional plaque
benefits, such as reducing inflammation (in the case of
IL-10 and LXR agonists) or promoting cholesterol efflux
(in the case of LXR agonists). Increased lipid efflux would
be expected to favourably affect the inflammatory state of
macrophages
43,76
and their ability to emigrate
5
in addition
to reducing plaque lipid content. In addition to LXR ago-
nists, increasing autophagy
5153,111
or ABCA1 and ABCG1
expression levels by inhibiting the microRNA miR-33
(REF.89) may help to target macrophage cholesterolefflux.
Another therapeutic strategy would be to reduce the
inflammatory state of plaque macrophages. One approach
to achieve this would be to polarize macrophages to the
M2 phenotype, as these cells might be particularly impor-
tant in the regressing atherosclerotic plaque as they have
several crucial properties: they secrete anti-inflamma-
tory factors and promote tissue remodelling and repair
through the induction of collagen formation and the
clearance of dying cells and debris; they secrete potent
anti-inflammatory factors such as IL-10 and reduce the
production of damaging reactive nitrogen species; and
they express high levels of MERTK and thereby increase
the efferocytosis of dying macrophages
112,113
. Thus, pro-
moting the M2 macrophage phenotype in plaques would
be expected to be beneficial in both atherosclerosis pro-
gression and regression, which is consistent with recent
studies showing that Ldlr
/
mice treated with IL-13 were
protected from atherosclerosis
88
and that M2 macrophages
are required for disease regression in the aortic transplant
model (Y. Vengrenyuk and E.A.F., unpublished observa-
tions). The manipulation of other factors that inhibit
M1 macrophage polarization may be similarly successful
in the context of atherosclerosis progression or regression.
Conclusions and future perspective
Macrophages are the central cells in atherosclerosis, and
the quantity and the phenotype of these cells in plaques
influence both disease progression and regression. Both
aspects of the disease are dynamic processes that rep-
resent a confluence of diverse metabolic and inflam-
matory pathways, and involve the entry of monocytes
into plaques and the retention, emigration and death of
lesional macrophages.
Important areas of future investigation include the
regulation and the quantitative effect of each of these
kinetic factors, the effects of other immune cells in the
atheroma on the properties of macrophages, and the
therapeutic manipulation of existing and newly discov-
ered factors that affect the lipid content of macrophages,
their number and their inflammatory phenotype.
Despite the necessity of performing mechanistic studies
in preclinical models, it will be important to relate these
findings to human pathophysiology. This is starting to
become a possibility, partly as a result of mining human
genetic studies (such as genome-wide association studies)
and various omic characterizations of human tissues.
A remaining challenge will be to use the present and ongo-
ing research to design clinical interventions that reduce
the unacceptably high risk of cardiovascular disease.
1. Moore,K.J. & Tabas,I. Macrophages in the
pathogenesis of atherosclerosis. Cell 145, 341355
(2011).
2. Randolph,G.J. Emigration of monocyte-derived cells
to lymph nodes during resolution of inflammation and
its failure in atherosclerosis. Curr. Opin. Lipidol. 19,
462468 (2008).
3. Nathan,C. & Ding,A. Nonresolving inflammation.
Cell 140, 871882 (2010).
4. Feig,J.E. etal. Reversal of hyperlipidemia
with a genetic switch favorably affects the content
and inflammatory state of macrophages in
atherosclerotic plaques. Circulation 123, 989998
(2011).
This study describes the use of Reversa mice as a
model of atherosclerosis regression.
5. Feig,J.E. etal. HDL promotes rapid atherosclerosis
regression in mice and alters inflammatory
properties of plaque monocyte-derived cells.
Proc. Natl Acad. Sci. USA 108, 71667171
(2011).
6. Llodra,J. etal. Emigration of monocyte-derived
cells from atherosclerotic lesions characterizes
regressive, but not progressive, plaques.
Proc. Natl Acad. Sci. USA 101, 1177911784
(2004).
This study shows the emigration of CD68
+
cells
from regressing atherosclerotic plaques.
7. Averill,L.E., Meagher,R.C. & Gerrity,R.G.
Enhanced monocyte progenitor cell proliferation in
bone marrow of hyperlipemic swine. Am. J.Pathol.
135, 369377 (1989).
8. Feldman,D.L., Mogelesky,T.C., Liptak,B.F. & Gerrity,R.G.
Leukocytosis in rabbits with diet-induced atherosclerosis.
Arterioscler. Thromb. 11, 985994 (1991).
9. Swirski,F.K. etal. Ly-6C
hi
monocytes dominate hyper-
cholesterolemia-associated monocytosis and give rise
to macrophages in atheromata. J.Clin. Invest. 117,
195205 (2007).
This study shows that Apoe
/
mice have a
monocytosis that is due to an increase in the
LY6C
hi
monocyte population.
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 719
2013 Macmillan Publishers Limited. All rights reserved
10. Tacke,F. etal. Monocyte subsets differentially employ
CCR2, CCR5, and CX3CR1 to accumulate within
atherosclerotic plaques. J.Clin. Invest. 117, 185194
(2007).
This study shows that Apoe
/
mice have a
monocytosis and it describes the chemokine
receptors that contribute to monocyte recruitment
in progressing plaques.
11. Yvan-Charvet,L. etal. ATP-binding cassette
transporters and HDL suppress hematopoietic stem
cell proliferation. Science 328, 16891693 (2010).
This study establishes the essential role of cellular
cholesterol efflux in suppressing haematopoietic
stem cell proliferation.
12. Ross,R. Atherosclerosis an inflammatory disease.
N.Engl. J.Med. 340, 115126 (1999).
13. Tabas,I., Williams,K.J. & Boren,J. Subendothelial
lipoprotein retention as the initiating process in
atherosclerosis: update and therapeutic implications.
Circulation 116, 18321844 (2007).
14. Glass,C.K. & Witztum,J.L. Atherosclerosis: the road
ahead. Cell 104, 503516 (2001).
15. Paulson,K.E. etal. Resident intimal dendritic cells
accumulate lipid and contribute to the initiation of
atherosclerosis. Circ. Res. 106, 383390 (2010).
16. Dutta,P. etal. Myocardial infarction accelerates
atherosclerosis. Nature 487, 325329 (2012).
17. Ley,K., Laudanna,C., Cybulsky,M.I. &
Nourshargh,S. Getting to the site of inflammation:
the leukocyte adhesion cascade updated. Nature Rev.
Immunol. 7, 678689 (2007).
18. Soehnlein,O. etal. Distinct functions of chemokine
receptor axes in the atherogenic mobilization and
recruitment of classical monocytes. EMBO Mol. Med.
5, 471481 (2013).
19. Woollard,K.J. & Geissmann,F. Monocytes in
atherosclerosis: subsets and functions.
Nature Rev. Cardiol. 7, 7786 (2010).
20. Weber,C. & Noels,H. Atherosclerosis: current
pathogenesis and therapeutic options.
Nature Med. 17, 14101422 (2011).
21. Combadiere,C. etal. Combined inhibition of CCL2,
CX3CR1, and CCR5 abrogates Ly6C
hi
and Ly6C
lo

monocytosis and almost abolishes atherosclerosis
in hypercholesterolemic mice. Circulation 117,
16491657 (2008).
22. Serbina,N.V. & Pamer,E.G. Monocyte emigration
from bone marrow during bacterial infection requires
signals mediated by chemokine receptor CCR2.
Nature Immunol. 7, 311317 (2006).
23. Landsman,L. etal. CX3CR1 is required for monocyte
homeostasis and atherogenesis by promoting cell
survival. Blood 113, 963972 (2009).
24. van Gils,J.M. etal. Endothelial expression of guidance
cues in vessel wall homeostasis dysregulation under
proatherosclerotic conditions. Arterioscler. Thromb.
Vasc. Biol. 33, 911919 (2013).
This study shows that neuronal guidance molecules
are differentially expressed on the endothelium in
athero-prone and athero-protected regions of the
vasculature.
25. van Gils,J.M. etal. The neuroimmune guidance
cue netrin-1 promotes atherosclerosis by inhibiting
the emigration of macrophages from plaques.
Nature Immunol. 13, 136143 (2012).
This work identifies netrin 1 as a retention signal
that blocks macrophage egress from inflamed vessel
walls in the presence of hypercholesterolaemia,
which leads to chronic vessel wall inflammation and
plaque progression.
26. Wanschel,A. etal. Neuroimmune guidance cue
semaphorin 3E is expressed in atherosclerotic
plaques and regulates macrophage retention.
Arterioscler. Thromb. Vasc. Biol. 33, 886893
(2013).
27. Moore,K.J. & Freeman,M.W. Scavenger receptors
in atherosclerosis: beyond lipid uptake. Arterioscler.
Thromb. Vasc. Biol. 26, 17021711 (2006).
28. Miller,Y.I. etal. Oxidation-specific epitopes are
danger-associated molecular patterns recognized
by pattern recognition receptors of innate immunity.
Circ. Res. 108, 235248 (2011).
29. Kunjathoor,V.V. etal. Scavenger receptors class A-I/II
and CD36 are the principal receptors responsible for
the uptake of modified low density lipoprotein leading
to lipid loading in macrophages. J.Biol. Chem. 277,
4998249988 (2002).
30. Podrez,E.A., Schmitt,D., Hoff,H.F. & Hazen,S.L.
Myeloperoxidase-generated reactive nitrogen
species convert LDL into an atherogenic form invitro.
J.Clin. Invest. 103, 15471560 (1999).
31. Kzhyshkowska,J., Neyen,C. & Gordon,S. Role of
macrophage scavenger receptors in atherosclerosis.
Immunobiology 217, 492502 (2012).
32. Maxfield,F.R. & Tabas,I. Role of cholesterol and
lipid organization in disease. Nature 438, 612621
(2005).
33. Kuchibhotla,S. etal. Absence of CD36 protects
against atherosclerosis in ApoE knock-out mice with
no additional protection provided by absence of
scavenger receptor A I/II. Cardiovasc. Res. 78,
185196 (2008).
34. Manning-Tobin,J.J. etal. Loss of SR-A and CD36
activity reduces atherosclerotic lesion complexity
without abrogating foam cell formation in
hyperlipidemic mice. Arterioscler. Thromb. Vasc. Biol.
29, 1926 (2009).
35. Tardif,J.C. Antioxidants: the good, the bad and the
ugly. Can. J.Cardiol. 22 (Suppl. B), 61B65B
(2006).
36. Boyanovsky,B.B., van der Westhuyzen,D.R. &
Webb,N.R. Group V secretory phospholipase A2-
modified low density lipoprotein promotes foam cell
formation by a SR-A- and CD36-independent process
that involves cellular proteoglycans.
J.Biol. Chem. 280, 3274632752 (2005).
37. Oorni,K. & Kovanen,P.T. Lipoprotein modification
by secretory phospholipase A2 enzymes contributes
to the initiation and progression of atherosclerosis.
Curr. Opin. Lipidol. 20, 421427 (2009).
38. Lind,L. etal. Circulating levels of secretory- and
lipoprotein-associated phospholipase A2 activities:
relation to atherosclerotic plaques and future
all-cause mortality. Eur. Heart J. 33, 29462954
(2012).
39. Kugiyama,K. etal. Circulating levels of secretory
typeII phospholipase A2 predict coronary events in
patients with coronary artery disease. Circulation
100, 12801284 (1999).
40. Kruth,H.S. Receptor-independent fluid-phase
pinocytosis mechanisms for induction of foam cell
formation with native low-density lipoprotein particles.
Curr. Opin. Lipidol. 22, 386393 (2011).
This is a discussion of studies that establish that
native LDL can contribute to foam cell formation
through its uptake via macrophage fluid-phase
pinocytosis.
41. Zhu,X. etal. Macrophage ABCA1 reduces
MyD88-dependent Toll-like receptor trafficking
to lipid rafts by reduction of lipid raft cholesterol.
J.Lipid Res. 51, 31963206 (2010).
This study shows that cholesterol enrichment of
lipid rafts promotes signalling via TLRs.
42. Mogilenko,D.A. etal. Endogenous apolipoprotein
A-I stabilizes ATP-binding cassette transporter A1
and modulates Toll-like receptor 4 signaling in
human macrophages. FASEB J. 26, 20192030
(2012).
43. Yvan-Charvet,L. etal. Increased inflammatory gene
expression in ABC transporter-deficient macrophages:
free cholesterol accumulation, increased signaling via
toll-like receptors, and neutrophil infiltration of
atherosclerotic lesions. Circulation 118, 18371847
(2008).
44. Jerome,W.G. Advanced atherosclerotic foam cell
formation has features of an acquired lysosomal
storage disorder. Rejuven. Res. 9, 245255
(2006).
45. Feng,B. etal. The endoplasmic reticulum is the site of
cholesterol-induced cytotoxicity in macrophages.
Nature Cell Biol. 5, 781792 (2003).
46. Tabas,I. Consequences and therapeutic implications
of macrophage apoptosis in atherosclerosis: the
importance of lesion stage and phagocytic efficiency.
Arterioscler. Thromb. Vasc. Biol. 25, 22552264
(2005).
47. Yvan-Charvet,L., Wang,N. & Tall,A.R. Role of HDL,
ABCA1, and ABCG1 transporters in cholesterol efflux
and immune responses. Arterioscler. Thromb. Vasc.
Biol. 30, 139143 (2010).
48. Spann,N.J. etal. Regulated accumulation of
desmosterol integrates macrophage lipid metabolism
and inflammatory responses. Cell 151, 138152
(2012).
49. Calkin,A.C. & Tontonoz,P. Transcriptional integration
of metabolism by the nuclear sterol-activated
receptors LXR and FXR. Nature Rev. Mol. Cell Biol.
13, 213224 (2012).
50. Singh,R. etal. Autophagy regulates lipid metabolism.
Nature 458, 11311135 (2009).
This study was the first to show a role for
autophagy in regulating lipid metabolism.
51. Ouimet,M. etal. Autophagy regulates cholesterol
efflux from macrophage foam cells via lysosomal acid
lipase. Cell. Metab. 13, 655667 (2011).
This study shows that autophagy regulates
cholesterol efflux in macrophage foam cells.
52. Liao,X. etal. Macrophage autophagy plays a
protective role in advanced atherosclerosis.
Cell. Metab. 15, 545553 (2012).
This study uncovers a protective role for the
autophagy process in atherosclerosis through
the regulation of plaque necrosis.
53. Razani,B. etal. Autophagy links inflammasomes
to atherosclerotic progression. Cell. Metab. 15,
534544 (2012).
This study shows a protective role for the
autophagy process in atherosclerosis through
the regulation of inflammasome activation.
54. Nakahira,K. etal. Autophagy proteins regulate
innate immune responses by inhibiting the release
of mitochondrial DNA mediated by the NALP3
inflammasome. Nature Immunol. 12, 222230
(2011).
55. Saitoh,T. etal. Loss of the autophagy protein
Atg16L1 enhances endotoxin-induced IL-1
production. Nature 456, 264268 (2008).
56. Duewell,P. etal. NLRP3 inflammasomes are required
for atherogenesis and activated by cholesterol
crystals. Nature 464, 13571361 (2010).
This study describes a role for the NLRP3
inflammasome in atherogenesis, thereby
uncovering a previously unappreciated role of
cholesterol crystals as key early initiators of
vascular inflammation.
57. Lim,R.S. etal. Identification of cholesterol crystals in
plaques of atherosclerotic mice using hyperspectral
CARS imaging. J.Lipid Res. 52, 21772186 (2011).
58. Gage,J., Hasu,M., Thabet,M. & Whitman,S.C.
Caspase-1 deficiency decreases atherosclerosis in
apolipoprotein E-null mice. Can. J.Cardiol. 28,
222229 (2012).
59. Usui,F. etal. Critical role of caspase-1 in vascular
inflammation and development of atherosclerosis in
Western diet-fed apolipoprotein E-deficient mice.
Biochem. Biophys. Res. Commun. 425, 162168
(2012).
60. Hornung,V. etal. Silica crystals and aluminum salts
activate the NALP3 inflammasome through
phagosomal destabilization. Nature Immunol. 9,
847856 (2008).
61. Freigang,S. etal. Nrf2 is essential for cholesterol
crystal-induced inflammasome activation and
exacerbation of atherosclerosis. Eur. J.Immunol. 41,
20402051 (2011).
62. Menu,P. etal. Atherosclerosis in ApoE-deficient
mice progresses independently of the NLRP3
inflammasome. Cell Death Dis. 2, e137 (2011).
63. Sheedy,F.J. etal. CD36 coordinates activation of
the NLRP3 inflammasome by facilitating the
intracellular nucleation of soluble to particulate
ligands in sterile inflammation. Nature Immunol.
14, 812820 (2013).
64. Niemi,K. etal. Serum amyloid A activates the NLRP3
inflammasome via P2X7 receptor and a cathepsin B-
sensitive pathway. J.Immunol. 186, 61196128
(2011).
65. Michelsen,K.S. etal. Lack of Toll-like receptor 4
or myeloid differentiation factor 88 reduces
atherosclerosis and alters plaque phenotype in mice
deficient in apolipoprotein E. Proc. Natl Acad. Sci.
USA 101, 1067910684 (2004).
66. Mullick,A.E., Tobias,P.S. & Curtiss,L.K. Modulation
of atherosclerosis in mice by Toll-like receptor 2.
J.Clin. Invest. 115, 31493156 (2005).
67. Stewart,C.R. etal. CD36 ligands promote sterile
inflammation through assembly of a Toll-like
receptor4 and 6 heterodimer. Nature Immunol. 11,
155161 (2010).
This study identifies a new TLR heterodimer,
TLR4TLR6, that is triggered by oxidized LDL
through CD36 and that promotes pro-inflammatory
signalling in macrophages.
68. Ding,Y. etal. Toll-like receptor 4 deficiency decreases
atherosclerosis but does not protect against
inflammation in obese low-density lipoprotein
receptor-deficient mice. Arterioscler. Thromb. Vasc.
Biol. 32, 15961604 (2012).
69. Kim,T.W. etal. The critical role of IL-1 receptor-
associated kinase 4-mediated NF-B activation in
modified low-density lipoprotein-induced inflammatory
gene expression and atherosclerosis. J.Immunol.
186, 28712880 (2011).
REVI EWS
720 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
70. Rekhter,M. etal. Genetic ablation of IRAK4 kinase
activity inhibits vascular lesion formation. Biochem.
Biophys. Res. Commun. 367, 642648 (2008).
71. Lutgens,E. etal. Deficient CD40TRAF6 signaling in
leukocytes prevents atherosclerosis by skewing the
immune response toward an antiinflammatory profile.
J.Exp. Med. 207, 391404 (2010).
72. Richards,M.R. etal. The LPS2 mutation in TRIF is
atheroprotective in hyperlipidemic low density
lipoprotein receptor knockout mice. Innate Immun.
19, 2029 (2013).
73. Bjorkbacka,H. etal. Reduced atherosclerosis in
MyD88-null mice links elevated serum cholesterol
levels to activation of innate immunity signaling
pathways. Nature Med. 10, 416421 (2004).
74. Bae,Y.S. etal. Macrophages generate reactive
oxygen species in response to minimally oxidized
low-density lipoprotein: toll-like receptor 4- and
spleen tyrosine kinase-dependent activation of
NADPH oxidase 2. Circ. Res. 104, 210218
(2009).
75. Seimon,T.A. etal. Atherogenic lipids and
lipoproteins trigger CD36TLR2-dependent
apoptosis in macrophages undergoing endoplasmic
reticulum stress. Cell. Metab. 12, 467482
(2010).
76. Zhu,X. etal. Increased cellular free cholesterol in
macrophage-specific Abca1 knock-out mice enhances
pro-inflammatory response of macrophages.
J.Biol. Chem. 283, 2293022941 (2008).
77. Adamson,S. & Leitinger,N. Phenotypic modulation
of macrophages in response to plaque lipids.
Curr. Opin. Lipidol. 22, 335342 (2011).
78. Chinetti-Gbaguidi,G. etal. Human atherosclerotic
plaque alternative macrophages display low
cholesterol handling but high phagocytosis because
of distinct activities of the PPAR and LXR pathways.
Circ. Res. 108, 985995 (2011).
79. Gallardo-Soler,A. etal. Arginase I induction by
modified lipoproteins in macrophages: a peroxisome
proliferator-activated receptor-/-mediated effect
that links lipid metabolism and immunity.
Mol. Endocrinol. 22, 13941402 (2008).
80. Kadl,A. etal. Identification of a novel macrophage
phenotype that develops in response to atherogenic
phospholipids via Nrf2. Circ. Res. 107, 737746
(2010).
This study identifies a new macrophage phenotype,
termed the Mox macrophage, that occurs in
macrophages that have been exposed to
atherogenic phospholipids.
81. Hansson,G.K. & Hermansson,A. The immune system
in atherosclerosis. Nature Immunol. 12, 204212
(2011).
82. Hanna,R.N. etal. The transcription factor NR4A1
(Nur77) controls bone marrow differentiation and the
survival of Ly6C

monocytes. Nature Immunol. 12,


778785 (2011).
This study identifies an important role for NR4A1
in M2 macrophage polarization.
83. Hanna,R.N. etal. NR4A1 (Nur77) deletion polarizes
macrophages toward an inflammatory phenotype and
increases atherosclerosis. Circ. Res. 110, 416427
(2012).
This study reports that the deletion of NR4A1
worsens atherosclerosis by increasing macrophage
polarization to the M1 phenotype.
84. Hamers,A.A. etal. Bone marrow-specific deficiency
of nuclear receptor Nur77 enhances atherosclerosis.
Circ. Res. 110, 428438 (2012).
This study reports that deletion of NR4A1
worsens atherosclerosis and inflammation.
85. Chao,L.C. etal. Bone marrow NR4A expression is
not a dominant factor in the development of
atherosclerosis or macrophage polarization in mice.
J.Lipid Res. 54, 806815 (2013).
In contrast to references 83 and 84, this study
finds no role for NR4A1 in macrophage
polarization or atherogenesis.
86. Liao,X. etal. Krppel-like factor 4 regulates
macrophage polarization. J.Clin. Invest. 121,
27362749 (2011).
This study identifies a key role for the transcription
factor KLF4 in directing M2 macrophage
polarization.
87. Sharma,N. etal. Myeloid Krppel-like factor 4
deficiency augments atherogenesis in ApoE
/
mice
brief report. Arterioscler. Thromb. Vasc. Biol. 32,
28362838 (2012).
This study shows that the deletion of KLF4 leads
to enhanced atherosclerosis in both chow-fed and
Western diet-fed Apoe
/
mice.
88. Cardilo-Reis,L. etal. Interleukin-13 protects from
atherosclerosis and modulates plaque composition
by skewing the macrophage phenotype. EMBO Mol.
Med. 4, 10721086 (2012).
89. Rayner,K.J. etal. Antagonism of miR-33 in mice
promotes reverse cholesterol transport and regression
of atherosclerosis. J.Clin. Invest. 121, 29212931
(2011).
90. Khallou-Laschet,J. etal. Macrophage plasticity in
experimental atherosclerosis. PLoS ONE 5, e8852
(2010).
91. Jenkins,S.J. etal. Local macrophage proliferation,
rather than recruitment from the blood, is a signature
of T
H
2 inflammation. Science 332, 12841288 (2011).
92. Zhang,M.Z. etal. CSF-1 signaling mediates
recovery from acute kidney injury. J.Clin. Invest.
122, 45194532 (2012).
93. Zhu,S.N., Chen,M., Jongstra-Bilen,J. &
Cybulsky,M.I. GM-CSF regulates intimal cell
proliferation in nascent atherosclerotic lesions.
J.Exp. Med. 206, 21412149 (2009).
94. Gerrity,R.G. & Naito,H.K. Lipid clearance from
fatty streak lesions by foam cell migration. Artery
8, 215219 (1980).
95. Ramkhelawon,B. etal. Hypoxia induces netrin-1
and unc5b in atherosclerotic plaques: mechanism
for macrophage retention and survival. Arterioscler
Thromb. Vasc. Biol. 33, 11801188 (2013).
96. Parathath,S. etal. Hypoxia is present in murine
atherosclerotic plaques and has multiple adverse
effects on macrophage lipid metabolism. Circ. Res.
109, 11411152 (2011).
97. Feig,J.E. etal. Regression of atherosclerosis is
characterized by broad changes in the plaque
macrophage transcriptome. PLoS ONE 7, e39790
(2012).
Through transcriptome analysis of macrophages
in progressing and regressing plaques, this work
identifies the genetic signature of macrophages
during the resolution of atherosclerotic
inflammation.
98. Trogan,E. etal. Laser capture microdissection analysis
of gene expression in macrophages from atherosclerotic
lesions of apolipoprotein E-deficient mice. Proc. Natl
Acad. Sci. USA 99, 22342239 (2002).
99. Yvan-Charvet,L. etal. ABCA1 and ABCG1 protect
against oxidative stress-induced macrophage
apoptosis during efferocytosis. Circ. Res. 106,
18611869 (2010).
100. Potteaux,S. etal. Suppressed monocyte recruitment
drives macrophage removal from atherosclerotic
plaques of Apoe
/
mice during disease regression.
J.Clin. Invest. 121, 20252036 (2011).
In this model of atherosclerosis regression, which
was achieved through the reconstitution of Apoe
/

mice with APOE, the authors show that reduced
monocyte recruitment and increased apoptotic
turnover of macrophages are crucial components
of atherosclerosis resolution.
101. Nagareddy,P.R. etal. Hyperglycemia promotes
myelopoiesis and impairs the resolution of
atherosclerosis. Cell. Metab. 17, 695708 (2013).
102. Raffai,R.L. & Weisgraber,K.H. Hypomorphic
apolipoprotein E mice: a new model of conditional
gene repair to examine apolipoprotein E-mediated
metabolism. J.Biol. Chem. 277, 1106411068 (2002).
103. Reis,E.D. etal. Dramatic remodeling of advanced
atherosclerotic plaques of the apolipoprotein
E-deficient mouse in a novel transplantation model.
J.Vasc. Surg. 34, 541547 (2001).
104. Rong,J.X. etal. Elevating high-density lipoprotein
cholesterol in apolipoprotein E-deficient mice
remodels advanced atherosclerotic lesions by
decreasing macrophage and increasing smooth
muscle cell content. Circulation 104, 24472452
(2001).
This study reports that atherosclerosis
regression that is induced by increased HDL
levels is characterized by decreasing macrophage
content and increasing smooth muscle cell
content.
105. Feig,J.E. etal. Statins promote the regression of
atherosclerosis via activation of the CCR7-dependent
emigration pathway in macrophages. PLoS ONE 6,
e28534 (2011).
106. Frias,J.C., Ma,Y., Williams,K.J., Fayad,Z.A. &
Fisher,E.A. Properties of a versatile nanoparticle
platform contrast agent to image and characterize
atherosclerotic plaques by magnetic resonance
imaging. Nano Lett. 6, 22202224 (2006).
107. Leuschner,F. etal. Therapeutic siRNA silencing in
inflammatory monocytes in mice. Nature Biotech.
29, 10051010 (2011).
108. Pan,H. etal. Programmable nanoparticle
functionalization for invivo targeting. FASEB J. 27,
255264 (2013).
109. Egawa,M. etal. Inflammatory monocytes recruited
to allergic skin acquire an anti-inflammatory M2
phenotype via basophil-derived interleukin-4.
Immunity 38, 570580 (2013).
110. Denney,L. etal. Activation of invariant NKT cells in
early phase of experimental autoimmune
encephalomyelitis results in differentiation of Ly6C
hi

inflammatory monocyte to M2 macrophages and
improved outcome. J.Immunol. 189, 551557
(2012).
111. Schrijvers,D.M., De Meyer,G.R. & Martinet,W.
Autophagy in atherosclerosis: a potential drug target
for plaque stabilization. Arterioscler. Thromb. Vasc.
Biol. 31, 27872791 (2011).
112. Thorp,E., Cui,D., Schrijvers,D.M., Kuriakose,G. &
Tabas,I. Mertk receptor mutation reduces
efferocytosis efficiency and promotes apoptotic cell
accumulation and plaque necrosis in atherosclerotic
lesions of apoe
/
mice. Arterioscler. Thromb. Vasc.
Biol. 28, 14211428 (2008).
113. Zizzo,G., Hilliard,B.A., Monestier,M. & Cohen,P.L.
Efficient clearance of early apoptotic cells by human
macrophages requires M2c polarization and MerTK
induction. J.Immunol. 189, 35083520 (2012).
114. Plump, A. S. et al. Severe hypercholesterolemia and
atherosclerosis in apolipoprotein E-deficient mice
created by homologous recombination in ES cells.
Cell 71, 343353 (1992).
115. Zhang, S. H., Reddick, R. L., Piedrahita, J. A. &
Maeda, N. Spontaneous hypercholesterolemia and
arterial lesions in mice lacking apolipoprotein E.
Science 258, 468471(1992).
116. Ishibashi, S. et al. Hypercholesterolemia in low density
lipoprotein receptor knockout mice and its reversal by
adenovirus-mediated gene delivery. J. Clin. Invest. 92,
883893 (1993).
117. Robbins, C. S. et al. Local proliferation dominates
lesional macrophage accumulation in atherosclerosis.
Nature Med. http://dx.doi.org/10.1038/nm.3258
(2013).
Acknowledgements
The work carried out in the authors laboratories related to this
Review is supported by the US National Institutes of Health
(grants R01 HL084312 and P01 HL098055 to E.A.F.; and
grants R01 R01HL117334 and R01HL108182 to K.J.M.).
Competing interests statement
The authors declare no competing financial interests.
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 721
2013 Macmillan Publishers Limited. All rights reserved
The term autophagy (from the Greek words auto
meaning self and phagein meaning to eat) refers
to a collection of diverse processes including
macroautophagy, microautophagy, chaperone-mediated
autophagy
1
and non-canonical autophagy
2
that enable
cells to digest their cytoplasmic contents in lysosomes.
Macroautophagy, which is autophagy in its strictest
form and which is the focus of this Review, depends
on specialized autophagy-related proteins (ATGs)
1

(BOX1) and is distinct from other cytoplasmic diges-
tive processes, including proteasomal degradation, as
a result of its ability to capture and to eliminate large
targets such as toxic protein aggregates, defunct or dis-
used organelles and invading microorganisms. Several
autophagy-related factors have also been recognized to
have autophagy-independent immune functions
3
, but
these are not discussed in thisReview.
Autophagy may be the primordial form of eukary-
otic innate immunity against invading microorganisms.
Our understanding of the immunological functions
of autophagy has dramatically increased in recent
years
4
, and it is now appreciated that, in mammals,
these primordial functions of autophagy have evolved
and have been incorporated into multiple innate and
adaptive immune pathways. In this Review, we dis-
cuss the four main roles of autophagy in immunity:
the direct elimination of microorganisms, the control
of inflammation, the control of adaptive immunity
through the regulation of antigen presentation and
lymphocyte homeostasis, and the secretion of immune
mediators (FIG.1).
The autophagy pathway
ATG-dependent autophagy. The principal morphologi-
cal feature of autophagy is the formation of endomem-
branous organelles, which are known as autophagosomes
(BOX1). An ensemble of ATG factors drives the formation
in the cytoplasm of the autophagic isolation membrane
(also known as the phagophore)
1
(BOX1). Beclin 1 (the
mammalian orthologue of the yeast protein ATG-6), the
serine/threonine protein kinase ULK1 (the mammalian
paralogue of the yeast protein ATG-1), autophagy-
related LC3 proteins and -aminobutyric acid recep-
tor-associated proteins (GABARAPs; the mammalian
paralogues of the yeast ATG-8 protein) are key regula-
tors of phagophore formation
1
(BOX1). A phagophore,
which is often depicted in models as an isolated crescent
in the cytoplasm, is transiently connected to and derived
from phosphatidylinositol-3-phosphate (PtdIns
3
P)-
positive domains of the endoplasmic reticulum (ER),
which are known as omegasomes
1
. Other cellular
compartments, such as the Golgi apparatus, the mito-
chondria and the plasma membrane-derived endocytic
organelles, also contribute to phagophore formation
1
.
Recent studies suggest that the ER-derived autophago-
somes form at ERmitochondria contact sites
5
. A pha-
gophore sequesters the captured cytoplasmic cargo
that is destined for autophagic disposal and, following
elongation and closure, an autophagosome is formed.
At this stage, the corresponding endomembranes are
commonly visualized as double membrane structures
1
.
The degradation of the captured cargo begins when the
double membrane autophagosome matures into a single
1
Department of Molecular
Genetics and Microbiology,
University of New Mexico
Health Sciences Center, 915
Camino de Salud, North East,
Albuquerque, New Mexico
87131, USA.
2
Laboratory of Host Defense,
World Premier International
Immunology Frontier
Research Center, Osaka
University, 31 Yamada-oka,
Suita, Osaka 5650871,
Japan.
Correspondence to V.D.
e-mail:
vderetic@salud.unm.edu
doi:10.1038/nri3532
Macroautophagy
The autophagy-related protein
(ATG)-dependent pathway,
sequestering cargo into
autophagosomes that fuse
with lysosomes, in which
the cargo is degraded.
Proteins, organelles (such as
mitochondria) and invading
microorganisms are selectively
degraded by macroautophagy.
Bulk cytoplasmic autophagy
occurs as a starvation
response.
Autophagy in infection, inflammation
and immunity
Vojo Deretic
1
, Tatsuya Saitoh
2
and Shizuo Akira
2
Abstract | Autophagy is a fundamental eukaryotic pathway that has multiple
effects on immunity. Autophagy is induced by pattern recognition receptors and,
through autophagic adaptors, it provides a mechanism for the elimination of
intracellular microorganisms. Autophagy controls inflammation through regulatory
interactions with innate immune signalling pathways, by removing endogenous
inflammasome agonists and through effects on the secretion of immune mediators.
Moreover, autophagy contributes to antigen presentation and to Tcell homeostasis,
and it affects Tcell repertoires and polarization. Thus, as we discuss in this Review,
autophagy has multitiered immunological functions that influence infection,
inflammation and immunity.
REVI EWS
722 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Microautophagy
Lysosomal import and
degradation of small
portions of the cytoplasm
that is independent of
autophagy-related proteins
(ATGs), and that is often
manifested as invaginations of
the lysososomal membrane
into the lysosomal lumen.
Chaperone-mediated
autophagy
The import and degradation of
soluble cytoplasmic proteins
by chaperone-dependent
direct translocation across
the lysosomal membrane.
membrane-delimited autolysosome. Autophagosomal
maturation is induced by the translocation of the SNARE
protein syntaxin 17 (REF.6) to the outer membrane of
the completed autophagosome. This facilitates fusion
of the autophagosome with lysosomes
5,6
and results
in the acidification of the autophagosome lumen, the
acquisition of lysosomal hydrolases and the degradation
of the cargo as well as the inner of the two membranes
1
.
Immune signalling and autophagy. A deficiency of
intracellular nutrients because of competition from
invading microorganisms is likely to have been one
of the primordial danger signals that was available
to eukaryotic cells to detect microbial invasion and to
eliminate pathogens through autophagy. Consistent
with this idea, metabolic signalling downstream of
amino acid starvation has been associated with anti-
microbial autophagy in response to bacterial
7
and viral
infection
8
(FIG.1a). Metabolic regulation by mammalian
target of rapamycin (mTOR), which inhibits autophagy,
and AMP-activated protein kinase (AMPK), which acti-
vates autophagy, has been covered in detail elsewhere
1
.
The ancient defence mechanism of autophagy has
been well integrated with other immune-sensing sys-
tems. Pattern recognition receptor (PRR) signalling
that is induced following the recognition of pathogen-
associated molecular patterns (PAMPs) and damage-
associated molecular patterns (DAMPs) can activate
autophagy (BOX 2); for example, Toll-like receptor 4
(TLR4) signalling leads to ubiquitylation of the ATG
beclin 1 by the E3 ligase TNF receptor-associated fac-
tor 6 (TRAF6)
9
. Ubiquitylated beclin 1 is released from
its inhibitor B cell lymphoma2 (BCL-2) and becomes
active
9
. TRAF6 also activates ULK1 through ubiquityla-
tion
10
and thereby controls both key pathways that lead
to autophagy (BOX1). TRAF6, ubiquitylated ULK1 and
the IB kinase (IKK) regulators TAK1-binding protein 2
(TAB2) and TAB3 also have roles in starvation-induced
autophagy (BOX 2), which supports the idea that a simi-
lar circuitry, with some variations, becomes induced
by nutritional and innate immune signals. In BOX2,
we propose a model by which immune and nutritional
triggers converge in common signalling pathways to
activate autophagy.
Box 1 | The autophagy pathway
Autophagy is controlled by the regulatory serine/threonine protein kinases ULK1 and ULK2 (which are the mammalian homologues of the yeast
autophagy-related protein 1 (ATG-1)) and by the lipid kinase activity of the classIII phosphatidylinositol-3-phosphate kinase (PtdIns
3
K) VPS34 (VPS-34 in
yeast), which forms a complex with beclin 1 (which is the mammalian orthologue of yeast ATG-6) and ATG14-like protein (ATG14L). ULK1, ULK2 and the
VSP34beclin 1ATG14L complex integrate upstream signals. These factors, which are engaged in the activation of autophagy during nutritional and
immune responses, induce the downstream ATG conjugation cascade
1
. This involves the association of the ATG5ATG12 conjugate with ATG16L1. The
ATG5ATG12ATG16L1 complex facilitates the addition of a phosphatidylethanolamine group to the carboxyl terminus of the mammalian paralogues
of ATG-8: LC3A, LC3B (which has been used as a marker for the identification
of autophagosomes in mammalian cells), LC3C, -aminobutyric acid
receptor-associated protein (GABARAP), GABARAP-like 1 (GABARAPL1)
and GABARAPL2 (REF.1). Lipidated mammalian paralogues of ATG-8 function
together with other factors to assemble, elongate and lead to the closure
of nascent autophagic organelles
1
(see the figure). The lipid kinase VPS34,
through its production of PtdIns
3
phosphate (PtdIns
3
P), has a dual function in
autophagy
1
. During initiation, VPS34-derived PtdIns
3
P is recognized by the
PtdIns
3
P-binding factors WD repeat domain phosphoinositide-interacting
protein 1 (WIPI1)WIPI4 (which are the mammalian paralogues of yeast
ATG-18), and these factors cooperate with ATG2 and ATG9 to form the
phagophore
1
. During the final maturation of autophagosomes into
autolysosomes, the attachment of a tail-anchored SNARE, syntaxin 17
(REF.6), to the autophagosomal membrane enables fusion with lysosomes.
At this stage, the VPS34beclin 1 complex contains ultraviolet radiation
resistance-associated gene protein (UVRAG) instead of ATG14L
1
.
Autophagy requires the rapid remodelling of endomembranes in the
cytoplasm, which involves the formation of transient PtdIns
3
P-positive
structures (known as omegasomes) on the endoplasmic reticulum (ER),
the ERmitochondria contact sites, the Golgi apparatus and the plasma
membrane-derived endosomal organelles. This happens independently of
transcriptional changes. However, transcriptional adjustments are necessary
for sustained autophagy, for instance, to replenish consumables such as LC3
and autophagic adaptors including sequestosome 1 (also known as p62), which
are degraded along with the captured cargo in autolysosomes. Transcriptional
adjustments are accomplished by transcription-positiveregulators (for
example, transcription factor EB (TFEB), forkhead box protein O3A (FOXO3A)
and NFE2-related factor 2 (NRF2)) and transcription-negative regulators (for
example, zinc finger protein with KRAB and SCAN domains 3 (ZKSCAN3; also
known as ZNF306), which coordinate autophagy with lysosomal, proteasomal,
lipolytic and oxidative stress response systems
138,139
.
HOPS, homotypic fusion and vacuole protein sorting.
Nature Reviews | Immunology
Donor membrane
ER omegasome
Golgi apparatus
ERmitochondria contact sites
Endosomes
Plasma membrane
Initiation
Elongation
and closure
Maturation
Degradation
Fusion
Isolation
membrane
LC3
Autophagosome
Autolysosome
ULK1
Beclin 1 VPS34
ATG14L
Beclin 1 VPS34
UVRAG
HOPS
ATG16L1
WIPI ATG2
ATG5 ATG12
LC3A
LC3B
LC3C
GABARAP
Syntaxin 17
GABARAPL1
GABARAPL2
ATG9
Lysosome
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 723
2013 Macmillan Publishers Limited. All rights reserved
Virus
Nature Reviews | Immunology
Bacteria
Amino acid
starvation
TLR
LC3
Ubiquitin
Galectin
a Elimination of microorganisms b Control of pro-inammatory signalling
c Adaptive immunity d Secretion of immune mediators
Regulated
secretion
Constitutive
secretion
Unconventional
secretion
mTOR NOX2 TRAF6
TRAF6
Endosome
Viral
ssRNA
SLR
SLR
Auto-
phago-
somes
Autolysosome
Particulate
antigen
Autolysosome
Lysosome
Autophagy
Autophagy
Cytosolic
proteins
MHC
class II
MHC
class I
Conventional
MHC class I
presentation
Unconventional
MHC class I
presentation
Citrullinated
antigen
LAP NOD2
miRNA
Proteins
Secretory
lysosomes
?
Cathepsin K
(in osteoclasts) and
lysozyme (in Paneth cells) IL-6 and IL-8
DAMPs (e.g. IL-1,
IL-18 and HMGB1)
LAP Xenophagy
ROS
NOD2
ATG16L1
Lysosome
Virus
DAMPs
K
+
eux
IPS1
RIG-I
Damaged
mitochondria
Leak of
lysosomal
protease
ROS mtDNA
TLR7 or
TLR8
Pro-inammatory
mediators
NLRP3
inammasome
Delivery of
PAMPs to PRRs
Failing
autophagy
IL-1
IL-18
AIM2
inammasome
ATG5
ATG12
AGO
ATGs
ATGs
DICER
B cell eects
HSC self renewal
B1 cell development
Plasma cell survival
and IgG secretion
T cell eects
Thymic selection of
T cell repertoire
T cell maturation
(clearance of
mitochondria and ER)
T cell survival afer TCR
activation
Immunological synapse
destabilization
Innate immune cell
eects
Control of activation
and release of IL-1 and
IL-1 (which inuences
T cell polarization to
T
H
17 cell phenotype)
Proteasome
ER
ER
Golgi apparatus
SLR
mRNA
REVI EWS
724 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved

Non-canonical autophagy
Macroautophagy that
has been reported to occur
independently of one or
more components of the
autophagy-related protein
(ATG) system. It should not be
confused with non-canonical
functions of ATGs, which
refers to the participation
of individual ATG factors
in processes other than
autophagy.
DAMPs, such as DNA complexes
11
, ATP
12
and high-
mobility group box1 protein (HMGB1)
13
, also activate
autophagy. HMGB1 derepresses beclin 1 by displacing
its negative regulator BCL-2 (BOX2), and can also extra-
cellularly activate autophagy by interacting with its cell
surface receptor RAGE (receptor for advanced glycation
end products).
Inflammatory cytokines are also involved in the
activation of autophagy (BOX 2). Indeed, the induc-
tion of autophagy by the pro-inflammatory cytokine
interleukin-1 (IL-1) is crucial for the control of
Mycobacterium tuberculosis in infected macrophages
14
.
IL-1 signals through the IL-1 receptor (IL-1R)
14
and
probably through the downstream recruitment of
TRAF6 and the subsequent TRAF6-dependent ubiqui-
tylation of beclin 1 (REF.9). T helper 1 (T
H
1) cell-derived
cytokines, such as interferon- (IFN), also induce
autophagy in effector cells; for example, these cytokines
induce autophagy in macrophages to enable them to
resist mycobacteria infection
1517
. IFN might activate
autophagy both through the function of immunity-
related GTPases
1517
and through the phosphorylation
of beclin 1 on Thr119 of its BH3 domain by death-
associated protein kinase1 (DAPK1). This results in
the dissociation of beclin 1 from its inhibitor BCL-2
(REF.18) and, together with other processes (BOX2), leads
to beclin 1 activation. In human macrophages, but not
in mouse macrophages, IFN cooperates with 1,25-dihy-
droxyvitaminD3 (also known as calcitriol) to induce
antimycobacterial autophagic activities
19
. Similarly,
tumour necrosis factor (TNF) has been shown to stimu-
late autophagy to restrict intracellular bacteria such as
Shigella spp. and Listeriaspp
20
.
By contrast, T
H
2 cell-associated cytokines, such as
IL-4 and IL-13, inhibit autophagy
16
. The signalling path-
ways that lead to the inhibition of autophagy by IL-4
and IL-13 are context dependent: they occur through
AKT signalling when autophagy is induced by starva-
tion, and are AKT independent and signal transducer
and activator of transcription 6 (STAT6) dependent
when autophagy is induced by IFN
16
. IL-10 can also
inhibit autophagy through AKT signalling
21
. Moreover,
STAT3, which transduces signals downstream of vari-
ous signals including IL-6, can inhibit autophagy
22
. This
inhibitory pathway does not involve transcriptional
signalling, but instead involves the binding of cyto-
plasmic STAT3 to protein kinase R (PKR; also known
as IFN-induced double-stranded RNA-activated protein
kinase)
22
. STAT3 interacts through its SH2 domain with
PKR and inhibits it. This prevents PKR from facilitating
autophagy through the hyperphosphorylation of eukary-
otic translation initiation factor 2 (EIF2; also known
as EIF2S1) and the subsequent inhibition of cellular and
viral protein synthesis
22
.
Reactive oxygen species (ROS) are classical anti-
microbial effectors, which have an important role in
immune signalling. ROS, and the oxidases that generate
them, affect autophagy
2327
(BOX2; FIG.1a). NADPH oxi-
dase and the autophagic machinery are connected via an
autophagy regulatory protein known as RUBICON (run
domain beclin 1-interacting and cysteine-rich-containing
protein), which interacts with beclin 1 and the PtdIns
3

kinase catalytic subunit VPS34. By physically disassociat-
ing from the autophagy inhibitory complex and associating
with the NADPH oxidase activating complex, RUBICON
activates two bactericidal mechanisms (that is, autophagy
and ROS production), although it is not known whether
the activation of these two mechanisms occurs simulta-
neously
26
. Nitric oxide inhibits autophagy by inactivating
Jun N-terminal kinase1 (JNK1) and IKK through direct
S-nitrosylation
27
. This prevents JNK1-dependent BCL-2
phosphorylation
27
and BCL-2 dissociation from beclin
1
28,29
and inhibits IKK-associated AMPK-dependent
autophagy initiation
27,30
.
Figure 1 | Four principal roles of autophagy in immunity. a | The role of autophagy
in the elimination of microorganisms is shown. An incoming microorganism can induce
autophagy by competing for nutrients or by stimulating innate immune receptors,
such as Toll-like receptors (TLRs). When the microorganism is taken up by phagocytosis
and remains in an intact vacuole, an autophagic process termed LC3-associated
phagocytosis (LAP) can promote the maturation of autophagosomes into autolysosomes.
Xenophagy of pathogens that enter the cytosol can be initiated by sequestosome 1-like
receptors (SLRs) or other mechanisms, including nucleotide-binding oligomerization
domain-containing protein 2 (NOD2)autophagy-related protein 16-like 1 (ATG16L1)
interactions. b | Several examples of the role of autophagy in the control of
pro-inflammatory signalling (see also FIG.3) are shown. Failure to remove SLRs by
autophagy can increase the levels of these receptors and the levels of pro-inflammatory
signalling. Autophagy can deliver cytoplasmic pathogen-associated molecular patterns
(PAMPs) to endocytic TLRs and can stimulate their activity. NOD-like receptors (NLRs;
such as NLRP3 (NOD-, LRR- and pyrin domain-containing protein 3)) and RIG-I-like
receptors (RLRs; such as absent in melanoma 2 (AIM2)) show complex positive and
negative co-regulation with autophagy: the ATG5ATG12 complex inhibits retinoic
acid-inducible geneI (RIG-I) signalling, and autophagy limits inflammasome activation
by removing damaged mitochondria, which then release the inflammasome activators
reactive oxygen species (ROS) and mitochondrial DNA. c | The role of autophagy in
adaptive immunity is shown. Autophagy can increase the MHC classII presentation
of cytoplasmic antigens, including self or viral antigens, as well as promoting the
citrullination of antigens. LAP can enhance the processing of particulate antigens
for MHC classII presentation. NOD2 enhances autophagic antigen presentation.
Autophagy may directly or indirectly affect MHC classI presentation by competing with
the proteasome for substrates, by influencing the peptidome pools through the control
of levels of components of microRNA (miRNA) machinery (for example, argonaute (AGO)
and DICER), or by supporting unconventional MHC classI presentation. In addition,
autophagy affects the self-renewal of haematopoietic stem cells (HSCs), B1 cell
development, plasma cell survival and IgG secretion. Autophagy affects Tcell survival
following Tcell receptor (TCR) activation, and it destabilizes the immunological synapse.
It also controls innate immune cell (such as macrophage) signalling through the release
of interleukin-1 (IL-1) and IL-1, which influence the polarization of Tcells into
Thelper 17 (T
H
17) cells. Autophagy also affects naive Tcell repertoire selection in the
thymus and the survival and function of maturing Tcells by removing the mitochondria
and endoplasmic reticulum (ER), thus ensuring calcium homeostasis. d | The role of
autophagy in the secretion of immune mediators is shown. Autophagy affects the quality
of regulated secretion from pre-stored granules. Autophagy affects the quality and the
quantity of the output of the constitutive secretory pathway (which is the conventional
pathway of protein secretion via the ER, the Golgi apparatus and the plasma membrane).
Autophagy supports a form of unconventional secretion that captures cytoplasmic
proteins for extracellular release. Note that secretory protein cargo in the regulated
and constitutive secretory pathways contains conventional leader peptides for
co-translational import into the ER lumen, whereas protein cargo that enters the
unconventional secretory pathway lacks leader peptides and does not enter the ER.
Dashed arrow indicates that this pathway remains to be defined. DAMPs, damage-
associated molecular patterns; HMGB1, high-mobility group box1 protein; IPS1, IFN
promoter stimulator protein 1; mtDNA, mitochondrial DNA; mTOR, mammalian target
of rapamycin; NOX2, NADPH oxidase 2; PRRs, pattern recognition receptors;
ssRNA, single-stranded RNA; TRAF6, TNF receptor-associated factor 6.
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 725
2013 Macmillan Publishers Limited. All rights reserved
SNARE
(Soluble NSF attachment
protein receptor). A member
of a class of proteins that
catalyse membrane fusion
and thus regulate organelle
identity and vesicular
trafficking. The membrane
fusion occurs through the
formation of a cognate Qa-,
Qb-, Qc- and R-SNARE
four-helix bundle, which
consists of SNAREs on donor
and acceptor membranes.
Mammalian target of
rapamycin
(mTOR). A serine/threonine
protein kinase that regulates
cell growth and metabolism.
mTOR is stimulated by growth
factor receptor and phosphati-
dylinositol-3,4,5-phosphate-
dependent signalling. It
responds to the availability of
nutrients (for example, amino
acids). Active mTOR inhibits
autophagy via the serine/
threonine protein kinase ULK1.
AMP-activated protein
kinase
(AMPK). A sensor of cellular
energetic state. It is activated
by increased AMP levels which
indicate decreased energy
status following hypoxia or
nutrient deprivation.
Xenophagy
The selective degradation
of intracellular pathogens
(such as bacteria or viruses)
through macroautophagy.
LC3-associated
phagocytosis
(LAP). A shared pathway that
involves conventional
phagocytosis and autophagy
at the maturation stage that is
mediated by the recruitment
of the autophagy protein LC3
(which is the mammalian
homologue of yeast ATG-8).
LAP results in a more robust
phagolysosome, which can
also function as a specialized
signalling compartment or
an antigen-presentation
compartment.
The connections between autophagy and immune
signalling seem to be surprisingly complex. From an
evolutionary perspective, these connections reflect the
integration of autophagy with several immune regulatory
systems.
Direct elimination of microorganisms
Autophagy intercepts pathogen invasion. The antimi-
crobial functions of autophagy provide a series of barri-
ers against invading microorganisms (FIG.1a). The first
antimicrobial function is xenophagy, which is the uptake
of intracellular microorganisms into double-membrane
autophagosomes
4
, and the second is LC3-associated
phagocytosis (LAP), which involves the engagement
of the autophagic machinery while the bacterium is
confined in the nascent and presumably intact phago-
some
24,31,32
(BOX3). Finally, a group of autophagic adap-
tors
33
, known as sequestosome 1-like receptors (SLRs), are
involved in eliminating microorganisms from the cyto-
plasm. SLRs recognize molecular tags (such as ubiqui-
tin, galectin and membrane phospholipid modifications)
present on invading microorganisms or on damaged
host membranes that are associated with the pathogen
and that physically recruit the autophagic machinery
3438

(FIG.2). The importance of autophagy in protecting the
cytoplasm from microbial invasion is highlighted by the
microbial countermeasures and adaptations that have
evolved to inhibit, to block specific stages or to com-
mandeer autophagy
39
(see BOX4 for recently discovered
examples).
Box 2 | Parallels between metabolic and immune signalling in autophagy activation
The model (see the figure) proposes that immune and nutritional signals converge to activate a similar cascade. Nutritional
triggers, such as starvation, are transduced via mammalian target of rapamycin (mTOR) and AMP-activated protein kinase
(AMPK). Starvation inhibits mTOR and activates AMPK, which in turn activates the serine/threonine protein kinase
ULK1 to phosphorylate beclin 1 (REF.140), activating molecule in BECN1-regulated autophagy protein 1 (AMBRA1)
10

and TAK1-binding protein 2 (TAB2)
141
. In the resting state, TAB2 and TAB3 bind beclin 1 and repress its activity
30,140
.
TAB2 activates TGF-activated kinase 1 (TAK1)
30
to further enhance AMPK activity. AMBRA1 (REF.10), along with
beclin 1 (REF.107), recruits TNF receptor-associated factor 6 (TRAF6). TRAF6 functions as an E3 ubiquitin ligase, generating
polyubiquitin chains that in turn stabilize and activate ULK1 (REF.10) and beclin 1 (REF.107), and that probably activate
the TAB2 and TAB3TAK1 complexes. Once fully amplified, the concomitant activation of ULK1 and beclin 1 leads to
autophagy through both branches
of autophagy regulatory kinases
that is, the protein kinase ULK1 (via
AMPK) and the lipid kinase VPS34
(via beclin 1) (not shown)
107
.
Immune signals activate
autophagy and engage at least
some components of the
pathways, including ULK1
(REF.108), beclin 1 (REF.142) and
TRAF6 (REF.107). Downstream of
the TAB2 and TAB3TAK1
complexes, IB kinases (IKKs)
affect autophagy independently of
nuclear factor-B
143
. TANK-binding
kinase1 (TBK1) also controls
autophagy
14,77
and promotes the
elimination of intracellular
microorganisms independently of
its role in typeI interferon (IFN)
activation (not shown)
144
. TBK1
controls both the capture of
autophagic cargo (for example,
microorganisms) via specialized
adaptor proteins
14,35,36,45
and the
maturation of autophagosomal
organelles into degradative
compartments
14,77
. TBK1 may also
contribute to autophagosome
formation at stages earlier than
autophagosome maturation
77
.
BCL-2, B cell lymphoma 2; DAPK1, death-associated protein kinase 1; HMGB1, high-mobility group box1 protein; IL-1,
interleukin-1; IRGM, immunity-related GTPase family M protein; JNK1, Jun N-terminal kinase 1; NLRP, NOD-, LRR- and pyrin
domain-containing protein; NOD2, nucleotide-binding oligomerization domain-containing protein 2; NOX2, NADPH oxidase 2;
P,phosphorylation; RAPTOR, regulatory-associated protein of mTOR; RIPK2, receptor-interacting serine/threonine-protein kinase 2;
ROS, reactive oxygen species; RUBICON, run domain beclin 1-interacting and cysteine-rich-containing protein; TLR4, Toll-like
receptor 4; Ub, ubiquitylation.
P
P
P
Nature Reviews | Immunology
Starvation AMPK TAK1
JNK1
ULK1
P
BCL-2
P
Beclin 1
AMBRA1
TRAF6
TAB2 and
TAB3
P
RAPTOR
mTOR
NOD2 RIPK2
IFN
HMGB1
Cardiolipin
RUBICON
TLR4
IL-1
CD40
NLRP3,
NLRP4 and
NLRP10
DAPK1
IRGM
NOX2 ROS
P
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Autophagy
REVI EWS
726 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Sequestosome 1-like
receptors
(SLRs). These are autophagic
adaptors that recognize
microbial targets and that link
them to autophagy machinery
by binding to mammalian
autophagy-related protein 8
(ATG-8) proteins (for example,
autophagy related LC3
proteins and -aminobutyric
acid receptor-associated
proteins (GABARAPs)).
Crohns disease
A form of chronic inflammatory
bowel disease that can affect
the entire gastrointestinal
tract, but is most common
in the terminal ileum. It is
characterized by transmural
inflammation, strictures and
granuloma formation.
To initiate autophagy, mammalian cells detect the
presence, location and extent of the cytoplasmic inva-
sion by a pathogen. Conventional PRRs can elicit
autophagic responses at different stages of the host
pathogen encounter
40
. First, TLRs and NOD-like recep-
tors (NLRs) detect released microbial products (that is,
PAMPs) very early following infection and this stimu-
lates autophagy. Second, autophagy can be initiated
during adhesion- and pathogen-induced uptake of bac-
teria by the host cell, or during active phagocytosis of
bacteria by macro phages
31,41,42
. Third, at stages after bac-
terial uptake, autophagy is induced following pathogen-
inflicted damage to the newly formed parasitophorous
vacuoles
35,36,38,4345
and on the escape of bacteria into the
cytoplasm
34,46,47
. Similarly, the initiation of autophagy
following virus infection occurs at various stages of the
virus life cycle
25,4857
(FIG.1a).
TLRs and autophagy cooperate in the response to
PAMPs. The early induction of autophagy downstream of
TLR stimulation ensures the prompt upregulation
ofantimicrobial activities, including the upregulation of
autophagy systems in advance of microbial invasion
58,59

(FIG. 1a). This and other triggers, such as IFN and
1,25-dihydroxyvitaminD3, may promote the expres-
sion and the early delivery of antimicrobial peptides
to the parasitophorous vacuoles
19,6062
, or they may
direct the autophagic response to the points of micro-
bial entry, as occurs during LAP
31
(FIG.1a). In addition,
autophagic membranes deliver cytoplasmic PAMPs,
such as single-stranded viral RNA, to the endosomal
lumen where they can make contact with the luminal
portion of endosomal TLRs, such as TLR7, to stimulate
other responses, including the typeI IFN response
50

(FIG.1b). Autophagy that is stimulated by TLRs, in con-
junction with LAP
53,63
, enhances antigen presentation
by dendritic cells (DCs) (FIG.1c). LAP also contributes
to the trafficking of TLR9 into specialized IFN signal-
ling compartments in plasmacytoid DCs (pDCs) (FIG.3).
Thus, TLRs and autophagy influence each other, which
amplifies the outputs of both systems in response to
microbial invasion.
NLRs interact with ATGs to localize autophagy. The
cooperation between NLRs and autophagy in antimi-
crobial defence is conserved from flies
64
to humans
41,42
.
NOD1 (nucleotide-binding oligomerization domain-
containing protein 1) and NOD2 detect muramyl pep-
tides in the cytoplasm and they direct the autophagic
machinery by recruiting ATG16-like 1 (ATG16L1) to
the plasma membrane at the site of bacterial entry
41

(FIG.1a). The NOD2-assisted localization of ATG16L1
at the plasma membrane is consistent with the role of
ATG16L1 in the formation of a portion of autophagic
precursors from the plasma membrane
65
(BOX1). Of
note, polymorphisms at the ATG16L1 and NOD2
loci have been associated with an increased risk of
Crohns disease
66
. Cells from donors who are homozy-
gous for the Crohns disease risk allele ATG16L1*300T
have a decreased capacity for autophagy induction in
response to the NOD2 agonist muramyl dipeptide.
The truncation of NOD2 that occurs in patients with
Crohns disease renders it cytoplasmic, which results in
the retention of ATG16L1 in the cytoplasm. This pre-
cludes ATG16L1 recruitment to bacterial entry sites
and thereby prevents timely and site-specific control of
autophagy.
Other NLRs also interact with the autophagic machin-
ery. NLRX1 and its interacting partner mitochondrial
Tu elongation factor (TUFM), which associates with
ATG5ATG12 complexes and with ATG16L1, promote
autophagy
67
(FIG.3b). NLRC4 (NOD-, LRR- and CARD-
containing protein 4), NLRP3 (NOD-, LRR- and pyrin
domain-containing protein 3), NLRP4 and NLRP10
interact with beclin 1 (REF.68) (BOX2). The recruitment
of NLPR4 to the plasma membrane during the phago-
cytosis of group A streptococci leads to its transient dis-
sociation from beclin 1. This enables the initiation of
beclin 1-mediated autophagic responses
68
. Collectively,
NLRs may gather autophagy factors in the vicinity of the
incoming microorganism (or resident mitochondria),
which leads to the activation of autophagy.
Nucleic acid sensors and autophagy. RIG-I-like receptors
(RLRs), including retinoic acid-inducible geneI (RIG-I),
melanoma differentiation-associated protein 5 (MDA5;
also known as IFIH1) and LGP2 (also known as DHX58),
recognize viral RNA to induce the production of typeI
IFNs and other pro-inflammatory cytokines. Unlike other
PRRs, RLRs seem to negatively regulate autophagy
67,69
.
However, downstream molecules that are engaged by
RLRs may have an RLR-independent (and possibly
competing) function in the induction of autophagy; for
example, downstream of nucleic acid sensing by RLRs,
the cytoplasmic adaptor protein stimulator of IFN genes
protein (STING) activates TANK-binding kinase 1
(TBK1) signalling and typeI IFN production, but can
also induce autophagy, possibly by the direct recogni-
tion of DNA or second messengers that are associated
with the presence of cytoplasmic DNA. Indeed, infec-
tion with double-stranded DNA viruses, such as herpes
simplex virus 1 (HSV1) or human cytomegalovirus,
induces autophagy that is dependent on STING
70,71
.
Similarly, bacterial DNA that is released during infection
Box 3 | LAP: a shared maturation step for autophagy and phagocytosis
Autophagy is often morphologically defined by the formation of double-membrane
autophagosomes in the cytoplasm. This is the geometrical consequence of the
formation of a phagosome that is derived from a pre-existing membranous intracellular
compartment, such as the endoplasmic reticulum (ER), rather than the consequence of
a functional requirement for a double membrane. An exception to this rule is the
formation of standard phagolysosomes by the autophagy-related protein (ATG) LC3,
through a process known as LC3-associated phagocytosis (LAP)
24,31,32,53
. LAP occurs
following the uptake of various extracellular targets: particles coated with Toll-like
receptor (TLR) agonists, phagocytosed dead cells, live epithelial cells that are
engulfed via entosis by neighbouring cells, and the Fc receptor-dependent uptake
of immune complexes. When a microorganism or a TLR ligand is taken up by
conventional phagocytosis
31,53
, the autophagic machinery enhances the maturation
of the conventional phagosomes through the same maturation pathway that is used
in internally formed autophagosomes. Thus, LAP uses beclin 1VPS34 complexes and
LC3 conjugation systems. LAP is independent of the serine/threonine protein kinase
ULK1 (REF. 32) ULK1 is only required to generate autophagosomes from internal
ER membranes during starvation.
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 727
2013 Macmillan Publishers Limited. All rights reserved
Nature Reviews | Immunology
PB1
SKICH
SKICH
SKICH
ZZ
GIR
CC
CC
CC CC CC
CC CC
CC
CC
UBA
UBA
Parasitophorous
vacuole
Bacteria
Galectin
PAMP
LC3C
LC3 or
GABARAP
NDP52
LRSAM1
(E3 ligase)
Optineurin
Isolation
membrane
-galactoside
LIR
LIR
CLIR
FW
PB1 ZZ
NLS
Sequestosome 1
NBR1
NDP52 (human)
NDP52 (mouse)
TAX1BP1
Optineurin
NES
a Sequestosome 1-like receptors b
LIR
X/(D,E,S,T)-X/(D,E,S,T)-X/(D,E,S,T)-W/F/Y-X/(D,E,S,T)-X-L/I/V
332 342
TRAF6-interacting region KIR
UBZ
UBZ
UBAN ZnF
UBZ
TBK1
RAB8B
TBK1
TBK1
RAB8B
Sequestosome 1
multimer
Ubiquitylation
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Ub
Ub
P
P
Ub
can induce autophagy via STING
45
(FIG.3e). Such bacte-
rial DNA can escape from M.tuberculosis-containing
phagosomes through pores that are introduced by the
bacterial secretion system ESX1 (REF.45). It has recently
become evident that, in response to cytoplasmic DNA,
host cell systems generate a second messenger, cyclic
GMPAMP (cGAMP)
72
(FIG.3e). This cyclic dinucleotide
is generated by host cGAMP synthase and activates the
typeI IFN pathway via STING
72
. It remains to be shown
whether cGAMP production in response to bacterial and
Figure 2 | Autophagy-mediated clearance of intracellular pathogens. a | Protein domains
of sequestosome 1-like receptors (SLRs) are shown. The LC3-interacting region (LIR) motif
of an SLR binds to autophagy-related LC3 proteins through its consensus sequence at
amino acids 332342 in sequestosome 1 (also known as p62). The conserved residues are
shown. X/(D,E,S,T) indicates that any amino acid (X) is allowed but that acidic (D,E) or
phosphorylatable amino acids (S,T) are often present (usually at least one or more within
the entire consensus sequence). The core LIR motif residues are aromatic pocket-filling W
(or F or Y) residues and aliphatic pocket-filling L (or I or V). They form an intermolecular
parallel -sheet with LC3 proteins or -aminobutyric acid receptor-associated proteins
(GABARAPs). The CLIR motif, which is a LIR motif that is specific for LC3C, lacks the aromatic
residue found in the LIR motif and, instead, uses hydrophobic contacts provided by
additional aliphatic residues located between the W and L position anchors to stabilize
interactions with LC3C. All human SLRs also contain a ubiquitin-binding domain (UBD):
UBA (as found in sequestosome 1 and NBR1) is a three-helix bundle UBD that has affinity
for monoubiquitin and K63 ubiquitin linkages; UBAN (as found in optineurin) is a parallel
coiled-coiled dimer UBD that has specificity for linear ubiquitin chains; and UBZ
(as found in nuclear dot protein52 (NDP52)) is a zinc finger -fold UBD that binds to
monoubiquitin and polyubiquitin. b|Amodel of cooperative action between different
SLRs and E3 ligases in bacterial targeting for xenophagy is shown. The schematic shows a
parasitophorous vacuole with glycosylated molecules (in this case -galactosides) facing
the lumen of the vacuole that contains a bacterium and that is experiencing membrane
damage. This membrane tear exposes -galactosides to galectins (for example, galectin 8)
which in turn bind to the galectin-interacting region (GIR) motif of NDP52. NDP52 also
directly interacts with the E3 ligase LRSAM1 and indirectly with the serine/threonine protein
kinase TANK-binding kinase 1 (TBK1), which interacts with optineurin. The CLIR motif of
NDP52 binds to LC3C, which is a proposed initiator in the LC3 and GABARAP cascade
during bacterial xenophagy. LRSAM1 or other E3 ubiquitin ligases polymerize ubiquitin at
molecular targets that are yet to be identified. The hypothetical model includes the putative
recognition of bacterial pathogen-associated molecular patterns (PAMPs) by LRSAM1
through its leucine-rich repeat domain. Ubiquitin tags are recognized by UBDs of NDP52,
optineurin and sequestosome 1. The LIR motif of optineurin is phosphorylated by TBK1 and
this improves LC3 and GABARAP binding. The UBA of sequestosome 1 is also phosphorylated
by TBK1 and this improves ubiquitin chain binding. As a consequence, the autophagic
isolation membrane initiates at the appropriate location and grows to capture the bacterium
and to eliminate it through autophagy. CC, coiled-coil domain; FW, four W domain (also
known as the NBR1 domain); KIR, KEAP1 interacting region; NES, nuclear export signal; NLS,
nuclear localization signal; P, phosphorylation; PB1, protein-binding domain 1; SKICH, skeletal
muscle and kidney enriched inositol phosphatase carboxyl homology domain; TAX1BP1,
TAX1-binding protein 1; Ub, ubiquitylation; ZnF, zinc finger domain; ZZ, ZZ-type ZnF domain.
REVI EWS
728 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Ubiquitin-binding domains
(UBDs). Domains that have
different specificities for
ubiquitin chains: UBA, which
is found in sequestosome 1,
favours K63 polyubiquitin
chains; UBZ, which is found
in nuclear dot protein 52
(NDP52), binds to
monoubiquitin and
polyubiquitin; and UBAN,
which is found in optineurin,
has specificity for linear
ubiquitin chains.
LC3-interacting region motif
(LIR motif). Canonical LIR
motif, which is X/(D,E,S,T)-X/
(D,E,S,T)-X/(D,E,S,T)-W/F/Y-X/
(D,E,S,T)-X-L/I/V, where X/
(D,E,S,T) indicates that any
amino acid (X) is allowed
but that acidic (D,E) or
phosphorylatable amino
acids (S,T) are often present.
Aromatic residues (W/F/Y)
occupy the aromatic pocket
and aliphatic side chains (L/I/V)
occupy an aliphatic pocket
present in all autophagy-
related protein 8 (ATG-8)
homologues. The LIR motif
forms an intermolecular
-sheet with ATG-8 homologues,
not discriminating between
mammalian paralogues of
yeast ATG-8.
CLIR motif
A variant of the LC3-interacting
region (LIR) motif (L-V-V),
which is found in nuclear
dot protein52 (NDP52).
The CLIR motif binds to the
LIR-interacting region in LC3C.
The CLIR motif lacks the
signature aromatic residue of
the LIR motif. Instead, it makes
compensatory hydrophobic
contacts with LC3C.
host-derived cytoplasmic DNA can induce autophagy.
Analogous cyclic dinucleotides that are secreted by
bacteria (for example, cyclic di-GMP or di-AMP) bind
to STING and can activate autophagy when they are
introduced into the host cell cytoplasm
45
(FIG.3e).
In conclusion, conventional PRRs recognize microbial
products at different stages of an infection and initiate
an autophagic response that contributes to the efficient
elimination of the invading microorganism. However,
in some instances, autophagy can suppress typeI IFN
responses, notably by inhibiting RLR signalling
67,69
. This
relationship paradoxically promotes viral replication, and
may explain why autophagy can have a pro-infectiverole.
SLRs clear microorganisms from the cytoplasm. If a path-
ogen escapes the autophagy barriers that are controlled
by conventional PRRs, it can still be captured in the cyto-
plasm or even in the cytosol. The autophagy-mediated
elimination of cytoplasmic pathogens depends on special-
ized adaptors known as SLRs
20,3436,46,73,74
(FIG.2a). SLRs,
which are named after the archetypical protein seques-
tosome 1 (also known as p62) include sequestosome 1
(REF.75), NBR1 (REF.76), optineurin
36
, nuclear dot protein
52 (NDP52; also known as CALCOCO2; it is full size
in humans but truncated in mice)
35
and an NDP52-like
receptor TAX1-binding protein 1 (TAX1BP1; also known
as CALCOCO3)
77
(FIG. 2a). SLRs have been shown to
affect inflammation in several ways
34,74,78
and, as they can
be consumed during autophagy
33
, it is possible that SLR
accumulation (following autophagy inhibition) or SLR
depletion (following autophagy activation) may modulate
inflammatory processes (FIG.1b).
SLRs contain one or more cargo recognition domains
(CRDs) that recognize ubiquitin-tagged
20,3436,45,46,73

or galectin-tagged
38,79
microbial or microorganism-
associated targets. Ubiquitin-binding domains (UBDs)
that are specific for different ubiquitin chains (UBA,
UBZ or UBAN)
80
are present in all known SLRs. SLR
CRDs that recognize tags other than ubiquitin include
a hook-like CRD (which is the galectin-interacting
region (GIR) motif ) that enables NDP52 to interact
with galectin 8 (REF.79) (FIG.2a). Galectin 8 links NDP52
to cytoplasmically exposed -galactoside glycans on
pathogen-damaged host membranes
38
(FIG.2b).
Moreover, all SLRs have an LC3-interacting region
motif (LIR motif), which is X/(D,E,S,T)-X/(D,E,S,T)-X/
(D,E,S,T)-W/F/Y-X/(D,E,S,T)-X-L/I/V, where X/
(D,E,S,T) indicates that any amino acid (X) is allowed
but that acidic (D,E) or phosphorylatable amino acids
(S,T) are often present (at least one or more within the
entire motif)
33
. A modified LIR, known as a CLIR motif
(L-V-V) has been identified in NDP52 (REF.81) and pos-
sibly in TAX1BP1 (REF.77) (FIG.2a). Phosphorylation of
the LIR motif and of the CRDs of SLRs modulates their
autophagic activities
14,36,82
(FIG.2b). The number of SLRs
and the type of unique or repetitive structures that are
recognized as tags may increase as more research is
carriedout.
The process of autophagic clearance of microorgan-
isms from the cytoplasm involves the sensing of DAMPs
and PAMPs, and often the function of multiple SLRs. An
example of autophagy-inducing DAMPs are the glycans
that become exposed following damage to host mem-
branes, as in the case of infection with Salmonellaspp.
38,79
.
Box 4 | Microbial countermeasures against autophagy
Microorganisms use a wide range of mechanisms to prevent, to counteract or to commandeer autophagy
39
.
These inhibitory mechanisms involve the targeting of beclin 1 (REFS49,145), the inhibition of autophagosomal
maturation
51,57
, the perforation of autophagosomal membranes to prevent acidification
44
, the proteolytic cleavage
of autophagy-related protein 8 (ATG8) to irreversibly remove carboxy-terminal lipid modifications
146
, and the
masking of epitopes or tags that are recognized by sequestosome 1-like receptors (SLRs)
46,147
. Autophagy may
even be activated to generate nutrients for invading microorganisms
148
.
With regard to bacteria, Listeria spp. proteins AktA
46
and InlK
147
interfere with recognition via host ubiquitin tags,
and Shigella spp. protein IcsB masks bacterial epitopes
149
, which are recognized by the autophagic machinery, either
directly or by recruiting cytoplasmic proteins. The Salmonella spp. deubiquitinase SseL removes ubiquitin tags
88
.
The Legionella spp. effector protein RavZ is injected into the host cytoplasm through the bacterial typeIV secretion
system and inhibits autophagy through irreversible deconjugation of mammalian homologues of ATG-8. This involves
the proteolysis of the C-terminal glycine in ATG-8 homologues, which prevents future phosphatidylethanolamine
modification
146
. Listeria spp. block autophagosome acidification through the pore-forming toxin listeriolysin O
44
.
Bacterial actin-based intracytoplasmic motility may be another factor in evading capture by autophagosomes.
This is supported by the role of septin scaffolds in enabling autophagy of cytoplasmic Shigella spp. as they start
to polymerize actin
20
.
Viruses also interfere with autophagy; for example, herpes simplex virus 1 infected cell protein 34.5 (ICP34.5)
49
,
influenza virus M2 protein
52
and HIV protein Nef
51,57
, all target beclin 1 to either completely block autophagy or
to inhibit autophagosomal maturation. Nef binds to the evolutionarily conserved domain of beclin 1 at the
same region as the endogenous inhibitor GAPR1 (Golgi-associated plant pathogenesis-related protein 1)
57
.
Mouse herpesvirus 68-encoded protein M11 (which is the viral homologue of B cell lymphoma 2) inhibits beclin 1
through its BH3 domain
145
, whereas Kaposi sarcoma-associated virus (KSHV) FLICE-like inhibitory protein (FLIP)
blocks ATG3 in the LC3 lipidation cascade
127
. HIV Nef, hepatitis C virus NS3 and measles virus Mev3 interact with
autophagy factor immunity-related GTPase family Mprotein (IRGM), which has consequences that are yet to be
fully understood
150
.
Finally, several microorganisms can harness autophagy for their own benefit, including using autophagy for
replication
39
; for example, the obligatory intracellular pathogen Anaplasma phagocytophilum uses its typeIV
secretion effector Ats1 to activate autophagy to supply nutrients
148
.
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 729
2013 Macmillan Publishers Limited. All rights reserved
Nature Reviews | Immunology
B cell
BCR
DNA-containing
antigen
a
d e
b c
pDC
Autophagy
Autophagy
Autophagy
Degradation
of BCL-10 by
autophagy
RLR signalling
Autophagy
Autophagy
Depolarized
mitochondria
ROS
IL-1 IL-1
Calpain Inammasome
Type I IFNs
Virus
Mitochondria
Cyclic di-AMP
Cyclic di-GMP
Bacteria
TLR9
LAP
B cell activation Type I IFN
Autoimmune plasma cells
ATG12
cGAMP
STING
TBK1
Type I IFN
ATG9
RIG-I
BCL-10
NF-B
MALT1
Beclin 1
VPS34
BCL-10
BCL-10
Sequestosome 1
MALT1
CARD9
CARD9 RUBICON
RUBICON
RIG-I Dectin
IPS1
ATG5
TUFM
NLRX1
Type I
IFN
GMP
DNA
cGAMP
synthase
AMP
Ub
Leucine-rich repeat
(LRR). A domain that is often
found in pattern recognition
receptors and that is involved
in pathogen-associated
molecular pattern recognition.
LRRs are repeats of L-X-X-L-X-
L-X-X-N-X-L or L-X-X-X-L-X-L-X-
X-C-X-X-L motifs (where X
represents any amino acid),
which form a horseshoe or
solenoid tertiary structure.
Moreover, microbial polymers, such as DNA, that are
present in the cytoplasm might function as autophagy-
inducing PAMPs, as occurs during M.tuberculosis infec-
tion
45
. In the cytoplasm, microorganisms may become
coated with ubiquitin
20,3436,45,46,73
,which may involve
specialized E3 ubiquitin ligases, such as LRSAM1, which
recognize Salmonella spp. through their leucine-rich
repeat (LRR) domains
83
. Following auto-ubiquitylation,
LRSAM1 ubiquitylates microbial targets that are yet to be
defined and possibly host molecules that are associated
with microorganisms. Any of these ubiquitin tags alone
or in combination may lead to the recruitment of SLRs
through their UBDs (FIG.2b).
There is further cooperation (FIG.2b) between the
recognition of the initial tears in microorganism-
harbouring vacuoles and ubiquitylation, as exemplified
in the case of Salmonella spp. infection. In this exam-
ple, LRSAM1 directly interacts with NDP52 (REF.83). In
human cells, NDP52 seems to function as an important
hub, as it can recognize LRSAM1 (REF.83), galectin 8
bound to -galactoside glycans
38,79
, ubiquitin tags
35
and
LC3C (also known as MAP1LC3C), which is required
for the acquisition of other LC3 proteins to mediate
autophagy
81
. The action of NDP52 is non-redundantly
reinforced by other SLRs that may be recruited to
Salmonella spp., such as optineurin (which localizes with
Figure 3 | Autophagy controls inflammatory processes. a | Autophagy promotes Toll-like receptor 9 (TLR9)
signalling in Bcells and typeI interferon (IFN) production by plasmacytoid dendritic cells (pDCs). b | The autophagy
protein complex autophagy-related protein 5 (ATG5)ATG12 inhibits RIG-I-like receptor (RLR) signalling by binding
to the caspase recruitment domains of retinoic acid-inducible geneI (RIG-I) and IFN promoter stimulatorprotein 1
(IPS1), which is the mitochondrial adaptor of RIG-I signalling. The NOD-like receptor X1 (NLRX1)-interacting partner
mitochondrial Tu elongation factor (TUFM) associates with the ATG5ATG12 complex to promote autophagy while
inhibiting RLR-dependent typeI IFN activation. c | Autophagy factors negatively regulate the caspase recruitment
domain-containing protein 9 (CARD9)B cell lymphoma 10 (BCL-10) mucosa-associated lymphoid tissue lymphoma
translocation protein 1 (MALT1) complex. RUBICON (run domain beclin 1-interacting and cysteine-rich-containing
protein), which is a binding partner and a negative regulator of beclin 1, inhibits CARD9, whereas sequestosome 1
leads to the degradation of BCL-10. d | Excessive production of reactive oxygen species (ROS) by depolarized
mitochondria that are not cleared by autophagy enhance RLR signalling. e | Viral, mitochondrial or bacterial DNA lead
to the activation of stimulator of IFN genes protein (STING), probably through cGAMP synthase and cyclic GMPAMP
(cGAMP) production, which increases the typeI IFN response. Autophagy removes sources of agonists that stimulate
STING, whereas autophagic factors (for example, ATG9) inhibit the activation of STING by affecting its cytoplasmic
translocation. Bacterial cyclic dinucleotides (di-AMP and di-GMP) can activate autophagy, thereby functioning as a
regulatory loop that amplifies the removal of infectious or endogenous irritants. BCR, B cell receptor; IL, interleukin;
LAP, LC3-associated phagocytosis; NK-B, nuclear factor-B; TBK1, TANK-binding kinase 1; Ub, ubiquitylation.
REVI EWS
730 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Mitophagy
A special form of autophagy
by which mitochondria (in a
damaged or depolarized
state) are engulfed by
autophagosomes and
degraded.
NDP52 in microdomains surrounding Salmonella spp.)
and sequestosome 1 (REF.36). These SLRs bind to ubiqui-
tin through their UBDs that have varied specificities for
ubiquitin chains
80
, and, through their LIR motifs, they
recruit the remaining LC3 proteins and GABARAPs to
mediate autophagosome formation. The affinity of SLRs
for LC3 and ubiquitin is further enhanced by phospho-
rylation of the LIR motif in optineurin and the UBD
in sequestosome 1 by TBK1 (FIG.2b). TBK1 is recruited
to these sites via protein complexes containing NDP52
(REF.35), optineurin
36
and RAB8B
14
, which is a small
GTPase that regulates membrane trafficking. The cargo
capture and autophagy is coordinated with maturation
into autolysosomes by RAB8B
14
. Notably, the NDP52
product that is encoded by the corresponding mouse
gene Calcoco2 is carboxy-terminally truncated (and lacks
motifs for ubiquitin and galectin binding) (FIG.2a) and
may be primarily expressed in embryonic tissues (see the
expressed sequence tags NCBI database). This suggests
that, in mice, the function of CALCOCO2 may be pri-
marily conferred through the recruitment of LRSAM1,
TBK1 and LC3, as the binding sites for these proteins are
retained in the truncated mouse protein. Alternatively,
NDP52 activities in mice may be compensated for by
the NDP52-like receptor CALCOCO3.
Other pathways to target cytoplasmic microorgan-
isms for autophagy. Autophagy-associated factors have
been shown to directly target microbial proteins
47
;
for example, ATG5 binds directly to a Shigella spp.
surface protein, VirG. The ATG5VirG interaction
only occurs when the VirG recognition epitope is not
masked by another Shigella spp. protein, IcsB (BOX4),
and requires the ATG5-binding partner TECPR1 (tec-
tonin -propeller repeat-containing protein 1) which
interacts with WIPI2 (WD repeat domain phospho-
inositide-interacting protein 2)
47
. WIPI2 is one of the
four mammalian PtdIns
3
P-binding ATG18 paralogues
and has a role in phagophore formation (BOX1). Of note,
the control of icsB-mutant Shigella spp. also requires
NOD1 activity to induce autophagy and to limit intra-
cellular growth
41
. Moreover, these mechanisms against
Shigellaspp. are further complemented by ubiquitylation
and SLR activity
20,34
. Taken together, these findings sup-
port the idea that autophagic defences are multilayered
and cooperative innature.
Other antibacterial pathways include the autophagy
factors beclin 1 (REF.84) and immunity-related GTPase
family M protein (IRGM) which directly bind to cardio-
lipin, a lipid that is only present in bacteria and mito-
chondria
17
. IRGM is necessary for the optimal induction
of autophagy
17
, is partially localized in mitochondria
(hence its affinity for cardiolipin)
17
and is a genetic
predisposition factor for Crohns disease
66,85,86
. The
E3ubiquitin-protein ligase SMURF1 is required for anti-
viral responses to Sindbis virus and to HSV1 (REF.87).
SMURF1 is involved in mitophagy
87
, which in some
respects resembles bacterial xenophagy. Surprisingly, the
HECT (homologous to the E6-AP C terminus) domain
of SMURF1 (and hence the region that has its E3 ligase
activity) is dispensable for mitophagy and, instead, its
phospholipid-binding domain C2 is required. In keep-
ing with the idea that membrane phospholipids or their
derivatives are potential tags or signalling intermedi-
ates, diacylglycerol has also been implicated in bacterial
autophagy
37
. Thus, the presence of an E3 ligase domain
in an autophagy-targeting factor does not necessarily
indicate that it ubiquitylates the cargo as a requirement
for selective autophagy.
In summary, we now understand in greater detail
how autophagy is activated in response to microbial
presence. Autophagy is controlled by nearly all classes
of PRRs and is also regulated by cytokines and recep-
tors that modulate innate and adaptive immunity.
Integration of the immune triggers is an obvious area
for continuing study, and it is already clear that the
antimicrobial functions of autophagy can be promoted
(for example, by T
H
1 cell-associated cytokines and
IL-1) or inhibited (for example, by T
H
2 cell-associated
cytokines) by immune mediators. At the intracellular
level, an idea has emerged of a graded, multitiered sys-
tem of autophagic responses that are commensurate
with the extent of microbial penetration. An interesting
recent development was the discovery of the hierarchi-
cal cooperation of several proteins in controlling anti-
microbial autophagy. However, gaps in our knowledge
remain, especially with regard to alternative, that is, not
ubiquitin-based, modes of target recognition. At pre-
sent, recognition of ubiquitin tags seems to be crucial
for autophagy, and this is further supported by evidence
of bacterial countermeasures that involve deubiquity-
lation
88
. Moreover, ubiquitin tags might only activate
autophagy after more specialized mechanisms of recog-
nition have failed. It remains to be determined whether
overly strong activation of antimicrobial autophagy
comes at a price, as exemplified by its interference with
typeI IFN signalling. Finally, although autophagy is a
mode of self defence that is available, in principle, to
any eukaryotic cell, the magnitude of its contribution
in complex organisms is cell type dependent, as shown
by the fact that it has an important role in protecting
neurons against HSV1 infection whereas it is absent
in keratinocytes of the vaginal mucosa
56
. Thus, at the
organismal level, tissue- and cell-specific engagement
of antimicrobial autophagy may be a keyfactor.
Control of inflammation
Genetic links with immune disorders. The role of
autophagy in inflammatory diseases was initially
established through genome-wide association stud-
ies (GWASs)
66
. Polymorphisms in autophagy-asso-
ciated genes, such as ATG16L1 and IRGM, are linked
to Crohns disease
66
. In addition to single nucleotide
polymorphisms, IRGM is an example of a gene dosage
(through copy number variants) correlation with a pre-
disposition to Crohns disease in human populations
85
.
One of the common IRGM polymorphisms in Crohns
disease leads to an escape from the negative regula-
tion of IRGM expression by a microRNA (miRNA)
86
.
A recent GWAS analysis of 75,000 individuals indicates
an overlap between susceptibility loci for inflammatory
bowel disease and mycobacterial infections
89
. Moreover,
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 731
2013 Macmillan Publishers Limited. All rights reserved
a link has been reported between Crohns disease and
the autophagy-targeting factor SMURF1 (REF. 89).
Furthermore, polymorphisms in ULK1 have also been
linked with Crohns disease
90
.
In addition, polymorphisms in autophagy-associated
genes have been associated with autoimmune dis-
orders. IRGM polymorphisms may be a risk factor
in systemic lupus erythematosus (SLE)
91
. Moreover,
GWASs have linked ATG5 variants with SLE
9294
and
asthma
95
. Rheumatoid arthritis has been associated
with variations in the PR domain-containing gene 1
(PRDM1)ATG5 intergenic region
96
. A new human
autophagy locus, which was first functionally identified
in Caenorhabditiselegans screens as epg5 and which was
shown in mice to be required for autophagosomal matu-
ration
97
, has been linked to the complex Vici syndrome
that includes immunodeficiency
98
. Thus, autophagy
shows clinical relevance as a result of genetic links with
immunological and inflammatory disorders.
Autophagy affects PRR-mediated typeI IFN signalling.
The autophagic machinery can amplify TLR signalling;
for example, autophagy enhances the delivery of cyto-
plasmic PAMPs to endosomal TLR7, which enables the
recognition of cytoplasmic viral replication intermedi-
ates and IFN production by pDCs
50
. However, the role
of autophagy in PAMPPRR amplification loops can
lead to aberrations and auto immunity. In the case of
TLR9, which normally responds to microbial unmethyl-
ated CpG DNA, the autophagic machinery can enhance
aberrant self DNA reactivity. This occurs via intracellular
trafficking events following Bcell receptor (BCR) stim-
ulation by self DNA-containing antigens
11
. Autophagy
promotes the fusion of DNA-containing antigen-loaded
BCR compartments and TLR9-containing endosomes
and leads to B cell hyper responsiveness
11
(FIG. 3a).
Similarly, in pDCs the autophagic machinery enhances
the trafficking of particulate self DNA-containing
immune complexes and of TLR9 to signalling-
competent compartments, which results in increased
IFN production
32
(FIG.3a). It is possible that these aber-
rant autophagic activities in Bcells and pDCs could
cooperate and lead to the generation of autoimmune
plasma cells (FIG.3a).
By contrast, there are examples of autophagy-related
factors that directly inhibit the formation or that sup-
press the activation of pro-inflammatory protein com-
plexes. The ATG5ATG12 complex negatively regulates
RLR signalling by directly binding to the caspase recruit-
ment domains (CARDs) of RIG-I and IFN promoter
stimulator protein 1 (IPS1; also known as MAVS, VISA
or CARDIF; it is a mitochondrial adaptor of RIG-I sig-
nalling)
69
(FIG.3b). RUBICON inhibits CARD9BCL-10
MALT1 (mucosa-associated lymphoid tissue lymphoma
translocation protein 1) signalling complexes by binding
to CARD9, which thereby terminates pro-inflammatory
signalling downstream of RIG-I or dectin 1 (REF.99)
(FIG.3c). The absence of autophagy amplifies RLR sig-
nalling through increased IPS1 levels as a result of an
accumulation of mitochondria, and increased pools of
depolarized mitochondria in the absence of mitophagy
are a source of ROS that also enhance RLR effects
25

(FIG.3d). NLRX1 promotes autophagy and inhibits the
RLR-dependent induction of typeI IFN signalling and
inflammation (FIG.3b); indeed, the two processes are
reciprocally affected by NLRX1 (REF.67). Finally, ATG9
negatively controls the trafficking of STING and sup-
presses the activation of TBK1 in typeI IFN signalling
in response to double-stranded DNA
100
(FIG.3e).
Collectively, these phenomena may represent feed-
back loops by which autophagy downregulates typeI
IFN responses following a period of productive induc-
tion or to increase the threshold for activation of typeI
IFN signalling. Alternatively, autophagic interference
with RLR signalling could reflect competition for limited
resources, such as TBK1, that are shared between typeI
IFN signalling and autophagy pathways
14,36,45,101
. These
phenomena, in addition to other factors that are not dis-
cussed here, could be an explanation for why autophagy
paradoxically enhances the replication of certain viruses.
Autophagy suppresses inflammasome activation. The
recognition that autophagy has an anti-inflammatory
function stems from the observation that the production
of IL-1 and IL-18 is increased in the absence of func-
tional ATG16L1 in a mouse model of Crohns disease
102
.
Inflammasomes are cytoplasmic complexes that respond
to PAMPs and DAMPs by inducing the proteolytic pro-
cessing and secretion of IL-1 and IL-18 (REF.103). An
inflammasome consists of pro-caspase 1, the adaptor
protein ASC and a sensor protein from the NLR family
(such as NLRP1, NLRP3, NLRC4, NLRP6 or NLRP12)
or from the PYHIN (pyrin and HIN domain-containing
protein) family (which includes absent in melanoma 2
(AIM2) and IFN-inducible protein 16 (IFI16))
103
.
Several convergent reports show that autophagy has a
negative role in inflammasome activation
104108
. Under
sterile conditions, autophagy clears the cytoplasm of
debris, protein aggregates and defective organelles that
can function as endogenous inflammasome agonists.
Studies suggest that basal levels of autophagy control the
set point for inflammasome activation
104,105
. If autophagy
is blocked, this leads to an accumulation of depolarized
mitochondria that leak endogenous inflammasome ago-
nists, such as mitochondrial DNA (which is detected,
at least partly, by AIM2) and ROS (which activate the
NLRP3 inflammasome)
104,105
(FIG.1b).
During infection, the removal of damaged mitochon-
dria may not be a passive process; for example, during
infection with influenza A virus, a NOD2receptor-
interacting serine/threonine protein kinase 2 (RIPK2)
pathway activates the autophagy factor ULK1 to main-
tain or to increase mitophagy and to thereby reduce
inflammasome activation
108
. This RIPK2-dependent
regulation of autophagy may complement the
RIPK2-independent action of NOD2 following the
recruitment of ATG16L1 to microorganisms
41
and both
processes may contribute to inflammatory syndromes
such as Crohns disease. Importantly, autophagy may
downregulate prolonged inflammasome activity by
removing aggregated inflammasome components
107
.
This depends on the recruitment of sequestosome 1 to
REVI EWS
732 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Exocyst
An evolutionarily conserved
protein complex that consists
of eight subunits and that is
best known for targeting
exocytic vesicles to sites of
docking and fusion at the
plasma membrane. It also
functions as a protein complex
assembly platform.
K63-ubiquitylated ASC
107
. A role for exocyst has been
implicated in the overall process
107
, but the process could
be self-starting, as sequestosome 1 associates with the
E3ligase TRAF6 (REF.78) that ubiquitylates beclin 1 to
initiate autophagy
9
; however, this has not been inves-
tigated. In summary, basal autophagy protects cells
from inadvertent inflammasome activation initiating
damaging sterile inflammation, whereas a deficiency in
autophagy causes increased IL-1 levels (FIG.1b).
Autophagy suppresses calpain-dependent IL-1 activa-
tion. Similarly to IL-1, IL-1 is synthesized as a cyto-
plasmic pro-form, and is processed by calpain or other
proteases before being actively secreted from cells or
passively released following cell death
109
. The activation
and the secretion of IL-1 involve caspase 1-dependent
and caspase 1-independent processes
109
. Autophagy-
defective (Atg5
fl/fl
LysMCre
+
) macrophages secrete
high levels of IL-1 through a calpain-dependent but
inflammasome-independent pathway
109
. This pheno-
type has been observed both invitro and invivo
45,109
.
The processing of pro-IL-1 by calpain is induced by
ROS that have been released from accumulated depo-
larized mitochondria in autophagy-deficient cells
109
.
Thus, ROS in autophagy-defective cells activate both the
inflammasome and the calpain pathways that lead to the
excess production of IL-1 and IL-1, respectively
104,109
.
IL-1 that has been released from ATG5-defective mac-
rophages has important consequences invivo, as it leads
to enhanced and prolonged T
H
17 cell responses in the
context of M.tuberculosis infection, which contributes to
lung tissue damage in a mouse model of tuberculosis
109
.
Autophagy and degradation of pro-inf lamma-
tory signalling factors. Autophagy factors inhibit
BCL-10-containing complexes
99
, and autophagy degrades
BCL-10 to reduce nuclear factor-B (NF-B) activation,
as shown in antigen-activated Tcells
110
. This occurs via
sequestosome 1 association with K63-polyubiquitylated
BCL-10 (FIG.3c). As mentioned above, BCL-10-containing
complexes are also inhibited by the binding of RUBICON
to CARD9 (REF.99). As RUBICON is a negative regulator
of autophagy, its translocation to BCL-10 complexes leads
to the direct inhibition of CARD9BCL-10MALT1
signalling and may also enhance the autophagic deg-
radation of these complexes (FIG.3c). NF-B signalling
may also be downregulated by autophagy via NSFL1C
cofactor p47, which is a protein that has a ubiquitin-
binding UBA domain and its orthologue in yeast binds
to ATG-8. This potential adaptor functions as a negative
regulator of IKK through the lysosomal (and presum-
ably autophagic) degradation of polyubiquitylated NF-B
essential modulator (NEMO)
111
. The targeting of NEMO
for autophagic degradation is specifically induced by the
murine cytomegalo virus protein M45; this suppresses
antiviral processes during infection
112
.
In summary, genetic and functional studies indicate
that autophagy is an anti-inflammatory mechanism
that affects numerous pathways. The most prominent
examples are the dampening of inflammasome acti-
vation and type I IFN signalling; however, broader
targeting of other inflammatory signalling pathways is
also becoming apparent. This is consistent with the idea
that autophagy eliminates microorganisms (and micro-
bial products or corpses an area that deserves more
investigation) and endogenous irritants to prevent exces-
sive inflammation. When autophagic clearance fails,
inflammation ensues as the bodys response to persistent
danger. This may be an explanation for the selection of
genetic alleles in autophagy-related factors that promote
inflammation at certain anatomical sites. This has been
recently shown in the case of the Crohns disease risk
factor ATG16L1, whereby a defect in autophagy may
paradoxically confer protection against uropathogenic
Escherichia coli (UPEC) by increasing the production
of pro-inflammatory cytokines and by promoting the
recruitment of innate immune cells to UPEC-infected
bladders in ATG16L1-hypomorphic mice
113
.
Autophagy in adaptive immunity
Autophagy in antigen presentation and Tcell responses.
Autophagy functions as a bulk topological inverter by
transporting proteins from the cytoplasm into the lumen
of antigen-processing compartments
54,114,115
(FIG.1c). This
property can be artificially exploited by fusing antigens
(such as influenza virus matrix protein 1) to LC3 to
enhance MHC classII-mediated antigen presentation
to CD4
+
Tcells
116
. The topological inversion principle
is more generally applicable as it supplies cytoplasmic
ligands to endosomal receptors; for example, autophagy
transfers viral RNA replication intermediates to lumenal
TLR7, as discussedabove.
The contributions of autophagy to MHC classII
presentation are not restricted to cytoplasmic pro-
teins and extend through LAP to exogenous antigens.
Autophagy augments the MHC classII-dependent pro-
cessing and presentation of phagocytosed extracellular
particulate antigens, as shown using PRR-stimulated
LAP of ovalbumin-coated latex beads in DCs
53
. PRR-
mediated enhancement of MHC class II presenta-
tion also occurs through the induction of autophagy
following NOD2 stimulation with bacterial muramyl
dipeptides
42
(FIG.1c).
Autophagy-enhanced MHC classII presentation
possibly has a role in the selection of naive Tcell rep-
ertoires in the thymus. In autophagy-deficient mice,
both positive and negative selection of CD4
+
Tcells
but not of CD8
+
T cells were affected, and this had
consequences for the selection of appropriate Tcell
receptor (TCR) repertoires and for the elimination
of autoreactive CD4
+
T cells
115
. Transplantation of
ATG5-deficient thymi into wild-type mice caused an
infiltration of autoreactive CD4
+
Tcells into multiple
organs and induced autoimmune colitis that was similar
to Crohns disease-related phenotypes
115
. These effects
have been attributed to the defective presentation of
cytoplasmic antigens by autophagy-deficient thymic
epithelial cells
115
. Other studies investigating DC func-
tion indicate that autophagy-dependent antigen pres-
entation is defective in DCs from patients with Crohns
disease that express the ATG16L1 or NOD2 disease risk
variants
42
. The contributions of autophagy-dependent
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 733
2013 Macmillan Publishers Limited. All rights reserved
MHCclassII antigen presentation in antimicrobial
defence are experimentally detectable in the context
of adaptive immunity against viral
54,114
and bacterial
63

infections. As a result of the enrichment of peptidy-
larginine deaminase in autophagosomes
117
, autophagy
promotes the presentation of citrullinated antigens that
may be relevant to autoreactivity (FIG.1c). This process
is constitutive in DCs and macrophages, but in Bcells it
only occurs following BCR stimulation
117
.
In contrast to this augmentation of MHC classII
antigen processing by autophagy, autophagic degrada-
tion promotes the disassembly of immunological syn-
apses
118
. Inhibition of ATG16L1 and IRGM expression
in DCs leads to hyperstable interactions with Tcells and
increases Tcell activation
118
. This activation might be
further enhanced in the Tcells by the lack of autophagic
degradation of BCL-10 (REF.110) (FIG.3c). Therefore,
although autophagy initially promotes MHC classII
antigen processing, at later stages it may help to down-
regulate the response. Consistent with this idea, during
cancer cell invasion the epithelial-to-mesenchymal transi-
tion is associated with an attenuation of immunological
synapses between target cells and cytotoxic lympho-
cytes in a process that is dependent on autophagy and
beclin1(REF.119).
Autophagy affects conventional MHC classI presen-
tation by competing with the proteasome for the deg-
radation of newly synthesized cytoplasmic proteins
120

(FIG. 1c). Autophagy may also contribute to uncon-
ventional MHC classI presentation pathways in ways
that are yet to be fully understood
121,122
. In principle,
autophagy could broadly affect MHC classI peptide rep-
ertoires; for example, NDP52-dependent autophagy tar-
gets DICER and argonaute (AGO)
123
, which are involved
in miRNA processing and function. This might correlate
with the abundance of MHC classI-associated peptides
that seem to be derived from transcripts containing
miRNA response elements (FIG.1c).
Autophagy in Tcell homeostasis and T
H
17 cell polari-
zation. In addition to its role in antigen presentation,
autophagy affects both the homeostasis and the func-
tion of Tcells (FIG.1c). Autophagy and clearance of
mitochondria are needed for normal haematopoietic
stem cell (HSC) maintenance and function, which
are essential for the production of both myeloid and
lymphoid progenitor cells
124
. After exiting the thymus,
naive T cells depend on autophagy and mitochon-
drial content reduction for their maturation
125
. Tcells
require maintenance of an ER that is capable of calcium
homeostasis
126
. Calcium ions are sequestered in ER-like
structures that accumulate in ATG7-deficient Tcells.
This curtails calcium fluxes in response to TCR engage-
ment
126
. In naive Tcells, autophagy is suppressed by
cellular FLICE-like inhibitory protein (CFLIP; also
known as CFLAR)
127,128
, but TCR signalling and CD28
co-stimulation induce autophagy in activated Tcells
129
.
Autophagy is a pro-survival process in activated Tcells
and counteracts the pro-apoptotic function of CD95
(also known as FAS) and CD95L (also known as FASL),
which are upregulated by TCR stimulation
128
.
Following Tcell activation, autophagy can influence
T
H
cell polarization, partly by controlling the inflamma-
tory function of innate immune cells; for example, the
excessive secretion of IL-1 and IL-1 from autophagy-
deficient macrophages, functions together with IL-6
and transforming growth factor- (TGF), to promote
T
H
17 cell responses
109
(FIG.1c). Indeed, the levels of IL-17
are increased in the lungs of mice with ATG5-deficient
myeloid cells during M. tuberculosis infection
109
.
CD4
+
Tcells show increased IL-17 expression when
lung-infiltrating leukocytes from mice with an ATG5
deletion in the myeloid lineage are restimulated with
mycobacterial antigens exvivo
109
. The increased T
H
17
cell response might also result from the increased dura-
bility of immunological synapses between Tcells and
autophagy-defective DCs
109,118
.
Autophagy in plasma cells and humoral immunity.
Although autophagy seems not to be important for the
survival of the majority of mature Bcells
130
, lymphoid pre-
cursor stages are affected by the absence of autophagy
131
.
Autophagy is needed for the survival of B1 cells, which
are a subset of self-renewing B cells that cannot be
replenished from the adult bone marrow
130
. Moreover,
autophagy has a role in ER maintenance in plasma cells
under conditions of high secretory demands. Somewhat
paradoxically, absence of autophagy leads to excessive
immunoglobulin secretion
132
. Autophagy is important
not only for the function of plasma cells but also for their
survival and homeostasis; it is especially important for
the preservation of the bone marrow plasma cell pool
and, thus, is relevant for the maintenance of long-lived
humoral immunity
132
(FIG.1c).
In summary, autophagy influences adaptive immune
responses through its effects on antigen presentation,
naive Tcell repertoire selection, Tcell homeostasis and
T
H
cell polarization. Remaining questions include whether
autophagy affects Tcell polarization beyond T
H
17 cell
responses for example, whether it affects regulatory
Tcells and other T cell subsets and how autophagy
in one cell type affects other cells and immune networks
in specific anatomical sites. Answers to these questions
should help to explain the complex phenotypes that are
associated with risk mutations in autophagygenes.
Secretion of immune mediators
In addition to its intracellular actions, autophagy influ-
ences the extracellular release of immune mediators
133135

(FIG.1d), some of which are stored as pre-made secretary
granules. The defective secretion of lysozyme from
Paneth cells in ATG16L1-hypomorphic mice results in an
intestinal phenotype that corresponds to that of patients
with Crohns disease
133
. Autophagy modulates the secre-
tion of low-molecular-mass immune mediators such as
ATP, which function extracellularly either directly or
following ectonucleotidase action on purinergic P2Y
and P2X receptors to promote immune cell chemotaxis
and inflammasome activation
135
. Autophagic processes
in senescent cells affect the secretion of IL-6 and IL-8
through the constitutive ER-to-Golgi apparatus secre-
tory pathway
136
. Autophagy reduces immunoglobulin
Citrullinated antigens
Citrullin is enzymatically
generated from arginine
residues by peptidylarginine
deaminase, which is an
enzyme that may be enriched
in autophagosomes. The
presence of autoantibodies in
patients with rheumatoid
arthritis correlates with the
levels of citrullinated antigens
(such as vimentin).
Epithelial-to-mesenchymal
transition
A reversal of the mesenchymal-
to-epithelial transition that
occurs during development.
Epithelial-to-mesenchymal
transition or the
dedifferentiation of epithelial
cells can have normal
physiological roles (such as
in wound healing) or can be
associated with fibrotic
pathologies and cancer.
miRNA response elements
Sequences on RNA
transcripts that have partial
complementarity to
microRNAs (miRNAs). miRNA
binding to the response
elements typically leads to
repression of target gene
expression.
REVI EWS
734 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
secretion from plasma cells and, thus, seems to be a
mechanism that prevents excessive antibody produc-
tion
132
. The effects of autophagy on secretory func-
tions are not limited to immune effectors: for example,
autophagy factors and LC3 have a role in the exocytosis
of secretory lysosomes, as has been observed for a form
of LAP during osteoclastic bone resorption
137
.
Autophagy may contribute to the unconventional
secretion of cytosolic proteins that have extracellular
immune functions. The exact mechanism by which this
occurs is not fully understood. IL-1 and IL-18 do not
have signal peptides that facilitate entry to the ER and
the conventional secretory pathway (ER to Golgi appa-
ratus to plasma membrane). Instead, they are delivered
following inflammasome activation to the extracellular
environment via unconventional secretion. Although
autophagy suppresses inflammasome activation under
normal nutrient-rich conditions, autophagy can contrib-
ute to the unconventional secretion of IL-1 under stress
conditions, for example, following nutrient starvation
106
.
It is possible that autophagy functions as a double-switch
mechanism to repress the inflammasome under basal con-
ditions but to temporarily increase its physiological out-
put in response to DAMPs or PAMPs during infection.
The pro-inflammatory role of autophagy also extends to
the unconventional secretion of IL-18 and the DAMP
HMGB1 (REF.106). This role depends on ATG factors
and on a specialized unconventional secretion regulator
Golgi reassembly-stacking protein (GRASP; for example
GRASP of 55 kDa (GRASP55)). GRASP is important not
only in unconventional secretion but also affects canonical
starvation-induced autophagy in mammalian cells
106
.
Our understanding of the molecular and cellular
mechanisms that are involved in the secretory role of
autophagy is still in its infancy. It nevertheless extends
the idea of the importance of autophagy in the intracel-
lular space to the extracellular milieu, where it affects
tissue organization, remodelling and the delivery of
extracellular mediators of immunity and inflammation.
Conclusions
Autophagy influences many aspects of innate and adap-
tive immunity. Indeed, it is possible that autophagy
evolved as one of the first antimicrobial defences in
eukaryotic cells, and that it was shaped early on in evolu-
tion from what may initially have arisen as a metabolic
and quality control pathway. Autophagy uses a distinct
set of adaptors, the SLRs, to eliminate invading micro-
organisms and, in turn, pathogens have evolved strategies
to evade autophagic capture. It seems that as evolution
has progressed, nearly all components of the innate
immune system, such as conventional PRRs and inflam-
masomes, have become integrated with autophagy. In the
chordate lineage, this has extended to adaptive immunity,
as is best shown in mammalian systems. As a result, in
humans, a failure in parts of the autophagic apparatus can
lead to inflammatory, autoimmune or general immune
disorders.
The current knowledge of immunological autophagy
is still in its infancy and important questions remain. As
IKKs and TBK1 regulate autophagy, why is autophagy
often at cross-purposes with typeI IFN-mediated inflam-
mation? Is the ubiquitylation of microbial components the
main process that leads to pathogen recognition by SLRs,
or does it result from a failure to detect microorganisms
by other means? How are the recognition of microbial
entry and the subsequent targeting of microorganisms
for autophagy integrated through the functions of host
membrane damage sensing, E3 ubiquitin ligases, and pro-
tein and lipid kinases? Are mitochondria and mitophagy
the present-day manifestation of an evolutionary battle
between autophagy and the bacterium endosymbiont
precursor to mitochondria? The increased understanding
of the roles of autophagy in secretion during inflamma-
tion and tissue remodelling is an important development.
In this context, the relationship between autophagy and
unconventional secretion, including the secretion of
DAMPs and of inflammasome-dependent substrates
such as IL-1, needs further work. Finally, although early
progress has been made in understanding the roles of
autophagy in adaptive immunity, this area warrants
more study. Some of the questions that have recently
been asked in this area refer to the role of autophagy in the
polarization of T
H
17 cells and possibly other Tcell subsets,
along with the role of autophagy in general lymphocyte
homeostasis and differentiation.
In summary, autophagy and immunity are fully
integrated and our continuing study of the interface
between these processes will be a fruitful area of sci-
entific inquiry for many years to come. In translational
terms, autophagy is undoubtedly an attractive target for
developing new treatments in inflammatory disorders
and autoimmunity, and may eventually be harnessed as
an anti-infective mechanism.
1. Mizushima,N., Yoshimori,T. & Ohsumi,Y. The role of
atg proteins in autophagosome formation. Annu.
Rev. Cell Dev. Biol. 27, 107132 (2011).
2. Codogno,P., Mehrpour,M. & Proikas-Cezanne,T.
Canonical and non-canonical autophagy: variations on
a common theme of self-eating? Nature Rev. Mol. Cell
Biol. 13, 712 (2012).
3. Virgin,H.W. & Levine,B. Autophagy genes in
immunity. Nature Immunol. 10, 461470 (2009).
4. Levine,B., Mizushima,N. & Virgin,H.W. Autophagy in
immunity and inflammation. Nature 469, 323335
(2011).
5. Hamasaki,M. etal. Autophagosomes form at
ER-mitochondria contact sites. Nature 495, 389393
(2013).
6. Itakura,E., Kishi-Itakura,C. & Mizushima,N. The
hairpin-type tail-anchored SNARE syntaxin 17 targets
to autophagosomes for fusion with endosomes/
lysosomes. Cell 151, 12561269 (2012).
7. Tattoli,I. etal. Amino acid starvation induced by
invasive bacterial pathogens triggers an innate host
defense program. Cell Host Microbe 11, 563575
(2012).
8. Shelly,S., Lukinova,N., Bambina,S., Berman,A. &
Cherry,S. Autophagy is an essential component of
Drosophila immunity against vesicular stomatitis
virus. Immunity 30, 588598 (2009).
9. Shi,C.S. & Kehrl,J.H. TRAF6 and A20 regulate lysine
63-linked ubiquitination of Beclin-1 to control
TLR4-induced autophagy. Sci. Signal. 3, ra42 (2010).
10. Nazio,F. etal. mTOR inhibits autophagy by controlling
ULK1 ubiquitylation, self-association and function
through AMBRA1 and TRAF6. Nature Cell Biol. 15,
406416 (2013).
11. Chaturvedi,A., Dorward,D. & Pierce,S.K. The Bcell
receptor governs the subcellular location of Toll-like
receptor 9 leading to hyperresponses to DNA-
containing antigens. Immunity 28, 799809 (2008).
12. Biswas,D. etal. ATP-induced autophagy is associated
with rapid killing of intracellular mycobacteria within
human monocytes/macrophages. BMC Immunol. 9, 35
(2008).
13. Tang,D. etal. Endogenous HMGB1 regulates
autophagy. J.Cell Biol. 190, 881892 (2010).
14. Pilli,M. etal. TBK-1 promotes autophagy-mediated
antimicrobial defense by controlling autophagosome
maturation. Immunity 37, 223234 (2012).
15. Gutierrez,M.G. etal. Autophagy is a defense
mechanism inhibiting BCG and Mycobacterium
tuberculosis survival in infected macrophages.
Cell 119, 753766 (2004).
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 735
2013 Macmillan Publishers Limited. All rights reserved
16. Harris,J. etal. T helper 2 cytokines inhibit autophagic
control of intracellular Mycobacterium tuberculosis.
Immunity 27, 505517 (2007).
17. Singh,S.B. etal. Human IRGM regulates autophagy and
cell-autonomous immunity functions through
mitochondria. Nature Cell Biol. 12, 11541165
(2010).
18. Zalckvar,E. etal. DAP-kinase-mediated phosphorylation
on the BH3 domain of beclin 1 promotes dissociation
of beclin 1 from Bcl-X
L
and induction of autophagy.
EMBO Rep. 10, 285292 (2009).
19. Fabri,M. etal. VitaminD is required for IFN-
mediated antimicrobial activity of human
macrophages. Sci. Transl. Med. 3, 104ra102 (2011).
This is a translationally important study showing
that vitaminD has a key cooperative role with IFN
in inducing antimicrobial autophagy in human cells,
whereas vitaminD metabolites are diminished in
tuberculosis and HIV.
20. Mostowy,S. etal. p62 and NDP52 proteins target
intracytosolic Shigella and Listeria to different
autophagy pathways. J.Biol. Chem. 286,
2698726995 (2011).
21. Van Grol,J. etal. HIV-1 inhibits autophagy in
bystander macrophage/monocytic cells throughSrc-
Akt and STAT3. PLoS ONE 5, e11733 (2010).
22. Shen,S. etal. Cytoplasmic STAT3 represses
autophagy by inhibiting PKR activity. Mol. Cell 48,
667680 (2012).
23. Scherz-Shouval,R. etal. Reactive oxygen species are
essential for autophagy and specifically regulate the
activity of Atg4. EMBO J. 26, 17491760 (2007).
24. Huang,J. etal. Activation of antibacterial autophagy
by NADPH oxidases. Proc. Natl Acad. Sci. USA 106,
62266231 (2009).
25. Tal,M.C. etal. Absence of autophagy results in
reactive oxygen species-dependent amplification
of RLR signaling. Proc. Natl Acad. Sci. USA 106,
27702775 (2009).
26. Yang,C.S. etal. Autophagy protein Rubicon mediates
phagocytic NADPH oxidase activation in response
to microbial infection or TLR stimulation.
Cell Host Microbe 11, 264276 (2012).
27. Sarkar,S. etal. Complex inhibitory effects of nitric
oxide on autophagy. Mol. Cell 43, 1932 (2011).
28. Maiuri,M.C. etal. Functional and physical interaction
between Bcl-X
L
and a BH3-like domain in Beclin-1.
EMBO J. 26, 25272539 (2007).
29. Wei,Y., Pattingre,S., Sinha,S., Bassik,M. & Levine,B.
JNK1-mediated phosphorylation of Bcl-2 regulates
starvation-induced autophagy. Mol. Cell 30, 678688
(2008).
30. Egan,D.F. etal. Phosphorylation of ULK1 (hATG1)
by AMP-activated protein kinase connects energy
sensing to mitophagy. Science 331, 456461
(2011).
31. Sanjuan,M.A. etal. Toll-like receptor signalling in
macrophages links the autophagy pathway to
phagocytosis. Nature 450, 12531257 (2007).
32. Henault,J. etal. Noncanonical autophagy is required
for type I interferon secretion in response to DNA-
immune complexes. Immunity 37, 986997 (2012).
This is an important study that differentiates LAP
from conventional autophagy by defining it at the
molecular level as a process that is independent of
ULK1 but that is dependent on beclin 1.
33. Johansen,T. & Lamark,T. Selective autophagy
mediated by autophagic adapter proteins.
Autophagy 7, 279296 (2011).
34. Dupont,N. etal. Shigella phagocytic vacuolar
membrane remnants participate in the cellular
response to pathogen invasion and are regulated by
autophagy. Cell Host Microbe 6, 137149 (2009).
35. Thurston,T.L., Ryzhakov,G., Bloor,S., von
Muhlinen,N. & Randow,F. The TBK1 adaptor and
autophagy receptor NDP52 restricts the proliferation
of ubiquitin-coated bacteria. Nature Immunol. 10,
12151221 (2009).
36. Wild,P. etal. Phosphorylation of the autophagy
receptor optineurin restricts Salmonella growth.
Science 333, 228233 (2011).
37. Shahnazari,S. etal. A diacylglycerol-dependent
signaling pathway contributes to regulation of
antibacterial autophagy. Cell Host Microbe 8,
137146 (2010).
38. Thurston,T.L., Wandel,M.P., von Muhlinen,N.,
Foeglein,A. & Randow,F. Galectin 8 targets damaged
vesicles for autophagy to defend cells against bacterial
invasion. Nature 482, 414418 (2012).
39. Deretic,V. & Levine,B. Autophagy, immunity, and
microbial adaptations. Cell Host Microbe 5, 527549
(2009).
40. Saitoh,T. & Akira,S. Regulation of innate immune
responses by autophagy-related proteins. J.Cell Biol.
189, 925935 (2010).
41. Travassos,L.H. etal. Nod1 and Nod2 direct
autophagy by recruiting ATG16L1 to the plasma
membrane at the site of bacterial entry. Nature
Immunol. 11, 5562 (2010).
42. Cooney,R. etal. NOD2 stimulation induces autophagy
in dendritic cells influencing bacterial handling and
antigen presentation. Nature Med. 16, 9097
(2010).
43. Birmingham,C.L., Smith,A.C., Bakowski,M.A.,
Yoshimori,T. & Brumell,J.H. Autophagy controls
Salmonella infection in response to damage to the
Salmonella-containing vacuole. J.Biol. Chem. 281,
1137411383 (2006).
44. Birmingham,C.L. etal. Listeriolysin O allows Listeria
monocytogenes replication in macrophage vacuoles.
Nature 451, 350354 (2008).
45. Watson,R.O., Manzanillo,P.S. & Cox,J.S.
Extracellular M.tuberculosis DNA targets bacteria
for autophagy by activating the host DNA-sensing
pathway. Cell 150, 803815 (2012).
This study shows that microbial DNA that is
released into the host cell cytosol can induce
autophagy via STING, which forms an important
intersection with typeI IFN signalling. It also shows
the invivo role of autophagy in protection against
M.tuberculosis.
46. Yoshikawa,Y. etal. Listeria monocytogenes ActA-
mediated escape from autophagic recognition.
Nature Cell Biol. 11, 12331240 (2009).
47. Ogawa,M. etal. A Tecpr1-dependent selective
autophagy pathway targets bacterial pathogens.
Cell Host Microbe 9, 376389 (2011). 48.
Talloczy,Z., Virgin,H.W.t. & Levine,B. PKR-dependent
autophagic degradation of herpes simplex virus
type1. Autophagy 2, 2429 (2006).
49. Orvedahl,A. etal. HSV-1 ICP34.5 confers
neurovirulence by targeting the beclin 1 autophagy
protein. Cell Host Microbe 1, 2335 (2007).
50. Lee,H.K., Lund,J.M., Ramanathan,B.,
Mizushima,N. & Iwasaki,A. Autophagy-dependent
viral recognition by plasmacytoid dendritic cells.
Science 315, 13981401 (2007).
51. Kyei,G.B. etal. Autophagy pathway intersects with
HIV-1 biosynthesis and regulates viral yields in
macrophages. J.Cell Biol. 186, 255268 (2009).
52. Gannage,M. etal. Matrix protein 2 of influenza A
virus blocks autophagosome fusion with lysosomes.
Cell Host Microbe 6, 367380 (2009).
53. Lee,H.K. etal. Invivo requirement for Atg5 in
antigen presentation by dendritic cells. Immunity 32,
227239 (2010).
54. Blanchet,F.P. etal. Human immunodeficiency virus-1
inhibition of immunoamphisomes in dendritic cells
impairs early innate and adaptive immune responses.
Immunity 32, 654669 (2010).
55. Orvedahl,A. etal. Autophagy protects against
sindbis virus infection of the central nervous system.
Cell Host Microbe 7, 115127 (2010).
56. Yordy,B., Iijima,N., Huttner,A., Leib,D. & Iwasaki,A.
A neuron-specific role for autophagy in antiviral
defense against herpes simplex virus. Cell Host
Microbe 12, 334345 (2012).
57. Shoji-Kawata,S. etal. Identification of a candidate
therapeutic autophagy-inducing peptide. Nature 494,
201206 (2013).
This study is a key step towards customizing the
induction of autophagy without using mTOR
inhibitors. The peptide that was designed on the
basis of the Nef-binding site on beclin 1 showed
efficacy in various infections and in other invitro
and invivo models.
58. Xu,Y. etal. Toll-like receptor 4 is a sensor for
autophagy associated with innate immunity.
Immunity 27, 135144 (2007).
59. Delgado,M.A., Elmaoued,R.A., Davis,A.S., Kyei,G.
& Deretic,V. Toll-like receptors control autophagy.
EMBO J. 27, 11101121 (2008).
60. Alonso,S., Pethe,K., Russell,D.G. & Purdy,G.E.
Lysosomal killing of Mycobacterium mediated by
ubiquitin-derived peptides is enhanced by
autophagy. Proc. Natl Acad. Sci. USA 104,
60316036 (2007).
61. Ponpuak,M. etal. Delivery of cytosolic components
by autophagic adaptor protein p62 endows
autophagosomes with unique antimicrobial
properties. Immunity 32, 329341 (2010).
62. Kim,B.H. etal. A family of IFN-inducible 65-kD
GTPases protects against bacterial infection. Science
332, 717721 (2011).
63. Jagannath,C. etal. Autophagy enhances the efficacy
of BCG vaccine by increasing peptide presentation in
mouse dendritic cells. Nature Med. 15, 267276
(2009).
64. Yano,T. etal. Autophagic control of listeria through
intracellular innate immune recognition in drosophila.
Nature Immunol. 9, 908916 (2008).
65. Moreau,K., Ravikumar,B., Renna,M., Puri,C. &
Rubinsztein,D.C. Autophagosome precursor
maturation requires homotypic fusion. Cell 146,
303317 (2011).
66. Wellcome Trust Case Control Consortium. Genome-
wide association study of 14,000 cases of seven
common diseases and 3,000 shared controls. Nature
447, 661678 (2007).
67. Lei,Y. etal. The mitochondrial proteins NLRX1 and
TUFM form a complex that regulates typeI interferon
and autophagy. Immunity 36, 933946 (2012).
68. Jounai,N. etal. NLRP4 negatively regulates
autophagic processes through an association with
beclin1. J.Immunol. 186, 16461655 (2011).
69. Jounai,N. etal. The Atg5 Atg12 conjugate associates
with innate antiviral immune responses. Proc. Natl
Acad. Sci. USA 104, 1405014055 (2007).
This is a key study that led to the appreciation of
the interference between autophagy and typeI IFN
signalling.
70. McFarlane,S. etal. Early induction of autophagy
in human fibroblasts after infection with human
cytomegalovirus or herpes simplex virus 1.
J.Virol. 85, 42124221 (2011).
71. Rasmussen,S.B. etal. Activation of autophagy by
-herpesviruses in myeloid cells is mediated by
cytoplasmic viral DNA through a mechanism
dependent on stimulator of IFN genes. J.Immunol.
187, 52685276 (2011).
72. Sun,L., Wu,J., Du,F., Chen,X. & Chen,Z.J.
Cyclic GMP-AMP synthase is a cytosolic DNA sensor
that activates the typeI interferon pathway.
Science 339, 786791 (2013).
73. Zheng,Y.T. etal. The adaptor protein p62/SQSTM1
targets invading bacteria to the autophagy pathway.
J.Immunol. 183, 59095916 (2009).
74. Yang,J.Q., Liu,H., Diaz-Meco,M.T. & Moscat,J.
NBR1 is a new PB1 signalling adapter in Th2
differentiation and allergic airway inflammation
invivo. EMBO J. 29, 34213433 (2010).
75. Bjorkoy,G. etal. p62/SQSTM1 forms protein
aggregates degraded by autophagy and has a
protective effect on huntingtin-induced cell death.
J.Cell Biol. 171, 603614 (2005).
This is a seminal study that led to the identification
of autophagic receptors and mechanisms of
selective autophagy in mammalian cells.
76. Kirkin,V. etal. A role for NBR1 in autophagosomal
degradation of ubiquitinated substrates. Mol. Cell 33,
505516 (2009).
77. Newman,A.C. etal. TBK1 kinase addiction in lung
cancer cells is mediated via autophagy of Tax1bp1/
Ndp52 and non-canonical NF-B signalling.
PLoS ONE 7, e50672 (2012).
78. Moscat,J. & Diaz-Meco,M. T. p62 at the crossroads
of autophagy, apoptosis, and cancer. Cell 137,
10011004 (2009).
79. Li,S. etal. Sterical hindrance promotes selectivity of
the autophagy cargo receptor NDP52 for the danger
receptor galectin-8 in antibacterial autophagy.
Science signaling 6, ra9 (2013).
80. Husnjak,K. & Dikic,I. Ubiquitin-binding proteins:
decoders of ubiquitin-mediated cellular functions.
Annu. Rev. Biochem. 81, 291322 (2012).
81. von Muhlinen,N. etal. LC3C, bound selectively by
a noncanonical LIR motif in NDP52, is required for
antibacterial autophagy. Mol. Cell 48, 329342
(2012).
This was an important step in defining the roles of
different mammalian ATG-8 paralogues in
autophagic antimicrobial defence.
82. Matsumoto,G., Wada,K., Okuno,M., Kurosawa,M.
& Nukina,N. Serine 403 phosphorylation of p62/
SQSTM1 regulates selective autophagic clearance of
ubiquitinated proteins. Mol. Cell 44, 279289
(2011).
83. Huett,A. etal. The LRR and RING domain protein
LRSAM1 is an E3 ligase crucial for ubiquitin-
dependent autophagy of intracellular Salmonella
Typhimurium. Cell Host Microbe 12, 778790 (2012).
This paper identifies an E3 ubiquitin ligase that
recognizes cytoplasmic bacteria, that ubiquitylates
autophagic targets and that cooperates with other
receptors to guide autophagic capture and
elimination of the invading microorganisms.
REVI EWS
736 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
84. Huang,W. etal. Crystal structure and biochemical
analyses reveal Beclin 1 as a novel membrane binding
protein. Cell Res. 22, 473489 (2012).
85. Craddock,N. etal. Genome-wide association study
of CNVs in 16,000 cases of eight common diseases
and 3,000 shared controls. Nature 464, 713720
(2010).
86. Brest,P. etal. A synonymous variant in IRGM alters a
binding site for miR-196 and causes deregulation of
IRGM-dependent xenophagy in Crohns disease.
Nature Genet. 43, 242245 (2011).
87. Orvedahl,A. etal. Image-based genome-wide siRNA
screen identifies selective autophagy factors. Nature
480, 113117 (2011).
88. Mesquita,F.S. etal. The Salmonella deubiquitinase
SseL inhibits selective autophagy of cytosolic
aggregates. PLoS Pathog. 8, e1002743 (2012).
89. Jostins,L. etal. Host-microbe interactions have
shaped the genetic architecture of inflammatory
bowel disease. Nature 491, 119124 (2012).
90. Henckaerts,L. etal. Genetic variation in the
autophagy gene ULK1 and risk of Crohns disease.
Inflamm. Bowel Dis. 17, 13921397 (2011).
91. Ramos,P.S. etal. A comprehensive analysis of
shared loci between systemic lupus erythematosus
(SLE) and sixteen autoimmune diseases reveals
limited genetic overlap. PLoS Genet. 7, e1002406
(2011).
92. Harley,J.B. etal. Genome-wide association scan in
women with systemic lupus erythematosus identifies
susceptibility variants in ITGAM, PXK, KIAA1542 and
other loci. Nature Genet. 40, 204210 (2008).
93. Han,J.W. etal. Genome-wide association study in a
Chinese Han population identifies nine new
susceptibility loci for systemic lupus erythematosus.
Nature Genet. 41, 12341237 (2009).
94. Yang,W. etal. Meta-analysis followed by replication
identifies loci in or near CDKN1B, TET3, CD80,
DRAM1, and ARID5B as associated with systemic
lupus erythematosus in Asians. Am. J.Hum. Genet.
92, 4151 (2013).
95. Martin,L.J. etal. Functional variant in the autophagy-
related 5 gene promotor is associated with childhood
asthma. PLoS ONE 7, e33454 (2012).
96. Raychaudhuri,S. etal. Genetic variants at CD28,
PRDM1 and CD2/CD58 are associated with
rheumatoid arthritis risk. Nature Genet. 41,
13131318 (2009).
97. Zhao,H. etal. Mice deficient in Epg5 exhibit selective
neuronal vulnerability to degeneration. J.Cell Biol.
200, 731741 (2013).
98. Cullup,T. etal. Recessive mutations in EPG5 cause
Vici syndrome, a multisystem disorder with defective
autophagy. Nature Genet. 45, 8387 (2013).
99. Yang,C.S. etal. The autophagy regulator Rubicon is
a feedback inhibitor of CARD9-mediated host innate
immunity. Cell Host Microbe 11, 277289 (2012).
100. Saitoh,T. etal. Atg9a controls dsDNA-driven
dynamic translocation of STING and the innate
immune response. Proc. Natl Acad. Sci. USA 106,
2084220846 (2009).
101. Li,S., Wang,L., Berman,M., Kong,Y.Y. & Dorf,M.E.
Mapping a dynamic innate immunity protein
interaction network regulating typeI interferon
production. Immunity 35, 426440 (2011).
102. Saitoh,T. etal. Loss of the autophagy protein
Atg16L1 enhances endotoxin-induced IL-1
production. Nature 456, 264268 (2008).
This is a key study that recognizes the
anti-inflammatory role of autophagy.
103. Rathinam,V.A., Vanaja,S.K. & Fitzgerald,K.A.
Regulation of inflammasome signaling. Nature
Immunol. 13, 333332 (2012).
104. Zhou,R., Yazdi,A.S., Menu,P. & Tschopp,J. A role
for mitochondria in NLRP3 inflammasome activation.
Nature 469, 221225 (2011).
This is a conceptually important study that shows
that housekeeping functions, such as the autophagic
clearance of depolarized mitochondria when they are
failing, can lead to endogenous sterile inflammation.
105. Nakahira,K. etal. Autophagy proteins regulate
innate immune responses by inhibiting the release
of mitochondrial DNA mediated by the NALP3
inflammasome. Nature Immunol. 12, 222230
(2011).
106. Dupont,N. etal. Autophagy-based unconventional
secretory pathway for extracellular delivery of IL-1.
EMBO J. 30, 47014711 (2011).
107. Shi,C.S. etal. Activation of autophagy by
inflammatory signals limits IL-1 production by
targeting ubiquitinated inflammasomes for
destruction. Nature Immunol. 13, 255263 (2012).
108. Lupfer,C. etal. Receptor interacting protein kinase
2-mediated mitophagy regulates inflammasome
activation during virus infection. Nature Immunol. 14,
480488 (2013).
109. Castillo,E.F. etal. Autophagy protects against
active tuberculosis by suppressing bacterial burden
and inflammation. Proc. Natl Acad. Sci. USA 109,
E3168E3176 (2012).
This paper is an invivo demonstration of the role
of autophagy in controlling both inflammation and
bacteria during infection.
110. Paul,S., Kashyap,A.K., Jia,W., He,Y.W. &
Schaefer,B.C. Selective autophagy of the adaptor
protein Bcl10 modulates Tcell receptor activation
of NF-B. Immunity 36, 947958 (2012).
111. Shibata,Y. etal. p47 negatively regulates IKK
activation by inducing the lysosomal degradation of
polyubiquitinated NEMO. Nature Commun. 3, 1061
(2012).
112. Fliss,P.M. etal. Viral mediated redirection of NEMO/
IKK to autophagosomes curtails the inflammatory
cascade. PLoS Pathog. 8, e1002517 (2012).
113. Wang,C. etal. Atg16L1 deficiency confers protection
from uropathogenic Escherichia coli infection invivo.
Proc. Natl Acad. Sci. USA 109, 1100811013
(2012).
114. Paludan,C. etal. Endogenous MHC classII processing
of a viral nuclear antigen after autophagy. Science
307, 593596 (2005).
115. Nedjic,J., Aichinger,M., Emmerich,J., Mizushima,N.
& Klein,L. Autophagy in thymic epithelium shapes the
T-cell repertoire and is essential for tolerance. Nature
455, 396400 (2008).
116. Schmid,D., Pypaert,M. & Munz,C. Antigen-loading
compartments for major histocompatibility complex
classII molecules continuously receive input from
autophagosomes. Immunity 26, 7992 (2007).
117. Ireland,J.M. & Unanue,E.R. Autophagy in antigen-
presenting cells results in presentation of
citrullinated peptides to CD4 Tcells. J.Exp. Med.
208, 26252632 (2011).
118. Wildenberg,M.E. etal. Autophagy attenuates
the adaptive immune response by destabilizing the
immunologic synapse. Gastroenterology 142,
14931503 e6 (2012).
119. Akalay,I. etal. Epithelial-to-mesenchymal transition
and autophagy induction in breast carcinoma promote
escape from T-cell-mediated lysis. Cancer Res. 73,
24182427 (2013).
120. Wenger,T. etal. Autophagy inhibition promotes
defective neosynthesized proteins storage in ALIS, and
induces redirection toward proteasome processing
and MHCI-restricted presentation. Autophagy 8,
350363 (2012).
121. Fiegl,D. etal. Amphisomal route of MHC classI
cross-presentation in bacteria-infected dendritic cells.
J.Immunol. 190, 27912806 (2013).
122. Tey,S.K. & Khanna,R. Autophagy mediates
transporter associated with antigen processing-
independent presentation of viral epitopes through
MHC classI pathway. Blood 120, 9941004
(2012).
123. Gibbings,D. etal. Selective autophagy degrades
DICER and AGO2 and regulates miRNA activity.
Nature Cell Biol. 14, 13141321 (2012).
124. Mortensen,M. etal. The autophagy protein Atg7 is
essential for hematopoietic stem cell maintenance.
J.Exp. Med. 208, 455467 (2011).
125. Pua,H.H., Guo,J., Komatsu,M. & He,Y.W.
Autophagy is essential for mitochondrial clearance in
mature T lymphocytes. J.Immunol. 182, 40464055
(2009).
126. Jia,W., Pua,H.H., Li,Q.J. & He,Y.W. Autophagy
regulates endoplasmic reticulum homeostasis and
calcium mobilization in T lymphocytes. J.Immunol.
186, 15641574 (2011).
127. Lee,J.S. etal. FLIP-mediated autophagy regulation in
cell death control. Nature Cell Biol. 11, 13551362
(2009).
128. He,M.X. & He,Y.W. A role for c-FLIP
L
in the regulation
of apoptosis, autophagy, and necroptosis in
Tlymphocytes. Cell Death Differ. 20, 188197
(2013).
129. Hubbard,V.M. etal. Macroautophagy regulates
energy metabolism during effector Tcell activation.
J.Immunol. 185, 73497357 (2010).
130. Miller,B.C. etal. The autophagy gene ATG5 plays an
essential role in B lymphocyte development.
Autophagy 4, 309314 (2008).
131. Arsov,I. etal. A role for autophagic protein beclin 1
early in lymphocyte development. J.Immunol. 186,
22012209 (2011).
132. Pengo,N. etal. Plasma cells require autophagy
for sustainable immunoglobulin production.
Nature Immunol. 14, 298305 (2013).
133. Cadwell,K. etal. A key role for autophagy and the
autophagy gene Atg16l1 in mouse and human
intestinal Paneth cells. Nature 456, 259263 (2008).
This is a key study that shows the general secretory
defects in autophagy-deficient cells that are of
consequence for inflammatory bowel diseases such
as Crohns disease.
134. Ushio,H. etal. Crucial role for autophagy in
degranulation of mast cells. J.Allergy Clin. Immunol.
127, 12671276 e6 (2011).
135. Michaud,M. etal. Autophagy-dependent anticancer
immune responses induced by chemotherapeutic
agents in mice. Science 334, 15731577 (2011).
136. Narita,M. etal. Spatial coupling of mTOR and
autophagy augments secretory phenotypes.
Science 332, 966970 (2011).
137. DeSelm,C.J. etal. Autophagy proteins regulate the
secretory component of osteoclastic bone resorption.
Dev. Cell 21, 966974 (2011).
138. Chauhan,S. & Boyd,D. ZKSCAN3 is a master
transcriptional repressor of autophagy.
Mol. Cell 50, 1628 (2013).
139. Settembre,C. etal. TFEB controls cellular lipid
metabolism through a starvation-induced
autoregulatory loop. Nature Cell Biol. 15, 647658
(2013).
140. Russell,R.C. etal. ULK1 induces autophagy by
phosphorylating Beclin-1 and activating VPS34 lipid
kinase. Nature Cell Biol. 15, 741750 (2013).
This is a study that functionally links two key
branches of autophagy regulators, the ULK1
branch and the beclin 1 branch.
141. Takaesu,G., Kobayashi,T. & Yoshimura,A.
TGF-activated kinase 1 (TAK1)-binding proteins
(TAB) 2 and 3 negatively regulate autophagy.
J.Biochem. 151, 157166 (2012).
142. Zalckvar,E., Berissi,H., Eisenstein,M. & Kimchi,A.
Phosphorylation of Beclin 1 by DAP-kinase promotes
autophagy by weakening its interactions with Bcl-2
and Bcl-XL. Autophagy 5, 720722 (2009).
143. Criollo,A. etal. Inhibition of autophagy by TAB2
and TAB3. EMBO J. 30, 49084920 (2011).
144. Radtke,A.L., Delbridge,L.M., Balachandran,S.,
Barber,G.N. & ORiordan,M.X. TBK1 protects
vacuolar integrity during intracellular bacterial
infection. PLoS Pathog. 3, e29 (2007).
145. Ku,B. etal. Structural and biochemical bases for the
inhibition of autophagy and apoptosis by viral BCL-2
of murine -herpesvirus 68. PLoS Pathog. 4, e25
(2008).
146. Choy,A. etal. The Legionella effector RavZ inhibits
host autophagy through irreversible Atg8
deconjugation. Science 338, 10721076 (2012).
This is an elegant recent example of how bacteria
defend themselves against autophagy by
preventing LC3 lipidation, which thus stalls the
autophagic response.
147. Dortet,L. etal. Recruitment of the major vault protein
by InlK: a Listeria monocytogenes strategy to avoid
autophagy. PLoS Pathog. 7, e1002168 (2011).
148. Niu,H., Xiong,Q., Yamamoto,A., Hayashi-Nishino,M.
& Rikihisa,Y. Autophagosomes induced by a bacterial
Beclin 1 binding protein facilitate obligatory
intracellular infection. Proc. Natl Acad. Sci. USA 109,
2080020807 (2012).
149. Ogawa,M. etal. Escape of intracellular Shigella
from autophagy. Science 307, 727731 (2005).
150. Gregoire,I.P. etal. IRGM is a common target of
RNA viruses that subvert the autophagy network.
PLoS Pathog. 7, e1002422 (2011).
Acknowledgements
We apologize to our colleagues whose work or specific studies
have not been cited. This is not a reflection of lack of interest
on the authors part or on the significance for the field but
was dictated by the article scope and reference number limi-
tations. V.D. is supported by US National Institutes of Health
grant AI042999 and by a grant from the Bill and Melinda
Gates Foundation.
Competing interests statement
The authors declare no competing financial interests.
DATABASES
Expressed sequence tags NCBI database:
http://www.ncbi.nlm.nih.gov/nucest
ALL LINKS ARE ACTIVE IN THE ONLINE PDF
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 737
2013 Macmillan Publishers Limited. All rights reserved
Considerable progress has been made both in under-
standing the basic immune mechanisms of kidney
disease and in translating these findings to clinical
therapies. Sophisticated animal studies combined
with the analysis of clinical samples have led to a pre-
cise knowledge of the autoimmune targets and of the
mechanisms responsible for kidney injury. Kidney
diseases are highly prevalent and cost-intensive,
but many discoveries in renal immunology are not
widely known in the immunological community,
although they are often relevant to diseases that affect
otherorgans.
In this Review, we discuss recent advances in our
understanding of immune-mediated kidney diseases,
emphasizing those of particular relevance to the wider
immunology community and those that have led to a
better understanding of basic immunological mechan-
isms. We have had to be selective in the topics consid-
ered and so have excluded a discussion of acute kidney
injury, kidney transplantation and alloimmunity, as
well as of systemic diseases with associated kidney
disease, such as type 2 diabetes and hypertension,
that are not primarily caused by the immune system,
despite the involvement of innate (and possibly adap-
tive) immune responses in the renal injury they cause.
Here, we discuss the innate immune mechanisms of
kidney injury and introduce novel concepts about the
role of the cellular immune responses that drive renal
disease. Moreover, we summarize recent discoveries
about complement- and antibody-mediated nephritis,
and we discuss kidney pathologies that are mediated
by renal autoantigen-specific antibodies, especially those
that are induced by crossreactive microorganism-specific
antibodies. Finally, we describe how the disruption of
kidney function and kidney pathologies can influence
systemic immune responses.
Kidney-resident immune cells
In the kidneys, toxic waste products of metabolism are
removed from the blood by nephrons. Each nephron
contains one glomerulus, which functions as a size-
selective filter that retains molecules above ~50 kDa
in the blood. Compounds of lower molecular mass
pass through the glomerular filter, enter the tubular
system and are excreted with the urine unless they
are re absorbed by the tubular epithelium (BOX1). The
kidneys produce several hormones that directly or
indirectly affect immune responses, including vita-
minD, which regulates bone homeostasis and phago-
cyte function, erythropoietin, which is induced in
response to hypoxia to regulate erythropoiesis, and
renin, which induces angiotensin and aldosterone to
regulate electrolyte balance, extracellular osmolarity
and blood pressure.
1
Institutes of Molecular
Medicine and Experimental
Immunology (IMMEI),
Rheinische Friedrich-
Wilhelms-Universitt,
Sigmund-Freud-Str. 25,
53105 Bonn, Germany.
2
III. Medizinische Klinik,
Universittsklinikum
Hamburg-Eppendorf,
Martinistrasse 52,
20246 Hamburg, Germany.
3
Medizinische Klinik und
Poliklinik IV, Ludwig-
Maximilians Universitt
Mnchen, Ziemssenstr. 1,
80336 Mnchen, Germany.
4
Clinical Institute of Pathology,
Medical University of Vienna,
Whringer Grtel 1820,
A-1090 Vienna, Austria.
e-mails: ckurts@web.de;
panzer@uke.de;
hjanders@med.uni-
muenchen.de; andrew.rees@
meduniwien.ac.at
All authors contributed
equally to this work.
doi:10.1038/nri3523
Published online
16 September 2013
The immune system and kidney
disease: basic concepts and clinical
implications
Christian Kurts
1
, Ulf Panzer
2
, Hans-Joachim Anders
3
and Andrew J.Rees
4
Abstract | The kidneys are frequently targeted by pathogenic immune responses against
renal autoantigens or by local manifestations of systemic autoimmunity. Recent studies in
rodent models and humans have uncovered several underlying mechanisms that can be
used to explain the previously enigmatic immunopathology of many kidney diseases. These
mechanisms include kidney-specific damage-associated molecular patterns that cause
sterile inflammation, the crosstalk between renal dendritic cells and Tcells, the development
of kidney-targeting autoantibodies and molecular mimicry with microbial pathogens.
Conversely, kidney failure affects general immunity, causing intestinal barrier dysfunction,
systemic inflammation and immunodeficiency that contribute to the morbidity and mortality
of patients with kidney disease. In this Review, we summarize the recent findings regarding
the interactions between the kidneys and the immune system.
REVI EWS
738 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Nephrons
Anatomically and functionally
independent kidney units that
each consist of one glomerulus
and one tubule. The nephron
delivers urine into collecting
ducts that empty into the renal
pelvis and, through the ureters,
into the urinary bladder.
Glomerulus
An anatomical structure that
is located in the kidney cortex
and that filters blood into the
tubular system.
Tubulointerstitium
The space between the tubuli
and glomeruli, which contains
capillaries, fibroblasts and
dendritic cells, and thus
is an important site for the
progression of nephritis.
Bacterial pyelonephritis
A bacterial infection of the
kidney, mostly due to
uropathogenic Escherichiacoli
that ascend through the
urethra, bladder and ureter
into the kidneys.
Under homeostatic conditions, the resident
immune cells of the kidneys include dendritic cells
(DCs) and macrophages, as well as a few lympho-
cytes
14
. DCs are restricted to the tubulointerstitium and
are absent from the glomeruli
1,2
. In mice, kidney DCs
are CD11c
+
CD11b
+
F4/80
+
CX
3
CR1
+
CD8

CD205

and
have a transcriptome that is typical of DCs resident
in various non-lymphoid tissues
5,6
. Kidney DCs are
derived from monocytes and from common DC pre-
cursors (CDPs), but in contrast with other organs,
some CDP-derived kidney DCs express CD64 (also
known asFcRI)
7
. Kidney DCs function as sentinels
in homeostasis, local injury and infection
3,8
. They rap-
idly produce neutrophil-recruiting chemokines dur-
ing bacterial pyelonephritis, which is the most prevalent
kidney infection
8
. Neutrophils can also be recruited by
tubular epithelial cells, but not as quickly as by DCs.
Mice lacking expression of CX
3
C-chemokine recep-
tor1 (CX
3
CR1) have a selective reduction in kidney
DC numbers
9
. There is also a high renal expression of
its ligand CX
3
C-chemokine ligand 1 (CX
3
CL1)
10
, which
suggests that the CX
3
CR1CX
3
CL1 chemokine pair
are important for DC recruitment to the kidney and
that CX
3
CR1 might be a specific therapeutic target to
modulate DC numbers in the kidneys. In renal ischae-
mia (which is relevant in kidney transplantation) and
in ureteral obstruction, renal DCs promote tissue
injury by producing pro-inflammatory cytokines
11,12
.
Basic leucine zipper transcriptional factor ATF-like3
(BATF3)-dependent CD103
+
tissue DCs, which can
cross-present antigens to CD8
+
Tcells, are rare and
their function in the kidney is unclear
13
. Macrophages
are preferentially found in the renal medulla and cap-
sule
1
and have homeostatic and repair functions
14
.
There are also mast cells in the kidney tubulointer-
stitium but their function is poorly understood
1517
.
In addition, the role of innate-like lymphocytes is
currently unclear. Finally, the renal lymph nodes rep-
resent a priming site for nephritogenic Tcells during
renal inflammation
18,19
.
Low-molecular-mass proteins can pass through the
glomerular filter but are reabsorbed and degraded by
tubular epithelial cells. However, some of these proteins
are captured by renal DCs or reach the renal lymph
nodes by lymphatic drainage within seconds after filtra-
tion
20
. Importantly, filtered proteins are concentrated in
Box 1 | Basic kidney anatomy and physiology
The kidneys purify toxic metabolic waste products from the blood in several hundred thousand functionally
independent units called nephrons. A nephron consists of one glomerulus and one double hairpin-shaped tubule
that drains the filtrate into the renal pelvis. The glomeruli located in the kidney cortex are bordered by the Bowmans
capsule. They are lined with parietal epithelial cells and contain the mesangium with many capillaries to filter the
blood. The glomerular filtration barrier consists of endothelial cells, the glomerular basement membrane and visceral
epithelial cells (also known as podocytes). All molecules below the molecular size of albumin (that is, 68 kDa) pass
the filter and enter the tubule, which consists of the proximal convoluted tubule, the loop of Henle and the distal
convoluted tubule. An intricate countercurrent system forms a high osmotic gradient in the renal medulla that
concentrates the filtrate. The tubular epithelial cells reabsorb water, small proteins, amino acids, carbohydrates
and electrolytes, thereby regulating plasma osmolality, extracellular volume, blood pressure and acidbase and
electrolyte balance. Non-reabsorbed compounds pass from the tubular system into the collecting ducts to form
urine. The space between the tubules is called the interstitium and contains most
of the intrarenal immune system, which mainly consists of dendritic cells,
but also of macrophages and fibroblasts.
Nature Reviews | Immunology
Kidney
Mesangial
cell
Podocyte
Bowmans capsule
Bowmans
space
Nephron
Endothelial
cell
Tubular
epithelial
cell
Glomerular
basement
membrane
Parietal
epithelial
cell
Proximal
convoluted
tubule
Ureter
Distal
convoluted
tubule
Glomerulus
Loop of
Henle
Ureter
Collecting
duct
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 739
2013 Macmillan Publishers Limited. All rights reserved
Tubules
Hairpin-like structures that
receive filtered blood. The
tubular epithelium reabsorbs
water, electrolytes, nutrients
and proteins. Each nephron
has a single tubule, which
defines proximal and distal
tubules as parts of the
nephron.
Chronic kidney disease
(CKD). Chronic (and often
progressive) impairment of
renal functions, such as blood
purification, barrier function
of the glomerular filter, water,
electrolyte and acidbase
homeostasis, endocrine
functions such as vitaminD
processing, erythropoietin
production and blood pressure
regulation.
Uraemia
End-stage chronic kidney
disease, the treatment of
which requires dialysis or
kidney transplantation.
Glomerulonephritis
A heterogeneous group of
immune-mediated kidney
diseases that initiate in the
glomeruli.
Podocyte
A visceral epithelial cell
that covers the glomerular
capillaries in the Bowmans
capsule. Podocytes are a
component of the glomerular
filtration barrier.
Fibrocytes
Monocyte-derived collagen-
producing cells that have been
suggested to contribute to
kidney fibrosis.
Kidney fibrosis
The end stage of chronic kidney
disease, when functional renal
tissue has been replaced by
fibrotic scar tissue and is usually
accompanied by uraemia.
the kidney proximal tubules, where >85% of the fluid is
reabsorbed. Thus, renal DCs and the renal lymph nodes
receive low-molecular-mass antigens from the circulation
at concentrations that are over tenfold higher than in any
other tissue. BATF3-dependent DCs in the renal lymph
nodes capture and cross-present these proteins to CD8
+

Tcells, which results in the programmed cell death1
ligand 1 (PDL1)-mediated deletion of these Tcells
21
.
Thus, the renal lymph nodes have a special role in the
development of immune tolerance against circulating
innocuous low-molecular-mass proteins, such as food
antigens and hormones.
Immune-mediated kidney disease
The kidneys are a frequent target of systemic immune and
autoimmune disorders, including systemic autoimmunity
and vasculitis, immune complex-related serum sickness
and complement disorders. This is partly related to the
size-selective and charge-dependent filtration process in
the glomeruli that promotes glomerular immune com-
plex deposition. In addition, immune responses against
kidney-derived autoantigens can cause autoimmune
kidney diseases.
In chronic kidney disease (CKD), low-molecular-
mass compounds accumulate in the body, which causes
uraemia. CKD affects approximately 10% of the Western
population and is a serious social and economic burden,
especially for those who progress to kidney failure and
that require dialysis or transplantation. The tissue injury
associated with CKD is commonly directly or indirectly
caused by the immune system (BOX2).
Direct immune-mediated injury often affects the glo-
meruli, at least initially, which causes different forms of
glomerulonephritis. Irreversible kidney damage occurs
when inflammation spreads to the tubulointerstitium
2224
.
Various mechanisms that cause this spreading have been
proposed: podocyte damage might facilitate leakage of the
glomerular filtrate and detachment of tubular cells from
their basement membrane
25
; destruction of glomerular
capillaries might restrict the perfusion of their downstream
tubulointerstitial capillaries and cause ischaemia
26
; pro-
inflammatory cytokines from inflamed glomeruli might
perfuse the tubulointerstitial capillaries and cause inflam-
mation
27
; reabsorption of abnormal amounts of protein
from the glomerular filtrate might induce stress responses
in tubular epithelial cells
28
; and glomerular antigens might
reach DCs in the adjacent tubulointerstitium, which in
turn might stimulate infiltrating Tcells to produce pro-
inflammatory cytokines
19
. Tubulointerstitial mono nuclear
cell infiltrates can contribute to continuing immuno-
pathology and to progressive tissue remodelling, which lead
to tubular atrophy and interstitial scarring, both by main-
taining local chronic inflammation and by recruiting fibro-
cytes
29
. The end state of CKD is kidney fibrosis a state in
which functional nephrons are replaced by fibrotictissue.
Immune-mediated CKD can be induced by immune
complex deposition, by innate immunity and by Tcells
that interact with kidney-resident immune cells.
Importantly, these immune mechanisms generally con-
tribute to the progression of CKD, even in non-immune-
initiated forms of the disease, and therefore there are
obvious implications for therapy.
Box 2 | Kidney disorders grouped by their involvement in immunity
Kidney disorders that are initiated and mainly mediated by an immune response
Renal infections with renotrophic pathogens, including uropathogenic Escherichia coli (UPEC), Hantan virus, BK virus,
Leptospira spp., Mycobacterium tuberculosis and HIV
Extrarenal infections with renal manifestations, including septic kidney injury, immune complex-mediated nephritis
(for example, post-infectious glomerulonephritis and endocarditis, hepatitis and virus-related immune complex
glomerulonephritis), interstitial nephritis and HIV nephropathy
Systemic autoimmunity against ubiquitous antigens with renal inflammation, including IgA nephropathy or Henoch
Schnlein purpura, lupus nephritis, Sjgrens syndrome, anti-neutrophil cytoplasmic antibody (ANCA)-associated
vasculitis, interstitial nephritis, secondary membranous nephropathy and antibody-mediated forms of atypical
haemolytic uraemic syndrome (aHUS)
Immune responses against renal antigens, including anti-glomerular basement membrane (anti-GBM) autoimmune
disease, the autoimmune disease primary membranous nephropathy and allograft rejection
Other systemic disorders that affect the kidneys and that have genetic (including, complement C3 glomerulonephritis
and aHUS) or unclear (including, minimal change disease and renal sarcoidosis) causes
Kidney disorders that involve renal inflammation as a secondary mechanism
Systemic autoimmunity against ubiquitous antigens with renal manifestations causing renal vascular obstruction
and ischaemia, including scleroderma renal crisis, panarteritis nodosa, giant cell vasculitis or phospholipid antibody
syndrome
Other systemic disorders that affect the kidney, including genetic disorders such as hereditary defects of GBM or
podocyte genes leading to focal segmental glomerulosclerosis and hereditary tubulopathies or polycystic disorders;
disorders driven by toxins, including Shiga toxin-producing Escherichia coli-induced HUS, drug- or contrast
media-induced kidney injury; crystal and paraprotein-related nephropathies; and disorders caused by metals or
food-borne toxins and toxic forms of focal segmental glomerulosclerosis
Disorders that affect haemodynamics and the vascular system can also affect the kidney, including atherosclerosis,
embolism, macro- or microvascular stenosis, shock, hepato-renal syndrome, thrombotic microangiopathy, eclampsia,
hyperfiltration-associated focal segmental glomerulosclerosis, global glomerulosclerosis
Obstructive nephropathy or renal amyloidosis
REVI EWS
740 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Nature Reviews | Immunology
Toxins, ischaemia and trauma
Activation
Renal
tissue
Apoptosis Necrosis
PRR
PRR-expressing
renal cell
Dendritic cell Endothelial cell Tubular epithelial cell Podocyte
DAMP
Macrophage Mesangial cell
Antigen
presentation
Migration
Type I IFNs, CXCL2,
IL-1 and IL-12
Acute kidney injury
and infections
Permeability
TNF, IL-6, chemokines
and IFN
Adhesion
molecules
IC-GN, diabetes and
HUS
Permeability
TNF, IL-6 and
chemokines
Proteinuria
Most glomerular
diseases
Permeability
TNF, IL-6 and
chemokines
Proteinuria
Acute kidney injury
and late-stage GN
ROS, IL-1,
TNF, IL-6 and
chemokines
Most kidney
diseases
TNF, IL-6,
chemokines
and IFN
IC-GN,
diabetes and
sepsis
Inflammasome
An intracellular complex
containing pattern recognition
receptors that activate
caspase1. Caspase 1
activation induces pyroptotic
cell death and interleukin-1
(IL-1) and IL-18 secretion.
Innate immune responses in CKD. Clinical entities of kid-
ney disease, such as post-ischemic and toxic acute kidney
injury, as well as nephropathies that are induced by dia-
betes, hypertension and crystal deposition, involve sterile
inflammation. As in other organs, sterile renal inflamma-
tion is induced by intrinsic damage-associated molecular
patterns (DAMPs) that are either released from dying
parenchymal cells or that are generated during extracellu-
lar matrix remodelling
3033
. The kidney hosts a large range
of different parenchymal cell types, including tubular
epithelial cells, and endothelial cells that express a subset
of Toll-like receptors (TLRs; that is, TLR1 to TLR6) and
inflammasome components, which suggests that these cells
can respond to DAMPs and that they can induce innate
immune responses and subsequent renal inflammation
34
.
However, NLRP3 (NOD-, LRR- and pyrin domain-con-
taining 3) inflammasome activation is limited to renal
mononuclear phagocytes. The resulting inflammation
depends on the nature of the stimulus (whether it is tran-
sient, repetitive or persistent) and the renal compartment
that is affected (FIG.1); for example, glomerular deposi-
tion of antibodies or immune complexes and the activa-
tion of complement and Fc receptor signalling drives the
several forms of immune complex glomerulonephritis
that have been described (BOX2; see below).
By contrast, ischaemia, toxins, crystals and urinary
outflow obstruction target the tubulointerstitial com-
partment, in which they drive sterile inflammation.
Renal tubular epithelial cells are highly susceptible to
intrinsic oxidative stress because of their high reabsorp-
tive and secretory activity and because their capillary
network is downstream of the glomerular capillaries,
which renders the medullary part of the tubulointer-
stitium susceptible to hypoxia, as occurs during renal
hypoperfusion and shock. During sepsis and ischae-
miareperfusion injury, necrotic tubular cells and
neutrophils release high-mobility group box1 protein
(HMGB1), histones, heat-shock proteins, hyaluronan,
fibronectin, biglycan and other DAMPs that activate
TLR2 and TLR4 on renal parenchymal cells and renal
DCs. Renal parenchymal cells and DCs then secrete
chemokines that promote an acute neutrophil-dependent
inflammatory response that mainly contributes to acute
kidney injury
3537
. Another important DAMP is ATP
that triggers sterile inflammation in the kidneys via
the NLRP3 inflammasome
38
. By contrast, adenosine
receptor A2a signalling inactivates DCs and abrogates
kidney injury
39
. The DAMP Tcell immunoglobulin
and mucin domain-containing protein1 (TIM1; also
known as kidney injury molecule 1) is induced on the
Figure 1 | Innate immune mechanisms in kidney inflammation. Renal cell necrosis or programmed forms of
inflammatory cell death release damage-associated molecular patterns (DAMPs) into the extracellular space, where
they activate pattern recognition receptors (PRRs). Renal dendritic cells and macrophages express numerous PRRs,
whereas PRR expression is limited on renal non-immune cells. PRR ligation activates the cell, which results in cell
type-specific consequences, such as the secretion of pro-inflammatory mediators that promote renal
immunopathology. In the glomerulus, PRR activation in mesangial cells also stimulates their proliferation, for example,
in mesangioproliferative forms of glomerulonephritis such as lupus nephritis, IgA nephropathy and hepatitis C
virus-associated glomerulonephritis. PRR activation of endothelial and epithelial cells (including podocytes and
tubular epithelial cells) in the glomerulus increases their permeability, which results in proteinuria, a clinically useful
biomarker of glomerular vascular permeability, inflammation and damage. Moreover, the activation of endothelial and
epithelial cells manifests as interstitial oedema and secretory dysfunction, for example, in septic acute kidney injury.
CXCL2, CXC-chemokine ligand 2; GN, glomerulonephritis; HUS, haemolytic uraemic syndrome; IC-GN, immune
complex glomerulonephritis; IFN, interferon; IL, interleukin; ROS, reactive oxygen species; TNF, tumour necrosis factor.
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 741
2013 Macmillan Publishers Limited. All rights reserved
Haemolytic uraemic
syndrome
(HUS). A group of diseases,
which are induced by infection
with Shiga toxin-producing
bacteria, or by genetic or
acquired defects in
complement regulators,
that are characterized by
microvascular injury and
thrombosis, which results
in haemolytic anaemia,
thrombocytopenia and
organ dysfunction (kidney
and often brain).
Thrombotic
thrombocytopenic purpura
(TTP). A rare life-threatening
disease, characterized by the
development of platelet
thrombi and microvascular
injury, which results from either
genetic or acquired defects of
the enzyme a disintegrin and
metalloproteinase with
thrombospondin motifs 13
(ADAMTS13), which has a
unique role in the homeostasis
of the coagulation system.
surface of tubular epithelial cells and binds to CD300b
(also known as CLM7) on myeloid cells, which drives
neutrophil recruitment to the post-ischemic kidney
31
.
The initial inflammatory response is amplified by infil-
trating neutrophils and later by LY6C
hi
macrophages,
which results in acute kidney injury. The cellular
pathophysiology of ischemic acute kidney injury has
recently been reviewed by others
40
.
Tubular cells are especially sensitive to the freely
filtered low-molecular-mass toxins that they reabsorb
from the tubular fluid. These toxins can accumu-
late and induce tubular cell necrosis and subsequent
TLR4-mediated tubulointerstitial inflammation
41
. The
high osmolarity and varying pH of urine promotes the
crystallization of small filtered molecules, such as uric
acid, calcium oxalate, calcium phosphate, myoglobin
and free immunoglobulin light chains in the tubules. The
crystals obstruct the tubules and directly injure the epi-
thelial cells that line them, which indirectly causes sterile
inflammation; examples of such crystalline nephropa-
thies include kidney stone disease, oxalate nephropathy,
acute urate nephropathy, adenine nephropathy, cysti-
nosis, rhabdomyolysis-induced acute kidney injury and
myeloma-associated cast nephropathy. A recently dis-
covered pathological mechanism of sterile renal inflam-
mation is that crystals that reach the tubulointerstitial
compartment can directly induce inflammation by
activating the NLRP3 inflammasome in renal DCs
34
.
In addition, urinary outflow obstruction causes renal
sterile inflammation through multiple mechanisms. It
remains to be clarified which kidney diseases will ben-
efit most from selective therapeutic blockade of these
aforementioned innate immune pathways. Persistent
renal inflammation is usually associated with epithelial
atrophy and aberrant mesenchymal cell repair, which is
known as glomerulosclerosis or interstitial fibrosis. The
direct contribution of innate immune responses to pro-
gressive fibrosis remains an area of debate
33,42
. In addi-
tion, NLRP3 has inflammasome-independent effects
in the tubular epithelium; for example, NLRP3 and
the adaptor molecule ASC are needed for SMAD2 and
SMAD3 phosphorylation in response to transforming
growth factor- receptor1 (TGFR1) signalling
4345
. As
TGFR1 signalling is an essential pathway for epithe-
lialmesenchymal transition and renal fibrosis, this non-
canonical effect of NLRP3 contributes to renal scarring.
Whether this process also contributes to other forms of
CKD remains to be studied.
Uromodulin (also known as TammHorsfall pro-
tein) is a kidney-specific molecule that is synthesized
by epithelial cells in the distal tubules and that is selec-
tively released into the tubular lumen. Uromodulin is
an adherent polymer that binds to particles, pathogens,
crystals and cytokines in the urine and facilitates their
elimination. Uromodulin deficiency aggravates uri-
nary tract infections, crystal aggregation and cytokine-
mediated luminal inflammation in the kidneys
46
.
Uromodulin leaks into the interstitium after tubular
injury and activates intrarenal DCs and blood monocytes
via TLR4 and the NLRP3 inflammasome in a DAMP-
like manner
47,48
. This provides another example of
endogenous molecules that function as immunostimula-
tory danger signals when they escape their normal physi-
ological compartment; uromodulin may also contribute
to the systemic inflammation associated withCKD.
Taken together, these findings show that non-infec-
tious triggers induce innate immune responses in the
kidney that can cause inappropriate immunopathol-
ogy. Distinct immune pathways contribute to certain
types of renal sterile inflammation such as the NLRP3
inflammasome in crystalline nephropathies. It remains
necessary to identify the predominant pathways in each
of the many different kidney diseases. Furthermore, the
non-canonical function of NLRP3 during TGF1R sig-
nalling that was first described in kidney disease not
only awaits validation in systemic immune regula-
tion but also deserves further study in different renal
epithelial celltypes.
Complement dysregulation and CKD. Recent advances
in complement biology have led to the reclassifica-
tion of glomerular diseases that are characterized by
complement deposition in the absence of concomitant
antibody deposition
49,50
. Complement C3 glomerulopa-
thies are caused by spontaneous and uncontrolled acti-
vation of the alternative complement pathway because
of mutations in the components or the molecules that
regulate it, such as factor B, factor H, factor I, mem-
brane cofactor protein and factor H-related proteins
5154
.
An autoimmune variant of C3 glomerulopathy is medi-
ated by an autoantibody (known as C3 nephritic factor)
that is specific for C3 convertase. C3 nephritic factor
stabilizes the C3 convertase, which leads to unrestrained
complement activation and the subsequent deposition
of C3 in the kidneys, which is accompanied by variable
pathomorphological findings (most often membrano-
proliferative changes). The importance of recognizing
C3 glomerulopathies as a separate clinical entity is
emphasized by initial reports that indicate the effec-
tiveness of treatment with the C5 inhibitor eculizumab
(Soliris; Alexion Pharmaceuticals)
5557
.
Thrombotic microangiopathy (TMA) is character-
ized by microvascular injury and thrombosis, which
results in haemolytic anaemia with erythrocyte frag-
mentation, thrombocytopenia and organ dysfunc-
tion. The kidney and brain are primarily affected by
this disease and the functional impairment in these
organs mainly determines the outcome of the patients.
The classification, pathogenesis and treatment strate-
gies of TMA remain controversial. Three major types of
TMA are commonly identified: two forms of haemolytic
uraemic syndrome (HUS), including Shiga toxin-
producing Escherichia coli-induced HUS (STEC-HUS)
and atypical HUS (aHUS), as well as thrombotic thrombo-
cytopenic purpura (TTP). Recent studies have improved
our knowledge of all three groups of disease.
Infection with Shiga toxin-producing E.coli, which
cause haemorrhagic enteritis, is the most common cause
of HUS in children. After translocation across the intes-
tinal epithelium, the Shiga toxin is transported in the cir-
culation by poorly defined mechanisms to capillary beds
in target organs. In the kidneys, Shiga toxin binds to the
REVI EWS
742 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Anti-neutrophil cytoplasmic
antibody
(ANCA). An autoantibody
that is commonly found in
pauci-immune focal necrotizing
glomerulonephritis.
Crescentic
glomerulonephritis
A rapidly progressive
form of glomerulonephritis
characterized by the
hyperproliferation of parietal
epithelial cells, which is driven
by Tcell and macrophage
infiltrates and by plasma
components leaking through
the glomerular filter.
Delayed-type
hypersensitivity
(DTH). An inappropriate
Tcell-initiated response to
self or foreign antigens that is
carried out by macrophages,
eosinophils or cytotoxic Tcells.
glycolipid receptor globotriaosylceramide (Gb3), which is
highly expressed on the glomerular endothelium, thereby
initiating the events that are responsible for microvascular
cell injury. Shiga toxin directly induces the expression of
P-selectin on human endothelial cells, and P-selectin then
binds to and activates complement C3 via the alternative
complement pathway, which leads to thrombus forma-
tion in the microvasculature
58
. This can be prevented
by treatment with a C3a receptor antagonist in a mouse
model of STEC-HUS
58
. Children with STEC-HUS have
complement hyperactivation
59
, and early reports docu-
ment marked improvement in small numbers of patients
shortly after treatment with eculizumab
60
. This is sup-
ported by a clinical study that used eculizumab during
the major STEC-HUS outbreak in northern Germany in
2011 (R.A.K.Stahl, personal communication).
Complement is also central to the pathogenesis of
aHUS, which is a rare group of disorders that includes
sporadic and familial diseases and that is often caused
by uncontrolled complement activation as a result of
innate or acquired defects in the regulatory components
of the complement system. In particular, mutations in
the genes that encode factor H, membrane cofactor pro-
tein, factorI and thrombomodulin have a crucial role
in aHUS
61
. Interestingly, the same mutations underlie
C3glomerulopathy (see above). Eculizumab has become
the first-line therapy in aHUS
62
. How similar and/or
identical defects in regulatory proteins of the alternative
complement pathway lead to a range of phenotypical
manifestations of systemic and renal disease remains to
be fully elucidated.
TTP has been linked to reduced activity of a disintegrin
and metalloproteinase with thrombospondin motifs13
(ADAMTS13), which results from either genetic or a
cquired defects, including the generation of ADAMTS13-
specific autoantibodies. Reduced ADAMTS13 activity
leads to the disruption of vonWillebrand factor-multimer
processing, the development of platelet thrombi and
microvascular injury
63
.
The major advances in the field of C3 glomerulopathy
and thrombotic microangiopathies now provide the basis
for a new pathogenesis-based disease classification, and
complement dysregulation is likely to be a general feature
in all of these disease entities. Most importantly, this gain
in understanding has resulted in the use of terminal com-
plement inhibition as a first-line therapy in aHUS, and
might also result in its use in the other forms of HUS in
certain circumstances in the future
61
. Moreover, hyperac-
tivation of C5a and its receptor may also be involved in
other renal autoimmune diseases such as anti-neutrophil
cytoplasmic antibody (ANCA)-associated vasculitis
64
.
Tcell responses targeting the kidney
DTH in crescentic glomerulonephritis. Glomerular cres-
cents, formed by proliferation of the glomerular pari-
etal epithelial cells and infiltrating leukocytes, are the
morphological hallmarks of the most aggressive form of
glomerulonephritis that progresses rapidly towards kid-
ney failure. Despite being first described 100years ago,
nephrotoxic nephritis remains one of the most widely
studied mouse models of crescentic glomerulonephritis.
It is induced by injecting mice with heterologous anti-
bodies specific for the glomerular basement membrane
(GBM) (Supplementary information S1 (table)). Injury
in this model was initially thought to be exclusively
mediated by antibodies
65
. Subsequent studies sug-
gested that there might also be roles for antigen-specific
Tcells
6668
, and Holdsworth and colleagues
69
established
that T cell-dependent delayed-type hypersensitivity
(DTH) responses to the heterologous immunoglobulins
deposited in the kidney were an underlying mechanism
of injury (FIG.2).
Recent studies showed the following sequence of
events to take place. In the first days following anti-
body injection, innate immune cells, including neutro-
phils, mast cells
15
and interleukin-17 (IL-17)-producing
Tcells
70
, mediate renal damage. Tcells specific for
the heterologous antibodies are simultaneously primed
in the lymphatic tissues and start entering the kidneys.
A first wave of T cells, starting 4days after nephritis
induction, consists of pathogenic T helper 17 (T
H
17)
cells expressing CC-chemokine receptor 6 (CCR6)
and retinoic acid receptor-related orphan receptor-t
(RORt)
71-74
. Their activity is controlled by CXC-
chemokine receptor 6 (CXCR6)-expressing regula-
tory invariant natural killer T (iNKT) cells, which
are recruited by immature renal DCs secreting CXC-
chemokine ligand 16 (CXCL16)
75
. If inflammation fails
to resolve, renal DCs eventually mature and recruit
CXCR3
+
T
H
1 cells by producing CXCL9 (REFS76,77).
T
H
1 cells encounter antigens presented by DCs in
the context of upregulated co-stimulatory molecules
and IL-12. Next, activated T
H
1 cells recruit more pro-
inflammatory cells, including monocytes and fibro-
cytes
29
, and stimulate mannose receptor-dependent
macrophages
78
to produce injurious mediators such as
tumour necrosis factor (TNF) and nitric oxide
69,72
. As
renal DCs are located in the interstitium but not within
the glomeruli, the stimulation of T
H
1 cells takes place in
the periglomerular space, adjacent to parietal epithelial
cells. The proliferative response of parietal epithelial
cells and immune cells contributes to the characteristic
glomerular crescents. CCR6
+
and CCR7
+
regulatory T
(T
Reg
) cells may still be able to control inflammation at
this stage
7981
. The severity of the initial injury deter-
mines the balance between pro-inflammatory and
anti-inflammatory Tcells in the tissue, and whether
kidney disease resolves or progresses to fibrosis. After
14days, host antibodies that have been raised against
the heterologous antibodies increasingly contribute to
kidneyinjury.
Although immunity in nephrotoxic nephritis is
directed against a different antigen than in human
crescentic glomerulonephritis, this model has been
instrumental in elucidating the mechanisms that drive
immune responses to glomerular antigens and has made
crucial contributions to the design of novel therapies.
However, the extent to which DTH is also responsi-
ble for human crescentic glomerulonephritis remains
uncertain. Furthermore, it would be desirable to study
whether these cellular immune mechanisms are also
relevant in other forms of glomerulonephritis.
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 743
2013 Macmillan Publishers Limited. All rights reserved
Nature Reviews | Immunology
IL-23R
Neutrophil
R
e
n
a
l

c
e
l
l

i
n

l
t
r
a
t
i
o
n
R
e
n
a
l

c
e
l
l

i
n

l
t
r
a
t
i
o
n
T cell T
H
17 cell
DC
CCR6
CCR2
Pro-inammatory responses a
b
Anti-inammatory responses
IL-17?
T
H
1 cell
CXCR3
CCR2 CX
3
CR1
Macrophage
TNF and
nitric oxide
IFN
Autologous
antibodies
Regulatory
iNKT cell
Immature DC
Acute inammation
Day 4
CXCR6
CCR6
CCR7
IL-4 and
IL-10
CXCL16
T
Reg
cell
1 week 2 week
Months?
1 week 2 week
Months?
IL-10 and
PDL1
Irreversible brosis
Glomerular
sclerosis
Tubulointerstitial
brosis
Fibrosis
Kidney
Kidney
injury
Tcell-mediated glomerular injury. The role of Tcells
in renal injury has long been controversial
6567
. A recent
study using transgenic mice showed that adoptively trans-
ferred CD4
+
T
H
cells and cytotoxic CD8
+
Tcells that are
specific for glomerular antigens can injure the kidneys
19
.
The resulting release of glomerular antigens starts a
vicious circle involving antigen capture and presentation
by renal DCs to T
H
cells, the production of chemokines
and cytokines, the recruitment of more CD8
+
Tcells
and macrophages, and increased renaldamage.
These findings, together with those in nephro-
toxic nephritis, emphasize the importance of crosstalk
between mature renal DCs and T
H
cells; in both cases
the removal of kidney DCs in mice by depletion
19,82
, by
CX
3
CR1 blockade or by genetic knockout
9,83
rapidly
reduced the mononuclear cell infiltration and halted
disease progression. Although the route by which glo-
merular antigens reach DCs in the tubulointerstitium
is still unclear, their ability to do so and to stimulate
T
H
cells may contribute to the spreading of glomerular
Figure 2 | Cellular immune response in experimental crescentic glomerulonephritis. The time-dependent changes
in the pro-inflammatory and anti-inflammatory functions of leukocyte subsets during the course of experimental crescentic
glomerulonephritis (a nephrotoxic nephritis model) are shown. a | The clinical outcome of the disease mainly depends on
the balance between pro-inflammatory and anti-inflammatory immune cells. Whether this concept is relevant to human
crescentic glomerulonephritis remains to be shown. Neutrophil recruitment to the kidney starts several hours after the
induction of nephrotoxic nephritis and is partly mediated by interleukin-17A (IL-17A)-producing Tcells, which are
activated by IL-23. The adaptive immune response is initiated by mature dendritic cells (DCs) that depend on CX
3
C-chemokine
receptor1 (CX
3
CR1) and CC-chemokine receptor 2 (CCR2). At earlier stages, immune responses that are mediated by
CCR6expressing Thelper 17 (T
H
17) cells predominate, whereas at later stages, CXC-chemokine receptor 3 (CXCR3)
+

T
H
1 cells are the prevailing mediators of renal tissue injury, as they produce cytokines such as interferon- (IFN), which
activate macrophages. In addition, host antibodies against the heterologous antibodies form intrarenal immune complexes
and thereby contribute to renal tissue damage. During the first days immature DCs attenuate crescentic glomerulonephritis
by attracting regulatory invariant natural killer T (iNKT) cells via the CXC-chemokine ligand 16 (CXCL16)CXCR6 axis, and
these cells produce IL-4 and IL-10 and thereby might reduce the destructive T
H
1 and T
H
17 cell responses. At a later stage,
CCR6
+
and CCR7
+
regulatory T (T
Reg
) cells are recruited into the inflamed kidney and protect against an overwhelming
T
H
1cell and T
H
17 cell-mediated immune response, at least partly through the local production of IL-10 and the expression
of programmed cell death 1 ligand 1 (PDL1). b | Periodic acid-Schiff (PAS) staining of kidney sections from patients with
acute crescentic glomerulonephritis shows glomerular and tubulointerstitial leukocyte infiltration. Irreversible kidney
damage occurs along with glomerular sclerosis and tubulointerstitial fibrosis when the inflammatory response persists.
IL-23R, IL-23 receptor; TNF, tumour necrosis factor. Image courtesy of U. Helmchen, Hamburg, Germany.
REVI EWS
744 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Proteinuria
The urinary loss of protein,
which has numerous clinical
consequences. Proteinuria is
also used as a biomarker for
renal filter dysfunction.
Anti-GBM disease
(Anti-glomerular basement
membrane disease; also known
as Goodpastures disease).
A severe form of crescentic
glomerulonephritis caused
by autoantibodies that are
specific for the NC1 domain of
the 3 chain of typeIV collagen
(3(IV)NC1) in the GBM.
Membranous nephropathy
A glomerulonephritis form
characterized by the
subepithelial deposition of
secretory phospholipase A2
receptor (PLA2R)-specific
antibodies, which leads to
podocyte injury and heavy
proteinuria. It is the most
common cause of the
nephrotic syndrome in adults.
Nephrotic syndrome
A syndrome characterized
by heavy proteinuria,
hypoalbuminaemia and
a loss of immunoglobulins,
which results in humoral
immunodeficiency, oedema,
hyperlipidaemia and
thrombosis. This syndrome
results from damage to the
glomerular filter, which
causes the loss of proteins
above 50 kDa in size from
the circulation.
injury to the tubulointerstitium
68
, and therefore may
represent a mechanism of kidney disease progression.
However, the relevance of these immune mechanisms
for human glomerulonephritis remains to be shown. In
particular, the role of cytotoxic T lymphocytes (CTLs)
in human nephritis is unclear. In addition, the (auto)
antigens presented to T
H
cells remain to be identified.
Finally, intrinsic renal cells, such as glomerular podo-
cytes
84
and tubular epithelial cells
85
, can also present anti-
gen to Tcells, but the invivo relevance of these processes
is unclear.
Proteinuria. Damage to the glomerular filtration bar-
rier causes protein to leak into the glomerular filtrate,
which results in abnormally high concentrations in the
urine: this is known as proteinuria. Proteinuria can itself
cause injury, which is mediated either by the properties
of specific proteins in the filtrate or simply through the
mass of filtered protein; for example, fibrin can induce
the proliferation of parietal glomerular epithelial cells
and thus can aggravate crescentic glomerulonephritis
86
.
Increased protein in tubular fluid enhances reabsorption
by the tubular epithelial cells and can overload their cata-
bolic capacity, which results in a lysosomal burst and the
release of cathepsins into the cytoplasm
28
. Filtered com-
plement components, especially properdin (also known
as factor P), contact the tubular epithelial cells and acti-
vate the alternative complement pathway that damages
tubular cells
87,88
. Tubulointerstitial DCs capture filtered
proteins, either directly or from tubular cells, and use
them to locally stimulate infiltrating CTLs or T
H
cells
82,89
.
Such presentation of antigens that would normally be
ignored may contribute to the infiltration of immune
cells into the tubulointerstitium and to the progression
of renal disease, but the relevance of this mechanism to
human kidney disease remains to be shown. Regardless
of the mechanisms involved, non-specifically reducing
proteinuria for example, by lowering glomerular fil-
tration pressure by the pharmacological inhibition of the
reninangiotensin system has become an important
therapeutic concept.
Antibody-dependent kidney diseases
Rodent studies have increased our understanding of
the nature of the immune responses in the kidneys and
how they are subverted to cause injury. Furthermore, the
examination of the patterns of immunoglobulin depo-
sition in the kidneys initiated the ultimately successful
search for autoantibodies in human anti-GBM disease and
membranous nephropathy and lead to the characterization
of the glomerular antigens they recognize.
Anti-GBM disease. Anti-GBM disease, formerly known
as Goodpastures disease, is a severe form of crescentic
glomerulonephritis that is caused by autoantibodies
specific for the non-collagenous 1 (NC1) domain of
the 3 chain of typeIV collagen (3(IV)NC1) in the
GBM
90,91
. TypeIV collagen in the GBM consists of 3, 4
and 5 chains, the NC1 domains of which form hexam-
ers that are stabilized by sulfilimine bonds
92
. Pathogenic
autoantibodies bind to two dominant epitopes on the
3(IV)NC1 domain (EA-3 and EB-3), and to a
homologous epitope on the 5(V1)NC1 domain
(EA-5)
92
. Although they are freely accessible in indi-
vidual NC1 domains, all three epitopes are hidden in
the hexamers and so are unavailable for antibody bind-
ing in the intact GBM. A conformational change in
NC1 hexamers within the GBM is required to expose
the epitopes and to facilitate autoantibody binding,
which then amplifies further conformational changes
and autoantibody binding. This may be an explana-
tion for the rapid development of the injury in this
disease. By contrast, GBM-specific alloantibodies that
develop after transplanting a normal kidney into 5(IV)
NC1-deficient mice recognize epitopes on the surface
of the NC1 hexamer and bind to them without the need
for conformational change
93
.
Susceptibility to anti-GBM disease is strongly influ-
enced by the HLA classII haplotype: over 80% of those
affected carry the HLA-DRB1*15:01 allele
94
. The direct
involvement of HLA-DRB1*15:01 in the specific auto-
immune response to 3(IV)NC1 has been confirmed
invitro using human Tcells
95,96
and in transgenic mice
that only express HLA-DRB1*15:01 (REF. 97). The natu-
rally processed 3(IV)NC1 peptides that were bound
to HLA-DRB1*15:01 on antigen-presenting cells have
been characterized
98
but Tcells from patients with anti-
GBM disease fail to respond to them. These peptides are
fairly resistant to antigen-processing enzymes, whereas
the four epitopes that are commonly recognized by the
patients Tcells are rapidly digested
95,96
. This may be an
explanation as to why NC1-specific autoreactive Tcells
in patients with this disease escape thymic deletion.
Rodent models of autoimmune anti-GBM disease
resemble the human clinical disease and are driven by
similar 3(IV)NC1 epitopes
90,97
, but DTH rather than
antibodies cause the severe injury, at least in mice
81,91
. It
remains to be seen whether the contribution of DTH to
injury in anti-GBM disease has been underestimated in
humans; indeed, T
H
1 cells that are specific for 3(IV)NC1
predominate in the acute phase of anti-GBM disease
in humans but they are replaced by an antigen-specific
IL-10-producing T
Reg
cell response that coincides with a
reduction in anti-GBM antibody levels and with reduced
disease activity
90,95
.
PLA2R-specific antibodies in membranous nephropa-
thy. Membranous nephropathy is a major cause of glo-
merulonephritis with nephrotic syndrome in adults. It is
characterized by the thickening of the GBM and the
deposition of immune complexes between the mem-
brane and the podocytes. Approximately 75% of cases
are idiopathic and 25% are secondary to a wide range
of causes, including neoplasia, infections, drugs and
systemic autoimmune disease. Classic studies using the
Heymann nephritis model of membranous nephropa-
thy (Supplementary information S1 (table)) showed that
circulating antibodies that are specific for megalin (also
known as LRP2) a protein that is expressed on the
surface of glomerular podocytes promote the forma-
tion of immune complexes in the kidneys
99
. However,
human podocytes lack megalin.
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 745
2013 Macmillan Publishers Limited. All rights reserved
IgA nephropathy
The most common form of
glomerulonephritis worldwide.
It is characterized by the
deposition of IgA-containing
immune complexes in the
mesangial compartment of
glomeruli, which leads to
mesangial cell-proliferative
lesions, haematuria and
proteinuria.
Pauci-immune focal
necrotizing
glomerulonephritis
(Pauci-immune FNGN).
A highly inflammatory form
of glomerulonephritis in which
glomerular immune complex
deposits are absent or scarce.
It is commonly associated with
small vessel vasculitis and with
anti-neutrophil cytoplasmic
antibodies.
NETosis
The formation and the release
of neutrophil extracellular
traps (NETs) by activated
neutrophils to ensnare
invading microorganisms.
NETs enhance neutrophil killing
of extracellular pathogens
while minimizing damage to
the host cells.
Humanized mice
Immunodeficient mice that
are engrafted with human
haematopoietic cells or tissues,
or mice that transgenically
express human genes.
The autoantigen in human idiopathic membra-
nous nephropathy was recently identified as secretory
phospholipase A2 receptor (PLA2R; also known as
CLEC13C) on podocytes
100
. PLA2R-specific auto-
antibodies, usually of the IgG4 subclass, were found in
the serum of 5070% of patients with primary mem-
branous nephropathy. Subsequent studies showed that
the levels of PLA2R-specific autoantibodies correlate
with the level of proteinuria and could possibly be used
to predict clinical outcome
100
and disease recurrence
after renal transplantation
101
. So far, there is no proof
that PLA2R-specific autoantibodies are pathogenic,
but a genome-wide association study has shown that
PLA2R1 polymorphisms influence susceptibility to idio-
pathic membranous nephropathy
102
. This study also
confirmed that there is a strong association between the
disease and certain HLA-DQA1 alleles, which suggests
that these HLA classII molecules may facilitate autoim-
munity against PLA2R
102
. However, as only 5070% of
patients with primary membranous nephropathy have
PLA2R-specific autoantibodies, additional autoantigens
remain to be identified. Moreover, the pathophysiological
role of PLA2R-specific autoantibodies is still unknown.
IgA nephropathy. IgA nephropathy is the most common
primary form of glomerulonephritis and is an impor-
tant cause of kidney failure. Recent studies suggest that
a multistep process is involved in the immunopatho-
genesis of this disease. Bcells from patients with IgA
nephropathy produce aberrantly glycosylated IgA
103
,
possibly as a consequence of the aberrant homing of
mucosal Bcells to the bone marrow, where they syn-
thesize under-galactosylated IgA. Patients with IgA
nephropathy develop autoantibodies against under-
galactosylated IgA, which might also cross-react with
mucosal microbial antigens, although this has not been
formally shown. These autoantibodies form immune
complexes in the circulation, which are then deposited in
the glomerular mesangium by mechanisms that are so far
incompletely understood
104
. The deposited immune com-
plexes induce the local expression of pro-inflammatory
mediators and growth factors, which activate mesangial
cells and enhance their secretion of extracellular matrix
proteins, which leads to glomerular sclerosis and loss
of renal function. The presence of IgG and IgA glycan-
specific autoantibodies was shown to correlate with
progressive disease in a large group of patients
105
, which
suggests that these glycan-specific autoantibodies are
potentially pathogenic. However, the factors that are
responsible for the synthesis of under-galactosylated
IgA, autoantibody generation, mesangial deposition of
immune complexes and injury remain elusive.
Lupus erythematosus. The extrarenal mechanisms
of lupus nephritis involve complex genetic variability
that compromises immune tolerance to nuclear auto-
antigens
106108
. The nucleic acid components of nuclear
autoantigens support this process via their TLR-
dependent autoadjuvant effects
109111
. As such, endoge-
nous nuclear particles are handled as viral particles and
induce interferon- signalling
112,113
, which is similar to
viral infections
114,115
. The link between systemic lupus
erythematosus and lupus nephritis is the production
of autoantibodies that bind to autoantigens in the kid-
neys; for example, a subset of double-stranded DNA
(dsDNA)-specific antibodies cross-react with annexin II
on the cell surface, in the cytoplasm and in the nucleus of
mesangial cells
116
, and also cross-react with nucleosomes
in the mesangium and in the glomerular capillary epi-
thelium
117
. The extent and the progression of glomeru-
lar immunopathology depends on the site of immune
complex formation, as this determines the predominant
glomerular cell type that is affected
118
(FIG.3).
Pauci-immune focal necrotizing glomerulonephritis.
Pauci-immune focal necrotizing glomerulonephritis (FNGN)
is a systemic autoimmune disease that is characterized
by crescentic glomerulonephritis. It typically occurs
in the context of systemic small vessel vasculitis and
autoantibodies that bind to neutrophil cytoplasmic
antigens specific for either myeloperoxidase (MPO) or
proteinase 3 (PR3; also known as myeloblastin)
119
. Most
patients with pauci-immune FNGN also have autoan-
tibodies to lysosome-associated membrane glycopro-
tein2 (LAMP2)
120,121
, although the frequency of these
antibodies is controversial
122
. All three target antigens
are released into injured glomeruli by infiltrating neutro-
phils after degranulation or through NETosis
123
. LAMP2 is
also expressed on the surface of the glomerular endothe-
lium
108
. Injury is thought to be autoantibody mediated,
not least because Bcell ablation with rituximab is a highly
effective treatment for pauci-immune FNGN
119
(TABLE1).
Despite this, deposits of immunoglobulin and comple-
ment components in pauci-immune FNGN are small and
restricted to necrotic areas of the kidneys. The role of
complement is being re-evaluated in PhaseI clinical trials
of complement inhibitors because patients with clinically
active disease have systemic complement activation
124,125
.
Finally, there is evidence that cell-mediated immunity is
also involved
126
: lymphocytes infiltrate the glomeruli and
the tubulointerstitium
127
, and there are circulating MPO-
specific and PR3-specific T
H
1 and T
H
17 cells in patients
with pauci-immune FNGN
126
. Furthermore, CD8
+
Tcells
are increased and express a transcriptomic signature that
correlates with the risk of disease relapse
128
.
Clinical
119
and genetic
129
studies combined with
invitro experiments
130
and rodent models
131
provide
compelling evidence that MPO-specific and PR3-specific
autoantibodies can be pathogenic. Mice that have been
injected with antibodies specific for MPO develop pauci-
immune FNGN, although injury is mild in most mouse
strains unless the antibody is administered together with
a neutrophil-activating factor such as TNF, C5a or IL-1
(REFS130,131). This facilitates binding of the antibodies
to circulating neutrophils and promotes their glomeru-
lar localization with the release of MPO
132
. Attempts to
induce pauci-immune FNGN in mice with PR3-specific
antibodies have been unsuccessful
130,131
, except in a sin-
gle report in which PR3-specific autoantibodies from a
patient with pauci-immune FNGN were injected into
humanized mice
133
. This possibly reflects the differences in
PR3 expression by human and mouse neutrophils
125,131
.
REVI EWS
746 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Nature Reviews | Immunology
Subepithelial immune
complex deposits
Podocyte injury
Large proteinuria
Membranous
nephropathy
Primary (PLA2R)
Secondary (lupus
nephritis class V)
Linear immune complex
deposits
Endothelial cell and
podocyte injury
CKD, proteinuria and
haematuria
Anti-GBM disease
Mesangial immune complex
deposits
Mesangial cell injury
Asymptomatic proteinuria
and microscopic haematuria
IgA nephropathy and lupus
nephritis class I and II
C3 deposition
Glomerular cell injury
Asymptomatic
proteinuria and
microscopic haematuria
C3 glomerulopathy and
aHUS
Pauci-immune
Vascular necrosis
CKD, proteinuria and
haematuria
Focal necrotizing
glomerulonephritis
and ANCA-associated
vasculitis
Subendothelial immune
complex deposits
Endothelial cell injury
CKD, proteinuria and
haematuria
Lupus nephritis class III
and IV
PLA2R-specic
antibodies
3(IV)NC1-
specic
antibodies
PR3
LAMP2
Neutrophil
MPO
ANCA
Mesangial
cell
Endothelial
cell
Podocyte
Podocyte
foot process
Antibodies that are specific for recombinant human
LAMP2 bind to the glomerular endothelium and cause
pauci-immune FNGN when they are injected into
Wistar-Kyoto rats
120
.
Mice that have been immunized with MPO develop
autoantibodies and DTH responses characterized by
T
H
1 and T
H
17 cells, but they remain healthy even in the
absence of autoimmune regulator (AIRE) which is
expressed by medullary thymic epithelial cells and which
promotes the expression of tissue-specific antigens
(including MPO) that regulate central tolerance to these
antigens and despite the abundance of MPO in thymic
myeloid cells
134,135
. Mice with autoimmunity to MPO
remain healthy but develop severe pauci-immune FNGN
in response to injection of GBM-specific antibodies at
levels below the threshold required to cause kidney tissue
Figure 3 | Local immune pathways in glomerulonephritis. Glomerular immunopathology often develops from
intraglomerular complement activation via the classical (immune complex-related) or alternative (immune complex-
independent) complement pathway. Immune complexes can form in different compartments of the glomerulus, which
determines the resulting histopathological lesion, as different glomerular cell types are primarily activated in each
compartment. The resulting histopathological lesions determine the classification of glomerulonephritis. Immune
complex deposition in the mesangium activates mesangial cells, which leads to mesangioproliferative glomerulopathies,
such as IgA nephropathy or lupus nephritis classI and II. Subendothelial immune complex deposits activate endothelial
cells, as seen in lupus nephritis classIII and IV. Subepithelial immune complex deposits preferentially activate the
visceral glomerular epithelium that is, podocytes and usually cause massive proteinuria, as these cells are
essential for the glomerular filtration barrier. As a result of the poor regeneration of podocytes compared with that
of the other glomerular cell types, podocyte loss leads to progressive membranous nephropathy and end-stage renal
disease. Primary membranous nephropathy mainly develops from autoimmunity against PLA2R, whereas secondary
forms of this nephropathy represent renal manifestations of systemic disorders such as lupus nephritis. Hence, the level
of proteinuria is an important prognostic biomarker and predictor of poor outcomes of glomerulopathies. Linear
immune complex deposits indicate antibody binding to autoantigens within the glomerular basement membrane (GBM),
for example, collagen IV antibodies in antiGBM disease. Anti-neutrophil cytoplasmic antibody (ANCA)-associated
glomerulonephritis develops in the absence of immune complex deposits (known as pauci-immune), as it is driven by
both ANCAs and cellular immunity. Complement component C3 glomerulopathies and atypical haemolytic uraemic
syndrome (aHUS) develop from the aberrant activation of the alternative complement pathway. The boxes list in
order the type of immune deposits, the glomerular structure that is primarily affected, the dominant clinical signs
and the related disorders for each mechanism. 3(IV)NC1, noncollagenous 1 (NC1) domain of the 3 chain of typeIV
collagen; CKD, chronic kidney disease; LAMP2, lysosome-associated membrane protein 2; MPO, myeloperoxidase;
PLA2R,secretory phospholipase A2 receptor; PR3, proteinase 3.
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 747
2013 Macmillan Publishers Limited. All rights reserved
Table 1 | Implementation of immunosuppressive or anti-inflammatory therapies in the treatment of kidney diseases
Target Drugs Effective in animal kidney
disease models?
Effective in human kidney disease?
IL-1 IL-1-specific antibody or
recombinant IL-1RA
Oxalate nephropathy, IgA
nephropathy and anti-GBM
disease
Unknown
IL-6 IL-6-specific antibody Lupus nephritis, anti-GBM
disease and immune complex
glomerulonephritis
Unknown
IL-17 IL-17-specific antibody Crescentic glomerulonephritis Unknown
TNF TNF-specific antibody
or
TNFRFc fusion protein
Lupus nephritis, anti-GBM and
ANCAs, glomerulonephritis,
glomerulosclerosis and acute
kidney injury
TNF-specific antibody was effective in severe lupus nephritis,
but had side effects
The TNF inhibitor etanercept (Enbrel; Amgen/Pfizer) was not
effective in ANCA-associated vasculitis
TGF TGF-specific antibody
that blocks TGF1
Renal scarring in diabetic
nephropathy
Clinical trials ongoing (NCT01113801*)
TWEAK TWEAK-specific
antibody
Lupus nephritis, lipid
nephropathy and crescentic
glomerulonephritis
Clinical trial ongoing in lupus nephritis (NCT01499355*)
CCR2 CCR2 antagonist Diabetic nephropathy,
hypertensive nephropathy and
crescentic glomerulonephritis
Clinical trial of ongoing in diabetic nephropathy (NCT01447147*)
CCR5 CCR5 antagonist Immune complex
glomerulonephritis and allograft
rejection
Unknown
TLR2 TLR2-specific antibody Acute kidney injury Clinical trial in delayed-kidney allograft function ongoing
(NCT01794663*)
Thymocytes Anti-thymocyte
globulin
Numerous immune disorders Kidney allograft rejection and graft-versus-host disease
Lymphocytes Anti-lymphocyte
globulin
Numerous immune disorders Kidney allograft rejection and graft-versus-host disease
CD52 (on
mature
lymphocytes)
CD52-specific
monoclonal antibody
Numerous immune disorders Clinical trials ongoing in ANCA-associated vasculitis
(NCT01405807*)
IL-2R (also
known as CD25)
IL-2R-specific antibody Allograft rejection Prevention of kidney allograft rejection

B7-1 (also known


as CD80)
CTLA4Fc fusion
protein
Allograft rejection and lupus
nephritis
Prevention of kidney allograft rejection in a PhaseII clinical trial
Negative results from a PhaseIII clinical trial is under debate and
further studies are ongoing (NCT00774852*)
CD20
+
Bcells CD20-specific antibody Lupus nephritis and anti-GBM
disease
Effective in refractory lupus nephritis (uncontrolled studies)
Not effective in LUNAR trial as an add-on to steroids and
mycophenolate mofetil
Effective in clinical trials for ANCA-associated vasculitis

(RAVE
and RITUXVAS trials)
Beneficial in observational studies of membranous
nephropathy; controlled clinical trials ongoing (NCT01508468*;
NCT01180036*)
Trials ongoing in steroid resistant focal glomerulosclerosis
(NCT01573533*; NCT00981838*; NCT00550342*)
BLYS (on Bcells) BLYS-specific antibody SLE, including lupus nephritis Effective in SLE but not specifically for severe lupus nephritis
(further trials ongoing)
Clinical trials ongoing in membranous nephropathy
(NCT01762852*; NCT01610492*)
BAFF (on Bcells) BAFF-specific antibody None reported Effective in SLE and clinical trials ongoing in lupus nephritis
(NCT01639339*)
C5 C5-specific antibody
or orally active C5aR
inhibitor
Anti-MPO FNGN Effective in atypical HUS, unclear data on effectiveness in
STEC-HUS and clinical trials ongoing in ANCA-associated
vasculitis (NCT01363388*)
ANCAs, anti-neutrophil cytoplasmic antibodies; BAFF, B cell-activating factor; BLYS, B lymphocyte stimulator; C5, complement component C5; C5aR, C5a
anaphylatoxin chemotactic receptor; CCR, CC-chemokine receptor; CTLA4, cytotoxic T lymphocyte antigen 4; FNGN, focal necrotizing glomerulonephritis;
GBM, glomerular basement membrane; HUS, haemolytic uraemic syndrome; IL, interleukin; IL-1RA, interleukin-1 receptor antagonist; MPO, myeloperoxidase;
SLE, systemic lupus erythematosus; STEC-HUS, Shiga toxin-producing Escherichia coli-induced haemolytic uraemic syndrome; TGF, transforming growth factor-;
TLR, Toll-like receptor; TNF, tumour necrosis factor; TNFR, tumour necrosis factor receptor; TWEAK, TNF-related weak inducer of apoptosis. *Identifier on
ClinicalTrials.gov.

Treatment approved by the US Food and Drug Administration.


REVI EWS
748 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Nature Reviews | Immunology
Hypoperfusion of
damaged nephrons
Renin, angiotensin
and aldosterone
Uraemia and
retention of
metabolic waste
Erythropoietin Renal anaemia
Immune dysregulation
and calcium and
bone loss
Intestinal barrier
dysfunction and
endotoxaemia
Loss of proteins with
immune functions (such
as immunoglobulin,
zinc-binding protein
and ferritin)
Hypertension
Immunosuppression
T
H
17 cell polarization
Systemic inammation
Chronic
oxidative stress
Vascular damage
and atherosclerosis
Infections
DC polarization
Sodium retention
Vitamin D
Uromodulin
Protein catabolism
Cytokine elimination
Complement turnover
Extensive
proteinuria
injury. Unexpectedly, injury is not caused by autoanti-
bodies, as it occurs in Bcell-deficient mice
134
. Instead, it is
caused by DTH
134
, as it can be transferred by Tcells
136
and
is abrogated in IL-17A-deficient mice
137
. Disease sever-
ity is modulated by forkhead box P3 (FOXP3)
+
T
Reg
cells,
which are induced by IL-10-producing mast cells that are
recruited to regional lymph nodes after immunization
withMPO
17
.
Neutrophil extracellular traps (NETs) are generated
in patients with FNGN
123
and have been suggested to
initiate the synthesis of autoantibodies to MPO. This is
consistent with the observation that the delivery of NETs
to mice either through direct injection or through
adoptive transfer of NET-pulsed DCs
125
induces
autoimmunity to MPO (and to DNA)
125
. However, the
administration of PR3 does not provoke pauci-immune
FNGN in rodents
120,121
.
The stimuli that initiate autoantibody synthesis in
pauci-immune FNGN remain unknown but have been
linked to infection since the earliest clinical descriptions
were made. Recent studies are beginning to suggest why:
nasal carriage of Staphylococcus aureus is associated with
clinical disease relapses
119
, and proteins that are derived
from this pathogen have been shown to induce Bcells
from patients with pauci-immune FNGN to produce
PR3-specific antibodies
138
. Some patients with auto-
immunity to PR3 have been reported to have anti-idio-
typic antibodies that bind to a peptide with a sequence
that is complementary to PR3 (REF.130). The comple-
mentary peptide is similar to staphylococcal and other
microbial proteins, and it has been suggested that these
proteins may function as molecular mimics. However,
these results have not been confirmed
139
. By contrast,
there is strong evidence for molecular mimicry between
LAMP2 and the bacterial adhesion protein FimH
120
.
Autoantibodies specific for LAMP2 commonly bind
to and cross-react with an epitope in FimH. Moreover,
immunization of WKY rats with FimH induces the
production of antibodies that bind to human and rat
LAMP2 and it promotes the development of pauci-
immune FNGN. This confirms the molecular mimicry
between the two molecules and suggests a pathogenic
role for LAMP2-specific autoantibodies. Detailed
prospective clinical analyses are now needed to deter-
mine the role of the molecular mimicry of LAMP2 in
pauci-immuneFNGN.
The effect of CKD on systemic immunity
The state of reduced renal function that results from
CKD causes marked alterations in the immune sys-
tem, including persistent systemic inflammation and
acquired immunosuppression
140
(FIG.4). Typical altera-
tions include increased systemic concentrations of pro-
inflammatory cytokines and acute phase proteins, such
as the pentraxins, as well as dysfunctional phagocytes,
Bcells and Tcells
141
. The persistent systemic inflamma-
tion contributes to bone loss, accelerated atherogenesis
and body wasting, whereas the immunosuppressed state
accounts for infectious complications, which together
determine the morbidity and the mortality that is asso-
ciated with CKD. The immune dysregulation was pre-
viously attributed to the effects of haemodialysis but is
now known to precede it and to persist afterwards
142
.
Several recently discovered consequences of the loss of
kidney functions on immune responses are described
below, which alone, or in concert, may affect general
immunity (FIG.4).
Uraemia. CKD results in the retention of low-
molecular-mass metabolites, such as phenylacetic
acid, homocysteine, various sulfates, guanidine com-
pounds and many others. These have inhibitory effects
on immune cell activation, promote leukocyte apop-
tosis and induce the oxidative burst in phagocytes
143
.
Chronic oxidative stress increases protein oxidation,
which reduces the activity of enzymes, cytokines and
antibodies, contributing to both general inflammation
and immune dysfunction in CKD. Moreover, oxidized
low-density lipoproteins attract and activate granulo-
cytes, and high-density lipoproteins, which are nor-
mally anti-atherogenic, are altered to lipoproteins with
pro-atherogenic properties
144
.
Figure 4 | Consequences of chronic kidney disease with potential effects on
systemic immunity. Chronic kidney disease (CKD) has several immediate
consequences (blue boxes), which are proposed to result in three main immunological
alterations (red boxes) through intermediate steps. First, chronic stimulation of the
reninangiotensinaldosterone system causes T helper 17 (T
H
17) cell polarization,
through dendritic cell (DC) polarization and possibly through sodium retention.
Second, uraemic intestinal barrier dysfunction, vitaminD deficiency and cytokine
accumulation (which may be due to impaired protein catabolism, reduced uromodulin
levels and chronic oxidative stress) result in systemic inflammation. Third, systemic
immunosuppression results from the uraemic accumulation of toxic metabolic waste,
the increased turnover of the components of the alternative complement pathway
because of impaired protein catabolism, and in cases of extensive proteinuria, the
urinary loss of proteins with immunological functions. This figure also integrates the
key clinical consequences of CKD, which include hypertension, vascular damage and
atherosclerosis, renal anaemia and bone loss (in bold). These mechanisms may alone
or in concert affect general immunity.
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 749
2013 Macmillan Publishers Limited. All rights reserved
Uraemia affects systemic immunity by causing intes-
tinal dysbiosis and by destabilizing the intestinal bar-
rier
140,145
(FIG.4). The metabolic consequences of uraemia
favour pathogen overgrowth, which can increase the
production of uraemic toxins inside the gut and can
reduce the production of immunoregulatory short-
chain fatty acids
146
. As in heart failure and liver cirrhosis,
uraemia-related hypervolaemia leads to intestinal wall
congestion, which impairs the intestinal wall barrier and
promotes the leakage of pathogen-associated molecular
patterns (PAMPs) into the circulation
140
. In fact, systemic
lipopolysaccharide (LPS) levels increase in patients with
CKD as renal function declines and are highest among
those on dialysis
147
. Intestinal PAMP leakage may not
only activate innate immune-mediated systemic inflam-
mation but also, paradoxically, could lead to concomi-
tant immunosuppression, through similar mechanisms
that account for endotoxin tolerance invitro and com-
pensatory anti-inflammatory syndrome in patients with
advanced sepsis
148,149
(FIG.4).
Renal protein catabolism. Proteins and polypeptides with
a molecular mass below 50 kDa pass into the glomeru-
lar filtrate and are reabsorbed and catabolized by the
tubular epithelium to enable amino acids to be recycled.
They consequently accumulate in the blood of patients
with CKD, reaching concentrations more than tenfold
higher than normal in severe cases, and they have marked
effects on immune function
143
. Examples of these effects
include the following: an accumulation of IgG light
chains (25 kDa in size) suppresses Bcell and granulocyte
function; increased concentrations of the MHC classI
component 2 microglobulin (45 kDa in size) aggregate
into amyloid fibrils; increased concentrations of leptin
(16 kDa in size) and the granulocyte protein resistin
(12 kDa in size) diminish phagocyte function; increased
levels of complement factor D (27 kDa in size) enhance
the activity of the alternative complement pathway and
generate immunosuppressive fragments (such as the
complement factor B Ba fragment; which, as a result of
its 33 kDa size, also accumulates in CKD on its own
150
);
the accumulation of retinol-binding proteins (21 kDa in
size) may influence the ratio of T
Reg
cells to T
H
17 cells; and
elevated levels of cytokines (typically 1040 kDa in size)
contribute to systemic inflammation (FIG.4).
In proteinuria, proteins larger than 50 kDa in size
are excreted in the urine. The loss of immunoglobulins,
complement factors, zinc-binding protein and trans-
ferrin contributes to the acquired humoral and cel-
lular immunodeficient state that predisposes patients
with nephrotic syndrome to bacterial infections (FIG.4).
Furthermore, several functional Tcell and macrophage
defects have been described in these patients
143
but their
functional relevance is unclear.
Kidney-derived hormones and hypertension.
VitaminD is activated by hydroxylation in the kidneys,
and declining levels in CKD lead to renal osteopathy.
VitaminD has immunosuppressive properties and low
levels predispose individuals to rheumatic disorders
151
.
These disorders are indeed more prevalent in CKD,
but it is unclear whether this is because low vitaminD
levels are pathogenic or because rheumatic diseases
cause CKD, or both. In addition, diseased kidneys
cannot produce sufficient quantities of erythropoietin,
resulting in the development of renal anaemia, which
contributes to oxidative stress that is induced by the
accumulation of uraemic toxins
152
; this is especially
common when anaemia is treated with iron, which
itself causes oxidativestress.
Blood levels of the blood pressure regulator renin are
increased in CKD as a result of the hypoperfusion of the
damaged nephrons (FIG.4). DCs express receptors for the
downstream mediator of renin, aldosterone, and respond
to aldosterone by promoting T
H
17 cell polarization
153
.
Aldosterone increases sodium reabsorption, and high
salt concentrations have recently been shown to maintain
T
H
17 cell polarization and to aggravate T
H
17 cell-driven
autoimmunity in mice
154,155
. IL-17 in turn increases
blood pressure by promoting vascular inflammation
156
.
Sodium retention also causes macrophages to produce
vascular endothelial growth factorC (VEGFC), which
induces neo-lymphangiogenesis in the skin to store the
salt
157
. This in turn increases extracellular volume and,
thus, blood pressure. Hypertension generally promotes
tissue inflammation, and tubulointerstitial nephritis is
known to raise blood pressure
158
. In summary, there are
complex feedback loops involving reninangiotensin
aldosterone stimulation, salt homeostasis, T
H
17 cells and
mononuclear phagocytes that may exacerbate hyperten-
sion and systemic inflammation, and that may promote
autoimmunity (FIG.4). The clinical implications of these
interactions warrant further studies.
Concluding remarks
Numerous discoveries have recently been made in the
field of renal immunology, which have clarified severe
and previously inexplicable kidney diseases; for exam-
ple, the identification of kidney-specific DAMPs, such
as uromodulin, that can drive sterile kidney inflam-
mation, or the identification of autoantigens that are
targeted in prevalent forms of glomerulonephritis, such
as PLA2R in membranous nephropathy. Knowledge
about relevant autoantigens is instrumental for the
design of non-invasive diagnostic procedures, such as
autoantibody assays. Progress has also been made in
understanding why the kidneys are frequent targets of
systemic autoimmunity, especially to injury by altered
antibodies, immune complexes and complement fac-
tors, and this has helped in implementing new treat-
ments in some cases. There are also anatomical and
physiological features that render the kidneys suscep-
tible to distinct forms of immune-mediated injury,
such as the high osmolarity of the renal medulla, which
favours crystal precipitation and inflammasome acti-
vation, or the constitutive renal protein catabolism of
tubular epithelial cells, which exposes them to Tcell
effector functions. Cellular immunity seems to require
more time to destroy the kidneys than it does to destroy
other tissues, making this organ a good site for basic
studies on immune cell crosstalk because immune cell
infiltrates can be observed over a longer time span.
Endotoxin tolerance
A transient state of
hyporesponsiveness of
the host or of cultured
macrophages and/or
monocytes to
lipopolysaccharide (LPS)
following previous exposure
to LPS.
REVI EWS
750 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Novel immune mechanisms that have been uncovered
during such studies some of which are discussed in
this Review may be relevant in the context of other
organ diseases. The revelation that the kidneys contrib-
ute to immune tolerance and that their detoxifying and
electrolyte-balancing activities ensure normal immune
effector cell function and intestinal microbial homeo-
stasis has been surprising. The kidney is the archetypal
organ of homeostasis and it is interesting to see that
this role now extends to the immunesystem.
Despite the progress that has been made, many
questions remain unanswered, some of which are
highlighted in this Review. Although the mechanisms
of kidney disease progression are increasingly well
understood, the factors that initiate these diseases often
remain unclear, for example, in IgA nephropathy, cres-
centic glomerulonephritis and membranous nephropa-
thy. However, the development of new therapies from
basic discoveries has already begun to affect clinical
practise in nephrology (TABLE1).
1. Kaissling,B. & Le Hir,M. Characterization and
distribution of interstitial cell types in the renal cortex
of rats. Kidney Int. 45, 709720 (1994).
2. Kruger,T. etal. Identification and functional
characterization of dendritic cells in the healthy
murine kidney and in experimental glomerulonephritis.
J.Am. Soc. Nephrol. 15, 613621 (2004).
3. Soos,T.J. etal. CX3CR1
+
interstitial dendritic cells
form a contiguous network throughout the entire
kidney. Kidney Int. 70, 591596 (2006).
4. Woltman,A.M. etal. Quantification of dendritic cell
subsets in human renal tissue under normal and
pathological conditions. Kidney Int. 71, 10011008
(2007).
5. Guilliams,M. etal. From skin dendritic cells to
a simplified classification of human and mouse
dendritic cell subsets. Eur. J.Immunol. 40,
20892094 (2010).
6. Miller,J.C. etal. Deciphering the transcriptional
network of the dendritic cell lineage. Nature Immunol.
13, 888899 (2012).
7. Schraml,B.U. etal. Genetic tracing via expression
history of DNGR-1 defines dendritic cells as a
hematopoietic lineage. Cell 154, 843858 (2013).
8. Tittel,A.P. etal. Functionally relevant neutrophilia in
CD11c diphtheria toxin receptor transgenic mice.
Nature Methods 9, 385390 (2012).
9. Hochheiser,K. etal. Exclusive CX
3
CR1-dependence
of kidney dendritic cells impacts glomerulonephritis
progression. J.Clin. Invest. http://dx.doi.org/10.1172/
JCI70143 (2013).
10. Kim,K.W. etal. Invivo structure/function and
expression analysis of the CX3C chemokine
fractalkine. Blood 118, e156e167 (2011).
11. Dong,X. etal. Resident dendritic cells are the
predominant TNF-secreting cell in early renal
ischemia-reperfusion injury. Kidney Int. 71, 619628
(2007).
12. Pindjakova,J. etal. Interleukin-1 accounts for
intrarenal Th17 cell activation during ureteral
obstruction. Kidney Int. 81, 379390 (2012).
13. Merad,M., Ginhoux,F. & Collin,M. Origin,
homeostasis and function of Langerhans cells and
other langerin-expressing dendritic cells. Nature Rev.
Immunol. 8, 935947 (2008).
14. Nelson,P.J. etal. The renal mononuclear phagocytic
system. J.Am. Soc. Nephrol. 23, 194203 (2012).
This paper summarizes the phenotypical range of
mononuclear phagocytes in healthy and diseased
kidneys.
15. Timoshanko,J.R., Kitching,R., Semple,T.J.,
Tipping,P.G. & Holdsworth,S.R. A pathogenetic
role for mast cells in experimental crescentic
glomerulonephritis. J.Am. Soc. Nephrol. 17,
150159 (2006).
16. Scandiuzzi,L. etal. Mouse mast cell protease-4
deteriorates renal function by contributing to
inflammation and fibrosis in immune complex-
mediated glomerulonephritis. J.Immunol. 185,
624633 (2010).
17. Gan,P.Y. etal. Mast cells contribute to peripheral
tolerance and attenuate autoimmune vasculitis.
J.Am. Soc. Nephrol. 23, 19551966 (2012).
18. Dong,X. etal. Antigen presentation by dendritic cells
in renal lymph nodes is linked to systemic and local
injury to the kidney. Kidney Int. 68, 10961108 (2005).
19. Heymann,F. etal. Kidney dendritic cell activation is
required for progression of renal disease in a mouse
model of glomerular injury. J.Clin. Invest. 119,
12861297 (2009).
This study shows that Tcells can induce glomerular
damage and that DC maturation drives
glomerulonephritis progression.
20. Lukacs-Kornek,V. etal. The kidney-renal lymph
node-system contributes to cross-tolerance against
innocuous circulating antigen. J.Immunol. 180,
706715 (2008).
21. Gottschalk,C. etal. Batf3-dependent dendritic cells in
the renal lymph node induce tolerance against circulating
antigens. J.Am. Soc. Nephrol. 24, 543549 (2013).
22. Bohle,A., Kressel,G., Muller,C.A. & Muller,G.A.
The pathogenesis of chronic renal failure. Pathol. Res.
Pract. 185, 421440 (1989).
23. Markovic-Lipkovski,J., Muller,C.A., Risler,T.,
Bohle,A. & Muller,G.A. Association of glomerular
and interstitial mononuclear leukocytes with different
forms of glomerulonephritis. Nephrol. Dial. Transplant.
5, 1017 (1990).
24. Risdon,R.A., Sloper,J.C. & De Wardener,H.E.
Relationship between renal function and histological
changes found in renal-biopsy specimens from
patients with persistent glomerular nephritis.
Lancet 2, 363366 (1968).
25. Kriz,W. & LeHir,M. Pathways to nephron loss
starting from glomerular diseases-insights from
animal models. Kidney Int. 67, 404419 (2005).
26. Bohle,A., Mackensen-Haen,S. & Wehrmann,M.
Significance of postglomerular capillaries in the
pathogenesis of chronic renal failure. Kidney Blood
Press. Res. 19, 191195 (1996).
27. Floege,J. & Grone,H.J. Progression of renal failure:
what is the role of cytokines? Nephrol. Dial.
Transplant. 10, 15751586 (1995).
28. Abbate,M., Zoja,C. & Remuzzi,G. How does
proteinuria cause progressive renal damage?
J.Am. Soc. Nephrol. 17, 29742984 (2006).
29. Niedermeier,M. etal. CD4
+
Tcells control the
differentiation of Gr1
+
monocytes into fibrocytes.
Proc. Natl Acad. Sci. USA 106, 1789217897
(2009).
30. Rock,K.L., Latz,E., Ontiveros,F. & Kono,H.
The sterile inflammatory response. Annu. Rev.
Immunol. 28, 321342 (2010).
31. Yamanishi,Y. etal. TIM1 is an endogenous ligand for
LMIR5/CD300b: LMIR5 deficiency ameliorates mouse
kidney ischemia/reperfusion injury. J.Exp. Med. 207,
15011511 (2010).
32. Anders,H.J. Toll-like receptors and danger signaling
in kidney injury. J.Am. Soc. Nephrol. 21, 12701274
(2010).
33. Rosin,D.L. & Okusa,M.D. Dangers within: DAMP
responses to damage and cell death in kidney disease.
J.Am. Soc. Nephrol. 22, 416425 (2011).
34. Mulay,S.R. etal. Calcium oxalate crystals induce
renal inflammation by NLRP3-mediated IL-1
secretion. J.Clin. Invest. 123, 236246 (2013).
This paper identifies the role of the NLRP3
inflammasome in crystal nephropathy.
35. Leemans,J.C. etal. Renal-associated TLR2
mediates ischemia/reperfusion injury in the kidney.
J.Clin. Invest. 115, 28942903 (2005).
36. Wu,H. etal. TLR4 activation mediates kidney
ischemia/reperfusion injury. J.Clin. Invest. 117,
28472859 (2007).
37. Allam,R. etal. Histones from dying renal cells
aggravate kidney injury via TLR2 and TLR4.
J.Am. Soc. Nephrol. 23, 13751388 (2012).
38. McDonald,B. etal. Intravascular danger signals
guide neutrophils to sites of sterile inflammation.
Science 330, 362366 (2010).
39. Li,L. etal. Dendritic cells tolerized with adenosine
A
2
AR agonist attenuate acute kidney injury.
J.Clin. Invest. 122, 39313942 (2012).
40. Bonventre,J.V. & Yang,L. Cellular pathophysiology
of ischemic acute kidney injury. J.Clin. Invest. 121,
42104221 (2011).
41. Zhang,B., Ramesh,G., Uematsu,S., Akira,S. &
Reeves,W.B. TLR4 signaling mediates inflammation
and tissue injury in nephrotoxicity. J.Am. Soc.
Nephrol. 19, 923932 (2008).
42. Babelova,A. etal. Biglycan, a danger signal that
activates the NLRP3 inflammasome via toll-like and
P2X receptors. J.Biol. Chem. 284, 2403524048
(2009).
43. Shigeoka,A.A. etal. An inflammasome-independent
role for epithelial-expressed Nlrp3 in renal ischemia-
reperfusion injury. J.Immunol. 185, 62776285
(2010).
44. Vilaysane,A. etal. The NLRP3 inflammasome
promotes renal inflammation and contributes to CKD.
J.Am. Soc. Nephrol. 21, 17321744 (2010).
45. Wang,W. etal. Inflammasome-independent NLRP3
augments TGF- signaling in kidney epithelium.
J.Immunol. 190, 12391249 (2013).
This paper shows that NLRP3 and the adaptor
protein ASC are needed for TGFR1 signalling,
which might contribute to kidney fibrosis.
46. Eddy,A.A. Scraping fibrosis: UMODulating renal
fibrosis. Nature Med. 17, 553555 (2011).
47. Saemann,M.D. etal. Tamm-Horsfall glycoprotein
links innate immune cell activation with adaptive
immunity via a Toll-like receptor-4-dependent
mechanism. J.Clin. Invest. 115, 468475 (2005).
48. Darisipudi,M.N. etal. Uromodulin triggers
IL-1-dependent innate immunity via the NLRP3
Inflammasome. J.Am. Soc. Nephrol. 23, 17831789
(2012).
49. Sethi,S. & Fervenza,F.C. Membranoproliferative
glomerulonephritis a new look at an old entity.
N.Engl. J.Med. 366, 11191131 (2012).
50. Fakhouri,F., Fremeaux-Bacchi,V., Noel,L.H.,
Cook,H.T. & Pickering,M.C. C3 glomerulopathy: a new
classification. Nature Rev. Nephrol. 6, 494499 (2010).
51. Chen,Q. etal. Combined C3b and factor B
autoantibodies and MPGN typeII. N.Engl. J.Med.
365, 23402342 (2011).
52. Martinez-Barricarte,R. etal. Human C3 mutation
reveals a mechanism of dense deposit disease
pathogenesis and provides insights into complement
activation and regulation. J.Clin. Invest. 120,
37023712 (2010).
53. Gale,D.P. etal. Identification of a mutation in
complement factor H-related protein 5 in patients of
Cypriot origin with glomerulonephritis. Lancet 376,
794801 (2010).
54. Bomback,A.S. & Appel,G.B. Pathogenesis of the
C3 glomerulopathies and reclassification of MPGN.
Nature Rev. Nephrol. 8, 634642 (2012).
55. Vivarelli,M., Pasini,A. & Emma,F. Eculizumab for the
treatment of dense-deposit disease. N.Engl. J.Med.
366, 11631165 (2012).
56. Daina,E., Noris,M. & Remuzzi,G. Eculizumab in a
patient with dense-deposit disease. N.Engl. J.Med.
366, 11611163 (2012).
57. Trachtman,H., Austin,C., Lewinski,M. & Stahl,R.A.
Renal and neurological involvement in typical
Shiga toxin-associated HUS. Nature Rev. Nephrol.
8, 658669 (2012).
58. Morigi,M. etal. Alternative pathway activation of
complement by Shiga toxin promotes exuberant C3a
formation that triggers microvascular thrombosis.
J.Immunol. 187, 172180 (2011).
59. Thurman,J.M. etal. Alternative pathway of
complement in children with diarrhea-associated
hemolytic uremic syndrome. Clin. J.Am. Soc. Nephrol.
4, 19201924 (2009).
60. Lapeyraque,A.L. etal. Eculizumab in severe
Shiga-toxin-associated HUS. N.Engl. J.Med. 364,
25612563 (2011).
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 751
2013 Macmillan Publishers Limited. All rights reserved
61. Noris,M., Mescia,F. & Remuzzi,G. STEC-HUS,
atypical HUS and TTP are all diseases of
complement activation. Nature Rev. Nephrol.
8, 622633 (2012).
62. Legendre,C.M. etal. Terminal complement inhibitor
eculizumab in atypical hemolytic-uremic syndrome.
N.Engl. J.Med. 368, 21692181 (2013).
This is a groundbreaking study showing that
terminal complement inhibition (using eculizumab)
abolishes complement-mediated thrombotic
microangiopathy and results in a time-dependent
improvement in renal function in patients with
aHUS.
63. Furlan,M. etal. von Willebrand factor-cleaving
protease in thrombotic thrombocytopenic purpura
and the hemolytic-uremic syndrome. N.Engl. J.Med.
339, 15781584 (1998).
64. Schreiber,A. etal. C5a receptor mediates neutrophil
activation and ANCA-induced glomerulonephritis.
J.Am. Soc. Nephrol. 20, 289298 (2009).
65. Dixon,F.J. What are sensitized cells doing in
glomerulonephritis? N.Engl. J.Med. 283, 536537
(1970).
66. Couser,W.G. Sensitized cells come of age: a new era
in renal immunology with important therapeutic
implications. J.Am. Soc. Nephrol. 10, 664665
(1999).
67. Bolton,W.K. What sensitized cells just might be doing
in glomerulonephritis. J.Clin. Invest. 109, 713714
(2002).
68. Sung,S.S. & Bolton,W.K. Tcells and dendritic
cells in glomerular disease: the new
glomerulotubular feedback loop. Kidney Int. 77,
393399 (2010).
69. Tipping,P.G. & Holdsworth,S.R. Tcells in crescentic
glomerulonephritis. J.Am. Soc. Nephrol. 17,
12531263 (2006).
70. Turner,J.E. etal. IL-17A production by renal
Tcells promotes kidney injury in crescentic GN.
J.Am. Soc. Nephrol. 23, 14861495 (2012).
71. Summers,S.A. etal. Th1 and Th17 cells induce
proliferative glomerulonephritis. J.Am. Soc. Nephrol.
20, 25182524 (2009).
72. Paust,H.J. etal. The IL-23/Th17 axis contributes
to renal injury in experimental glomerulonephritis.
J.Am. Soc. Nephrol. 20, 969979 (2009).
73. Turner,J.E. etal. CCR6 recruits regulatory Tcells
and Th17 cells to the kidney in glomerulonephritis.
J.Am. Soc. Nephrol. 21, 974985 (2010).
74. Steinmetz,O.M. etal. The Th17-defining
transcription factor RORt promotes
glomerulonephritis. J.Am. Soc. Nephrol. 22,
472483 (2011).
75. Riedel,J.H. etal. Immature renal dendritic cells
recruit regulatory CXCR6
+
invariant natural killer
T cells to attenuate crescentic GN. J.Am. Soc.
Nephrol. 23, 19872000 (2012).
This paper is one of the first descriptions of
regulatory iNKT cells having an anti-inflammatory
effect. It shows that immature renal DCs attract
these cells via the CXCL16CXCR6 axis
to attenuate crescentic glomerulonephritis.
76. Panzer,U. etal. Chemokine receptor CXCR3 mediates
T cell recruitment and tissue injury in nephrotoxic
nephritis in mice. J.Am. Soc. Nephrol. 18,
20712084 (2007).
77. Menke,J. etal. CXCL9, but not CXCL10, promotes
CXCR3-dependent immune-mediated kidney
disease. J.Am. Soc. Nephrol. 19, 11771189
(2008).
78. Chavele,K.M. etal. Mannose receptor interacts with
Fc receptors and is critical for the development of
crescentic glomerulonephritis in mice. J.Clin. Invest.
120, 14691478 (2010).
79. Wolf,D. etal. CD4
+
CD25
+
regulatory Tcells inhibit
experimental anti-glomerular basement membrane
glomerulonephritis in mice. J.Am. Soc. Nephrol. 16,
13601370 (2005).
This is the first study to describe a role for
T
Reg
cells in kidney disease.
80. Paust,H.J. etal. Regulatory Tcells control the
Th1 immune response in murine crescentic
glomerulonephritis. Kidney Int. 80, 154164
(2011).
81. Ooi,J.D. etal. Endogenous foxp3
+
T-regulatory cells
suppress anti-glomerular basement membrane
nephritis. Kidney Int. 79, 977986 (2011).
82. Hochheiser,K. etal. Kidney dendritic cells become
pathogenic during crescentic glomerulonephritis with
proteinuria. J.Am. Soc. Nephrol. 22, 306316
(2011).
83. Feng,L. etal. Prevention of crescentic
glomerulonephritis by immunoneutralization of the
fractalkine receptor CX
3
CR1 rapid communication.
Kidney Int. 56, 612620 (1999).
84. Goldwich,A. etal. Podocytes are nonhematopoietic
professional antigen-presenting cells. J.Am. Soc.
Nephrol. 24, 906916 (2013).
85. Li,S., Kurts,C., Kontgen,F., Holdsworth,S.R. &
Tipping,P.G. Major histocompatibility complex
classII expression by intrinsic renal cells is required
for crescentic glomerulonephritis. J.Exp. Med. 188,
597602 (1998).
86. Ryu,M. etal. Plasma leakage through glomerular
basement membrane ruptures triggers the
proliferation of parietal epithelial cells and crescent
formation in non-inflammatory glomerular injury.
J.Pathol. 228, 448494 (2012).
87. Neale,T.J., Tipping,P.G., Carson,S.D. &
Holdsworth,S.R. Participation of cell-mediated
immunity in deposition of fibrin in glomerulonephritis.
Lancet 2, 421424 (1988).
88. Zaferani,A. etal. Identification of tubular heparan
sulfate as a docking platform for the alternative
complement component properdin in proteinuric renal
disease. J.Biol. Chem. 286, 53595367 (2011).
89. Macconi,D. etal. Proteasomal processing of albumin
by renal dendritic cells generates antigenic peptides.
J.Am. Soc. Nephrol. 20, 123130 (2009).
90. Ooi,J.D., Holdsworth,S.R. & Kitching,A.R.
Advances in the pathogenesis of Goodpastures
disease: From epitopes to autoantibodies to effector
Tcells. J.Autoimmun. 31, 295300 (2008).
91. Reynolds,J. Strain differences and the genetic basis of
experimental autoimmune anti-glomerular basement
membrane glomerulonephritis. Int. J.Exp. Pathol. 92,
211217 (2011).
92. Pedchenko,V. etal. Molecular architecture of the
Goodpasture autoantigen in anti-GBM nephritis.
N.Engl. J.Med. 363, 343354 (2010).
This paper clarifies the crucial importance of
autoantigen conformation for autoantibody
binding in anti-GBM disease.
93. Olaru,F. etal. Proteolysis breaks tolerance toward
intact 345(IV) collagen, eliciting novel anti-
glomerular basement membrane autoantibodies
specific for 345NC1 hexamers. J.Immunol. 190,
14241432 (2013).
94. Phelps,R.G. & Rees,A.J. The HLA complex in
Goodpastures disease: A model for analyzing
susceptibility to autoimmunity. Kidney Int. 56,
16381653 (1999).
95. Cairns,L.S. etal. The fine specificity and cytokine
profile of T-helper cells responsive to the 3 chain
of typeIV collagen in Goodpastures disease.
J.Am. Soc. Nephrol. 14, 28012812 (2003).
96. Zou,J. etal. Healthy individuals have Goodpasture
autoantigen-reactive Tcells. J.Am. Soc. Nephrol.
19, 396404 (2008).
97. Ooi,J.D. etal. The HLA-DRB1*15: 01-restricted
Goodpastures T cell epitope induces GN. J.Am. Soc.
Nephrol. 24, 419431 (2013).
This study shows the crucial role of HLA-DR*15:01
in the autoimmune response to the 3(IV)NC1,
which is the target of autoimmunity in anti-GBM
disease.
98. Phelps,R.G., Jones,V.L., Coughlan,M., Turner,A.N.
& Rees,A.J. Presentation of the Goodpasture
autoantigen to CD4 Tcells Is influenced more by
processing constraints than by HLA classII peptide
binding preferences. J.Biol. Chem. 273,
1144011447 (1998).
99. Kerjaschki,D. & Farquhar,M.G. Immunocytochemical
localization of the Heymann nephritis antigen (GP330)
in glomerular epithelial cells of normal Lewis rats.
J.Exp. Med. 157, 667686 (1983).
100. Beck,L.H.Jr etal. M-type phospholipase A2 receptor
as target antigen in idiopathic membranous
nephropathy. N.Engl. J.Med. 361, 1121 (2009).
This seminal paper identifies PLA2R as the
enigmatic autoantigen in membranous
nephropathy.
101. Stahl,R., Hoxha,E. & Fechner,K. PLA2R
autoantibodies and recurrent membranous
nephropathy after transplantation. N.Engl. J.Med.
363, 496498 (2010).
102. Stanescu,H.C. etal. Risk HLA-DQA1 and PLA
2
R1
alleles in idiopathic membranous nephropathy.
N.Engl. J.Med. 364, 616626 (2011).
103. Suzuki,H. etal. IgA1-secreting cell lines from patients
with IgA nephropathy produce aberrantly glycosylated
IgA1. J.Clin. Invest. 118, 629639 (2008).
104. Suzuki,H. etal. Aberrantly glycosylated IgA1 in IgA
nephropathy patients is recognized by IgG antibodies
with restricted heterogeneity. J.Clin. Invest. 119,
16681677 (2009).
This is a seminal paper indicating that glycan-
specific autoantibodies are associated with IgA
nephropathy and may represent a disease-specific
marker and potential therapeutic target.
105. Berthoux,F. etal. Autoantibodies targeting galactose-
deficient IgA1 associate with progression of IgA
nephropathy. J.Am. Soc. Nephrol. 23, 15791587
(2012).
106. Goodnow,C.C. Multistep pathogenesis of
autoimmune disease. Cell 130, 2535 (2007).
107. Hakkim,A. etal. Impairment of neutrophil
extracellular trap degradation is associated with lupus
nephritis. Proc. Natl Acad. Sci. USA 107, 98139818
(2010).
108. Bosch,X. Systemic lupus erythematosus and the
neutrophil. N.Engl. J.Med. 365, 758760 (2011).
109. Marshak-Rothstein,A. & Rifkin,I.R. Immunologically
active autoantigens: the role of toll-like receptors in
the development of chronic inflammatory disease.
Annu. Rev. Immunol. 25, 419441 (2007).
110. Guiducci,C. etal. TLR recognition of self nucleic acids
hampers glucocorticoid activity in lupus. Nature 465,
937941 (2010).
This study shows that TLR7 and TLR9 signalling
drives lupus nephritis by rendering DCs resistant
to steroids.
111. Pawar,R.D. etal. Inhibition of Toll-like receptor-7
(TLR-7) or TLR-7 plus TLR-9 attenuates
glomerulonephritis and lung injury in experimental
lupus. J.Am. Soc. Nephrol. 18, 17211731 (2007).
112. Savarese,E. etal. U1 small nuclear ribonucleoprotein
immune complexes induce typeI interferon in
plasmacytoid dendritic cells through TLR7.
Blood 107, 32293234 (2006).
113. Garcia-Romo,G.S. etal. Netting neutrophils are
major inducers of typeI IFN production in pediatric
systemic lupus erythematosus. Sci. Transl. Med.
3, 73ra20 (2011).
114. Theofilopoulos,A.N., Baccala,R., Beutler,B. &
Kono,D.H. Type I interferons (/) in immunity and
autoimmunity. Annu. Rev. Immunol. 23, 307336
(2005).
115. Migliorini,A. & Anders,H.J. A novel pathogenetic
concept-antiviral immunity in lupus nephritis.
Nature Rev. Nephrology 8, 183189 (2012).
116. Yung,S., Cheung,K.F., Zhang,Q. & Chan,T.M.
Anti-dsDNA antibodies bind to mesangial annexin II in
lupus nephritis. J.Am. Soc. Nephrol. 21, 19121927
(2010).
117. Mortensen,E.S. & Rekvig,O.P. Nephritogenic
potential of anti-DNA antibodies against necrotic
nucleosomes. J.Am. Soc. Nephrol. 20, 696704
(2009).
118. Weening,J.J. etal. The classification of
glomerulonephritis in systemic lupus erythematosus
revisited. J.Am. Soc. Nephrol. 15, 241250 (2004).
119. Chen,M. & Kallenberg,C.G. ANCA-associated
vasculitides advances in pathogenesis and
treatment. Nature Rev. Rheumatol 6, 653664
(2010).
120. Kain,R. etal. Molecular mimicry in pauci-immune
focal necrotizing glomerulonephritis. Nature Med. 14,
10881096 (2008).
This paper describes molecular mimicry of
bacterial antigens underlying pauci-immune FNGN.
121. Kain,R. etal. High prevalence of autoantibodies to
hLAMP-2 in anti-neutrophil cytoplasmic antibody-
associated vasculitis. J.Am. Soc. Nephrol. 23,
556566 (2012).
122. Roth,A.J. etal. Anti-LAMP-2 antibodies are not
prevalent in patients with antineutrophil cytoplasmic
autoantibody glomerulonephritis. J.Am. Soc. Nephrol.
23, 545555 (2012).
123. Kessenbrock,K. etal. Netting neutrophils in
autoimmune small-vessel vasculitis. Nature Med. 15,
623625 (2009).
124. Gou,S.-J., Yuan,J., Chen,M., Yu,F. & Zhao,M.-H.
Circulating complement activation in patients with
anti-neutrophil cytoplasmic antibody-associated
vasculitis. Kidney Int. 83, 129137 (2013).
125. Schreiber,A. & Kettritz,R. The neutrophil in
antineutrophil cytoplasmic autoantibody-associated
vasculitis. J.Leukoc. Biol. http://dx.doi.org/10.1189/
jlb.1012525 (2013).
126. Abdulahad,W.H., Lamprecht,P. & Kallenberg,C.G.
T-helper cells as new players in ANCA-associated
vasculitides. Arthritis Res. Ther. 13, 236 (2011).
REVI EWS
752 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
127. Velden,J. etal. Renal IL-17 expression in human
ANCA-associated glomerulonephritis. Am. J.Physiol.
Renal Physiol. 302, F1663F1673 (2012).
128. McKinney,E.F. etal. A CD8
+
Tcell transcription
signature predicts prognosis in autoimmune disease.
Nature Med. 16, 586591 (2010).
129. Lyons,P.A. etal. Genetically distinct subsets within
ANCA-associated vasculitis. N.Engl. J.Med. 367,
214223 (2012).
130. Jennette,J.C., Falk,R.J., Hu,P. & Xiao,H.
Pathogenesis of antineutrophil cytoplasmic
autoantibody-associated small-vessel vasculitis.
Annu. Rev. Pathol. 8, 139160 (2013).
131. Heeringa,P. & Little,M.A. Invivo approaches to
investigate ANCA-associated vasculitis: lessons and
limitations. Arthritis Res. Ther. 13, 204 (2011).
132. Kuligowski,M.P. etal. Antimyeloperoxidase
antibodies rapidly induce 4-integrin-dependent
glomerular neutrophil adhesion. Blood 113,
64856494 (2009).
133. Little,M.A. etal. Anti-proteinase 3 anti-neutrophil
cytoplasm autoantibodies recapitulate systemic
vasculitis in mice with a humanized immune system.
PLoS ONE 7, e28626 (2012).
134. Ruth,A.J. etal. Anti-neutrophil cytoplasmic
antibodies and effector CD4
+
cells play
nonredundant roles in anti-myeloperoxidase
crescentic glomerulonephritis. J.Am. Soc. Nephrol.
17, 19401949 (2006).
135. Tan,D.S. etal. Thymic deletion and regulatory T cells
prevent antimyeloperoxidase GN. J.Am. Soc. Nephrol.
24, 573585 (2013).
136. Ooi,J.D. etal. The immunodominant
myeloperoxidase T-cell epitope induces local
cell-mediated injury in antimyeloperoxidase
glomerulonephritis. Proc. Natl Acad. Sci. USA 109,
E2615E2624 (2012).
137. Gan,P.Y. etal. Th17 cells promote autoimmune
anti-myeloperoxidase glomerulonephritis. J.Am. Soc.
Nephrol. 21, 925931 (2010).
138. Tadema,H. etal. Bacterial DNA motifs trigger ANCA
production in ANCA-associated vasculitis in remission.
Rheumatology (Oxford) 50, 689696 (2011).
139. Tadema,H., Kallenberg,C.G., Stegeman,C.A. &
Heeringa,P. Reactivity against complementary
proteinase-3 is not increased in patients with
PR3-ANCA-associated vasculitis. PLoS ONE 6,
e17972 (2011).
140. Vaziri,N.D. CKD impairs barrier function and alters
microbial flora of the intestine: a major link to
inflammation and uremic toxicity. Curr. Opin. Nephrol.
Hypertens. 21, 587592 (2012).
141. Lech,M., Rommele,C. & Anders,H.J. Pentraxins in
nephrology: C-reactive protein, serum amyloid P and
pentraxin-3. Nephrol. Dial. Transplant. 28, 803811
(2013).
142. Carrero,J.J. & Stenvinkel,P. Inflammation in
end-stage renal disease what have we learned in
10years? Semin. Dial. 23, 498509 (2010).
143. Cohen,G. & Horl,W.H. Immune dysfunction in
uremia an update. Toxins 4, 962990 (2012).
144. Weichhart,T. etal. Serum amyloid A in uremic HDL
promotes inflammation. J.Am. Soc. Nephrol. 23,
934947 (2012).
145. Anders,H.J., Andersen,K. & Stecher,B.
The intestinal microbiota, a leaky gut, and
abnormal immunity in kidney disease. Kidney Int. 83,
10101016 (2013).
146. Meyer,T.W. & Hostetter,T.H. Uremic solutes from
colon microbes. Kidney Int. 81, 949954 (2012).
147. McIntyre,C.W. etal. Circulating endotoxemia:
a novel factor in systemic inflammation and
cardiovascular disease in chronic kidney disease.
Clin. J.Am. Soc. Nephrol. 6, 133141 (2011).
148. Hotchkiss,R.S., Coopersmith,C.M., McDunn,J.E. &
Ferguson,T.A. The sepsis seesaw: tilting toward
immunosuppression. Nature Med. 15, 496497
(2009).
149. Stearns-Kurosawa,D.J., Osuchowski,M.F.,
Valentine,C., Kurosawa,S. & Remick,D.G.
The pathogenesis of sepsis. Annu. Rev. Pathol.
6, 1948 (2011).
150. Oppermann,M. etal. Elevated plasma levels of
the immunosuppressive complement fragment
Ba in renal failure. Kidney Int. 40, 939947
(1991).
151. Pelajo,C.F., Lopez-Benitez,J.M. & Miller,L.C.
VitaminD and autoimmune rheumatologic disorders.
Autoimmun Rev. 9, 507510 (2010).
152. Vaziri,N.D. Oxidative stress in uremia: nature,
mechanisms, and potential consequences.
Semin. Nephrol. 24, 469473 (2004).
153. Herrada,A.A. etal. Aldosterone promotes
autoimmune damage by enhancing Th17-mediated
immunity. J.Immunol. 184, 191202 (2010).
154. Kleinewietfeld,M. etal. Sodium chloride drives
autoimmune disease by the induction of pathogenic
T
H
17 cells. Nature 496, 518522 (2013).
155. Wu,C. etal. Induction of pathogenic T
H
17 cells by
inducible salt-sensing kinase SGK1. Nature 496,
513517 (2013).
156. Madhur,M.S. etal. Interleukin 17 promotes
angiotensin II-induced hypertension and vascular
dysfunction. Hypertension 55, 500507 (2010).
157. Machnik,A. etal. Macrophages regulate salt-
dependent volume and blood pressure by a vascular
endothelial growth factor-C-dependent buffering
mechanism. Nature Med. 15, 545552 (2009).
158. Harrison,D.G., Marvar,P.J. & Titze,J.M. Vascular
inflammatory cells in hypertension. Front. Physiol.
3, 128 (2012).
Acknowledgements
We apologize to all colleagues whose work could not be cited
or discussed in greater detail due to space restrictions. The
authors are supported by the German Research foundation
(DFG Klinische Forschergruppe 228, SFB704 and TR57,
Gradui ertenkol l eg 1202 and Excel l ence Cl uster
ImmunoSensation) and the EU Consortia INTRICATE and
REDDSTAR.
Competing interests statement
The authors declare no competing financial interests.
SUPPLEMENTARY INFORMATION
See online article: S1 (table)
ALL LINKS ARE ACTIVE IN THE ONLINE PDF
REVI EWS
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 753
2013 Macmillan Publishers Limited. All rights reserved
Non-invasive cell tracking is an emerg-
ing approach for imaging cells in their
native environment. The phenotype of
an immune cell is partly defined by the
patterns and expression levels of cell
surface molecules and secreted factors.
Thesemolecules are commonly assayed
invitro using techniques such as flow
cytometry and immunohistochemistry.
However, determining which cell sur-
face molecules and secreted factors are
present, and at what levels, under vari-
ous culture conditions or disease states,
is only one piece of the puzzle. A more
challenging question to answer relates
to the biological roles of these molecules
invivo. Non-invasive imaging of the traf-
ficking patterns of phenotypically defined
populations of immune cells can have an
important role in answering this question.
Moreover, immune cells are increasingly
being used as next-generation therapeu-
tics to treat chronic conditions such as
autoimmune disease and cancer. A com-
mon requirement for the development of
nearly all cell therapies is a non-invasive
means to visualize the biodistribution of
cells following injection. Imaging of cell
trafficking can provide crucial information
regarding the persistence and the motility
of transferred cells, as well as the optimal
routes of delivery and the therapeutic
doses for individuals. On the regulatory
side, emerging therapies using immune
cells can be slow to gain regulatory
approval partly because clinical research-
ers are challenged to verify where the cells
traffic to immediately after inoculation
and where they migrate to over time. Cell
tracking in vivo can potentially provide this
information and might help to overcome
regulatory barriers.
Inflammation is a hallmark of many
of the major diseases on which biomedi-
cal research commonly focuses, such as
autoimmune diseases, neurological dis-
orders, transplant rejection and cancer.
Measuring the effectiveness of treatment
in terms of the inflammatory burden
can be challenging. Conventional meth-
ods often include invasive biopsies and
histology or imaging methods that are
nonspecific for inflammation or that lack
quantitative measures. There is a need
for improved inflammation-specific
imaging diagnostics, as well as surro gate
biomarkers of inflammation, that
could enable researchers to determine
the efficacy of an anti-inflammatory
therapy safely, quickly, quantitatively and
in a longitudinal manner. There is also a
need for pharmacological safety profiling
to detect off-target inflammatory side
effects in preclinical and clinical drug
trials. Vital imaging can help to steer the
decision-making process at the preclinical
and clinical trial stages; it can facilitate
smaller, less costly trials by enabling the
enrolment of fewer patients. Imaging can
potentially yield a rich data set from each
patient in terms of inflammation severity
and its time course in a three-dimensional
anatomical context.
Given the clear need for invivo cell
tracking, much progress has been made in
this area in recent years. Imaging methods
using radionuclides have traditionally
been used for the non-invasive imaging
of leukocytes. However, technologies
using magnetic resonance imaging (MRI)
(BOX1) are now emerging, and the field
is experiencing a rapid expansion in the
development of new imaging probes and
genetically encoded reporters that enable
the visualization of specific cell popula-
tions and molecular events invivo in both
animals and humans. These new capa-
bilities have been made possible by next-
generation, non-toxic cell labelling probes
and by MRI methods. MRI has the advan-
tage that it does not use ionizing radiation
and can safely image deep tissues at high
resolution.
In this Innovation article, we describe
emerging methods and applications of
MRI-based tracking of immune cells.
These methods use exogenous cell labels
that are comprised of iron oxide nano-
particles or perfluorocarbon (PFC) nano-
emulsions (FIG.1), or genetically-encoded
MRI reporters (FIG.2). The field of MRI-
based cell tracking is rapidly evolving, and
in the future we believe it will be possible
to visualize not only the location and the
number of cells invivo but also more sophis-
ticated biological processes, such as immune
cell viability and activation status. These
advances indicate a future in which MRI
platforms will be used to elucidate a wide
range of human cellular immunological
processes invivo.
I NNOVATI ON
Tracking immune cells invivo using
magnetic resonance imaging
Eric T.Ahrens and Jeff W.M.Bulte
Abstract | The increasing complexity of invivo imaging technologies, coupled
with the development of cell therapies, has fuelled a revolution in immune cell
tracking invivo. Powerful magnetic resonance imaging (MRI) methods are now
being developed that use iron oxide- and
19
F-based probes. These MRI
technologies can be used for image-guided immune cell delivery and for the
visualization of immune cell homing and engraftment, inflammation, cell
physiology and gene expression. MRI-based cell tracking is now also being
applied to evaluate therapeutics that modulate endogenous immune cell
recruitment and to monitor emerging cellular immunotherapies. These recent
uses show that MRI has the potential to be developed in many applications to
follow the fate of immune cells invivo.
PERSPECTIVES
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 755
2013 Macmillan Publishers Limited. All rights reserved
Available invivo cell tracking techniques
Much of our current knowledge of immune
cell trafficking has come from static end
points that have been obtained using
conventional light microscopy and flow
cytometry. In recent years, a wide range
of imaging modalities have become avail-
able that can image immune cells in their
native environment; however, each of these
techniques has inherent advantages and
limitations (TABLE1). About a decade ago, the
field saw a revolution with the introduction
of intravital, two-photon microscopy, which
enables precise determination of the move-
ments of immune cells within lymph nodes
and tumours
1
. However, because of limited
tissue opacity, it is not possible to non-
invasively look into tissues at a depth beyond
~1 mm using two-photon microscopy.
Deeper tissue penetration and whole-body
non-invasive tomography of small animals
can be achieved by capturing the photons
emitted by fluorescent or bioluminescent
probes. Such bioluminescence imaging has
become one of the most widely used tech-
niques for non-invasive tracking of immune
cells and inflammation in small animals,
but it cannot be used in larger animals or
clinically because of the physical limits of
light penetration in deep tissues.
Box 1 | Magnetic resonance imaging
The signal used for magnetic resonance imaging (MRI) is derived from endogenous mobile water protons (
1
H) or fluorinated molecules (such as
19
F) that
are present or introduced in the subject. When the subject is placed in a large static magnetic field, the magnetic moment associated with
1
H or
19
F
tends to align along the direction of the magnetic field. The
1
H or
19
F nuclei are perturbed from this equilibrium by pulsed radio-frequency radiation.
Following the removal of the radio-frequency radiation, the nuclei recover to equilibrium and induce a transient voltage in a receiver antenna; this
transient voltage constitutes the nuclear magnetic resonance (NMR) signal. The physical properties of a specific tissue, such as the density of nuclei, the
nuclear spinlattice relaxation time (T1) and the spinspin relaxation time (T2), often determine the amount of signal that is available. The alignment of
the nuclei along the magnetic field direction is not instantaneous, but occurs gradually over a period that is parameterized by the time constant T1. T2 is
the characteristic time constant for which nuclei remain in phase with each other, and its value is reflected in the duration of the transient NMR signal.
MRI-based cell tracking involves detecting cells that exhibit a differential signal. The MRI signal can be controlled in four ways, as discussed below.
Positive contrast agents containing paramagnetic metals
Paramagnetic contrast agents primarily affect T1. Most often, T1 contrast agents contain Gd
3+
that is chelated to a low-molecular-mass molecule to
limit toxicity. The surrounding water protons rapidly exchange with the complex, which results in a reduction of T1 and an increase in signal intensity
(positive contrast) of Gd
3+
-labelled cells on T1-weighted magnetic resonance images.
Negative contrast agents containing superparamagnetic iron oxides
Superparamagnetic iron oxide (SPIO) contrast agents primarily affect T2 by virtue of their iron oxide crystals, which have a strong magnetic moment.
These agents commonly consist of small crystalline particles of ferrous and ferric oxides (FeOFe
2
O
3
) that are coated with dextran. These particulates
strongly perturb the magnetic field that they are in proximity to. Surrounding water molecules subsequently experience a highly inhomogeneous
magnetic field, which results in a local signal loss (negative contrast) of SPIO-labelled cells on T2-weighted magnetic resonance images.
Molecular probes that induce chemical
exchange saturation transfer
Certain protons that are loosely bound at
specific chemical sites, such as amide protons,
have a slightly different resonance frequency
than water protons, from which the MRI signal is
always collected. When these labile non-water
protons are irradiated with a saturation pulse
at their specific off-resonance frequency, the
protons lose the ability to create an MRI signal.
The saturated labile non-water protons
exchange position with the water protons,
which leads to a loss of MRI signal of the
labelled cells on chemical exchange saturation
transfer (CEST) images. The CEST irradiation
can be turned on and off (see the figure).
Molecular probes containing
19
F
The physical principles behind detection and
image formation are the same for both
1
H
and
19
F MRI. Unlike metal ion-based magnetic
resonance contrast agents, which are
detected through their indirect effects on
the surrounding water protons, the
19
F probe
functions as a tracer agent in that
19
F MRI
directly detects the
19
F nuclei that are
associated with the labelled cells, with
no background. The
19
F signal is directly
proportional to the number of fluorine atoms
and the number of labelled cells that are
present (see the figure). A composite
19
F and
1
H
image is generally constructed, which shows
the regions containing labelled cells within
their anatomical context.
Nature Reviews | Immunology
Unlabelled
cell
Gd
3+
-
labelled cell
Positive
contrast
CEST on CEST o
Hot spot
Negative
contrast
SPIO-
labelled cell
19
F-labelled
cell
CEST-labelled cell
MRI signal MRI detection
Water protons
H H
O
Exchange of water
protons with the
Gd
3+
-containing
complex
H H
O
H H
O
H N
C
C
F
F F F F
F F F F
F F
F F F
H H
O
F
Water protons sense
perturbed magnetic
eld around SPIO
particles
Exchange of water protons
with saturated labile
protons from a reporter
Direct detection of
19
F nuclei inside
labelled cells
SPIO
H
N
N
N
H
O
O O
O
O
O
O
O
O
O
O
Gd
3+
PERSPECTI VES
756 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Nature Reviews | Immunology
Ex vivo
labelling
In situ
labelling
Cell isolation
and culture
Cell collection
and washing
Co-incubation with
labelling agent
Injection of
labelled cells
Intravenous
injection of
labelling agent
Isolation of
individual organs
Uptake by
phagocytic cells
of the RES
19
F Inammation
index
19
F MRI and/or
1
H MRI scan
Composite
19
F and
1
H image
1
H image
NMR spectroscopy
SPIO
nanoparticle
or
PFC
emulsion
SPIO
nanoparticle
or
PFC
emulsion
Macrophage
L
i
v
e
r
S
p
l
e
e
n
L
y
m
p
h
n
o
d
e
PFC
hot spot
SPIO
negative
contrast
Radionuclide labelling of cells is the old-
est technique for tracking immune cells in
larger animals and humans
2
; for example,
111
In-oxine labelling of autologous white
blood cells is used as a diagnostic agent for
the imaging of occult inflammation and
infection in humans and is currently the
only cell tracking technique approved by the
United States Food and Drug Administration.
Nuclear imaging is particularly useful for
whole-body distribution studies, as there is
no backgroundsignal.
Reporter genes became available in
the field of nuclear medicine in the early
1990s following the development of the
thymidine kinase enzyme that is derived
from herpes simplex virus (HSV)
3
. After
administration of positron-emitting sub-
strates, such as
18
FIAU (1-(2-deoxy-2-[
18
F]-
-D-arabinofuranosyl)-5-iodouracil)
or
18
FHBG (9-[4-[
18
F]fluoro-3-
(hydroxymethyl)butyl]guanine), HSV
thymidine kinase phosphorylates the
radioactive probe and is responsible for its
prolonged retention in transfected cells. In a
few studies, this approach has been used to
monitor Tcell trafficking
4,5
, including the
use of positron emission tomography (PET)
to visualize cytotoxic Tcell homing to the
tumour in a patient with glioma
6
.
However, there are cytotoxicity and
patient safety concerns that limit the use
of radionuclide-based methods for cell
tracking. Moreover, radionuclide-based
techniques are unable to provide anatomical
imaging by themselves, and these scans must
be combined, for example, with computed
tomography (CT) or MRI, which adds to
the methodological complexity and the cost.
In addition, for the diagnostic imaging of
inflammation using
111
In-oxine labelling
of white blood cells, there is a considerable
cost associated with the invasive leukopho-
resis procedure and exvivo labelling of the
patients cells that is necessary before reinfu-
sion. For applications in which the trans-
ferred cells, having been labelled exvivo, are
expected to persist in the body for extended
periods of time (for example, in Tcell and
stem cell therapies), the effect of radio-
cytotoxicity on cell viability is a concern.
Figure 1 | Schematic showing exvivo and insitu labelling of cells
with magnetic resonance nanoparticle contrast agents. Cell label-
ling can be achieved exvivo in pre-selected and cultured cells by adding
the labelling reagent superparamagnetic iron oxide (SPIO) nanopar-
ticles or perfluorocarbon (PFC) emulsion directly to the media fol-
lowed by co-incubation. Often the labelling reagent is complexed with a
cationic transfection agent before it is added to the cell culture to label
non-phagocytic cells. Self-delivering formulation s of PFC emulsions
have also been devised that do not require a transfection agent.
Alternatively, electroporation can be used to label cells in culture. After
collection and washing, labelled cells are then administered to the sub-
ject. In addition, the labelling agent can be intra venously injected; in this
in situ labelling approach, the agent is intrinsically taken up by
phagocytic cells of the reticuloendothelial (RES) system, particularly by
monocytes and macrophages, which then accumulate at sites of inflam-
mation. For both exvivo and insitu labelling approaches an
1
H magnetic
resonance imaging (MRI) scan is then carried out to visualize the anat-
omy. The resulting images have a decreased signal in regions containing
SPIO-labelled cells (known as negative contrast). In the case of PFC
probes, a
19
F MRI scan is also acquired in the same imaging session. A
composite, pseudo-coloured
19
F and
1
H image is then constructed, in
which labelled cells appear in the
19
F colour channel (known as hot spot
contrast). By quantifying the
19
F MRI signal in individual organs using
using nuclear magnetic resonance (NMR) spectroscopy an inflammation
index can be calculated, which is a direct correlation between the
19
F
signal and the number of labelled macrophages.
PERSPECTI VES
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 757
2013 Macmillan Publishers Limited. All rights reserved
Ligand
Nature Reviews | Immunology
Fe
2+
Fe
2+
Fe
2+
Iron-binding
protein
SPIO
Vector carrying
reporter gene
Reporter gene
Phosphorylation
Translation
Transcription
Imino proton on
thymidine analogue
Ligand-conjugated
SPIO probe
Genetically engineered cell
P
P
Cell uptake
Exogenous
substrate
Exogenous
substrate
Endogenous
substrate (Fe
2+
)
No substrate
required
HSV thymidine
kinase
Overexpressed
receptor
LRP
R
R O
O H
N
N
H
Inaddition, the finite half-life of the radio-
isotopes that are used precludes extended
longitudinal studies. These limitations of
radionuclide-based techniques have helped
to stimulate the development of alternative
imaging methods, particularly MRI-based
methods (TABLE2).
Metal ion-based MRI contrast agents
Positive and negative contrast agents.
Intravascular MRI contrast agents that
incorporate metal ions (such as Gd
3+
) are
commonly used as positive agents to enhance
the signal of lesions in clinical scans. The
paramagnetic metal ions in these agents
accelerate the relaxation of nearby water
protons (
1
H), and this signal enhancement
can be detected by
1
H MRI (BOX1) as a local-
ized increase of image contrast. MRI contrast
agents can also be used to label immune cells
for cell tracking studies (TABLE2). After cell-
ular internalization by simple co-incubation,
electroporation or conjugation to cell-
labelling moieties, paramagnetic metal ions
such as Mn
2+
and Gd
3+
can provide cells
with positive contrast
7
. Positive contrast is
generally preferred for image interpretation
because the signal of the underlying anatomi-
cal structure is enhanced; the alternative,
negative contrast (see below), decreases the
signal, which makes the tissue invisible at
times. However, labelling with paramagnetic
metal ions only provides modest sensitivity
in terms of the number of cells that can be
detected (TABLE2). Particulate formulations
of metal ions, such as superparamagnetic
iron oxide (SPIO) nanoparticles, create a very
strong localized magnetic field disturbance
in the MRI scanner because of the synergistic
magnetic alignment of the individual iron
ions. This localized magnetic perturbation
affects nearby water protons, which causes a
very strong negative contrast, in which the
areas containing SPIO particles become dark
on the images
8,9
. Immune cells can be labelled
with SPIO nanoparticles using two main
approaches (FIG.1): sorted or mixed cell popu-
lations can be labelled exvivo by incubation
with SPIO nanoparticles in culture media; or
non-selected phagocytic cell populations (for
example, macrophages) can be labelled in
situ following intravenous injection of SPIO
nanoparticles. In many cases, SPIO particles
have been conjugated to fluorochromes,
which creates dual-mode agents that enable
the validation of invivo MRI cell tracking
by post-mortem light microscopy
10,11
.
Labelling immune cells exvivo. Inspired by
clinical studies using radionuclide-based
tracking of tumour-infiltrating lymphocytes,
one of the first experimental applications of
exvivo SPIO labelling of cells was to label
lymphocytes (FIG.3a), first by labelling the
cell membrane
12
, followed by labelling intra-
cellular endosomes
13,14
. Membrane labelling
can be achieved by targeting antibodies or
peptides to specific cell surface epitopes,
whereas intracellular labelling requires the
use of cell-penetrating peptides or trans-
fection agents if cells have a low level of
phagocytosis or macropinocytosis activity.
Intracellular labelling of lymphocytes has
proven to be challenging, owing to the fairly
small cytoplasmic volume of non-activated
cells and the low phagocytic activity. The
most widely used method for SPIO labelling
of immune cells is to complex the negatively
charged SPIO particles with various cationic
transfection agents, such as poly-l-lysine
or protamine sulphate, followed by
co-incubation with cells in culture
1517
.
Receptor-mediated endocytosis using spe-
cific SPIOantibody conjugates, such as
conjugates that are specific for CD11c, has
also been reported for the exvivo labelling
of both immature and mature dendritic cells
(DCs)
18
. Electroporation of the imaging
agent into immune cells is also feasible
19
.
MRI-based immune cell tracking using
SPIO particles has been applied to many
types of preclinical study, ranging from the
tumour homing of cytotoxic Tcells
20
and
natural killer cells
21
to the organ-specific
homing of autoimmune Tcells
22,23
(FIG.3b)
and to studies investigating the migra-
tory patterns of DCs used in cancer vac-
cines
2426
. Iron oxide particles with a large,
micrometre-sized diameter have been used
to extracellularly label lymphocytes with
greater sensitivity than SPIO nanoparticles
27
.
As these iron oxide particles have a tenfold
larger radius than SPIO nanoparticles, the
labelled cells become much more magnetic,
which enables the detection of single cells.
Figure 2 | The development of MRI reporter genes. Certain nucleic acid-based reporters encode
cell surface receptors that bind to specific ligands that are conjugated to superparamagnetic iron
oxide (SPIO) to render them magnetic resonance imaging (MRI)-detectable. They require delivery of
the ligandSPIO complex as an exogenous substrate. An example of this approach is an engineered
transferrin receptor that binds transferrin-ligated SPIO. Another exogenous substrate-based approach
involves a reporter gene encoding the thymidine kinase enzyme derived from herpes simplex virus
(HSV). This enzyme phosphorylates thymidine analogues, which causes them to remain trapped in
transduced cells. When thymidine analogues are used that are rich in specific (imino) protons,
chemical exchange saturation transfer (CEST) MRI can be used to track the transduced cells.
Other approaches use endogenous substrates, whereby the encoded reporter binds iron (Fe
2+
) that
is naturally present in the body as the contrasting metal ion. Finally, a CEST reporter (such as lysine-
rich protein (LRP)) can be used that does not require a substrate, as the protein itself contains
multiple labile amide protons that can be saturated and detected.
PERSPECTI VES
758 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
However, the clinical translation potential of
these fairly large particles remains uncertain.
When dedicated gradient insert coils are
used to enhance the performance of the
MRI scanner, SPIO-labelled macrophages
can also be imaged at the single-cell level
28
.
Labelling macrophages invivo. Labelling of
cells in vivo following systemic injection of
SPIO particles (FIG. 1) is generally achieved by
nanoparticles being taken up by the phago-
cytic cells of the reticuloendothelial system
(RES), including circulating blood mono-
cytes and tissue macrophages (and in much
smaller numbers, neutrophils and DCs),
which are frequently found at inflamma-
tory sites. This approach has been broadly
applied to image inflammatory events in
both preclinical and clinical settings
2935
.
In a clinical context, this approach has been
most widely used to image lymph nodes for
tumour staging
36
and to detect atheroscle-
rotic plaques
37
. Immune cell labelling in situ
can also be achieved by the direct injection
of SPIO agents into a tissue. In a mouse
study, SPIO labelling of DCs was accom-
plished following vaccination with irradiated
labelled tumour cells from which antigens
and SPIO were co-captured by the DCs.
Labelled DCs could be visualized homing to
the sentinel lymph node several days after
this so-called magnetovaccination
26
(FIG.3c).
Concerns and limitations. It is important
that any cell labelling protocol that is used
for imaging does not substantially alter the
immunological properties of the cells, and
that it does not cause marked cytotoxicity
or changes in the phenotype and the func-
tion of the cells. Changes in the phenotype
of immune cells after labelling could reduce
the efficacy of immunotherapeutic strate-
gies. Overall, very few studies have noted an
adverse effect of SPIO labelling on cell func-
tion or phenotype. In one study
38
, following
SPIO labelling in culture, macrophages
were found to have altered cytokine produc-
tion, which was shifted towards an anti-
inflammatory, less responsive phenotype.
In other studies, the cytokine profile, surface
expression of phenotypic markers, migratory
capacity and ability to present antigens did
not differ in SPIO-labelled DCs compared
with in unlabelled DCs
26,39,40
. SPIO particles
are biodegradable and, depending on the cell
type, are broken down quickly; for example,
when SPIO particles are taken up by Kpffer
cells in the liver, the iron is metabolized
and can be found in haemoglobin as soon
as one week after administration
41
. Thus,
SPIO-based cell labelling is best suited to
short-term imaging studies; for example,
MRI-guided injection, in which the infusion
of cells and their initial engraftment can be
imaged in real-time
10
, is viewed as a key
early clinical use of SPIO-based cell tracking.
There are other limitations to SPIO-
based cell tracking that are also common to
essentially all nanoparticle-based and PFC
nanoemulsion-based (see below) imaging
reagents: in mitotic cells, cell division and
subsequent dilution of the intracellular label
can potentially limit long-term cell visuali-
zation; cell death can lead to dispersion of
the reagent and loss of MRI detectability;
the imaging reagent could potentially be
transferred to resident phagocytes (such as
macrophages); and if a large number of these
labelled phagocytes remain in a region of
interest, false positive contrast couldresult.
PFC cell labels and
19
F MRI detection
In an effort to design next-generation cell
tracking probes, there has recently been
considerable interest in the use of PFC nano-
emulsions. PFC emulsions can be used to
track cells using exvivo and insitu cell label-
ling approaches that are analogous to those
used for SPIO nanoparticles (FIG.1). PFC-
based cell tracking enables high specificity
cell detection and the quantification of cells.
PFC-labelled cells are detected using
19
F MRI
(BOX1), whereas metal ion-based contrast
agents (such as SPIO particles) are detected
using
1
H MRI. The PFC functions as a tracer
agent rather than as a contrast agent, as
19
F
MRI directly detects the
19
F nuclei that are
associated with the labelled cells. Importantly,
because of the extremely low concentration of
naturally occurring
19
F in the body,
19
F MRI
has no background signal from the hosts
tissues and therefore only labelled cells are
observed. Thus, false positive cell detection is
unlikely, which overcomes one of the major
limitations of metal ion-based cell labelling
approaches. Moreover, quantification of the
19
F MRI signal is directly related to the appar-
ent number of cells in the regions of interest
or, alternatively, to the inflammation severity.
PFCs are among the most biologically
inert organic molecules that have ever been
produced
42
. They are water insoluble and
immiscible in cell membranes and, for cel-
lular use, they must be formulated into
colloidal suspensions, such as nanoemul-
sions, ideally with a small droplet diameter
(<200 nm)
43
. There are no known enzymes
that metabolize PFCs invivo and they
are not degraded at typical lysosomal pH
values
42
; thus, cell labelling with PFCs can
potentially be long lasting.
Table 1 | Comparison of non-invasive imaging modalities for invivo cell tracking
Imaging technique Resolution* Tissue
penetration
depth*
Sensitivity of cell
detection*
Possibility of
longitudinal
studies*
Used generally
in the clinic?
Used in clinical
cell tracking?
MRI +++ +++ +++ ++ Yes Yes
SPECT + +++ ++ + Yes Yes
PET + +++ ++ +++ Yes Yes
CT or X-ray +++ +++ + + Yes No
Ultrasound imaging ++ ++ ++ + Yes No
BLI + + ++ +++ No No
Fluorescence imaging
or NIR
+ + ++ ++ No No
2PLSM +++ + +++ ++ No No
MPI
84
+ +++ ++ ++ No No
2PLSM, two-photon laser scanning microscopy; BLI, bioluminescence imaging; CT, computed tomography; MPI, magnetic particle imaging; MRI, magnetic
resonance imaging; NIR, near-infrared imaging; PET, positron emission tomography; SPECT, single-photon emission computed tomography. *Strengths and
weaknesses are given using a relative scale in which + = poor, ++ = moderate and +++ = excellent.
PERSPECTI VES
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 759
2013 Macmillan Publishers Limited. All rights reserved
Labelling immune cells exvivo. The idea
of using PFCs as tracer agents for
19
F MRI
followed shortly after the discovery of
1
H
MRI
44
, but their use in cell tracking is a
relatively recent development
45,46
. Different
nanoemulsion formulations have been
designed for exvivo (FIG.3d) and insitu cell
labelling
43
. For exvivo labelling, an impor-
tant innovation has been the formulation
of nanoemulsions that intracellularly label
cells in culture without the use of transfec-
tion methods
47
through the inclusion
of charged moieties on the nano emulsion
droplet surface thereby improving
overall cell viability and the ease of use
of PFCs. Numerous invitro studies have
investigated the effects of PFC labelling on
cellular pheno type and function in primary
immune cells, for example, using mouse
DCs
45
and Tcells
48
. The most detailed
study so far involved PFC-labelled pri-
mary human DCs
49
; cells were assayed for
viability, maturation phenotype, cytokine
production, Tcell stimulatory capacity and
chemotaxis, and no differences in these
parameters were observed between labelled
and unlabelled cells invitro.
Tracking immune cells invivo. Tracking
cells in vivo using PFCs was first used to vis-
ualize DC migration in mice
45
. Several Tcell
studies have used PFC-based cell tracking
to examine early pancreas inflammation in
non-obese diabetic (NOD) mice
50
, which
are an acute inflammation model
48
, and the
biodistribution of mucin 1 (MUC1)-specific
Tcells in inflammatory bowel disease
(IBD)
51
. A surprising finding of the IBD
study was the localization of MUC1-specific
Tcells in the pancreatic duct, which indi-
cates that pancreatitis, which is often diag-
nosed in patients with IBD, might be a true
extra-intestinal manifestation of the disease.
In addition, primary human DCs that are
relevant to immunotherapeutic clinical
trials have been labelled using PFCs and
longitudinally tracked invivo into draining
lymph nodes in NOD severe combined
immunodeficient (SCID) mice
49
.
Imaging of macrophage recruitment to
inflammatory sites in vivo could help to
elucidate the host inflammatory response,
could provide considerable diagnostic
value for a wide range of diseases and
could be used as a surrogate marker and
to monitor therapeutic interventions. By
using PFC nanoemulsions that have been
formulated for intravenous injection and
that have a long blood half-life, nanoemul-
sion droplets enter into the RES and are
intrinsically taken up insitu by monocytes
and macrophages and, to a lesser extent,
by neutrophils and DCs. As these insitu-
labelled cells participate in inflammatory
events in the body, the result is
19
F accumu-
lation at inflammatory sites. Importantly,
the invivo
19
F MRI signal can be quantified
in inflammatory sites, and this signal is lin-
early proportional to the macrophage bur-
den
52,53
.
19
F labelling in situ has been widely
used to visualize inflammation in a large
number of preclinical models of human
disease, such as IBD
53
(FIG.3e), bacterial
infection
54
, organ transplant rejection
55,56
,
experimental allergic encephalomyelitis
52,57
,
peripheral nerve inflammation
58
, pulmo-
nary inflammation
59
and cardiac and
cerebral ischaemia
60
.
Extending the use of PFC emulsions.
Similar to dual-mode SPIO agents, PFC
nanoemulsions that can be detected by both
19
F MRI and fluorescence have also been
devised
45,47
; these reagents contain a bright
fluorescent dye that is directly conjugated
to the PFC molecule prior to nanoemulsion
formulation, which thereby ensures that the
originating
19
F MRI and fluorescent signals
are coincident in labelled cells. The fluo-
rescent moiety enables the identification of
the fate and the phenotype of labelled cells
using flow cytometry, fluorescence micro-
scopy or invivo optical imaging; for exam-
ple, dual-mode fluorescent PFC emulsions
have been used to label primary murine
CD4
+
Tcells that were then adoptively
transferred into wild-type mice and imaged
in multiple lymph nodes using MRI, fol-
lowed by analysis using optical methods
47
.
Near-infrared dyes have also been incor-
porated into PFC emulsions, which enables
invivo imaging at moderate tissue depths
61
.
Moreover, fluorescent PFC emulsions have
considerable potential as an optical-only cell
labelling probe because of their brightness,
low toxicity and long retention time incells.
Conventional
19
F nuclear magnetic
resonance (NMR) spectroscopy can also be
used to assay the biodistribution of PFC-
labelled immune cells with high sensitivity
in fixed, intact tissue panels.
19
F NMR can
measure the total cell count (or cell den-
sity) of labelled cells in tissue samples; for
example, NMR cytometry has been used
in Tcell biodistribution studies in rodent
IBD and diabetes models
50,51,62
. As NMR
is non-destructive, the same tissues can
then be processed for histology to further
refine the tissue analysis.
19
F-capable NMR
spectrometers are ubiquitous in research
laboratories and are commonly used for
molecular structure determination, there-
fore the implementation of NMR cytometry
is fairly straightforward. NMR instru-
mentation can also be used to rapidly and
quantitatively assay inflammation in intact,
excised tissue samples, which yields an
inflammatory index
52
that is proportional
to the macrophageburden.
Table 2 | Overview of available invivo MRI-based cell tracking techniques
Probe MRI technique Type of contrast Minimum number
of detectable cells
Quantification
of cell number?
Clinical trial approval?
Gd
3+
or Mn
2+

labelling
T1-weighted
1
H MRI* Positive ~10
3
No No
SPIO labelling T2-weighted
1
H MRI

Negative 1 No Yes (in the Netherlands, China,


Switzerland, Czech Republic,
Israel, Poland and United Kingdom)
PFC labelling
19
F MRI Hot spot (also
known as tracer)
10
3
10
5
Yes Yes (in USA)
Metal-binding
reporter genes
T2-weighted
1
H MRI

Negative 10
4
No No
CEST agents and
reporter genes
1
H CEST MRI Differential (can be
colour encoded)
10
4
No No
CEST, chemical exchange saturation transfer; MRI, magnetic resonance imaging; PFC, perfluorocarbon; SPIO, superparamagnetic iron oxide. *T1 is the nuclear
spinlattice relaxation time.

T2 is the spinspin relaxation time.


PERSPECTI VES
760 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Nature Reviews | Immunology
c
a b
d
e
R
ac
ac
dc
a
dc
Future directions in immune cell tracking
Current MRI-based cell tracking technolo-
gies are poised to have a broad impact on
basic and biomedical research, but there are
still substantial efforts underway to devise
next-generation cell imaging technologies.
MRI reporter genes. One important aim of
these efforts is the development of robust
MRI reporter genes. An ideal reporter
gene would not be diluted by cell mitosis
andwould be cleared rapidly after cell
death and could potentially elucidate cell
viability, activation status and/or differentia-
tion status. In recent years, researchers have
been exploring schemes using exogenous
(that is, injected) substrates, endogenous
substrates or those that do not require a
substrate (FIG.2) in their quest to define
optimal nucleic acid-based MRI reporters.
The first exogenous substrate approach
involves genetically encoded, cell surface
ligands that bind to MRI-detectable probes;
for example, suitable ligands could include
biotin
63
or the transferrin receptor
64
, for use
with a probe that consists of a SPIO nano-
particle bound to streptavidin or transfer-
rin, respectively. Alternatively, transfection
of cells with the HSV thymidine kinase
reporter gene used for PET imaging can
be used to accumulate a thymidine ana-
logue probe in these cells that is detectable
by chemical exchange saturation transfer
(CEST) MRI
65
(BOX1). One of the challenges
of using ligandprobe-based reporter
schemes is the efficient delivery of the
probe to the tissues of interest.
The second class of MRI reporters use
endogenous substrates (FIG.2); several dis-
tinct technologies have been developed in
this area. In one approach, transgenic or
vector technologies are used to induce cells
to express genes that encode engineered
iron-binding proteins, for example, those in
the ferritin family
6669
. Following transgene
expression insitu, the ferritin outer protein
shell sequesters physiologically available
iron, and the endogenous formation of
an iron oxide crystal in the ferritin core
renders the complex paramagnetic, which
produces MRI contrast. Reporters that do
not require a substrate have more recently
been developed that involve the expression
of amide-rich proteins that can be detected
by CEST MRI
70
. So far, MRI reporter genes
have not been used for immune cell tracking
and their use has been mostly limited to
tumour cells
65,70
and to stem cells
69
.
Sensing the extracellular and intracellular
environment. Another key area of innova-
tion in the use of MRI to track cells has been
the development of probes that can sense
enzymatic activity and/or changes in the
microenvironment, for example, changes in
the local pH. Using arginine-filled liposomes
as pH-sensitive contrast agents, changes in
tissue pH that are associated with an immune
response can potentially be detected by
CEST MRI
71,72
. In addition, there are invivo
MRI probe technologies to detect enzymes
that are overexpressed following cellular
activation, such as protein kinase A (PKA).
Following PKA-mediated phosphorylation
of its peptide substrates, which can function
as CEST contrast agents, a 50% change in the
CESTMRI signal was observed
73
.
MRI measurements of intracellular
oximetry have also been used to assay cell
viability and metabolism
74
. Certain PFC
molecules used for invivo cytometry readily
Figure 3 | Tracking immune cells with MRI using SPIO nanoparticles and PFC emulsions.
a | Intracellular labelling of mononuclear cells with magnetoliposomes is shown
13
. The electron
micrograph shows superparamagnetic iron oxide (SPIO) particles in secondary lysosomes (small
arrows) and in primary lysosomes that are fusing with endosomes (large arrow). The scale bar rep-
resents 200 nm. b|Imaging of a non-obese diabetic (NOD)severe combined immunodeficient
(SCID) mouse pancreas ex vivo is shown following the adoptive transfer of SPIO-labelled Tcells
22
.
Image hypointensities represent infiltrating Tcells that are observed at 24hours post transfer.
c|Imaging of invivo antigen capture and trafficking of dendritic cells (DCs) is shown
26
. Sentinel DCs
were labelled insitu by intradermal injection of unlabelled (dashed arrow) or SPIO-labelled (solid
arrow) irradiated cancer cells, which function as a vaccine. Following phagocytosis of both SPIO
particles and tumour antigens in a process known as magnetovaccination, the hypointense DCs
migrate into the medulla of the draining popliteal lymph node, as observed on day 8. d|An electron
micrograph of a perfluorocarbon (PFC)-labelled DC is shown. Numerous bright spots (PFC droplets)
are observed inside the cell. Particles appear as smooth spheroids
45
. Arrowheads indicate vesicles.
The scale bar represents 200 nm. e|Inflammatory bowel disease (IBD) in an interleukin-10 (Il10)
/

mouse model was visualized using insitu PFC labelling and
19
F magnetic resonance imaging (MRI)
invivo
53
. The
19
F image (pseudo-colour) is shown on the far left, the composite
1
H and
19
F image is
shown in the middle and a three-dimensional rendering of the invivo
19
F MRI data from the abdomen
is shown on the far right. A reference tube that contains PFC emulsion (R) was placed alongside the
torso of the mouse. The images through the abdomen show PFC accumulation (indicating inflam-
mation) in the ascending colon (ac) and descending colon (dc), where (a) is the anus. Part a is repro-
duced, with permission, from REF.13 (1993) John Wiley and Sons. Part b is reproduced, with
permission, from REF.22 (2002) John Wiley and Sons. Part c is reproduced, with permission, from
REF.26 (2009) American Association for Cancer Research. Part d is reproduced, with permission, from
REF.45 (2005) Macmillan Publishers Ltd. All rights reserved. Part e is reproduced, with permission,
from REF.52 (2012) Macmillan Publishers Ltd. All rights reserved.
PERSPECTI VES
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 761
2013 Macmillan Publishers Limited. All rights reserved
dissolve paramagnetic oxygen, which alters
the MRI property (that is, the relaxation
time) of the
19
F signal in a manner that lin-
early increases with the absolute partial pres-
sure of oxygen
75
. In a rodent brain glioma
model, intracellular oximetry technology has
recently been used to monitor the apoptosis
dynamics of tumour cells as a result of their
interaction with CD8
+
Tcells that are specific
for glioma-associated antigens
76
.
Tracking cellcell interactions. MRI also
has the potential to visualize multiple cell
populations invivo to follow the spatial
dynamics of cellcell interactions, such as
antigen presentation to B cells or Tcells,
and macrophage-mediated recruitment
of lymphocytes; for example, exvivo cell-
labelling agents containing PFC molecules
with different
19
F NMR chemical shifts
(that is, molecular signatures) can be used
to label different immune cell subsets.
Following injection of these labelled sub-
sets, invivo multicolour magnetic reso-
nance images can be constructed, which
show the distribution of each cell species
according to its unique
19
F spectral signa-
ture
77
(a process known as multichannel
invivo cytometry). Alternatively, discrete
immune cell populations could be labelled
with two different modes of contrast, such
as a positive contrast agent with a negative
contrast agent (TABLE2), a metal-based con-
trast agent with a CEST agent
78
, or multiple
CEST agents of different off-resonance
frequencies, which enables the generation
of multiple colours
79
.
Prospects and challenges for translation
The use of MRI-based cell tracking in
clinical trials is still in its infancy. Only a
few small clinical pilot studies using SPIO-
based cell tracking have been reported
so far
80
. In the first clinical trial to use
MRI-based cell tracking, an immunothera-
peutic DC vaccine was labelled exvivo
with SPIO nanoparticles, and ultrasound
imaging was used to guide the delivery of
these cells into the inguinal lymph nodes
in patients with melanoma
40
. Using MRI it
was shown that the target lymph node was
missed in four out of eight patients; this
low success rate was not detected when
111
In-oxine-labelled DCs were tracked
using low-resolution radioscintigraphy
40
.
In another study, peripheral blood mono-
nuclear cells were labelled exvivo with
SPIO particles, injected systemically and
visualized at sites of cutaneous inflam-
mation
81
. As a setback for further clinical
use, the SPIO agents that have been used
for cell labelling ferumoxides injectable
solution (Feridex; Berlex laboratories in
the United States (also known as Endorem
in Europe)) and ferucarbotran (Resovist;
Bayer Schering Parma AG) are no
longer being manufactured because of
economic considerations. There are other
experimental SPIO agents available but
they would need large investments in
order to be developed as a clinical MRI
contrastagent.
Instead, there has recently been a tech-
nological push to use PFC nanoemulsions
for clinical MRI-based cell tracking
49
.
PFC labelling is clinically attractive because
of the very low toxicity of the probes, the
exceptionally high cell specificity in images
and the cell quantification capacity invivo.
The sensitivity of PFC labelling to detect
small cell numbers is potentially less than
for SPIO agents (TABLE2), but this disad-
vantage may be offset by the fact that there
is no background signal in
19
F images, cell
detection is unambiguous (in other words,
there are no false positives) and images can
be readily quantitated to yield apparent cell
numbers in regions of interest. The techni-
cal barriers that are associated with the
implementation and the quantification of
19
F MRI on a clinical scanner are surmount-
able; prior studies have shown the feasibil-
ity of efficient data collection schemes and
the quantification of PFC signal intensity
at clinical field strengths
82
.
Our view is that MRI-based cell track-
ing will routinely be used in the future for
clinical trials to monitor therapeutic cell
delivery and inflammation. We expect
that MRI-based cell tracking will provide
unexpected results regarding how cells
behave invivo following delivery and that
this information will be crucial to obtain
clinical success. Moreover, we foresee real-
time SPIO-based MRI-guided immune cell
delivery as a routine application of cell
tracking, in conjunction with MRI-
compatible injection catheters
83
. Finally,
the development of cell tracking software
for automated analysis of MRI scans may
further facilitate image interpretation and
intercomparison studies.
Conclusion
Overall, MRI-based cell tracking has great
potential, not only in basic immunological
research but also to guide the development
of experimental immune cell-based thera-
pies. SPIO-based cell tracking has already
been used in the clinic and PFC cell labels,
which are detected using
19
F MRI, are cur-
rently being incorporated into clinical
trials in the United States; these methods
are likely to gain momentum in the future.
In the meantime, emerging MRI-based
cell tracking approaches could enable the
visualization of cell survival, the activation
status of immune cells and the identification
of cellcell interactions invivo in ways that
were previously impossible.
Eric T.Ahrens is at the Department of Radiology,
University of California at San Diego, San Diego,
California 920930695, USA.
Jeff W.M.Bulte is at the Russell H. Morgan
Department of Radiology and Radiological Science,
Division of MR Research, Department of Oncology,
Department of Biomedical Engineering,
Department of Chemical & Biomolecular Engineering,
and Cellular Imaging Section and Vascular Biology
Program, Institute for Cell Engineering,
The Johns Hopkins University School of Medicine,
217 Traylor Building, 720 Rutland Avenue,
Baltimore, Maryland 21205, USA.
e-mails: eta@ucsd.edu;
jwmbulte@mri.jhu.edu
doi:10.1038/nri3531
Published online 10 September 2013
1. Bousso,P. & Moreau,H.D. Functional
immunoimaging: the revolution continues.
Nature Rev. Immunol. 12, 858864 (2012).
2. Thakur,M.L., Lavender,J.P., Arnot,R.N.,
Silvester,D.J. & Segal,A.W. Indium-111-labeled
autologous leukocytes in man. J.Nuclear Med. 18,
10141021 (1977).
3. Tjuvajev,J.G. etal. Imaging the expression of
transfected genes invivo. Cancer Res. 55,
61266132 (1995).
4. Koehne,G. etal. Serial invivo imaging of the targeted
migration of human HSV-TK-transduced antigen-
specific lymphocytes. Nature Biotech. 21, 405413
(2003).
5. Dubey,P. etal. Quantitative imaging of the Tcell
antitumor response by positron-emission tomography.
Proc. Natl Acad. Sci. USA 100, 12321237 (2003).
6. Yaghoubi,S.S. etal. Noninvasive detection of
therapeutic cytolytic Tcells with [
18
F]-FHBG PET in a
patient with glioma. Nature Clin. Pract. Oncol.
6, 5358 (2009).
7. Aoki,I. etal. Cell labeling for magnetic resonance
imaging with the T-1 agent manganese chloride.
NMR Biomed. 19, 5059 (2006).
8. Hawrylak,N. etal. Nuclear-magnetic-resonance
(NMR) imaging of iron oxide-labeled neural
transplants. Exp. Neurol. 121, 181192 (1993).
9. Bulte,J.W.M. & Kraitchman,D.L. Iron oxide MR
contrast agents for molecular and cellular imaging.
NMR Biomed. 17, 484499 (2004).
10. Gorelik,M. etal. Use of MR cell tracking to evaluate
targeting of glial precursor cells to inflammatory tissue
by exploiting the very late antigen-4 docking receptor.
Radiology 265, 175185 (2012).
11. Cromer Berman,S.M. etal. Cell motility of neural
stem cells is reduced after SPIO-labeling, which is
mitigated after exocytosis. Magn. Reson. Med. 69,
255262 (2013).
12. Bulte,J.W. etal. Specific MR imaging of human
lymphocytes by monoclonal antibody-guided
dextran-magnetite particles. Magn. Reson. Med.
25, 148157 (1992).
13. Bulte,J.W. etal. Selective MR imaging of labeled
human peripheral blood mononuclear cells by
liposome mediated incorporation of dextran-magnetite
particles. Magn. Reson. Med. 29, 3237 (1993).
14. Yeh,T.C., Zhang,W., Ildstad,S.T. & Ho,C.
Intracellular labeling of T-cells with superparamagnetic
contrast agents. Magn. Reson. Med. 30, 617625
(1993).
15. Frank,J.A. etal. Clinically applicable labeling of
mammalian and stem cells by combining
superparamagnetic iron oxides and transfection
agents. Radiology 228, 480487 (2003).
PERSPECTI VES
762 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
16. Arbab,A.S. etal. Labeling of cells with ferumoxides-
protamine sulfate complexes does not inhibit function
or differentiation capacity of hematopoietic or
mesenchymal stem cells. NMR Biomed. 18, 553559
(2005).
17. Thu,M.S. etal. Self-assembling nanocomplexes by
combining ferumoxytol, heparin and protamine for cell
tracking by magnetic resonance imaging. Nature Med.
18, 463467 (2012).
18. Ahrens,E.T., Feili-Hariri,M., Xu,H., Genove,G. &
Morel,P.A. Receptor-mediated endocytosis of iron-
oxide particles provides efficient labeling of dendritic
cells for invivo MR imaging. Magn. Reson. Med. 49,
10061013 (2003).
19. Walczak,P. etal. Magnetoelectroporation: improved
labeling of neural stem cells and leukocytes for cellular
magnetic resonance imaging using a single FDA-
approved agent. Nanomedicine 2, 8994 (2006).
20. Kircher,M.F. etal. Invivo high resolution three-
dimensional imaging of antigen-specific cytotoxic
T-lymphocyte trafficking to tumors. Cancer Res. 63,
68386846 (2003).
21. Daldrup-Link,H.E. etal. Invivo tracking of
genetically engineered, anti-HER2/neu directed
natural killer cells to HER2/neu positive mammary
tumors with magnetic resonance imaging.
Eur. Radiol. 15, 413 (2005).
22. Moore,A. etal. MRI of insulitis in autoimmune
diabetes. Magn. Reson. Med. 47, 751758
(2002).
23. Anderson,S.A. etal. Magnetic resonance imaging of
labeled T-cells in a mouse model of multiple sclerosis.
Ann. Neurol. 55, 654659 (2004).
24. Baumjohann,D. etal. Invivo magnetic resonance
imaging of dendritic cell migration into the draining
lymph nodes of mice. Eur. J.Immunol. 36,
25442555 (2006).
25. Rohani,R. etal. Invivo cellular MRI of dendritic cell
migration using micrometer-sized iron oxide (MPIO)
particles. Mol. Imaging Biol. 13, 679694 (2011).
26. Long,C.M., van Laarhoven,H.W., Bulte,J.W. &
Levitsky,H.I. Magnetovaccination as a novel method
to assess and quantify dendritic cell tumor antigen
capture and delivery to lymph nodes. Cancer Res. 69,
31803187 (2009).
27. Shapiro,E.M., Medford-Davis,L.N., Fahmy,T.M.,
Dunbar,C.E. & Koretsky,A.P. Antibody-mediated cell
labeling of peripheral Tcells with micron-sized iron
oxide particles (MPIOs) allows single cell detection by
MRI. Contrast Media Mol. Imaging 2, 147153
(2007).
28. Heyn,C. etal. Invivo magnetic resonance imaging of
single cells in mouse brain with optical validation.
Magn. Reson. Med. 55, 2329 (2006).
29. Chan,T.W., Eley,C., Liberti,P., So,A. & Kressel,H.Y.
Magnetic resonance imaging of abscesses using lipid-
coated iron oxide particles. Investigative Radiol. 27,
443449 (1992).
30. Dousset,V. etal. MR imaging of relapsing multiple
sclerosis patients using ultra-small-particle iron oxide
and compared with gadolinium. AJNR Am.
J.Neuroradiol. 27, 10001005 (2006).
31. Vellinga,M.M. etal. Pluriformity of inflammation
in multiple sclerosis shown by ultra-small iron oxide
particle enhancement. Brain 131, 800807 (2008).
32. Sosnovik,D.E. & Nahrendorf,M. Cells and iron oxide
nanoparticles on the move: magnetic resonance
imaging of monocyte homing and myocardial
inflammation in patients with ST-elevation myocardial
infarction. Circ. Cardiovasc. Imaging 5, 551554
(2012).
33. Gaglia,J.L. etal. Noninvasive imaging of pancreatic
islet inflammation in type1A diabetes patients.
J.Clin. Invest. 121, 442445 (2011).
34. Luciani,A. etal. Adipose tissue macrophages: MR
tracking to monitor obesity-associated inflammation.
Radiology 263, 786793 (2012).
35. Kanno,S. etal. Macrophage accumulation associated
with rat cardiac allograft rejection detected by
magnetic resonance imaging with ultrasmall
superparamagnetic iron oxide particles. Circulation
104, 934938 (2001).
36. Harisinghani,M.G. etal. Noninvasive detection of
clinically occult lymph-node metastases in prostate
cancer. New Engl. J.Med. 348, 24912499
(2003).
37. Kooi,M.E. etal. Accumulation of ultrasmall
superparamagnetic particles of iron oxide in human
atherosclerotic plaques can be detected by invivo
magnetic resonance imaging. Circulation 107,
24532458 (2003).
38. Siglienti,I., Bendszus,M., Kleinschnitz,C. & Stoll,G.
Cytokine profile of iron-laden macrophages:
implications for cellular magnetic resonance imaging.
J.Neuroimmunol. 173, 166173 (2006).
39. Verdijk,P. etal. Sensitivity of magnetic resonance
imaging of dendritic cells for invivo tracking of cellular
cancer vaccines. Int. J.Cancer 120, 978984 (2007).
40. de Vries,I.J.M. etal. Magnetic resonance tracking of
dendritic cells in melanoma patients for monitoring of
cellular therapy. Nature Biotech. 23, 14071413
(2005).
41. Weissleder,R. etal. Superparamagnetic iron oxide:
pharmacokinetics and toxicity. AJR Am. J.Roentgenol.
152, 167173 (1989).
42. Riess,J.G. Oxygen carriers (blood substitutes)
raison detre, chemistry, and some physiology. Chem.
Rev. 101, 27972919 (2001).
43. Janjic,J.M. & Ahrens,E.T. Fluorine-containing
nanoemulsions for MRI cell tracking. Wiley Interdiscip.
Rev. Nanomed. Nanobiotechnol 1, 492501 (2009).
44. Holland,G.N., Bottomley,P.A. & Hinshaw,W.S. F-19
magnetic-resonance imaging. J.Magn. Reson. 28,
133136 (1977).
45. Ahrens,E.T., Flores,R., Xu,H.Y. & Morel,P.A.
Invivo imaging platform for tracking
immunotherapeutic cells. Nature Biotech. 23,
983987 (2005).
46. Bulte,J.W.M. Hot spot MRI emerges from the
background. Nature Biotech. 23, 945946 (2005).
47. Janjic,J.M.Srinivas,M., Kadayakkara,D.K. &
Ahrens,E.T. Self-delivering nanoemulsions for dual
fluorine-19 MRI and fluorescence detection.
J.Am. Chem. Soc. 130, 28322841 (2008).
48. Srinivas,M. etal. Invivo cytometry of antigen-specific
t cells using 19F MRI. Magn. Reson. Med. 62,
747753 (2009).
49. Helfer,B.M. etal. Functional assessment of human
dendritic cells labeled for invivo F-19 magnetic
resonance imaging cell tracking. Cytotherapy 12,
238250 (2010).
50. Srinivas,M., Morel,P.A., Ernst,L.A., Laidlaw,D.H. &
Ahrens,E.T. Fluorine-19 MRI for visualization and
quantification of cell migration in a diabetes model.
Magn. Reson. Med. 58, 725734 (2007).
51. Kadayakkara,D.K. etal. Inflammation driven by
overexpression of the hypoglycosylated abnormal
mucin 1 (MUC1) links inflammatory bowel disease
and pancreatitis. Pancreas 39, 510515 (2010).
52. Ahrens,E.T., Young,W.B., Xu,H. & Pusateri,L.K.
Rapid quantification of inflammation in tissue samples
using perfluorocarbon emulsion and fluorine-19
nuclear magnetic resonance. Biotechniques 50, 229
234 (2011).
53. Kadayakkara,D.K., Ranganathan,S., Young,W.B. &
Ahrens,E.T. Assaying macrophage activity in a
murine model of inflammatory bowel disease using
fluorine-19 MRI. Lab Invest. 92, 636645 (2012).
54. Hertlein,T. etal. Visualization of abscess formation in
a murine thigh infection model of Staphylococcus
aureus by F-19-magnetic resonance imaging (MRI).
PLoS ONE 6, e18246 (2011).
55. Hitchens,T.K. etal. 19F MRI detection of acute
allograft rejection with invivo perfluorocarbon
labeling of immune cells. Magn. Reson. Med. 65,
11441153 (2011).
56. Flogel,U. etal. Noninvasive detection of graft
rejection by invivo
19
F MRI in the early stage.
Am. J.Transplant. 11, 235244 (2011).
57. Noth,U. etal. Perfluoro-15-crown-5-ether labelled
macrophages in adoptive transfer experimental
allergic encephalomyelitis. Artif. Cells Blood Substit.
Immobil. Biotechnol. 25, 243254 (1997).
58. Weise,G., Basse-Luesebrink,T.C., Wessig,C.,
Jakob,P.M. & Stoll,G. Invivo imaging of
inflammation in the peripheral nervous system
by
19
F MRI. Exp. Neurol. 229, 494501 (2011).
59. Ebner,B. etal. Early assessment of pulmonary
inflammation by
19
F MRI invivo. Circ. Cardiovasc.
Imag. 3, U202U109 (2010).
60. Flogel,U. etal. Invivo monitoring of inflammation
after cardiac and cerebral ischemia by fluorine
magnetic resonance imaging. Circulation 118,
140148 (2008).
61. OHanlon,C.E., Amede,K.G., OHear,M.R. &
Janjic,J.M. NIR-labeled perfluoropolyether
nanoemulsions for drug delivery and imaging.
J.Fluor. Chem. 137, 2733 (2012).
62. Morel,P.A. etal. Gene expression analysis of
dendritic cells that prevent diabetes in NOD mice:
analysis of chemokines and costimulatory molecules.
J.Leukoc. Biol. 90, 539550 (2011).
63. Tannous,B.A. etal. Metabolic biotinylation of cell
surface receptors for invivo imaging. Nature Methods
3, 391396 (2006).
64. Weissleder,R. etal. Invivo magnetic resonance imaging
of transgene expression. Nature Med. 6, 351355
(2000).
65. Bar-Shir,A. etal. Transforming thymidine into a
magnetic resonance imaging probe for monitoring gene
expression. J.Am. Chem. Soc. 135, 16171624 (2013).
66. Ahrens,E.T. Contrast agents for magnetic resonance
imaging and methods related thereto. US Patent
10/384,496 (2003).
67. Genove,G., DeMarco,U., Xu,H.Y., Goins,W.F. &
Ahrens,E.T. A new transgene reporter for invivo
magnetic resonance imaging. Nature Med. 11,
450454 (2005).
68. Cohen,B., Dafni,H., Meir,G., Harmelin,A. &
Neeman,M. Ferritin as an endogenous MRI
reporter for noninvasive imaging of gene expression
in C6 glioma tumors. Neoplasia 7, 109117 (2005).
69. Iordanova,B. & Ahrens,E.T. Invivo magnetic
resonance imaging of ferritin-based reporter
visualizes native neuroblast migration. Neuroimage
59, 10041012 (2012).
70. Gilad,A.A. etal. Artificial reporter gene providing
MRI contrast based on proton exchange. Nature
Biotech. 25, 217219 (2007).
71. Sherry,A.D. & Woods,M. Chemical exchange
saturation transfer contrast agents for magnetic
resonance imaging. Annu. Rev. Biomed. Eng. 10,
391411 (2008).
72. Chan,K.W. etal. MRI-detectable pH nanosensors
incorporated into hydrogels for invivo sensing of
transplanted-cell viability. Nature Mater. 12,
268275 (2013).
73. Airan,R.D. etal. MRI biosensor for protein kinase A
encoded by a single synthetic gene. Magn. Reson.
Med. 68, 19191923 (2012).
74. Kadayakkara,D.K., Janjic,J.M., Pusateri,L.K.,
Young,W.B. & Ahrens,E.T. Invivo observation of
intracellular oximetry in perfluorocarbon-labeled
glioma cells and chemotherapeutic response in the
CNS using fluorine-19 MRI. Magn. Reson. Med. 64,
12521259 (2010).
75. Sotak,C.H. etal. A new perfluorocarbon for use
in fluorine-19 magnetic resonance imaging and
spectroscopy. Magn. Reson. Med. 29, 188195 (1993).
76. Zhong,J., Sakaki,M., Okada,H. & Ahrens,E.T.
Invivo intracellular oxygen dynamics in murine brain
glioma and immunotherapeutic response of cytotoxic
Tcells observed by fluorine-19 magnetic resonance
imaging. PLoS One 8, e59479 (2013).
77. Partlow,K.C. etal. F-19 magnetic resonance
imaging for stem/progenitor cell tracking with multiple
unique perfluorocarbon nanobeacons. FASEB J. 21,
16471654 (2007).
78. Gilad,A.A. etal. Feasibility of concurrent dual
contrast enhancement using CEST contrast agents
and superparamagnetic iron oxide particles.
Magn. Reson. Med. 61, 970974 (2009).
79. Liu,G. etal. Invivo multicolor molecular MR imaging
using diamagnetic chemical exchange saturation
transfer liposomes. Magn. Reson. Med. 67,
11061113 (2012).
80. Bulte,J.W. Invivo MRI cell tracking: clinical studies.
AJR Am. J.Roentgenol. 193, 314325 (2009).
81. Richards,J.M.J. etal. Clinical cell tracking of
mononuclear cells using magnetic resonance imaging
and superparamagnetic particles of iron oxide.
Br. J.Surg. 98, 44 (2011).
82. Keupp,J. etal. Simultaneous dual-nuclei imaging for
motion corrected detection and quantification of 19F
imaging agents. Magn. Reson. Med. 66, 11161122
(2011).
83. Aarntzen,E.H. etal. Invivo tracking techniques for
cellular regeneration, replacement, and redirection.
J.Nucl. Med. 53, 18251828 (2012).
84. Bulte,J.W. etal. MPI cell tracking: what can we learn
from MRI? Proc. Soc. Photo-Opt. Instrum. Eng. 7965,
79650z (2011).
Acknowledgements
The authors are supported by US National Institutes of Health
grants R01-CA134633, P41-EB0019772, R01-NS045062,
R01-EB007825, R01-DA026299 and U54-CA151838, the
Maryland Stem Cell Research Foundation and the California
Institute for Regenerative Medicine.
Competing interests statement
The authors declare competing financial interests: see Web
version for details.
PERSPECTI VES
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 763
2013 Macmillan Publishers Limited. All rights reserved
For more than a century, immunologists
have devised theories to provide descriptions,
explanations and predictions about how an
immune response is induced. Such interest in
theories in a field that is primarily based on
molecular biology is somewhat uncommon
1
,
but it has undoubtedly been very fruitful.
However, having been confronted with the
difficulties that are associated with some of
these theories
2
, several immunologists might
be tempted to abandon their ambition to
build what they see as grand theories, pre-
ferring instead to focus on the elucidation
of the molecular mechanisms of immune
responses. Nevertheless, theories are essential
to interpret observations and to develop new
research hypotheses (BOX1). Therefore, we are
convinced that the development of theories
should remain associated with advances in
immunology. In this Essay, we propose a new
theoretical framework that aims to explain
how immune cells integrate spatial, temporal,
quantitative and qualitative signals to mount
an immune response. We start with the
case of natural killer (NK) cells and then we
extend our framework to other immunecells.
The dynamics of NK cell reactivity
NK cells are innate lymphoid cells that pro-
duce cytokines and that can be cytolytic
,3,4
.
They distinguish their cellular targets from
healthy cells via a panel of activating and
inhibitory cell surface receptors, as well
as via adhesion molecules. Most NK cell
activating receptors recognize cell surface
ligands that are induced on modified cells,
for example, following tumour transforma-
tion, microbial infection, or physical and
chemical assaults
5
. This recognition of
dysregulated self components enables NK
cells to be selectively activated by host cells
that have undergone molecular modifica-
tions and that are stressed in some way.
Prototypical examples of NK cell surface
receptors that function according to this
mode of recognition include natural killer
group 2 member D (NKG2D) and NK cell
p30-related protein (NKp30; also known
as NCR3). By contrast, NK cell inhibitory
receptors recognize cell surface molecules
that are readily expressed in the steady
state on most cells, such as killer cell
immunoglobulin-like receptors (KIRs) and
LY49 molecules that recognize MHC classI
molecules in humans and mice, respectively.
This mode of recognition enables NK cells
to be selectively activated by cells that have
received stress signals, which leads to the
downregulation of constitutively expressed
molecules, such as MHC classI molecules
following cytomegalovirus infection; it is
thus referred to as missing self recognition
6
.
By integrating activating and inhibitory sig-
nals after encountering a given inter acting
cell, NK cells thereby spare normal cells
and contribute to the elimination of cells
expressing strongly modified motifs (FIG.1a).
Importantly, the responsiveness of NK
cells decreases in the case of chronic engage-
ment of various activating receptors, such as
NKG2D, human activating KIRs (for exam-
ple, KIR2DS1) or mouse activating LY49
molecules (for example, LY49H) (FIG.1b).
Similarly, whereas NK cells developing in
an MHC classI-sufficient environment can
eliminate MHC classI-negative cells, NK
cells that develop in or that are transferred to
an MHC classI-deficient environment are no
longer able to do so
710
(FIG.1c). Thus, NK cells
detect sudden modifications in their environ-
ment, but when modified motifs are long
lasting, NK cells adapt to them by ceasing
to be responsive. These observations lay the
foundations for a new theoretical framework
the discontinuity theory which pro-
poses that immune cells induce an effector
response when there is a discontinuity in the
molecular motifs with which their receptors
interact, whereas they tend to become
tolerant to continuously expressedmotifs.
Extension to other immune cells
Several families of innate immune cells also
respond to a sudden modification of their
ligands. Macrophages, in particular, respond
not only to microorganisms but also to
modified host cells
1113
. The molecular tar-
gets in such cases include pre-expressed host
molecules that show biochemical modifica-
tions, such as oxidation in the sugar chains
of glycoproteins and glycolipids, molecules
that show neo-expression as a result of cel-
lular stress (for example, thrombospondin1)
or intracellular molecules that relocalize to
the cell membrane after injury (for example,
phosphatidylserine, DNA, endoplasmic
reticulum-resident proteins or F-actin)
14,15
.
Crucially, macrophages also adapt to long-
lasting motifs in their environment. This
adaptation is shown by the well-known
phenomenon of endotoxin tolerance a
process by which cells that are exposed to
low concentrations of endotoxin become
hyporesponsive to further challenges with
endotoxin, thereby avoiding the potentially
harmful effects caused by the continual
presence of pro-inflammatory stimuli
16
.
Along the same lines, phagocytosis
of potential target cells is blocked when
macro phages detect CD47 (a transmem-
brane protein that is ubiquitously expressed
by host cells) via the inhibitory receptor
signal-regulatory protein1 (SIRP1; also
known as SHPS1). As a consequence, when
a macrophage encounters a CD47-deficient
ESSAY
The speed of change: towards a
discontinuity theory of immunity?
Thomas Pradeu, Sbastien Jaeger & Eric Vivier
Abstract | Immunology though deeply experimental in everyday practice is
also a theoretical discipline. Recent advances in the understanding of innate
immunity, how it is triggered and how it shares features that have previously been
uniquely ascribed to the adaptive immune system, can contribute to the
refinement of the theoretical framework of immunology. In particular, natural
killer cells and macrophages are activated by transient modifications, but adapt
to long-lasting modifications that occur in the surrounding tissue environment.
This process facilitates the maintenance of self-tolerance while permitting
efficient immune responses. In this Essay we extend this idea to other components
of the immune system and we propose some general principles that lay the
foundations for a unifying theory of immunity the discontinuity theory.
According to this theoretical framework, effector immune responses (namely,
activated responses that lead to the potential elimination of the target antigen)
are induced by an antigenic discontinuity; that is, by the sudden modification of
molecular motifs with which immune cells interact.
PERSPECTI VES
764 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
erythrocyte, it phagocytoses it, as the
inhibitory macrophage receptor SIRP1 is
no longer engaged by CD47. However, mac-
rophages from CD47-deficient mice do not
phagocytose CD47-deficient erythrocytes
17
.
Therefore, the discontinuity theory applies
very well to macrophages, which sense
changes in their environment but adapt to
these changes when they become persistent.
Could the same concept be extended
to other immune cells, such as B cells and
Tcells? The activation of these lymphocytes
is admittedly much more complex, particu-
larly because they undergo a thorough selec-
tion process in primary lymphoid organs
where, in theory, potentially self-reactive
lymphocytes are removed. However, several
data hint at a possible discontinuity-based
mechanism of activation in B cells and
Tcells. It is particularly well established that
conventional B cells and Tcells are anergized
when an antigen is chronically or constantly
present
18,19
. For instance, several chronic
infections cause functional exhaustion or
deletion of antigen-specific conventional
Tcells
20,21
. Similarly, the chronic expo-
sure of non-conventional mouse Tcells
(V5V1
+
) to SKINT1 (putative selection
and upkeep of intraepithelial T cells pro-
tein 1 homolog), which is a prototypical
member of a novel family that is related to
butyrophilin-like molecules, leads to their
hyporesponsiveness
22
. All of these observa-
tions may contribute to a renewal of the tun-
able activation threshold model, according
to which lymphocytes sense relative changes,
rather than the absolute strength of the
stimulation, and respond only if a sufficiently
rapid increase in their level of stimulation
occurs
23,24
. More work will undoubtedly be
needed to determine whether the discontinu-
ity theory applies to B cells and Tcells. For
the moment, suffice it to say that this theory
primarily aims to explain innate immune
responses, which are key to understanding
immune responses in general
25
.
Lastly, an area in which the framework dis-
cussed in this Essay seems particularly relevant
is that of tumour immunology. It is now well
established that tumours give rise to effector
immune responses
2630
. Although tumour
cells are, genetically speaking, self cells, they
are transformed cells that possess several fea-
tures that distinguish them from healthy host
cells, particularly the expression of abnormal
motifs at their cell surface
28,31
. Effector immune
responses to tumours can be explained by
the appearance of these strongly modified pat-
terns. Conversely, several experiments showed
that a persistent tumour antigen diminishes
the efficacy of the antitumour response
32,33
.
The core of the discontinuity theory
Antigenic discontinuity. We believe that
a very simple theoretical framework can
unify the diverse immunological data that is
discussed above using a single explanatory
principle. We call this framework the discon-
tinuity theory. Briefly, it states that effector
immune responses are induced by an anti-
genic discontinuity; that is, by the appearance
of molecular motifs that are qualitatively or
quantitatively different from those with which
the immune system has regularly interacted
so far. Note that an antigen is defined here
as any entity with which immune receptors,
innate or adaptive, interact. The criterion
of immunogenicity that we use is antigenic
discontinuity itself, and not the origin of the
antigen; indeed, a discontinuity can appear
at the level of self motifs (for example, a
tumour), as well as at the level of non-self
motifs (for example, a microorganism)
34
.
From this point of view, the framework pre-
sented in this Essay echoes the concepts of
immune surveillance
3537
and altered self
38,39
,
and incorporates these ideas in a more com-
prehensive perspective. More generally, the
idea that antigenic discontinuity is immuno-
genic can be related to the common and well
supported principle that biological entities
react to variations in stimuli, for example, as
has been shown for sight
4042
.
A response initiated by a variation in the
quantities of antigens with respect to time.
The central statement of the discontinuity
theory is that an effector immune response
occurs as a function of a variation in the
quantities of antigen (dQ) with respect to
time (dt) that is, the value dQ/dt (BOX2
details how the variation can be modelled
with respect to time). Indeed, two key factors
in the induction of an immune response
are the speed at which antigens appear and
the duration of antigen exposure. In addi-
tion, two other aspects must be taken into
account. First, the spatial aspect: the localiza-
tion of the immune interaction is crucial, as
immune responses are regulated by the local
environment, which is often characterized
by the prevalence of different cell types (for
example, the gut or the skin are very differ-
ent in terms of their cellular composition).
Therefore, what is important is the local
antigenic discontinuity, because a given anti-
gen can be immunogenic or not according to
its localization. Second, the structural aspect:
the degree of molecular difference between
usual antigens with which immune receptors
interact and the antigen under consideration
is important.
Now, why should the immune response
be understood as a function of the variation
in quantities of antigen with respect to time?
In many cases, antigens that progressively
appear do not represent a strong modifica-
tion and hence are not eliminated
43,44
. Motifs
that change at a very high rate can make
protective immune responses impossible, as
happens with many parasites and viruses,
including with HIV
4547
. Indeed, in these
conditions, the time that is required for the
generation of a protective immune response
is greater than the time that is needed for
the pathogens to mutate. As a consequence,
immune responses to these pathogens are
constantly inadequate. The duration of
antigen exposure is also crucial, as cells that
have been exposed to chronic stimulations
become desensitized by various mechanisms.
Desensitization is by no means specific to
immunity. In all living objects, the detection
of changes in the environment of a cell (for
example, the recognition of a stimulus by a
cell surface receptor) is associated with the
initiation of regulatory pathways that are
either intrinsic or extrinsic to the stimulated
Box 1 | The importance of theories in biology
Theoretical thinking is often criticized for its excessive abstraction and its lack of use in the
everyday life of scientists. However, it has long been recognized that having a theoretical
framework is indispensable for conducting fruitful research, and that even scientific observations
and experiments that seem the most objective are always, at least to some extent, influenced by
theoretical views
53,64,65
. A theory can be defined as an organized set of testable explanatory and
predictive statements. Some biological theories take the form of hierarchical mechanisms
1
, but
even these make the formulation of explanations and predictions possible. A purely descriptive
mechanism (for example, at the molecular level) would not give rise to true explanations and
predictions, and as such it would not significantly contribute to the advancement of science. The
formulation of theories helps scientists to determine what to investigate, to interpret what they
observe and to develop a specific perspective. It also makes fruitful comparisons between different
scientific fields possible
66
. Finally, it encourages scientists to make bold predictions, with the risk of
being proved wrong, but also with the conviction that the refutation of incorrect theories is useful.
Therefore, because theories are crucial in the formulation of explanations and in the construction
of new experiments, it seems highly desirable that all fields of biology, including molecular ones
such as immunology, aim to construct new and testable theories.
PERSPECTI VES
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 765
2013 Macmillan Publishers Limited. All rights reserved
Nature Reviews | Immunology
NK cell
MHC class I
Modied
host cell
KIR
Healthy host cell
Healthy
host cell
a Sudden appearance of activating ligands
b Chronic presence of activating ligands c Chronic absence of MHC class I recognition
KIR
NKG2D ligand
NKG2D
NKG2D
ligand
NKG2D
Ignorance
or tolerance
Infection or
transformation
NK cell-mediated
cytotoxicity
Modied
host cell
Modied
host cell
Eliminated
cell
Activation
Hyporesponsiveness Hyporesponsiveness
No MHC class IKIR
interaction
+
cell. Desensitization of signalling pathways is
a common cell-intrinsic regulatory mecha-
nism and is a property that results from
feedback and/or feedforward loops
48
. The
mechanisms used to achieve desensitization
include cell surface receptor internalization
(for example, Bcells and Tcells internalize
their antigen receptors following antigen
recognition) and the degradation of key
downstream signalling elements. Other cell-
intrinsic regulatory circuits include
the activation of enzymatic pathways that
counteract the signalling cascade initiated
by the initial stimulus; for example, immuno-
receptor tyrosine-based inhibition motif
(ITIM)-containing receptors recruit and
activate phosphatases that antagonize the
activating kinases that are triggered by
immunoreceptor tyrosine-based activation
motif (ITAM)-containing receptors
49
. There
are also several cell extrinsic regulatory path-
ways, such as regulatory T (T
Reg
) cells that
negatively control Tcell activation
5052
. More
ambitious developments of the theory, such
as considering more refined modes of dis-
continuity detection mechanisms or feedback
and feedforward loops, could be explored
but each of these refinements will also need
efforts on the experimental side to generate
suitable (dynamic) results, to identify crucial
questions and to progressively refine our
knowledge. In particular, whether the extrin-
sic control exerted by T
Reg
cells on effector
cells is also controlled by mechanisms that
are described by the discontinuity theory has
to be experimentally addressed to offer an
adequate perspective of this issue. However,
this shows what is perhaps the most valu-
able benefit of using a theory: investigations
can be focused on the most pertinent and
informative hypotheses.
The main claims made by the disconti-
nuity theory can be summarized as follows.
The sudden appearance of a structurally
different motif induces a strong effector
immune response; it ends when the antigen
is eliminated and may lead to the genera-
tion of memory cells (FIG.2a). An initially
unusual but persistent motif will first trigger
an effector immune response, but then the
chronic presence of the motif will lead to the
termination of the response (FIG.2b). The slow
appearance of a structurally different motif
leads to a very limited reaction and eventually
to the tolerance of this motif (FIG.2c).
Why is the discontinuity theory useful?
A unifying framework. A new theoretical
framework is accepted only if it offers both
a new way of seeing things and original
experimental predictions (BOX1). A major
driving force in the development of new
scientific theories is the need for increased
unifying power: scientists often seek frame-
works that are more comprehensive than the
preceeding theories
53
. Therefore, it is impor-
tant to emphasize that the discontinuity
theory gathers under a simple explanation
a range of phenomena that have tradition-
ally received distinct and heterogeneous
explanations; for example, effector immune
responses to tumour cells find ad hoc expla-
nations in the selfnon-self theory, through
the idea of the altered self
39
and the danger
theory
54
, whereas the discontinuity theory
states that tumours induce effector immune
responses because they have unusual motifs.
When discontinuities in self components
arise in host tissues (for example, during
tumour development or cellular stress), an
effector immune response does occur, and
when non-self components are presented
to the immune system in a progressive,
continuous way (for example, in the case
of commensal bacteria
55,56
), they often
induce regulatory mechanisms
57
and are not
destroyed by the immune system. Overall,
the discontinuity framework should not be
perceived as opposing previous theories (in
particular, the self versus non-self and the
danger theories) but rather as complement-
ing and unifyingthem.
Predictions and challenges. The most
important challenge is to determine how
the discontinuity theory could help in
the current field of immunology from a
Figure 1 | Natural killer cell tuning. a | Natural killer (NK) cells can distinguish their targets (that is,
infected or transformed host cells) from healthy host cells on the basis of the engagement of activating
receptors (such as natural killer group 2 member D (NKG2D); shown by the plus sign) and the lack of
engagement of killer inhibitory receptors (KIRs; shown by the minus sign). In this example, the recogni-
tion of MHC classI molecules on healthy cells by the inhibitory KIRs expressed on human NK cells induces
a dominant inhibitory signal in NK cells. By contrast, host cells can be modified as a result of various
assaults, such as tumour transformation or microbial infection. In these circumstances, molecules
induced or upregulated at the surface of the target cell, engage activating receptors, such as NKG2D,
which leads to NK cell activation and eventually to the elimination of the target cells. b|Although
NK cells are activated by modified cells expressing ligands for activating receptors, the chronic presence
of activating ligands reprogrammes NK cells to become hypo responsive. c | The absence of MHC classI
expression on target cells releases NK cells from the inhibition induced by KIRs, but NK cells adapt to
the long-lasting absence of MHC classI recognition. When MHC classI or KIR molecules are not
present, or when MHC classI and KIR molecules do not interact, NK cells adapt to this environment
by increasing their threshold of reactivity, which leads to NK cell hyporesponsiveness.
PERSPECTI VES
766 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
Nature Reviews | Immunology
R(t) =
R(t) = T
R(t) = (E(t), E(t1), E(t2), ... , E(t))
dE(t)
dt
E(t) =
E(t) E(th)
1

h
T(x) =

1 + e
(x)
E(t) = E(t) E(t1)

i =
i = 0
|E(ti)|
120
100
100
100
100
150
50
0
200
200
300 400 500
130
160
190
220
80
60
40
20
0
0
0
T
(
x
)
R
a
t
i
o

b
o
u
n
d
/
f
r
e
e

i
n
d
o
-
1
Biotin-anti-
NKG2D (CX5)
a
b
c
d
e
Cell activity
Time
1 2 3 4 5 6 7 8 9 10 11 12 13 14
0
Time
401 801 1201
E(t)
R(t)
1601 2001 2401 2801
Streptavidin
100
150
50
0
200
0
Time
401 801 1201 1601 2001 2401 2801
100
150
50
0
200
0
Time
401 801 1201 1601 2001 2401 2801
100
150
50
0
200
250
0
Time
401 801 1201 1601 2001 2401 2801
15
practical point of view. A key prediction
made by the discontinuity theory is that
chronicity diminishes the intensity of the
immune response. Indeed, according to
the above analysis of the dynamics of the
immune response, components that persist
for a notable period of time in the organ-
ism will tend to become increasingly toler-
ated by the immune system. However, it is
important to mention that there are other
mechanisms involved in tolerance and that
acute antigenic challenges might also be
tolerogenic
58
. Nevertheless, the discontinu-
ity theory leads to several original and test-
able predictions.
Box 2 | A mathematical model of the discontinuity theory
As can be seen in the following figures, our model accurately reproduces the
behaviours that are qualitatively shown in FIG.2.
Discontinuity refers to a perturbation of any environmental signal detected by
the system in time and space. As we are interested in the variations in the inputs
rather than in their values, R(t) (the output variable that corresponds to the
immune response intensity that is triggered by the cell) will essentially be a
function of the derivatives of E(t) (the input variable that encompasses all external
signals that are sensed by an immune cell) rather than of E(t) itself. Thus,
Nature Reviews | Immunology
R(t) =
R(t) = T
R(t) = (E(t), E(t1), E(t2), ... , E(t))
dE(t)
dt
E(t) =
E(t) E(th)
1

h
T(x) =

1 + e
(x)
E(t) = E(t) E(t1)

i =
i = 0
|E(ti)|
120
100
100
100
100
150
50
0
200
200
300 400 500
130
160
190
220
80
60
40
20
0
0
0
T
(
x
)
R
a
t
i
o

b
o
u
n
d
/
f
r
e
e

i
n
d
o
-
1
Biotin-anti-
NKG2D (CX5)
a
b
c
d
e
Cell activity
Time
1 2 3 4 5 6 7 8 9 10 11 12 13 14
0
Time
401 801 1201
E(t)
R(t)
1601 2001 2401 2801
Streptavidin
100
150
50
0
200
0
Time
401 801 1201 1601 2001 2401 2801
100
150
50
0
200
0
Time
401 801 1201 1601 2001 2401 2801
100
150
50
0
200
250
0
Time
401 801 1201 1601 2001 2401 2801
15
We considered the values of the variables at discrete time points rather than
over a continuum of time, and we suggested manipulating the discrete
equivalent of a derivative quantity, the finite differences defined by:
Nature Reviews | Immunology
R(t) =
R(t) = T
R(t) = (E(t), E(t1), E(t2), ... , E(t))
dE(t)
dt
E(t) =
E(t) E(th)
1

h
T(x) =

1 + e
(x)
E(t) = E(t) E(t1)

i =
i = 0
|E(ti)|
120
100
100
100
100
150
50
0
200
200
300 400 500
130
160
190
220
80
60
40
20
0
0
0
T
(
x
)
R
a
t
i
o

b
o
u
n
d
/
f
r
e
e

i
n
d
o
-
1
Biotin-anti-
NKG2D (CX5)
a
b
c
d
e
Cell activity
Time
1 2 3 4 5 6 7 8 9 10 11 12 13 14
0
Time
401 801 1201
E(t)
R(t)
1601 2001 2401 2801
Streptavidin
100
150
50
0
200
0
Time
401 801 1201 1601 2001 2401 2801
100
150
50
0
200
0
Time
401 801 1201 1601 2001 2401 2801
100
150
50
0
200
250
0
Time
401 801 1201 1601 2001 2401 2801
15
where h denotes the time point. As a convention, we will take a time point equal
to 1, thus in our case:
Nature Reviews | Immunology
R(t) =
R(t) = T
R(t) = (E(t), E(t1), E(t2), ... , E(t))
dE(t)
dt
E(t) =
E(t) E(th)
1

h
T(x) =

1 + e
(x)
E(t) = E(t) E(t1)

i =
i = 0
|E(ti)|
120
100
100
100
100
150
50
0
200
200
300 400 500
130
160
190
220
80
60
40
20
0
0
0
T
(
x
)
R
a
t
i
o

b
o
u
n
d
/
f
r
e
e

i
n
d
o
-
1
Biotin-anti-
NKG2D (CX5)
a
b
c
d
e
Cell activity
Time
1 2 3 4 5 6 7 8 9 10 11 12 13 14
0
Time
401 801 1201
E(t)
R(t)
1601 2001 2401 2801
Streptavidin
100
150
50
0
200
0
Time
401 801 1201 1601 2001 2401 2801
100
150
50
0
200
0
Time
401 801 1201 1601 2001 2401 2801
100
150
50
0
200
250
0
Time
401 801 1201 1601 2001 2401 2801
15
In addition, as a central point of the discontinuity theory is the idea of
adaptation, we took into account memory effects, so that R(t) is not only a
function of E(t), where the current time point is t, but also depends on E values
at previous time points:
Nature Reviews | Immunology
R(t) =
R(t) = T
R(t) = (E(t), E(t1), E(t2), ... , E(t))
dE(t)
dt
E(t) =
E(t) E(th)
1

h
T(x) =

1 + e
(x)
E(t) = E(t) E(t1)

i =
i = 0
|E(ti)|
120
100
100
100
100
150
50
0
200
200
300 400 500
130
160
190
220
80
60
40
20
0
0
0
T
(
x
)
R
a
t
i
o

b
o
u
n
d
/
f
r
e
e

i
n
d
o
-
1
Biotin-anti-
NKG2D (CX5)
a
b
c
d
e
Cell activity
Time
1 2 3 4 5 6 7 8 9 10 11 12 13 14
0
Time
401 801 1201
E(t)
R(t)
1601 2001 2401 2801
Streptavidin
100
150
50
0
200
0
Time
401 801 1201 1601 2001 2401 2801
100
150
50
0
200
0
Time
401 801 1201 1601 2001 2401 2801
100
150
50
0
200
250
0
Time
401 801 1201 1601 2001 2401 2801
15
where denotes the memory parameter.
In this first version, we thus propose the following function for R(t):
Nature Reviews | Immunology
R(t) =
R(t) = T
R(t) = (E(t), E(t1), E(t2), ... , E(t))
dE(t)
dt
E(t) =
E(t) E(th)
1

h
T(x) =

1 + e
(x)
E(t) = E(t) E(t1)

i =
i = 0
|E(ti)|
120
100
100
100
100
150
50
0
200
200
300 400 500
130
160
190
220
80
60
40
20
0
0
0
T
(
x
)
R
a
t
i
o

b
o
u
n
d
/
f
r
e
e

i
n
d
o
-
1
Biotin-anti-
NKG2D (CX5)
a
b
c
d
e
Cell activity
Time
1 2 3 4 5 6 7 8 9 10 11 12 13 14
0
Time
401 801 1201
E(t)
R(t)
1601 2001 2401 2801
Streptavidin
100
150
50
0
200
0
Time
401 801 1201 1601 2001 2401 2801
100
150
50
0
200
0
Time
401 801 1201 1601 2001 2401 2801
100
150
50
0
200
250
0
Time
401 801 1201 1601 2001 2401 2801
15
where T(x) is a transfer function that models the response capacity of the cell.
Indeed, most if not all experimental data about immune responses induced by
increasing activation intensities indicate a threshold at lower intensities and a
saturation phenomenon at higher intensities.
We thus suggest using a sigmoidal transfer function given by:
Nature Reviews | Immunology
R(t) =
R(t) = T
R(t) = (E(t), E(t1), E(t2), ... , E(t))
dE(t)
dt
E(t) =
E(t) E(th)
1

h
T(x) =

1 + e
(x)
E(t) = E(t) E(t1)

i =
i = 0
|E(ti)|
120
100
100
100
100
150
50
0
200
200
300 400 500
130
160
190
220
80
60
40
20
0
0
0
T
(
x
)
R
a
t
i
o

b
o
u
n
d
/
f
r
e
e

i
n
d
o
-
1
Biotin-anti-
NKG2D (CX5)
a
b
c
d
e
Cell activity
Time
1 2 3 4 5 6 7 8 9 10 11 12 13 14
0
Time
401 801 1201
E(t)
R(t)
1601 2001 2401 2801
Streptavidin
100
150
50
0
200
0
Time
401 801 1201 1601 2001 2401 2801
100
150
50
0
200
0
Time
401 801 1201 1601 2001 2401 2801
100
150
50
0
200
250
0
Time
401 801 1201 1601 2001 2401 2801
15
where the coefficient will set the amplitude order of the response R(t) while
and will determine the saturation range and the threshold at lower values of
E(t). They will eventually have to be set using experimental data and will
depend on the cell type that is being considered, but it is not of great
importance at this qualitative level. R(t) represents all properties that we
wanted to model. Indeed, R(t) will depend only on E(t) variations: it will be high when these variations are acute and it will return to zero as soon as
E(t) is stable on the previous time points, which will correspond to the adaptation phenomenon.
The transfer function T(x) will model both the activation threshold and the saturation effects of the cell reactivity, and will set the amplitude of the
response through the parameter . In this example, =100, =1 and =5 (see the figure, part a).
Comparison of the model with experimental data is shown (see the figure, part b). The upper panel shows calcium flux changes in freshly isolated
mouse splenic natural killer (NK) cells measured by flow cytometry as previously described
67
. Briefly, biotinylated natural killer group 2 member D
(NKG2D) monoclonal antibodies (clone CX5) were added to the cells followed by streptavidin to induce NKG2D crosslinking. The level of stimulation is
shown as the ratio of indo-1 violet to indo-1 blue. The lower panel shows the immune cell response profile R(t) predicted (at 3000 time points) by the
model for an entry function E(t)=250(1 e
0.3t
), which fits with the experimental data shown in the upper panel. This response profile was obtained with
parameter values: =100, =250, =0.03 and =30. In the lower panel of the figure, part b and in the figure, parts ce, the numbers on the y axis are the
values of R(t) (red line) and E(t) (black line).
Immune cell responses R(t), as a function of a Poissonian input profile [E(t)=te
t
] that could be associated with typical dynamics of external
perturbations, such as infections and any type of injection, are shown (see the figure, part c).
Immune cell responses R(t), as a function of a sigmoid input profile [E(t)=/(1+e
t
)] that could be associated with typical dynamics of endogenous
perturbations (for example, interleukin production) are shown (see the figure, part d).
Immune cell responses R(t), as a function of an oscillating input profile [E(t)=/(1+cos(t))] modelling an intermittent antigen exposure are shown
(see the figure, part e).
PERSPECTI VES
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 767
2013 Macmillan Publishers Limited. All rights reserved
Nature Reviews | Immunology
Time
a Sudden appearance of an unusal antigen
b Initially unusual but persistent antigen
c Slow appearance of an unusual antigen
d Intermittent appearance of an antigen
Antigen
concentration
Immune
response
First, autoimmunity which is a chronic
disorder resulting from the long-lasting
activation of the immune system could be
used as the main counterexample. However,
our framework predicts two possibilities for
the maintenance of chronic autoimmune
disorders: that autoantigens change over the
course of the illness, which resembles the
antigenic drift observed in influenza virus
infection
59
, and/or that the presentation of
autoantigens (that is, their contextual inter-
action with immune effectors) is not in fact
continuous, but rather that it is intermittent
(FIG.2d). This hypothesis obviously remains
to be experimentallytested.
Second, some substantial modifications
in the host for example, those that occur
during gestation or puberty in mammals, or
during metamorphosis in insects or amphib-
ians do not seem to be immunogenic,
which is in contrast with the discontinuity
theory. Our prediction is that the absence
of a destructive immune response in these
circumstances can be explained by the fact
that the immune system only interacts
with small quantities of these modified anti-
gens and interacts with these antigens in a
progressive manner
60
, which involves regu-
latory rather than activating processes, as
documented in the case of metamorphosis
61
.
Finally, a recent study challenges the
rationale for the use of incomplete Freunds
adjuvant (IFA) emulsions for cancer vaccina-
tion. The study found that IFA emulsions
promote antigen persistence at the vaccina-
tion site, thereby inducing Tcells that have
an exhausted phenotype and that eventually
die
62
at these sites rather than promoting
their mobilization to the tumour sites where
they are needed. In addition, repeated oral
administration of antigens can induce toler-
ance and hence desensitize children with egg
allergy
63
. These studies are in remarkable
agreement with the discontinuitytheory.
Conclusions and perspectives
In an attempt to assess the discontinuity
theory, we sought to develop a model that
translates the main concept of this theoretical
framework into a mathematical formalism. As
described above, the fundamental idea
is that an effector immune response is
induced by a discontinuity in the steady state,
which is considered to be the self-referential
environment. Therefore, we generated a
dynamic model with an input variable E(t)
that encompasses all external signals that
are sensed by an immune cell and an output
variable R(t) that quantifies the immune
response intensity that is triggered by the
cell (BOX2). Despite its simplicity, this model
perfectly fits the principles of the disconti-
nuity theory, and presents them in a more
formal, and therefore testable, way. In the
current model, multiple signals are modelled
by a single input variable; although we have
also developed a more complex version with
a multidimensional variable describing dif-
ferent stimuli, we cannot at present assess
its accuracy because of a lack of data and, as
such, we will not discuss it in this article.
Therefore, in this Essay, we have tried
to lay the foundations for a new theory to
explain how the immune system integrates
spatial, qualitative, quantitative and temporal
signals to generate an appropriate immune
response. We suggest that the capacity of the
immune system to perceive sudden antigenic
modifications has been shaped by natural
selection. This means that the immune sys-
tem usually efficiently deals with acute but
not with long-lasting disruptions (such as
chronic infections and cancers). This view
of the immune system gives rise to several
predictions, as shown above through a few
examples. So far, the discontinuity theory can
apply to several biological scales, the cell and
the populations of cells, as well as to the entire
organism. Experimental testing will obviously
inform the exact scope of this theory. More
generally, there is little doubt that this theoreti-
cal framework will need to be amended and
improved; the best contribution we can make is
to invite others to use the formal and empirical
tools proposed here to challengeit.
Thomas Pradeu is at the Universit Paris-Sorbonne,
Philosophy Department, 1 rue Victor Cousin,
75005 Paris, France.
Sbastien Jaeger is at the Centre dImmunologie de
Marseille-Luminy (CIML), Aix Marseille Universit,
UM2, Campus de Luminy, 13288 Marseille, France; at
INSERM, U 1104, 13288 Marseille, France; and at
Centre national de la recherche scientifique,
UMR7280, 13288 Marseille, France.
Eric Vivier is at the Centre dImmunologie de Marseille-
Luminy (CIML), Aix Marseille Universit, UM2, Campus
de Luminy, 13288 Marseille, France; at INSERM, U
1104, 13288 Marseille, France; at Centre national de
la recherche scientifique, UMR7280, 13288 Marseille,
France; and at the Assistance Publique des Hpitaux
de Marseille, Hpital de la Conception,
Marseille 13385, France.
Correspondence to T.P.and E.V.
e-mails: thomas.pradeu@paris-sorbonne.fr;
vivier@ciml.univ-mrs.fl
doi:10.1038./nri3521
Published online 2 September 2013
Figure 2 | The induction of an immune response according to the discontinuity theory. The
discontinuity theory states that the key to the induction of an immune response is antigenic differ-
ence in a time-dependent context. a | If structurally different motifs suddenly appear (that is, there
is a strong quantitative difference with respect to time), then a vigorous immune response occurs,
possibly followed by the generation of memory cells. b | In the case of a motif that is initially unusual
but persists over time, the effector immune response is rapidly extinguished. c | If immune receptors
interact with motifs that change very progressively (that is, there is weak quantitative variation with
respect to time), then the immune response is weak and the motifs become tolerated. d | Finally, if
a structurally different motif appears in an intermittent way, then a very strong and long-lasting
immune response occurs.
PERSPECTI VES
768 | OCTOBER 2013 | VOLUME 13 www.nature.com/reviews/immunol
2013 Macmillan Publishers Limited. All rights reserved
1. Craver,C. F. in Blackwell Guide to the Philosophy of
Science (eds Machamer,P.K. and Silberstein,M.),
5579 (Blackwell, 2001).
2. Greenspan,N.S. Conceptualizing immune
responsiveness. Nature Immunol. 8, 57 (2007).
3. Vivier,E. etal. Innate or adaptive immunity? The
example of natural killer cells. Science 331, 4449
(2011).
4. Orr,M.T. & Lanier,L.L. Natural killer cell education
and tolerance. Cell 142, 847856 (2010).
5. Gasser,S. & Raulet,D.H. Activation and self-tolerance
of natural killer cells. Immunol. Rev. 214, 130142
(2006).
6. Krre,K., Ljunggren,H.G., Piontek,G. & Kiessling,R.
Selective rejection of H-2-deficient lymphoma variants
suggests alternative immune defence strategy.
Nature 319, 675678 (1986).
7. Hglund,P. etal. Recognition of 2-microglobulin-
negative (2m-) T-cell blasts by natural killer cells
from normal but not from 2m- mice:
nonresponsiveness controlled by 2m- bone marrow
in chimeric mice. Proc. Natl Acad. Sci. USA 88,
1033210336 (1991).
8. Liao,N.S., Bix,M., Zijlstra,M., Jaenisch,R. &
Raulet,D. MHC classI deficiency: susceptibility to
natural killer (NK) cells and impaired NK activity.
Science 253, 199202 (1991).
9. Joncker,N.T., Shifrin,N., Delebecque,F. &
Raulet,D.H. Mature natural killer cells reset their
responsiveness when exposed to an altered MHC
environment. J.Exp. Med. 207, 20652072
(2010).
10. Elliott,J.M., Wahle,J.A. & Yokoyama,W.M. MHC
classI-deficient natural killer cells acquire a licensed
phenotype after transfer into an MHC classI-sufficient
environment. J.Exp. Med. 207, 20732079 (2010).
11. Savill,J., Dransfield,I., Gregory,C. & Haslett,C. A
blast from the past: clearance of apoptotic cells
regulates immune responses. Nature Rev. Immunol. 2,
965975 (2002).
12. Stuart,L.M. & Ezekowitz,R.A. Phagocytosis and
comparative innate immunity: learning on the fly.
Nature Rev. Immunol. 8, 131141 (2008).
13. Green,D.R., Ferguson,T., Zitvogel,L. & Kroemer,G.
Immunogenic and tolerogenic cell death. Nature Rev.
Immunol. 9, 353363 (2009).
14. Jeannin,P., Jaillon,S. & Delneste,Y. Pattern recognition
receptors in the immune response against dying cells.
Curr. Opin. Immunol. 20, 530537 (2008).
15. Ahrens,S. etal. F-actin is an evolutionarily conserved
damage-associated molecular pattern recognized by
DNGR-1, a receptor for dead cells. Immunity 36,
635645 (2012).
16. Foster,S.L., Hargreaves,D.C. & Medzhitov,R.
Gene-specific control of inflammation by TLR-induced
chromatin modifications. Nature 447, 972978
(2007).
17. Wang,H. etal. Lack of CD47 on nonhematopoietic cells
induces split macrophage tolerance to CD47
null
cells.
Proc. Natl Acad. Sci. USA 104, 1374413749 (2007).
18. Ramsdell,F. & Fowlkes,B.J. Maintenance of invivo
tolerance by persistence of antigen. Science 257,
11301134 (1992).
19. Goodnow,C.C. etal. Altered immunoglobulin
expression and functional silencing of self-reactive B
lymphocytes in transgenic mice. Nature 334, 676682
(1988).
20. Virgin,H.W., Wherry,E.J. & Ahmed,R. Redefining
chronic viral infection. Cell 138, 3050 (2009).
21. Utzschneider,D.T. etal. T cells maintain an exhausted
phenotype after antigen withdrawal and population
reexpansion. Nature Immunol. 14, 603610 (2013).
22. Turchinovich,G. & Hayday,A.C. Skint-1 identifies a
common molecular mechanism for the development of
interferon--secreting versus interleukin-17-secreting
T cells. Immunity 35, 5968 (2011).
23. Grossman,Z. & Paul,W.E. Adaptive cellular
interactions in the immune system: the tunable
activation threshold and the significance of
subthreshold responses. Proc. Natl Acad. Sci. USA 89,
1036510369 (1992).
24. Grossman,Z., Min,B., Meier-Schellersheim,M. &
Paul,W.E. Concomitant regulation of T-cell activation
and homeostasis. Nature Rev. Immunol. 4, 387395
(2004).
25. Schenten,D. & Medzhitov,R. The control of adaptive
immune responses by the innate immune system.
Adv.Immunol. 109, 87124 (2011).
26. Shankaran,V. etal. IFN and lymphocytes prevent
primary tumour development and shape tumour
immunogenicity. Nature 410, 11071111 (2001).
27. Galon,J. etal. Type, density, and location of immune
cells within human colorectal tumors predict clinical
outcome. Science 313, 19601964 (2006).
28. Smyth,M.J., Dunn,G.P. & Schreiber,R.D. in
Advances in Immunology (ed. Allison, J. P.) 150
(Elsevier, 2006).
29. Raulet,D.H. & Guerra,N. Oncogenic stress sensed
by the immune system: role of natural killer cell
receptors, Nature Rev. Immunol. 9, 568580 (2009).
30. Kraman,M. etal. Suppression of antitumor immunity
by stromal cells expressing fibroblast activation
protein-. Science 330, 827830 (2010).
31. Pardoll,D. Does the immune system see tumors as
foreign or self? Annu. Rev. Immunol. 21, 807839
(2003).
32. den Boer,A.T. etal. The tumoricidal activity of
memory CD8
+
T cells is hampered by persistent
systemic antigen, but full functional capacity is
regained in an antigen-free environment.
J.Immunol. 172, 60746079 (2004).
33. Drake,C.G., Jaffee,E. & Pardoll,D.M. Mechanisms of
immune evasion by tumors. Adv. Immunol. 90, 5181
(2006).
34. Pradeu,T. & Carosella,E.D. On the definition of a
criterion of immunogenicity. Proc. Natl Acad. Sci. USA
103, 1785817861 (2006).
35. Burnet,F.M. Cancer: a biological approach.
Br.Med.J. 1, 17 (1957).
36. Thomas,L. in Cellular and humoral aspects of the
hypersensitive states (Lawrence, 1959).
37. Burnet,F.M. Immunological surveillance (Pergamon,
1970).
38. Zinkernagel,R.M. & Doherty,P.C. Immunological
surveillance against altered self components by
sensitised T lymphocytes in lymphocytes
choriomeningitis. Nature 251, 547548 (1974).
39. Houghton,A.N. Cancer antigens: immune recognition
of self and altered self. J.Exp. Med. 180, 14 (1994).
40. Laughlin,S.B. The role of sensory adaptation in
the retina. J.Exp. Biol. 146, 3962 (1989).
41. Wark,B., Lundstrom,B.N. & Fairhall,A. Sensory
adaptation. Curr. Opin. Neurobiol. 17, 423429
(2007).
42. Bray,D., Levin,M.D. & Morton-Firth,C.J.
Receptor clustering as a cellular mechanism to
control sensitivity. Nature 393, 8588 (1998).
43. Henrickson,S.E. etal. T cell sensing of antigen dose
governs interactive behavior with dendritic cells and
sets a threshold for T cell activation. Nature Immunol.
9, 282291 (2008).
44. Apostolou,I. & von Boehmer,H. Invivo instruction of
suppressor commitment in naive T cells. J.Exp. Med.
199, 14011408 (2004).
45. Kyes,S., Horrocks,P. & Newbold,C. Antigenic
variation at the infected red cell surface in malaria.
Annu. Rev. Microbiol. 55, 673707 (2001).
46. MacGregor,P., Szor,B., Savill,N.J. &
Matthews,K.R. Trypanosomal immune evasion,
chronicity and transmission: an elegant balancing act.
Nature Rev. Micro 10, 431438 (2012).
47. Ndungu,T. & Weiss,R.A. On HIV diversity. AIDS 26,
12551260 (2012).
48. Csiksz-Nagy,A. & Soyer,O.S. Adaptive dynamics
with a single two-state protein. J.R.Soc. Interface
5, S41S47 (2008).
49. Daron,M., Jaeger,S., Du Pasquier,L. & Vivier,E.
Immunoreceptor tyrosine-based inhibition motifs:
a quest in the past and future. Immunol. Rev. 224,
1143 (2008).
50. Hori,S., Nomura,T. & Sakaguchi,S. Control of
regulatory Tcell development by the transcription
factor Foxp3. Science 299, 10571061 (2003).
51. Fontenot,J.D., Gavin,M.A. & Rudensky,A.Y.
Foxp3 programs the development and function of
CD4
+
CD25
+
regulatory T cells. Nature Immunol.
4, 330336 (2003).
52. Khattri,R., Cox,T., Yasayko,S.-A. & Ramsdell,F.
An essential role for Scurfin in CD4
+
CD25
+
T
regulatory cells. Nature Immunol. 4, 337342
(2003).
53. Kuhn,T. S. The structure of scientific revolutions
(Univ. of Chicago Press, 1962).
54. Matzinger,P. The danger model: a renewed sense of
self. Science 296, 301305 (2002).
55. Hooper,L.V., Littman,D.R. & Macpherson,A.J.
Interactions between the microbiota and the immune
system. Science 336, 12681273 (2012).
56. Lathrop,S.K. etal. Peripheral education of the
immune system by colonic commensal microbiota.
Nature 478, 250254 (2011).
57. Belkaid,Y. & Oldenhove,G. Tuning
microenvironments: induction of regulatory T cells
by dendritic cells. Immunity 29, 362371 (2008).
58. Shklovskaya,E. etal. Langerhans cells are
precommitted to immune tolerance induction.
Proc. Natl Acad. Sci. USA 108, 1804918054
(2011).
59. Carrat,F. & Flahault,A. Influenza vaccine: The
challenge of antigenic drift. Vaccine 25, 68526862
(2007).
60. Pradeu,T. The limits of the self: immunology and
biological identity 131183 (Oxford Univ. Press, 2012).
61. Du Pasquier,L. & Bernard,C.C.A. Active suppression
of the allogeneic histocompatibility reactions during
the metamorphosis of the clawed toad Xenopus.
Differentiation 16, 131183 (1980).
62. Hailemichael,Y. etal. Persistent antigen at vaccination
sites induces tumor-specific CD8
+
T cell sequestration,
dysfunction and deletion. Nature Med. 19, 465472
(2013).
63. Burks,A.W. etal. Oral immunotherapy for treatment
of egg allergy in children. N.Engl. J.Med. 367,
233243 (2012).
64. Bernard,C. Experimental medicine. (original text
published in 1865). (Transaction, 1999).
65. Popper,K.R. Logik der forschung: zur
erkenntnistheorie der modernen naturwissenschaft.
(J. Springer, 1935).
66. National Research Council. The role of theory in
advancing 21st-century biology: catalyzing
transformative research. [online], http://www.nap.edu/
openbook.php?record_id=12026) (The National
Academies, 2008).
67. Guia, S. et al. Confinement of activating receptors at
the plasma membrane controls natural killer cell
tolerance. Sci. Signal. 4, ra21 (2011).
Acknowledgements
We thank H. Brailly, L. Du Pasquier, J. Ewbank, M. Fougereau,
P. Kourilsky, L. Pri, B. Malissen and A. Trautmann for their
constructive comments, S. Guia and S. Ugolini for their exper-
imental data on calcium flux, and C. Chapple for his editorial
help. E.V.s laboratory is supported by the European Research
Council (THINK Advanced Grant) and by institutional grants
from Institut National de la Sant et de la Recherche
Mdicale, Le Centre national de la recherche scientifique and
Aix Marseille to Centre dImmunologie de Marseille-Luminy.
T.P. and E.V. are scholars of the Institut Universitaire de
France.
Competing interests statement
The authors declare competing financial interests: see Web
version for details.
PERSPECTI VES
NATURE REVIEWS | IMMUNOLOGY VOLUME 13 | OCTOBER 2013 | 769
2013 Macmillan Publishers Limited. All rights reserved